Login Register

Access the GIFSQ Portal

Select your user type to log in or register a new account.

Student Portal

Access your food safety courses and exams.

Instructor Portal

Manage courses, view student submissions, and grade quizzes.

Company Portal

Manage corporate setup, view employee logs, and access QA services.

GIFSQ ID: FTN-0092

← All subjects

Robotics and Automation in Food Processing

Unit 1: Why Robots in Food Processing

Note

Unit 1: Why Robots in Food Processing

A question to start with

A biscuit factory runs three shifts a day. On the packing line, workers place biscuits into trays, tray after tray, hour after hour. The biscuits are warm, soft, and easy to break. Some trays leave the line with a missing biscuit. Some leave with a cracked one. The line stops when a worker needs a break, slows down near the end of a long shift, and stops again when a new worker needs training. One morning the manager asks a plain question: what would change if a machine did the picking and placing, and people did the checking, the fixing, and the thinking? This unit is about that question. It looks at why food factories use robots, what robots do well, what they do badly, and how people and machines share the work.

A robot in a food factory is a machine that can be programmed to move, pick up, place, cut, sort, or pack food, and to repeat the same movement again and again. Automation is the wider idea: machines, sensors, and controls working together so that a process runs with less direct hand work at each step. A robot is one part of automation. A conveyor, a sensor that weighs a pack, a camera that checks a label, and a computer that records the result are other parts. Together they form a system, and the system is only as good as its weakest part.

Food is not like metal or plastic. It is soft, wet, sticky, uneven, and it changes with temperature and time. A biscuit fresh from the oven is not the same as a biscuit that has cooled for an hour. A tomato is not the same shape as the tomato beside it. A piece of dough rises, spreads, and sticks. This is why robots in food processing are interesting: they must handle living, changing material with care, speed, and cleanliness, and they must do it in a place that is washed down every day.

Questions to think about

  • If a robot can place biscuits faster than a person, why do many food factories still use people on the packing line?
  • A food is soft, sticky, and different in every piece. What does that demand from a machine that has to pick it up?
  • When a line stops, who loses more: the factory that owns a robot, or the factory that relies on hand work alone?
Think about the biscuit line before you read on

The robot wins on speed and repeatability, but only when the biscuits arrive in a steady stream, in a known position, and in a known condition. If the biscuits are broken, stuck together, or spread across the belt in a random way, the robot needs help from sensors, guides, and people. Hand work wins on judgement: a person sees a cracked biscuit at once, feels a sticky dough, and adjusts without being told. The best lines use both, with the robot doing the fast repeatable move and the person doing the watching, checking, and fixing.

What a robot does on a food line

A robot on a food line does four kinds of work, and most food robots do at least one of them. First, it moves food: it picks a piece from a belt and places it in a tray, a box, or a mould. Second, it processes food: it cuts, slices, portions, decorates, or assembles. Third, it packs food: it fills a pouch, closes a carton, stacks boxes on a pallet, or wraps a pallet for storage. Fourth, it checks food: with a camera or a sensor, it looks for a missing piece, a wrong colour, a broken seal, or a wrong label, and it rejects the pack that fails the check.

Each of these jobs has a simple test: does the machine do it faster, more evenly, or more safely than the hand method it replaces? Speed matters because food lines run fast, and a slow step holds up every step after it. Evenness matters because buyers expect every pack to look the same and weigh the same. Safety matters because food must stay clean, and because some jobs are cold, hot, wet, heavy, or repetitive in ways that tire people and cause injury over time.

A robot also brings a record. Every pick, every reject, every stop can be counted. That count tells the manager where the line loses time, which product breaks most often, and when a tool needs cleaning or changing. Hand work can give the same information, but only if someone writes it down at the time, and on a fast line that rarely happens in full.

Quick self-check

Question: A robot stacks boxes on a pallet all day without a break. What is the main gain, and what is the main risk?

The main gain is steady speed and no fatigue: every box lands in the same place, and the pace does not fall at the end of a shift. The main risk is that a fault repeats itself: if the first box is in the wrong place, every box after it may be wrong too, until someone notices and stops the line.

Question: Why does a robot need a sensor as well as an arm?

An arm alone can only repeat a fixed movement. Food arrives in different places and in different shapes, so the robot needs a sensor, such as a camera, to find each piece before it picks it up. Without sensing, the robot picks at empty air or crushes the piece it was meant to save.

Why food factories started using robots

Food factories did not adopt robots because robots are new or impressive. They adopted them because of five pressures that show up in almost every food plant.

The first pressure is speed and volume. Demand rises and falls, and a line that must pack a large number of units in a short time needs a pace that hands alone cannot hold for a whole shift. A robot holds its pace from the first minute to the last.

The second pressure is consistency. A buyer who opens a pack of biscuits, a tray of ready meals, or a bag of cut vegetables expects the same weight, the same count, and the same look every time. A robot that is set up well gives that evenness, pack after pack.

The third pressure is hygiene. Hands touch food, and every touch is a chance for contamination. A robot tool that is designed for food contact, cleaned on schedule, and kept away from the floor reduces the number of touches between people and the open product. In cold rooms, in wet areas, and near raw meat or fish, fewer touches also means fewer chances for microbes to move from one surface to another.

The fourth pressure is worker safety and comfort. Some food jobs are hard on the body: lifting heavy sacks, stacking heavy boxes, working in a freezer, standing in steam, or repeating the same wrist movement thousands of times a day. Moving those jobs to a machine protects people from strain and from cold, heat, and wet work, and it frees people for jobs that need judgement.

The fifth pressure is labour supply. Food factories often run early mornings, nights, weekends, and seasonal peaks, when workers are hard to find and keep. A robot does not solve a labour shortage by itself, but it lets a smaller team run a line, and it keeps the line running when a team member is absent.

Five pressures that push a food factory toward robots

1. Speed and volume: more packs in less time
↓
2. Consistency: every pack the same weight and look
↓
3. Hygiene: fewer hand touches on open food
↓
4. Worker safety: less lifting, cold, steam, and repetition
↓
5. Labour supply: a line that runs with a smaller team

The diagram shows the five pressures in order. No single pressure decides the case for a robot on its own. A good decision weighs all five against the cost, the food, and the skill of the team that will run the machine.

Quick self-check

Question: A small bakery makes a different bread each day in small batches. Which of the five pressures is weakest there, and why?

Speed and volume is weakest, because the batches are small and the product changes daily. A robot set up for one bread would need re-setting for the next, and the time spent changing over could eat up the time saved. Consistency and hygiene still matter, but the case for a robot is weaker than in a large plant making one product all day.

Question: How can a robot improve hygiene if the robot itself must be cleaned?

It improves hygiene by cutting the number of hand touches on open food and by using smooth, washable surfaces that are cleaned on a set schedule. The gain is lost if cleaning is skipped, so the cleaning plan is part of the robot plan, not an afterthought.

What robots do well, and what they do badly

Robots are strong at repeatable movement. They move to the same point, with the same force, at the same speed, thousands of times. They do not get tired, bored, or distracted. They work in cold rooms, in steam, and in places where dust or noise makes hand work unpleasant. They lift heavy loads without strain, and they keep a steady pace through a long shift.

Robots are weak at judgement. A person sees a biscuit that is cracked, a tomato that is bruised, a pack that is torn, and acts at once. A robot sees only what its sensor is set to see. If the fault is new, or the sensor is dirty, or the light has changed, the robot may miss it or reject good food. Robots are also weak at handling change: a new product shape, a new pack size, or a softer batch of dough can need a new tool, a new program, or a new setting. That change takes time, skill, and testing.

Robots are weak at soft, variable food. A hard metal part has a fixed shape and a fixed size. A piece of bread, a fillet of fish, a soft cheese, or a ripe fruit changes shape when it is gripped, and it changes again when it warms up or dries out. Gripping too hard bruises or breaks it. Gripping too softly drops it. The gripper, the tool at the end of the arm that holds the food, is often the hardest part of the whole system to get right, and it is where most food robot projects succeed or fail.

Robots also need a steady supply. A picking robot that waits for food wastes its speed. A packing robot that waits for trays wastes its speed. The belt, the feeder, the tray supply, and the product flow must all match the robot, or the robot stands idle while the line starves.

Quick self-check

Question: Why is a gripper often harder to design than the robot arm?

The arm is a standard machine with known movement. The gripper must touch food that is soft, wet, sticky, and different in every piece, and it must hold the piece firmly without bruising, tearing, or dropping it. That contact is where the food meets the machine, so every property of the food matters there.

Question: A robot rejects good packs because the light in the room changed in the afternoon. What does that tell you?

It tells you the inspection depends on light, and the light was not controlled. The fix is not a better robot: it is steady lighting, a clean camera lens, and a check that is tested under morning and afternoon light before the line is accepted.

People and robots: sharing the work

The best food lines do not replace people. They move people to the work that people do best. People watch, judge, taste, smell, fix, clean, set up, and improve. Robots lift, repeat, and hold a pace. When the split is right, the line is faster and the work is better for the people on it.

This split shows up in job roles. An operator starts the line, watches the flow, clears a jam, and calls for help when the fault is beyond a simple fix. A setter changes the tool and the program when the product changes, and tests the first packs before the line runs at full speed. A maintenance worker cleans, lubricates, checks wear, and replaces parts before they fail. A quality checker samples packs, tastes, measures, and records results that a camera cannot judge. None of these jobs disappears when a robot arrives. They change, and they need training.

Training matters more than the machine. A team that does not understand the robot will stop it at the first fault and wait for help, and the line will lose more time than it gains. A team that understands the controls, the cleaning routine, the common faults, and the safe way to enter the robot area will keep the line running and will spot a small problem before it becomes a stop. Every robot project should include training time, practice time, and simple written steps for starting, stopping, cleaning, and clearing a jam.

There is also a human question that a technical plan should not ignore: people worry about their jobs when a robot arrives. A factory that explains what will change, which jobs will move, and what new skills will be taught keeps trust. A factory that installs a machine in silence loses it. The unit on costs and payback returns to this point, because the value of a robot depends on the people who run it staying with the plant and learning to run it well.

Quick self-check

Question: A line runs faster with a robot, but stops longer at each fault because nobody on the shift knows how to clear it. Is the robot a gain?

Not yet. Speed only helps when the line is running. If every fault waits for one expert, the lost time can cancel the speed gain. Training the shift team to clear common faults is part of making the robot pay.

Where robots fit, and where they do not

Robots fit best where the product is steady, the volume is high, the movement repeats, and the environment is hard for people. Palletising boxes, packing identical trays, placing identical biscuits, sorting by size, and sealing cartons are classic fits. The product is known, the target position is known, and the same movement repeats all day.

Robots fit less well where the product changes every hour, the batch is tiny, the food is very soft or very variable, or the job needs touch, smell, or taste to judge quality. A kitchen that makes a different dish each day, a plant that runs short trial batches, or a line that handles very ripe fruit may gain more from better hand tools, better layout, and better training than from a robot.

Between those two ends sits the middle ground where most food plants live: a mix of steady products and changing products, of high volume and short runs. In that middle ground the right question is not "robot or people" but "which step, and how much". One robot at the end of the line, stacking boxes, may pay back quickly while the packing of the food itself stays by hand. One vision system that rejects broken packs may pay back while the filling stays manual. Starting with one step, learning from it, and then adding the next step is safer than automating a whole line at once.

Where a robot fits best

Best fit: steady product, high volume, same movement, hard environment
↓
Middle ground: automate one step, learn, then add the next
↓
Weak fit: tiny batches, changing product, very soft food, judgement by touch or taste

Quick self-check

Question: A plant packs one soup in one size all year. Another plant packs twenty soups in small batches. Which plant should automate first?

The first plant, because the product is steady and the volume is high, so the set-up cost is spread over many identical packs. The second plant would spend much of its time changing over, and may gain more from flexible hand packing with good guides and checks.

A simple way to judge a robot idea

Before buying anything, a food technologist can test a robot idea with five plain questions. First, what exact job will the robot do, and what will it not do? Second, what is the food like at that step: its size, shape, softness, stickiness, temperature, and how much it varies? Third, how will the food arrive: in a steady stream, in a known position, or scattered at random? Fourth, who will run it, clean it, fix it, and change it over, and what training will they need? Fifth, what happens when it stops: how long does a stop last, and who can restart it?

If any of those questions has no answer, the project is not ready. A supplier can answer the machine questions, but only the plant can answer the food questions, because only the plant knows its product, its line, and its people. That is why a trial matters: run the real food, at real speed, in real conditions, and count the good packs, the damaged packs, and the stops. A trial with the real product is worth more than a video of a different product running perfectly somewhere else.

The rest of this subject builds on this unit. Unit 2 names the types of robots and what each one does. Unit 3 looks closely at handling, picking, and packing, where the gripper meets the food. Unit 4 follows a whole processing line from raw material to pack. Unit 5 explains machine vision and inspection. Unit 6 covers safety, hygiene, and working beside robots. Unit 7 brings the money together: costs, payback, and choosing the right automation.

Try it yourself

Visit a food shop or watch a food production video and pick one packed food, such as a tray of biscuits, a bag of salad, or a box of cereal. Write down the five judgement questions from the section above and answer each one for that product in two or three sentences. Then decide: would a robot fit the packing step, the sorting step, or neither? Give one reason for your choice that comes from the food itself, not from the machine. Compare your answer with a classmate and note where your views of the same food differ.

Key ideas to carry forward

  • A robot is a programmable machine that moves, picks, places, cuts, sorts, or packs; automation is the wider system of machines, sensors, and controls working together.
  • Food is soft, wet, sticky, uneven, and changing, so food robots must handle variable material with care, speed, and cleanliness.
  • Robots are pushed by speed and volume, consistency, hygiene, worker safety, and labour supply.
  • Robots are strong at repeatable movement and steady pace, and weak at judgement, at handling change, and at gripping soft variable food.
  • The gripper is often the hardest part: it must hold food firmly without bruising, tearing, or dropping it.
  • People move to watching, judging, fixing, cleaning, setting up, and improving when a robot takes the repetitive move.
  • Robots fit best with a steady product, high volume, and a repeating movement; they fit least with tiny batches, constant change, and judgement by touch or taste.
  • Judge any robot idea by the job, the food, the way food arrives, the people who will run it, and what happens when it stops.
  • A trial with the real food at real speed counts more than a video of a different product.

Glossary

Automation

Machines, sensors, and controls working together so a process runs with less direct hand work at each step. Food tech example: a line where a conveyor, a filler, a sealer, and a checkweigher run together is automation, even if no robot arm is present.

Checkweigher

A machine that weighs each pack on a moving belt and rejects packs that are too light or too heavy. Food tech example: a checkweigher after a biscuit packer rejects packs missing a biscuit.

Consistency

How alike each pack or piece is in weight, count, size, and look. Food tech example: biscuits from a well-set line have the same count and the same weight in every pack.

Contamination

Unwanted material or microbes getting into food. Food tech example: a hand touching open salad after touching a dirty surface can cause contamination.

End effector

The tool at the end of a robot arm that does the work, such as a gripper, a suction cup, or a cutter. Food tech example: a suction cup end effector lifts a flat cracker without squeezing its edges.

Gripper

The part of a robot that holds food while it is moved. Food tech example: a soft finger gripper holds a bread roll without squashing it.

Hygiene

Practices that keep food clean and free from contamination. Food tech example: washing hands, cleaning tools on schedule, and keeping raw and ready-to-eat food apart are hygiene practices.

Operator

A person who runs a machine or a line, watches it, and deals with simple faults. Food tech example: the operator starts the packing line, watches the trays fill, and clears a jammed tray.

Palletising

Stacking boxes or bags in a set pattern on a pallet for storage or transport. Food tech example: a robot palletiser stacks cartons of canned soup in layers on a wooden pallet.

Programmable

Able to be given a set of instructions that can be changed for a new task. Food tech example: a robot can be reprogrammed to place biscuits in a tray of six instead of a tray of twelve.

Repeatability

The ability of a machine to return to the same position and repeat the same movement again and again. Food tech example: a robot with good repeatability places every jar lid in the same spot on every jar.

Sensor

A device that detects something, such as presence, weight, temperature, colour, or position, and sends a signal. Food tech example: a sensor on a belt tells the robot that a tray has arrived and is ready to be filled.

Setter

A worker who changes a machine over for a new product, fits a new tool, and tests the first packs. Food tech example: the setter fits a new gripper and tests it on the new biscuit size before the line runs at full speed.

Throughput

The amount of product that passes through a line or a machine in a given time. Food tech example: a packing line with a throughput of 60 packs a minute fills 60 packs every minute when it is running well.

Trial

A test run of a machine with the real product, at real speed, in real conditions, before a purchase is decided. Food tech example: a trial runs the plant's own soft cheese through a picking robot to see how many pieces are damaged.

References

  • Irudayaraj, J., Food Processing Operations Modeling, Design, and Analysis, 2nd edition, 2002.
  • Singh, R. P., Computer Applications in Food Technology, 1996.

Unit 2: Types of Robots and What Each One Does

Note

Unit 2: Types of Robots and What Each One Does

A question to start with

A food plant needs three jobs done. One job is to lift heavy sacks of flour from a pallet and tip them into a mixer. A second job is to pick single strawberries from a fast belt and place each one gently into a punnet, without bruising the fruit. A third job is to move sealed cartons along a line, stack them on a pallet, and wrap the pallet. One robot cannot do all three jobs well. The sack job needs strength. The strawberry job needs speed and a gentle touch. The carton job needs a long reach and a steady pattern. This unit names the main types of robots used in food processing, explains how each one moves, and shows which job each type fits.

A robot is named by how it moves, not by what it looks like. The main parts are the same in most types: a base that fixes the robot to the floor or to a frame, joints that bend or turn, links that join the joints, a wrist at the end, and an end effector that does the work, such as a gripper, a suction cup, a cutter, or a spray head. A controller is the computer that tells the joints when and how far to move. A program is the set of instructions the controller follows. The number of axes is the number of directions or joints the robot can move in: more axes usually mean more freedom to reach around an object, and also more to program and to keep safe.

The shape of the work space matters as much as the shape of the robot. The work envelope is the space the robot can reach. A robot with a long reach can serve a wide belt or a tall pallet, but it also needs a larger guarded area. A robot with a short reach is faster in a small space but cannot serve a wide line. Choosing a robot starts with the space, the load, the speed, and the food, not with a catalogue picture.

Questions to think about

  • Why would a plant choose a small fast robot for strawberries and a large strong robot for flour sacks, when both jobs are "pick and place"?
  • A robot has six axes. What does that freedom give it, and what does it cost in programming and safety?
  • If a robot can reach across a whole line, why is a long reach not always better?
Think about the three jobs before you read on

The sack job is about load and lift: the robot must carry weight without shaking, so it needs a strong arm and a firm grip. The strawberry job is about speed and touch: the robot must move fast, see each berry, and grip it softly, so it needs a light arm, a camera, and a soft gripper. The carton job is about reach and pattern: the robot must place each carton in a set place on a pallet, so it needs a steady arm with a known pattern. One arm cannot be best at all three, because strength, speed, and reach pull the design in different directions.

Articulated robots: the familiar arm

An articulated robot is the arm most people picture when they hear the word robot. It has a base, a shoulder, an elbow, and a wrist, joined in a chain, usually with six axes. It moves like a human arm, reaching up, down, sideways, and around an object. It can tilt a tool, turn a pack over, or reach into a box from above and from the side.

In food plants, articulated robots do heavy lifting, palletising, loading mixers and ovens, handling trays, and placing packs into cartons. They are strong, they have a long reach, and they can work in a large envelope. Their weakness is speed over short moves: a big arm takes time to stop and start, and a big arm moving fast needs a large guarded area, because its sweep is wide. They are also harder to clean in place, because the arm has many joints and folds where food, flour, and water can collect. For that reason, food-grade articulated robots are built with smooth covers, sealed joints, and washable surfaces, and they are placed so that the cleaning team can reach around them.

An articulated robot is a good choice when the load is heavy, the reach is long, the movement needs tilting and turning, and the same job repeats for a long run. It is a poor choice when the job is a very fast short pick from a moving belt, where a lighter robot would win.

Quick self-check

Question: Why does an articulated robot need a larger guarded area than a small fast robot?

Because its arm is long and its sweep is wide. When the arm turns, the elbow and the load travel through a large space, and a person standing in that space could be struck. The guard must cover the whole envelope, not only the place where the gripper works.

Question: When would you pick an articulated robot over a lighter type?

When the load is heavy, the reach is long, or the tool must tilt and turn, such as tipping a sack into a mixer or stacking cartons on a tall pallet. Strength and reach matter more there than short-move speed.

SCARA robots: fast, flat, and precise

A SCARA robot has a rigid vertical axis and two horizontal arms that move in a flat plane, like a human arm moving across a table. SCARA stands for a selective compliance arm: it is stiff in the vertical direction and gives a little in the horizontal direction, which helps when a part must be pushed into a place that is not exactly aligned. Most SCARA robots have four axes.

SCARA robots are fast and precise over short moves in a flat plane. In food plants they place lids, insert trays, pick small packs from a belt, assemble cartons, and do fast pick and place where the target is at a known height. They are easy to program for a flat task, and they take up less floor space than a large articulated arm. Their weakness is reach in the vertical direction and the ability to tilt: a SCARA robot does not reach around an object or turn it over in the way an articulated arm can. It works best when the food or pack sits flat, at a known height, and the move is mostly sideways.

Quick self-check

Question: A line needs to place a lid on a tub that sits flat on a belt. Why might a SCARA robot fit?

Because the move is flat, short, and precise, and the height is known. The SCARA arm moves fast across the belt, lowers the lid straight down, and repeats the same move. It does not need to tilt or reach around the tub.

Question: Why is a SCARA robot a poor choice for stacking cartons on a tall pallet?

Because the pallet is tall and the cartons must be placed at many heights and in a pattern that needs reach and tilting. A SCARA robot works in a flat plane and cannot serve a tall stack well.

Delta robots: speed on a moving belt

A delta robot, also called a parallel robot or a spider robot, hangs from a frame above the belt. Three arms work in parallel, joined to a small moving platform that carries the gripper. It moves very fast in short up-and-down and sideways moves, and it is often used with a camera that finds each piece on a moving belt. Delta robots are the fastest pickers in food plants, and they are common in biscuit, confectionery, bakery, and ready-meal packing.

A delta robot is light, so it can accelerate and stop quickly without shaking the frame. Its work envelope is a dome shape under the frame, wide at the bottom and narrow at the top. It is built for light loads: single biscuits, chocolates, cheese portions, pieces of fruit, or small packs. It is not built for heavy sacks or large cartons. Its gripper is usually a suction cup or a soft finger that can pick one piece at a time from a moving belt and place it in a tray that is also moving. The belt, the camera, and the robot must be timed together, so that the robot knows where the piece will be when the gripper arrives.

Hygiene is a strong point for delta robots in food use. They hang above the product, they have few horizontal surfaces where water can sit, and food-grade versions are built in stainless steel with smooth surfaces that wash down well. Their weakness is load and reach: they cannot lift heavy items, and they cannot reach far outside their dome.

How a delta robot picks from a moving belt

Pieces move along the belt in a random pattern
↓
Camera finds each piece and sends its position
↓
Delta robot moves to the piece and picks it up
↓
Robot places the piece in a tray or pack
↓
Tray moves on, and the robot picks the next piece

Quick self-check

Question: Why does a delta robot need a camera, while a palletising robot often does not?

Because the pieces on a fast belt arrive in random places, so the robot must find each one before it picks it. A palletising robot places cartons in a fixed pattern in known places, so it can repeat a set program without searching for each carton.

Question: What happens to a delta robot line if the belt speed rises but the camera is not re-timed?

The robot picks at the place where the piece was, not where it is now. Picks miss, pieces are dropped or crushed, and good food is wasted. Belt speed, camera timing, and robot speed must match.

Cartesian and gantry robots: straight lines and large areas

A Cartesian robot moves along straight lines, usually on three axes: left and right, forward and back, up and down. A gantry robot is a large Cartesian robot mounted on a frame that spans a wide area, such as a whole palletising zone or a storage area. The tool hangs from the frame and moves in straight lines to any point in the area.

Cartesian and gantry robots are simple to program, because the movement is in straight lines and the position is easy to describe. They are strong and stable, and they can serve a large area without a wide swinging arm. In food plants they load and unload trays in ovens and freezers, move pallets, handle large boxes, and work in cold stores where a simple straight-line move is safer than a swinging arm. Their weakness is speed over short moves and flexibility: they do not tilt and turn like an articulated arm, and a large gantry frame takes up space and needs a clear area under it.

A gantry robot is a good choice when the work area is wide and flat, the loads are heavy, and the movement is mostly straight lines. It is a poor choice when the tool must tilt, turn a pack over, or work in a tight space around other machines.

Quick self-check

Question: Why is a gantry robot simple to program?

Because it moves in straight lines along three axes, so each position is a set of three numbers: how far along, how far across, and how high. There is no elbow or wrist angle to work out, so the program is easier to write and to check.

Collaborative robots: working beside people

A collaborative robot, often called a cobot, is designed to work in the same space as people, without a full guard or fence, for certain tasks. It has sensors that detect contact, it moves more slowly, and it stops or gives way when it touches a person or an unexpected object. It is usually smaller and lighter than a large industrial robot, and it is often used for a single simple job, such as placing packs in a carton, holding a tray, or feeding a machine.

Cobots suit small plants and short runs, because they are easy to move, easy to set up, and easy to teach by guiding the arm through the movement by hand. In food plants they help with packing, labelling, and handling light trays, and they let a person and a robot share a bench: the robot does the repeat move, the person does the check and the changeover. Their limits are load and speed: a cobot that is safe to touch cannot lift a heavy sack or move at the speed of a delta robot. Safety also depends on the tool and the load, not only on the robot: a sharp cutter or a hot tray on a cobot is still a hazard, and the risk must be checked for the whole job, not only for the arm.

A cobot is not a cheaper version of an industrial robot. It is a different tool for a different job: shared space, light loads, and easy changeover. The unit on safety returns to this point in full, because a cobot without a risk check is not safe by name alone.

Quick self-check

Question: A cobot stops when it touches a person. Does that make any tool on the cobot safe?

No. The stop protects against the arm hitting a person, but the tool and the load can still hurt. A sharp blade, a hot tray, or a heavy box on a cobot is still a hazard, and the risk check must cover the tool, the load, and the food, not only the arm.

Question: When would a cobot be a better choice than a fenced industrial robot?

When the load is light, the job is simple, the product changes often, and a person needs to work in the same space. The cobot is easy to move and re-teach, and it does not need a large fenced area.

Other machines that work like robots

Not every automated machine in a food plant is a robot arm, and a food technologist should know the neighbours. An automated guided vehicle moves pallets and materials around a plant without a driver, following a path on the floor or a signal from the plant. A conveyor system moves food and packs between steps, and a smart conveyor can sort, space, and turn packs. A pick and place machine in a packing hall may use a simple pneumatic arm rather than a full robot, and for one fixed movement that simple arm can be cheaper and easier to keep.

Machine vision is covered in Unit 5, but it belongs here too: a camera and a computer that find, measure, and sort food are often the sense organ of a robot, and sometimes the vision system alone, with a jet of air that blows a bad piece off the belt, does the job without any arm at all. A food technologist who only thinks in robot arms will miss these simpler tools, and a simple tool that fits the job is always better than a robot that does not.

Matching the robot to the job

Heavy load, long reach, tilt and turn: articulated robot
↓
Fast flat pick, known height: SCARA robot
↓
Very fast light pick from a moving belt: delta robot
↓
Wide area, heavy load, straight lines: gantry robot
↓
Light load, shared space, frequent change: collaborative robot

Choosing between types: load, speed, reach, and food

A choice between robot types comes down to four measures, plus the food itself. Load is the weight the robot must carry, including the tool and the food. Speed is how many picks or places it must do in a minute, at the real pace of the line. Reach is how far the tool must travel, and how high. Accuracy is how close the tool must land to the target, every time. A robot that is strong enough but too slow will hold up the line. A robot that is fast enough but too weak will drop the load. A robot that is accurate enough on paper but cannot reach the far side of the belt will fail on the first day.

The food adds its own measures. A soft fruit needs a gentle grip and a light touch. A sticky dough needs a tool that releases cleanly. A hot product needs a tool that can take heat. A wet product needs a tool and an arm that can be washed. A product that varies in size needs a sensor that can find it and a gripper that can adjust. These food measures often decide the choice before the catalogue measures do, because a robot that cannot hold the food cannot do the job at any speed.

A practical way to choose is to write the job down in one paragraph, with the load, the speed, the reach, the accuracy, the food, and the space, and then to ask for a trial with the real product. A supplier who will not trial the real food is a warning sign. A trial that counts damaged pieces, missed picks, and stops is the evidence that matters, not a speed figure measured with a different product.

Quick self-check

Question: A robot is rated for a high speed with a light plastic block. Why might it be slower with real food?

Because real food is soft, variable, and sometimes sticky, so the gripper must move more gently, the camera must find each piece, and the release must be clean. The catalogue speed was measured with a fixed shape that is easy to grip, not with food that changes.

Question: What is the first measure to check when a robot must lift sacks of flour?

Load: the weight of the sack plus the weight of the tool. If the robot cannot carry that load safely, speed and reach do not matter, and a stronger type must be chosen.

Try it yourself

Pick three foods from a shop: a heavy bag of rice, a punnet of soft berries, and a carton of milk. For each one, choose the robot type from this unit that fits best, and write two sentences giving the reason: one sentence about the food, and one about the robot. Then write one sentence for each food naming the robot type that would fit worst, and why. If you can, watch a short video of a food packing line and note which robot type you see, and whether the choice matches the reasons you wrote down.

Key ideas to carry forward

  • A robot is named by how it moves: base, joints, links, wrist, end effector, controller, and program.
  • The work envelope is the space a robot can reach; a long reach needs a larger guarded area.
  • Articulated robots are strong, with a long reach and the ability to tilt and turn; they suit heavy lifting and palletising.
  • SCARA robots are fast and precise in a flat plane at a known height; they suit lids, trays, and small packs.
  • Delta robots are the fastest light pickers from a moving belt, and they need a camera to find each piece.
  • Cartesian and gantry robots move in straight lines over a wide area; they suit heavy loads and simple straight moves.
  • Collaborative robots work beside people for light jobs with frequent change, but the tool and the load must still be risk-checked.
  • Automated guided vehicles, conveyors, simple pick and place arms, and vision with an air jet are neighbours that may fit better than a robot arm.
  • Choose by load, speed, reach, and accuracy, plus the food: softness, stickiness, temperature, wetness, and variation.
  • A trial with the real food at real speed is the evidence that matters.

Glossary

Actuator

A device that makes a robot joint move, such as a motor. Food tech example: the motors in a delta robot drive its three arms to pick biscuits from a belt.

Articulated robot

A robot arm with a shoulder, elbow, and wrist in a chain, usually with six axes, that moves like a human arm. Food tech example: an articulated robot stacks cartons of soup on a pallet.

Axis

One direction or joint a robot can move in. Food tech example: a six-axis robot can move its tool in six ways, including turning and tilting.

Cartesian robot

A robot that moves its tool along straight lines on three axes: left and right, forward and back, up and down. Food tech example: a Cartesian robot lifts trays in and out of a freezer.

Collaborative robot

A robot designed to work in the same space as people, with sensors that stop it on contact, for light tasks. Food tech example: a collaborative robot places packs in a carton while a worker checks them at the same bench.

Controller

The computer that tells a robot when and how far to move, following its program. Food tech example: the controller runs the palletising program and stops the robot when a sensor reports a fault.

Delta robot

A fast light robot hung from a frame above a belt, with three arms working in parallel. Food tech example: a delta robot picks single chocolates from a moving belt and places them in a box.

End effector

The tool at the end of a robot arm that does the work, such as a gripper, a suction cup, or a cutter. Food tech example: a suction cup end effector lifts a flat cracker without squeezing its edges.

Gantry robot

A large Cartesian robot on a frame spanning a wide area, moving its tool in straight lines. Food tech example: a gantry robot moves pallets across a cold store.

Joint

A point where two parts of a robot arm meet and move relative to each other. Food tech example: the elbow joint of an articulated robot bends to lower a sack into a mixer.

Link

A rigid part of a robot arm that joins two joints. Food tech example: the link between the shoulder and the elbow carries the load of a carton.

Payload

The weight a robot can carry, including its tool and the food or pack it holds. Food tech example: a robot with a payload of 10 kg cannot safely lift a sack plus a gripper that together weigh more than that.

Program

The set of instructions a robot controller follows to do its job. Food tech example: a palletising program tells the robot where to place each carton in each layer.

SCARA robot

A robot with two horizontal arms moving in a flat plane, fast and precise at a known height. Food tech example: a SCARA robot places lids on tubs moving along a belt.

Work envelope

The space a robot can reach with its tool. Food tech example: the work envelope of a palletising robot must cover the whole pallet, from the lowest layer to the top layer.

References

  • Irudayaraj, J., Food Processing Operations Modeling, Design, and Analysis, 2nd edition, 2002.
  • Singh, R. P., Computer Applications in Food Technology, 1996.

Unit 3: Robotic Handling, Picking and Packing

Note

Unit 3: Robotic Handling, Picking and Packing

A question to start with

A line packs soft cheese portions into trays. The portions are cut, they sit on a belt, and they must land in the tray in neat rows, with no gaps and no damaged corners. A person does this well: she sees each portion, picks it up by its firm side, sets it down gently, and leaves a small space for the next one. A robot that tries the same job faces a plain problem: how do you pick up something soft without squashing it, and how do you put it down in exactly the right place when every portion is a slightly different size? This unit is about handling, picking, and packing: the contact between the machine and the food, and the steps that turn a stream of loose pieces into a neat, sealed pack.

Handling means moving food without damaging it and without letting it fall, stick, or change shape. Picking means taking one piece, or a set number of pieces, from a stream or a pile. Placing means setting the piece down in a set position. Packing means putting the placed food into its tray, pouch, carton, or box, and getting it ready for sealing and labelling. Each step has its own faults: a pick can bruise, a place can be off by a few millimetres, a pack can be short by one piece. A good handling system controls all three, and it counts what it does, so that a fault is found at once.

The belt is where handling starts. Food arrives on a belt in one of three ways: in a neat row, in a random scatter, or in a pile. A neat row is easiest: the robot knows where the next piece will be. A random scatter needs a camera to find each piece. A pile needs a feeder that separates the pieces before the robot can pick them. The way food arrives decides the gripper, the camera, and the speed, so handling design starts with the belt, not with the robot.

Questions to think about

  • A soft cheese portion and a hard biscuit need different grippers. What is different about the food, and what must the gripper do differently?
  • Why does a robot place pieces in a tray more evenly than a person, yet still miss pieces that a person would catch?
  • If a tray leaves the line with one piece missing, is that a picking fault, a placing fault, or a counting fault, and how would you tell?
Think about the cheese portions before you read on

The cheese is soft, so the gripper must spread its force over a wide area and grip the firm side, not the cut face. The biscuit is hard and flat, so a suction cup on its top face works well. The robot places evenly because it repeats a set position, but it misses faults a person would catch because it only checks what its sensor is set to check. A missing piece in a tray can come from any of the three steps: the robot never picked it, it picked it and dropped it, or it placed it but the count was wrong. Counting the picks, the places, and the finished trays separately shows which step failed.

Grippers: the tool that touches the food

The gripper is the most important part of a handling system, because it is the only part that touches the food. There are four main families of grippers in food plants, and many grippers mix two families in one tool.

Mechanical finger grippers use two or more fingers that close around a piece. They suit firm food with a known shape, such as a tub, a bottle, a carton, or a whole fruit with a firm skin. The fingers can be shaped to the product, and they can be lined with soft material to spread the force. Their risk is pressure: too much force bruises or breaks the food, too little lets it slip.

Suction grippers use a vacuum cup that sticks to a flat, smooth surface. They suit flat food such as biscuits, crackers, cheese slices, and flat packs. They lift without squeezing the edges, and they release cleanly when the vacuum is switched off. Their weakness is surface: a rough, crumbly, wet, or curved surface will not seal, and a porous food lets air leak in so the suction fails. A suction cup that works on a smooth cracker may fail on a rough bread roll.

Soft grippers use flexible fingers, a soft balloon, or a flexible pouch that wraps around the piece. They suit soft, uneven food such as fruit, bread rolls, and baked goods, because they spread the force and follow the shape. They are gentler than hard fingers, but they are slower to close and they wear out faster, so they need checking and replacing on schedule.

Needle and clamp grippers use thin pins or a clamp that holds the food by its firm part. They suit some meats, fish, and bakery products where a finger or a suction cup cannot get a hold. Their risk is marking: a needle leaves a hole, and a clamp leaves a mark, so they are used only where the mark is hidden or does not matter to the buyer.

Choosing a gripper starts with the food: its size, shape, weight, softness, surface, stickiness, and temperature. Then comes the move: how fast, how far, and how the piece must be turned. A gripper that is tested with one batch may fail on the next batch if the food is softer, wetter, or stickier, so the trial must use the softest, wettest, and stickiest product the line will see, not only the best batch.

Quick self-check

Question: Why does a suction cup fail on a rough bread roll but work on a flat cracker?

Because suction needs a seal. A flat smooth cracker lets the cup seal and hold a vacuum. A rough curved roll lets air leak in around the edge, so the vacuum fails and the roll drops.

Question: A finger gripper bruises peaches. What two changes could help, and why?

First, line the fingers with soft material and shape them to the peach, so the force spreads over a wider area. Second, lower the closing force and slow the close, so the grip is firm enough to hold but not hard enough to bruise.

Picking from a belt: finding, timing, and placing

Picking from a moving belt is a timing problem. The belt moves, the piece moves with it, and the robot must meet the piece at the right place at the right time. A camera or a sensor finds the piece, the controller works out where it will be when the gripper arrives, the gripper picks it, and the robot places it in a tray or pack that may itself be moving. This is called tracking: the robot follows the moving target instead of waiting for it to stop.

Three things make belt picking work. First, singulation: the pieces must be separated so the robot can pick one at a time. A feeder, a vibrating tray, or a spaced belt separates pieces that arrive in a pile. Second, sensing: the camera must see the piece clearly, in steady light, without glare from a wet surface or a shiny pack. Third, timing: the belt speed, the camera, and the robot must agree, so that a piece found at one point is picked at the point the robot expects.

Picking also needs a plan for the pieces the robot cannot pick. Some pieces are broken, stuck together, or in the wrong orientation. The system must decide: leave them, reject them, or send them back for another pass. A line that has no plan for bad pieces will either pack them or stop. A good line has a reject path: a jet of air, a diverter, or a separate belt that carries the bad pieces away, and a count of how many were rejected and why.

Quick self-check

Question: Pieces arrive on a belt in a pile, stuck together. What should be fixed first: the robot, the camera, or the feeder?

The feeder. The robot cannot pick one piece from a pile, and the camera cannot see pieces hidden under others. Separating the pieces first, with a feeder or a spaced belt, makes the camera and the robot work as designed.

Question: Why must the belt speed, the camera, and the robot agree?

Because the piece keeps moving after the camera sees it. If the belt is faster than the robot expects, the gripper arrives at an empty place. If it is slower, the gripper arrives late and may crush the piece.

Placing in trays: patterns, orientation, and fill

Placing is more than dropping a piece into a tray. The piece must land in the right cell, facing the right way, with the right spacing, and the tray must be full before it moves on. Orientation matters: a biscuit with a pattern on top must face up, a cheese portion must sit on its flat side, a piece of fruit must sit with its stem away from the pack edge. A robot that places a piece upside down has not done its job, even if the count is right.

Trays add their own problems. A tray may be slightly out of position, a cell may be dirty, or a tray may be missing. The robot must check that the tray is present and in place before it places food in it. A sensor that finds the tray, or a guide that holds it in place, prevents food landing on an empty belt. The fill must also be counted: a tray that needs twelve pieces must leave with twelve, and a tray with eleven must be rejected or topped up, not sent on as a full tray.

Speed and care pull in opposite directions at the placing step. A fast place can bounce a soft piece out of its cell. A slow place holds up the line. The answer is usually a controlled release: the gripper lowers the piece into the cell, releases it gently, and lifts away without dragging. For sticky food, the gripper may need a push-off or a puff of air to release cleanly, because a piece that sticks to the gripper will be placed late or in the wrong cell.

From a loose piece to a full tray

Pieces separated on the belt
↓
Camera finds each piece and checks it
↓
Gripper picks one piece with the right force
↓
Robot places the piece in the right tray cell, facing the right way
↓
Tray count checked: full trays move on, short trays are rejected or topped up

Quick self-check

Question: A tray needs twelve biscuits and leaves with eleven. How would you find out which step failed?

Count the picks, the places, and the finished trays separately. If picks are twelve and places are eleven, a piece was dropped. If picks are eleven, the feeder or the belt starved the robot. If places are twelve but the tray has eleven, a piece bounced out after placing.

Packing: pouches, cartons, cases, and pallets

Packing has four levels, and a robot may work at any of them. Primary packing is the pack that touches the food: a pouch, a tray with film, a tub, or a bottle. Secondary packing is the carton or box that holds several primary packs. Tertiary packing is the case or the pallet that holds several cartons for storage and transport. Each level has its own robot job, its own gripper, and its own check.

At primary level, the robot often works with a forming, filling, and sealing machine: it places food in a pouch or tray, the machine seals it, and a checkweigher and a camera check the weight, the seal, and the label. The robot must not damage the seal area, because a crumb or a drop of liquid in the seal will make the pack leak.

At secondary level, the robot places primary packs into a carton in a set pattern. The carton must be open, held in place, and not torn by the robot. At tertiary level, the robot, usually an articulated or gantry robot, stacks cartons on a pallet in a set pattern, with layers that interlock so the pallet is stable. A pallet that is stacked badly will lean, fall, or damage the cartons at the bottom. The pattern, called the pallet pattern, is part of the program, and it changes with the carton size.

Case packing and palletising are the most common robot jobs in food plants, because the loads are heavy, the pattern is fixed, and the job repeats all day. Primary packing of soft food is harder, because the food is variable and the pack must be sealed cleanly. A plant that automates case packing first often learns the skills it needs for primary packing later.

Quick self-check

Question: Why is a crumb in the seal area a packing fault, not only a cosmetic fault?

Because the seal must be airtight to keep the food safe and to hold its shelf life. A crumb in the seal leaves a gap, air and microbes can enter, and the pack may leak or spoil early.

Question: Why does the pallet pattern matter for a pallet that will be moved and stored?

Because the pattern decides whether the pallet is stable. Layers that interlock spread the load and stop the stack leaning or falling. A bad pattern puts weight on weak cartons and can crush them at the bottom.

Changeover: the time between products

Changeover is the time spent changing a line from one product to another: changing the gripper, the tray, the program, the carton size, and the label, and testing the first packs. Changeover decides whether a robot pays for a plant that makes many products in small batches. A robot that packs fast but takes an hour to change over may be slower over a whole day than a hand line that changes over in ten minutes.

Good changeover is designed in, not added later. Grippers that fit with one fixing, programs that are saved and named clearly, trays that are colour coded, and simple written steps all cut changeover time. A trial changeover, timed with a stopwatch, shows where the time goes: waiting for a tool, searching for a program, adjusting a guide by hand, or testing packs that fail. The slowest step is the one to fix first.

Changeover also needs a first-pack check. After every change, the first packs are checked in full: count, weight, orientation, seal, and label. Only when the first packs pass does the line run at full speed. A line that skips the first-pack check will make bad packs at full speed until someone notices.

Faults, jams, and keeping the line running

Handling lines stop for a small set of reasons, and knowing them helps a team fix them fast. A jam happens when food or a pack sticks, folds, or lands in the wrong place. A missed pick happens when the gripper fails to hold the piece. A double pick happens when two pieces stick together and the robot picks both. A dropped piece happens when the grip is too soft or the move is too fast. A misplace happens when the tray is out of position or the robot program is wrong.

Each fault has a plain fix. A jam is cleared, the cause is found, and the guide or feeder is adjusted. A missed pick is checked against the gripper, the suction, and the surface of the food. A double pick is checked against the feeder: pieces that arrive stuck together need separating before picking. A dropped piece is checked against the grip force and the speed of the move. A misplace is checked against the tray position and the program. Writing the fault, the cause, and the fix in a simple log lets the next shift fix the same fault faster, and lets the manager see which fault costs the most time.

Quick self-check

Question: The robot picks two biscuits at once. Is the gripper the first thing to change?

Usually not. Two biscuits picked at once usually means they arrived stuck together or overlapping, so the feeder or the spacing is the cause. Fix the separation first, then check the gripper if the fault continues.

Try it yourself

Take a pack of biscuits or a tray of fruit from a shop. Open it and look at the pieces: are they all facing the same way, are they all the same size, is any piece broken, is any cell empty? Write down what you see, and then write a short picking and placing plan for that product in five lines: how the pieces would arrive, what gripper you would use, how the camera would find each piece, how the tray would be checked, and what would happen to a broken piece. Compare your plan with a classmate: where do you agree on the gripper, and where do you differ?

Key ideas to carry forward

  • Handling is moving food without damaging it; picking takes one piece, placing sets it in a set position, packing puts it in its tray, pouch, carton, or box.
  • The way food arrives, neat row, random scatter, or pile, decides the gripper, the camera, and the speed.
  • Finger grippers suit firm food, suction grippers suit flat smooth food, soft grippers suit soft uneven food, needle and clamp grippers suit food where a mark does not matter.
  • Belt picking is a timing problem: singulation, sensing, and timing must agree, or picks miss.
  • Placing needs the right cell, the right orientation, the right spacing, and a count check before the tray moves on.
  • Packing has four levels: primary pack, secondary carton, tertiary case and pallet, each with its own robot job and check.
  • A crumb or liquid in a seal can break the seal and spoil the pack.
  • Changeover time decides whether a robot pays for a plant that makes many products in small batches.
  • Faults have plain causes: jams, missed picks, double picks, drops, and misplaces each point to a step that can be checked and fixed.

Glossary

Changeover

The time and work spent changing a line from one product to another, including tools, programs, trays, and labels. Food tech example: changing a biscuit line from a tray of six to a tray of twelve needs a new gripper setting and a new program.

Double pick

When a robot picks two pieces at once instead of one. Food tech example: two cheese slices stuck together are picked as one, so the tray ends up short.

Feeder

A device that separates pieces and feeds them one at a time to a picking point. Food tech example: a vibrating feeder separates biscuits that arrive in a pile so the robot can pick them singly.

Fill

The number of pieces placed in a tray or pack. Food tech example: a tray with a fill of twelve must leave the line with twelve biscuits, not eleven.

Gripper

The part of a robot that holds food while it is moved. Food tech example: a soft finger gripper holds a bread roll without squashing it.

Orientation

The direction a piece faces when it is placed. Food tech example: biscuits are placed pattern side up, so every pack looks the same when it is opened.

Pallet pattern

The set arrangement of cartons in each layer on a pallet. Food tech example: a pallet pattern that interlocks layers stops the stack leaning during transport.

Primary packing

The pack that touches the food directly, such as a pouch, tray, tub, or bottle. Food tech example: a sealed pouch of dried fruit is primary packing.

Secondary packing

The carton or box that holds several primary packs. Food tech example: a carton holding twelve pouches of soup is secondary packing.

Singulation

Separating pieces so they can be picked one at a time. Food tech example: a spaced belt singulates chocolates that arrive in a pile.

Suction gripper

A gripper that uses a vacuum cup to hold a flat smooth surface. Food tech example: a suction gripper lifts a flat cracker by its top face without squeezing its edges.

Tertiary packing

The case or pallet that holds several cartons for storage and transport. Food tech example: a pallet stacked with cartons of canned soup is tertiary packing.

Tracking

Following a moving piece on a belt so the robot meets it at the right place and time. Food tech example: a delta robot tracks a biscuit moving along a belt and picks it without stopping the belt.

Tray

A shaped container with cells that hold food pieces in a set pattern. Food tech example: a tray with twelve cells holds twelve cheese portions in neat rows.

References

  • Irudayaraj, J., Food Processing Operations Modeling, Design, and Analysis, 2nd edition, 2002.
  • Singh, R. P., Computer Applications in Food Technology, 1996.

Unit 4: Automation in Processing Lines: From Raw Material to Pack

Note

Unit 4: Automation in Processing Lines: From Raw Material to Pack

A question to start with

A tomato sauce line starts with crates of tomatoes, and ends with sealed jars on a pallet. Between those two ends, the tomatoes are washed, sorted, crushed, heated, mixed, filled, sealed, labelled, checked, and packed. If one step runs faster than the next, tomatoes pile up. If one step stops, every step after it starves. If the heating step runs too hot, the sauce burns. If the filling step runs too fast, the jars overflow. The question is not how to automate one step, but how to make all the steps work together as one line. This unit follows a processing line from raw material to pack, and shows how automation joins the steps, controls them, and keeps the product safe and even.

A processing line is a chain of steps, each step changing the food a little: cleaning, sorting, cutting, mixing, heating, cooling, filling, sealing, and packing. Automation in a line means three things working together. First, machines do the work at each step. Second, conveyors and transfer points move the food from one step to the next without dropping, spilling, or mixing batches. Third, controls watch the line: sensors measure temperature, flow, weight, level, and speed, and a control system adjusts the machines to hold those measures steady. When the three work together, the line runs at a steady pace, with steady quality, and with a record of what happened.

The line must also be planned as a whole. A fast filler at the end cannot fix a slow washer at the start. A precise heater cannot fix a mixer that feeds it unevenly. The slowest step, called the bottleneck, sets the pace of the whole line. Finding the bottleneck, and fixing it first, is the first rule of line automation.

Questions to think about

  • If the filler runs twice as fast as the heater, what happens to the sauce between the two steps, and why does it matter?
  • A line makes a perfect sauce but the jars are half full. Is that a processing fault or a filling fault, and how would you prove it?
  • Why does a control system need a sensor before it can control anything?
Think about the tomato sauce line before you read on

If the filler is faster than the heater, the filler waits, or the sauce waits in a tank and cools, thickens, or separates while it waits. Half-full jars are a filling fault if the sauce leaving the heater is correct: the filling step is not putting the right amount in each jar. A control system needs a sensor because it cannot adjust what it cannot measure. The sensor tells the control what the temperature, flow, or level is now, and the control changes the heater, the pump, or the belt to bring it back to the set point.

Reception and intake: the first automation

Automation starts before the food reaches the line. At reception, raw material arrives in crates, sacks, tanks, or bulk loads. Automated intake weighs the load, records the supplier and the batch, and moves the material into storage or into the first step. A weighbridge, a tank level sensor, or a load cell under a hopper gives the weight without a person writing it down by hand, and the record links the batch to the supplier, the date, and the time.

Storage is also automated. Silos hold flour, grain, or sugar, and level sensors show how much is inside. Cold stores hold milk, meat, or fruit at a set temperature, and temperature sensors record the cold chain. A control system can move material from storage to the line on a schedule, so that the line is fed steadily and the oldest stock is used first. This matters for food safety: a record that shows which batch went into which production run lets a plant trace a fault back to its source.

Intake must also protect the line from what should not enter it. A metal detector or a screen at intake catches stones, metal, and other foreign material before it reaches a pump, a cutter, or a filler, where it could damage the machine or end up in a pack. A line that lets foreign material in will pay for it later, in damaged machines and in rejected packs.

Quick self-check

Question: Why record the supplier and batch at reception, not only the weight?

Because the record links the finished product back to its source. If a fault is found later, the plant can trace which batch it came from, which other products used the same batch, and which supplier to check.

Question: What does a level sensor in a silo prevent?

It prevents the line running dry without warning, and it prevents overfilling. The line can be fed steadily, and the team knows when to order more before the silo is empty.

Cleaning, sorting, and preparation

The first steps on the line prepare the raw material: washing, sorting, peeling, cutting, and trimming. Automation here is about evenness and gentleness. A washer must clean every piece without bruising it. A sorter must separate good pieces from bad ones, and separate by size, without damaging the good pieces. A cutter must cut every piece to the same size, because size affects cooking time, drying time, and the look of the pack.

Sorting uses the senses a person would use, but at line speed. A camera sorts by colour and shape. A sensor sorts by weight. An air jet blows a bad piece off the belt. A vibrating table spreads pieces so the camera can see each one. Sorting is covered in detail in Unit 5, but its place in the line matters here: sorting early, before cooking or mixing, saves energy and time, because bad pieces are removed before they are heated, mixed, and packed.

Preparation also sets the flow. A buffer tank, a hopper, or a short conveyor between two steps holds a small amount of food, so that a short stop in one step does not stop the whole line. Buffers are small by design: a large buffer holds food for a long time, and food that waits can cool, dry, separate, or grow microbes. The buffer must be sized to the food and to the step, and it must be cleaned on schedule.

Quick self-check

Question: Why sort bad tomatoes out before cooking, not after?

Because cooking, mixing, and packing a bad tomato wastes energy, time, and pack material, and a bad tomato in the mix can spoil the flavour and the safety of the whole batch. Removing it early saves the work that would have been spent on it.

Question: What is the risk of a buffer tank that is too large?

Food waits in it for a long time, so it can cool, thicken, separate, or grow microbes. A buffer should hold only enough to smooth a short stop, not enough to store the product.

Mixing, heating, cooling, and holding

The middle of the line changes the food: mixing blends ingredients, heating cooks or pasteurises, cooling brings the food back to a safe temperature, and holding keeps it at a set temperature for a set time. Automation here is about control: holding temperature, time, flow, and mixing speed steady, and recording them.

A mixing system uses a recipe: the amounts of each ingredient, the order they are added, and the mixing time and speed. An automated system weighs or meters each ingredient, adds it in order, and records the batch. This gives the same recipe every time, and it gives a record that the recipe was followed. A hand-mixed batch can vary with the person, the day, and the hurry. An automated batch varies less, provided the meters are checked and the recipe is correct.

Heating is controlled by a sensor and a control loop. A temperature sensor measures the food, the controller compares the reading with the set point, and the heater is turned up or down to close the gap. This loop, called feedback control, runs all the time. A chart or a log records the temperature, so that the plant can prove the food was heated as required. Cooling works the same way, with a sensor and a cooler. Holding time is controlled by flow: the speed of the pump and the size of the holding tube fix how long the food stays at temperature.

The control system must also know what to do when something goes wrong. If the temperature falls, the flow should stop or the product should be diverted back for re-heating, not sent on as finished product. This divert step is a safety feature, not an extra: it keeps under-heated food out of the pack.

How feedback control holds a temperature steady

Sensor measures the food temperature
↓
Controller compares the reading with the set point
↓
Heater is turned up or down to close the gap
↓
Sensor measures again, and the loop repeats

Quick self-check

Question: Why must an automated recipe be checked, not only trusted?

Because the system follows the recipe it was given. If the recipe is wrong, or a meter is wrong, the system will make the same wrong batch again and again. Checking the first batch, and checking the meters on schedule, catches the error before it repeats.

Question: The temperature falls below the set point. What should the line do with the food that passed through at that time?

It should divert that food back for re-heating or hold it for a decision, not send it on as finished product. Food that was not heated as required cannot be assumed safe, even if it looks normal.

Filling, sealing, and labelling

Filling puts the right amount in each pack. A filler can work by volume, by weight, or by counting pieces. A volumetric filler puts the same volume in each pack, which is fast but can vary in weight if the food is not even. A weight filler weighs each pack and adjusts the fill, which is slower but more accurate. A piece counter counts pieces into each pack. The choice depends on the food and on what the label promises: a pack sold by weight must be filled by weight or checked by weight after filling.

Sealing closes the pack. Heat sealing melts a film or a pouch edge to make an airtight join. A lid is pressed or screwed on. A can is seamed. The seal must be complete, with no food, liquid, or fold in the seal area. A seal checker, a camera, or a pressure test finds packs with a weak seal. A pack with a failed seal must be rejected, because it will leak, let air in, and spoil early.

Labelling puts the name, the weight, the date, the batch code, and other information on the pack. An automated labeller prints and applies the label in the right place, and a camera checks that the label is present, straight, and readable. A wrong label is a serious fault: it can give the wrong date, the wrong weight, or the wrong information about ingredients, and it can send a product to the wrong place. The label check is a quality check and a safety check at the same time.

Quick self-check

Question: Why is a pack sold by weight usually checked by a checkweigher after filling?

Because the label promises a weight, and only a weighing can prove the pack holds that weight. A volume filler can vary in weight if the food is uneven, so the checkweigher catches packs that are too light or too heavy.

Question: A label is printed but stuck on crooked, covering the date. Is that a fault?

Yes. The date must be readable, and a label that covers it or sits crooked can hide information the buyer needs. The label check should reject packs where the label is missing, wrong, or unreadable.

Conveyors, transfer points, and line balance

Conveyors move food and packs between steps. A belt conveyor carries packs on a flat belt. A roller conveyor carries cartons on rollers. A chain conveyor carries heavy loads. A pneumatic conveyor moves powder or grain through a pipe with air. A screw conveyor moves powder, paste, or granules through a tube. Each type suits a different food, and each type can damage the food if it is wrong: a drop from one belt to another can break a biscuit, a fast screw can crush a soft piece, a rough belt can mark a pack.

Transfer points, where food moves from one conveyor to the next, are where most damage and most jams happen. A good transfer is level, smooth, and guided, with no gap for a piece to fall into and no drop that breaks it. A transfer that is left as a gap will cause the same jam every day.

Line balance is the work of matching the speed of every step to the slowest step, so that no step starves and no step floods. A line that is balanced runs steadily, with small buffers that smooth short stops. A line that is not balanced piles up before the slow step and stands empty after it. Balancing is done by timing each step, finding the bottleneck, and fixing it first: by speeding the slow step, by adding a second machine at that step, or by slowing the fast steps to match.

A line from raw material to pack

Reception: weigh, record, store
↓
Preparation: wash, sort, cut, trim
↓
Processing: mix, heat, cool, hold
↓
Filling and sealing: right amount, airtight seal
↓
Labelling and checking: weight, seal, label, date
↓
Packing and palletising: carton, case, pallet

Control systems: from a switch to a whole plant

Control in a food plant has layers. At the lowest layer, a single machine has its own controls: a switch, a sensor, and a motor. At the next layer, a programmable logic controller, called a PLC, runs a step or a machine: it reads sensors, follows a program, and drives motors, valves, and heaters. At the next layer, a supervisory system watches the whole line: it shows the operator the state of each step, records temperatures and counts, sounds an alarm when a measure leaves its set range, and stores the record for later checking. At the top layer, a plant system links production to orders, stock, and traceability: which batch went into which run, how many packs were made, and how many were rejected.

The operator's screen is the face of the control system. It should show, in plain form, what is running, what is stopped, what the key measures are, and what alarm is sounding and why. A screen that shows too much, or that uses unclear words, will be ignored in a hurry. A screen that shows the right few measures, with a clear alarm and a clear action, will be used.

Alarms deserve care. An alarm that sounds too often, or for a small fault, will be ignored, and a real alarm will then be missed. Alarms should be set for measures that matter, with a clear limit, and with a named action: stop, divert, check, or call. The log of alarms is a useful record: it shows which step alarms most, and where the line needs fixing.

Cleaning in place and changeover on a line

A processing line must be cleaned, and automation must allow for cleaning, not fight it. Cleaning in place, called CIP, is a system that pumps cleaning solutions through pipes, tanks, and machines without taking them apart. It follows a set cycle: rinse, wash with a cleaning solution, rinse again, and sometimes a sanitising step. The cycle is timed and recorded, and the line is not started again until the cleaning record is complete and the rinse water is clear. A line that cannot be cleaned in place must be taken apart, cleaned by hand, and put back together, which takes longer and risks a wrong reassembly.

Changeover on a line is like changeover on a packing machine, but larger: the recipe changes, the filler setting changes, the label changes, and the line must be cleaned between products that should not mix. A product that contains a common allergen, such as milk, egg, peanut, or wheat, must not leave traces in the next product. The cleaning between those products is a safety step, and it must be proved, not assumed. The first packs after a changeover are checked in full before the line runs at full speed.

Quick self-check

Question: Why is cleaning in place recorded, not only done?

Because the record proves the line was cleaned, for how long, and with what, before the next batch started. Without the record, a later fault cannot be traced to a cleaning failure or cleared of it.

Question: A line changes from a product with milk to a product without milk. What is the main risk, and what is the fix?

The main risk is milk traces left in the line ending up in the milk-free product. The fix is a full clean between the products, with a check that the cleaning worked, before the new product starts.

Records, traceability, and the line that can prove itself

An automated line produces a record as it runs: temperatures, times, weights, counts, rejects, alarms, cleaning cycles, and batch numbers. That record is part of the product. It lets the plant prove that the food was heated as required, that the packs were weighed, that the labels were checked, and that the cleaning was done. It also lets the plant find a fault: if a batch of packs is light, the record shows whether the filler or the checkweigher was at fault, and when the fault started.

Traceability is the link from the finished pack back to the raw material, through every step. A batch code on the pack points to the production run, the run points to the recipe and the raw material batches, and those point to the supplier and the reception record. An automated system that records the batch at each step makes that link easy. A system that records only at the end makes it hard, and a system that records nothing makes it impossible.

The record also helps the plant improve. Counts of stops, rejects, and alarms show where the line loses time and product. A manager who reads those counts each week can fix the worst fault first, instead of fixing the fault that was noticed last. That is the quiet gain of line automation: not only faster work, but work that can be seen, counted, and improved.

Try it yourself

Choose a food with a simple line, such as bread, jam, or pasteurised milk. Draw the line as a chain of six to eight steps, from raw material to pack, on paper or in a table. For each step, write the machine, the sensor or check that watches it, and the record it produces. Then mark the step you think is the bottleneck, and write one sentence saying why. Compare your line with a classmate: did you both put the same step as the bottleneck, and what evidence would settle the difference?

Key ideas to carry forward

  • A processing line is a chain of steps, and the slowest step, the bottleneck, sets the pace of the whole line.
  • Automation joins machines, conveyors, and controls: machines do the work, conveyors move the food, controls hold the measures steady.
  • Intake weighs, records, and stores raw material, and links each batch to its supplier and date.
  • Sorting early saves the energy and time that would be spent on bad pieces later.
  • Mixing follows a recipe, heating and cooling follow a feedback loop, and holding time is fixed by flow.
  • If a temperature falls, the product must be diverted, not sent on as finished product.
  • Filling can be by volume, by weight, or by count; a pack sold by weight must be checked by weight.
  • Sealing must be airtight and free of food in the seal; labelling must be present, correct, and readable.
  • Transfer points cause most damage and jams; line balance matches every step to the slowest step.
  • Control has layers: machine controls, PLC, supervisory system, and plant system, with a clear operator screen and alarms that matter.
  • Cleaning in place and cleaning between products, especially where allergens are involved, are safety steps that must be recorded and proved.
  • Records give traceability from the finished pack back to the raw material, and counts of stops and rejects show where to improve.

Glossary

Batch

A set amount of food made in one run under the same conditions. Food tech example: one mixing tank of sauce made from one recipe is a batch.

Batch code

A code printed on a pack that links it to the production run it came from. Food tech example: a batch code on a jar of sauce lets the plant find the day, the recipe, and the raw material used for that jar.

Bottleneck

The slowest step in a line, which sets the pace of the whole line. Food tech example: if the filler can fill 60 jars a minute but the labeller can label only 40, the labeller is the bottleneck.

Buffer

A small store between two steps that holds food briefly so a short stop in one step does not stop the whole line. Food tech example: a small tank between the heater and the filler holds sauce during a short filler stop.

Cleaning in place

Cleaning pipes, tanks, and machines by pumping cleaning solutions through them without taking them apart. Food tech example: a milk line is cleaned in place with a timed rinse and wash cycle after each run.

Feedback control

Control where a sensor measures a result, a controller compares it with a set point, and the machine is adjusted to close the gap. Food tech example: a temperature sensor and a heater hold a sauce at its set temperature by feedback control.

Flow

The amount of food moving through a pipe or a line in a given time. Food tech example: the flow of milk through a heater fixes how long the milk stays hot.

Holding time

The time food stays at a set temperature, or in a set place, as part of a process. Food tech example: milk held hot for a set time in a holding tube is pasteurised.

Line balance

Matching the speed of every step in a line to the slowest step, so no step starves and no step floods. Food tech example: a line is balanced when the washer, the filler, and the labeller all run at the pace of the slowest step.

Programmable logic controller

A computer that runs a machine or a step, reading sensors and driving motors, valves, and heaters by a program. Food tech example: a PLC runs a filling machine, starting and stopping the filler from level sensors.

Recipe

The set amounts, order, times, and settings used to make a product. Food tech example: a sauce recipe lists the ingredients, the amounts, the mixing order, and the heating temperature.

Set point

The target value a control system tries to hold, such as a temperature or a speed. Food tech example: the set point for a pasteuriser is the temperature the milk must reach and hold.

Supervisory system

A system that watches a whole line, shows its state to the operator, records measures, and sounds alarms. Food tech example: a supervisory screen shows the temperature of each heater and alarms when one leaves its set range.

Traceability

The ability to follow a finished pack back through every step to its raw material and supplier. Food tech example: a batch code lets a jam maker trace a jar back to the fruit delivery it came from.

Transfer point

A place where food or packs move from one conveyor or machine to the next. Food tech example: the gap between two belts where biscuits drop from one to the next is a transfer point.

References

  • Irudayaraj, J., Food Processing Operations Modeling, Design, and Analysis, 2nd edition, 2002.
  • Singh, R. P., Computer Applications in Food Technology, 1996.

Unit 5: Machine Vision and Robotic Inspection

Note

Unit 5: Machine Vision and Robotic Inspection

A question to start with

A line packs dried fruit. The pieces move fast on a belt, hundreds every minute. Some pieces are the right colour and size. Some are too dark, which means they are over-dried. Some are broken. Some are not fruit at all: a small stone, a piece of stem, a bit of plastic from a torn bag. A person cannot look at every piece at that speed, and a person who tries will miss pieces and tire quickly. A camera can look at every piece, but a camera alone does not know what a good piece looks like. This unit is about machine vision: how a camera, a light, and a computer learn to see food, judge it, and tell a robot or a reject device what to do.

Machine vision is a system that takes a picture of food or a pack, measures something in that picture, and makes a decision from the measurement. The system has four parts: lighting, a camera, a computer with software, and an action, such as a reject, a sort, or a robot pick. If any part is weak, the whole system is weak. Good lighting matters as much as a good camera. A clear picture taken in bad light is still a bad picture.

Inspection is the use of vision, or of other sensors, to check food and packs against a standard. The standard is a plain description of what is acceptable: the right colour, the right size, the right shape, the right label, the right seal, the right weight. The inspection system must be taught that standard, tested with real good and real bad pieces, and checked often, because light, dust, and product change can all move the result.

Questions to think about

  • A camera sees colour. What does it need to be taught before it can tell a good piece of dried fruit from a bad one?
  • Why would a vision system that works in the morning fail in the afternoon, with no change to the food?
  • If a vision system rejects good food, is that as serious as accepting bad food, and why?
Think about the dried fruit before you read on

The camera needs examples: pictures of good pieces and pictures of bad pieces, so the software can learn the difference in colour, size, and shape that marks a bad piece. It can fail in the afternoon because the light changed: sunlight through a window, a lamp warming up, or a shadow can change the colour the camera sees. Rejecting good food wastes product and money, but accepting bad food can send a stone, a broken piece, or a wrong label to a buyer, which is a safety and a trust fault. Both errors matter, and a good system counts both.

Light, camera, and picture: the seeing part

Light is the first part of vision, and it is the part most often ignored. Food must be lit evenly, with no bright spots and no shadows, so that the camera sees the same colour and shape for the same piece every time. A light that flickers, a window that lets in sunlight, or a lamp that is dirty will change the picture. For that reason, vision systems use their own lights, in a housing that blocks outside light, and the lights are cleaned and checked on schedule.

The camera takes the picture. A simple camera takes a colour picture, like a phone camera. A monochrome camera takes a picture in shades of grey, which can be enough for size and shape, and it is often faster. Some systems use other kinds of light: infrared light can show moisture or a bruise under the skin that normal light hides, and X-ray can show a stone or a piece of metal inside a pack. The choice of camera depends on what must be seen: colour, size, shape, a label, a seal, or something hidden inside.

The picture is made of pixels, small squares of colour or brightness. The software measures the pixels: how many pixels the piece covers, which tells its size; what colour the pixels are, which tells ripeness or browning; where the edges are, which tells shape. A good picture has enough pixels to measure the smallest fault that matters. A picture with too few pixels will miss small stones, small cracks, and small print errors.

Quick self-check

Question: Why does a vision system use its own lights in a housing, instead of the room light?

Because room light changes with the time of day, the weather, and the lamps. The system's own lights give the same light every time, so the same food gives the same picture, and the decision does not change with the sunlight.

Question: A system must find a small stone in dried fruit. What does the picture need?

It needs enough pixels to show the stone, and a light and a camera that make the stone look different from the fruit, in colour or in brightness. If the stone and the fruit look the same in the picture, no software can separate them.

What vision measures: colour, size, shape, and surface

Vision systems measure four things, and most food checks use at least two of them. Colour tells ripeness, browning, cooking, and freshness: a green tomato is not ripe, a brown apple slice is oxidised, a pale biscuit is under-baked. Size tells whether a piece will fit its pack and cook evenly: a large piece and a small piece in the same pack will cook at different rates. Shape tells whether a piece is whole, broken, or wrongly cut: a broken biscuit, a bent chip, or a misshapen fruit can be found by its outline. Surface tells whether a piece has a crack, a spot, a mould patch, or a missing topping: the surface pattern is measured from the picture, and a patch that does not match the normal pattern is flagged.

Each measure needs a limit. Colour needs a range: too dark is bad, too light is bad, the middle is good. Size needs a smallest and a largest acceptable size. Shape needs a description of the outline that counts as whole. Surface needs a rule for how large a spot can be before the piece is rejected. These limits are set by testing real good and real bad pieces, not by guessing. The limits are written down, so that a later change can be traced and the system can be set back to its tested limits if someone changes them.

A single measure can mislead. A dark piece may be a good piece of a dark variety, or a bad piece that is burnt. A small piece may be a good piece of a small variety. For that reason, good systems use two or more measures together: colour plus size, or shape plus surface. A piece is rejected only when the combined measures say it is bad.

Quick self-check

Question: Why use colour plus size, rather than colour alone, to judge a tomato?

Because colour alone can mislead: a small green tomato and a large green tomato are both unripe, but a small red tomato may be a good small variety while a large red tomato is a good large variety. Size and colour together give a fuller picture of ripeness and type.

Question: Who should set the limits for a vision check, and how?

The food technologist and the quality team should set them, by testing real good and real bad pieces and recording the limits that separate them. Limits set by guessing, or by one person alone, will reject good food or accept bad food.

Inspection of packs: label, seal, fill, and code

Vision also inspects packs, and pack inspection is often where vision pays back first, because the pack is a fixed shape with printed information. A camera checks that the label is present, that it is the right label, that it is straight, and that the date and batch code are printed and readable. Optical character reading is the name for software that reads printed letters and numbers from a picture: it can read a date code and check that the date is correct, and it can read a batch code and check that it matches the run.

A camera checks the seal by looking at its width, its straightness, and whether food or liquid is caught in it. A camera checks the fill by looking at the level in a jar or the count in a tray. A camera checks the cap by looking at whether it is present, straight, and fully closed. Each of these checks is a picture with a limit, and each limit is tested with real good and real bad packs.

Pack inspection must be fast, because the line is fast, and it must be reliable, because a missed wrong label or a missed failed seal can reach a buyer. A common design is to put the camera where the pack is held still for a moment, or where the belt is steady, so the picture is sharp. A blurred picture, taken while the pack shakes, will give a wrong decision, and the system will reject good packs or accept bad ones.

Quick self-check

Question: A date code is printed but smudged. Should the vision system accept the pack?

No. A date that cannot be read cannot be checked, and a buyer cannot read it either. The pack should be rejected, and the printer should be checked, because a smudged code points to a printer fault that will affect more packs.

Question: Why check the label and the seal with a camera, when a person could check them?

Because a person cannot check every pack at line speed without tiring and missing packs. A camera checks every pack, at the same standard, and it records the result, so the check is complete and can be proved later.

From picture to action: sorting, rejecting, and picking

A picture is only useful if it leads to an action. In sorting, the vision system sends a signal to a device that moves the piece: an air jet blows a bad piece off the belt, a flap diverts it to another belt, or a robot picks it and places it in the right tray. In rejecting, the system sends a signal to a reject device that pushes a bad pack off the line into a reject bin. In picking, the system sends the position of each good piece to a robot, as in Unit 3, so the robot can pick it.

Timing matters in all three actions. The piece keeps moving after the picture is taken, so the system must work out where the piece will be when the air jet, the flap, or the robot reaches it. If the timing is wrong, the jet misses, the flap diverts a good piece, or the robot picks at empty air. The belt speed, the camera, and the action device must be timed together, and the timing must be tested at the real line speed, not at a slow test speed.

The action must also be counted. A count of rejects, sorted by reason, shows what is going wrong: too many dark pieces points to the dryer, too many broken pieces points to the cutter, too many wrong labels points to the labeller. A system that rejects without counting hides the cause of the fault.

From a picture to an action

Light the food evenly, block outside light
↓
Camera takes a picture of each piece or pack
↓
Software measures colour, size, shape, and surface
↓
Decision: good or bad, by the set limits
↓
Action: sort, reject, or send the position to a robot, and count the result

Quick self-check

Question: A vision system rejects many packs, but nobody knows why. What is missing?

A count by reason. The system should record why each pack was rejected: wrong label, failed seal, low weight, or unreadable date. Without that count, the team cannot tell which machine is at fault.

Other inspection tools: X-ray, metal detection, and weighing

Vision is not the only inspection tool, and some faults cannot be seen in a normal picture. X-ray inspection looks inside a pack: it can find a stone, a piece of glass, or a piece of metal hidden inside the food, and it can check the fill level in a sealed pack. It works by passing X-rays through the pack and measuring how much is absorbed: dense material absorbs more, and the picture shows it as a darker or brighter area. X-ray systems are used where the risk of a hidden foreign body is high, and they are checked with test pieces of a known size, so that the smallest piece they must find is proved to be found.

Metal detection uses a magnetic field to find metal in food. It is common at the end of a line, after packing, and it rejects packs that contain metal. It is checked with test pieces of metal of a known size, passed through the detector on a regular schedule. A metal detector that is not checked can drift and miss metal, or reject good packs.

Weighing is also inspection. A checkweigher, described in Unit 1, weighs each pack and rejects packs that are too light or too heavy. A light pack may be missing a piece. A heavy pack may have a double fill. Weighing is simple, fast, and reliable, and it catches faults that a camera cannot see, because a camera cannot weigh.

The best inspection uses more than one tool, because each tool sees a different fault. A camera sees colour and label. An X-ray sees inside. A metal detector sees metal. A checkweigher sees weight. Together they cover more faults than any one tool alone.

Quick self-check

Question: Why check an X-ray system with a test piece of a known size?

Because the check proves the system can find a piece of that size, every time. Without a test piece, nobody knows whether the system is still sensitive enough, or whether it has drifted and now misses small pieces.

Question: A camera says a pack is good, but the checkweigher says it is too light. Which one is right?

Both may be right about what they measure: the camera sees a good label and a good seal, and the weigher finds a missing piece inside. The pack is bad because it is light, even though it looks good. That is why more than one check is needed.

Testing a vision system: good pieces, bad pieces, and false results

A vision system must be tested before it is trusted, and the test must use real food and real packs, not only pictures from a catalogue. The test set has three groups: good pieces that must be accepted, bad pieces that must be rejected, and borderline pieces that are near the limit. The system is run on all three, and the results are counted.

Two errors matter. A false reject is a good piece that is rejected. It wastes product, and too many false rejects will make the team switch the system off. A false accept is a bad piece that is accepted. It sends bad food to a buyer, and too many false accepts mean the system is not doing its job. A good system keeps both errors low, and the balance between them is a decision: where safety is at stake, a false reject is better than a false accept, and the limits are set to catch every bad piece even if a few good pieces are rejected with them.

Testing does not end at installation. Light changes, lenses get dirty, products change, and software is updated. A vision system should be checked on a schedule, with the same test pieces, and the results recorded. A check that is done only when the system fails is too late: the line may have packed bad food for hours before the failure was noticed.

Two errors in every inspection

False reject: a good piece is rejected, product is wasted
↓
False accept: a bad piece is accepted, bad food reaches a buyer
↓
Test with good, bad, and borderline pieces, and count both errors

Quick self-check

Question: Why test a vision system with borderline pieces, not only clear good and clear bad pieces?

Because borderline pieces are where the system is most likely to be wrong. A system that passes clear cases may still fail on pieces near the limit, and those are the pieces that will be wrongly accepted or wrongly rejected on the line.

Try it yourself

Take ten pieces of one fruit or vegetable, such as apples or potatoes, and sort them by eye into good, bad, and borderline, using colour, size, shape, and surface. Write down the rule you used for each group in one sentence. Then ask a classmate to sort the same ten pieces using your rules, without hearing your decisions. Compare the two sorts: where did you agree, and where did you differ? What does that tell you about writing a clear limit for a vision system?

Key ideas to carry forward

  • Machine vision takes a picture, measures it, and makes a decision: lighting, camera, software, and action must all work together.
  • Good lighting is even, steady, and shielded from outside light; bad light changes the picture and the decision.
  • Vision measures colour, size, shape, and surface; each measure needs a limit set by testing real good and real bad pieces.
  • Pack inspection checks the label, the date and batch code, the seal, the fill, and the cap, and a smudged or unreadable code is a reject.
  • A picture must lead to an action: sort, reject, or send a position to a robot, and the action must be timed to the moving belt.
  • X-ray sees inside a pack, metal detection finds metal, and a checkweigher finds a light or heavy pack; each tool sees a different fault.
  • Two errors matter: a false reject wastes good food, and a false accept sends bad food to a buyer.
  • A vision system is tested with good, bad, and borderline pieces, and checked on a schedule with the same test pieces.

Glossary

False accept

When an inspection system accepts a bad piece or pack. Food tech example: a stone in dried fruit passes the camera check and reaches a pack.

False reject

When an inspection system rejects a good piece or pack. Food tech example: a good biscuit is rejected because a shadow changed its colour in the picture.

Infrared

Light beyond red that the eye cannot see, which can show moisture or a bruise under the skin. Food tech example: infrared light shows a bruise inside an apple that normal light hides.

Inspection

Checking food or packs against a standard, using vision or other sensors. Food tech example: a camera inspects each jar for a missing label before it is packed.

Machine vision

A system that takes a picture, measures it with software, and makes a decision from the measurement. Food tech example: a vision system measures the colour of each biscuit and rejects biscuits that are too dark.

Metal detector

A device that uses a magnetic field to find metal in food or in a pack. Food tech example: a metal detector at the end of a line rejects a pack that contains a small piece of metal.

Optical character reading

Software that reads printed letters and numbers from a picture. Food tech example: a vision system reads the date code on a pack and checks that the date is correct.

Pixel

A small square of colour or brightness that makes up a picture. Food tech example: a picture with more pixels can show a smaller crack in a biscuit.

Reject

To remove a bad piece or pack from the line, by a jet of air, a flap, or a robot. Food tech example: a checkweigher rejects a pack that is too light, pushing it off the belt into a reject bin.

Test piece

A piece of a known size and material used to check that an inspection system can find it. Food tech example: a small metal test piece is passed through a metal detector on a schedule to prove the detector still works.

X-ray inspection

Inspection that passes X-rays through a pack to see dense material hidden inside. Food tech example: X-ray inspection finds a small stone inside a sealed bag of rice.

Sorting

Separating food into groups by a measure such as colour, size, or shape. Food tech example: a sorter separates ripe tomatoes from unripe ones by colour.

References

  • Irudayaraj, J., Food Processing Operations Modeling, Design, and Analysis, 2nd edition, 2002.
  • Singh, R. P., Computer Applications in Food Technology, 1996.

Unit 6: Safety, Hygiene and Working Beside Robots

Note

Unit 6: Safety, Hygiene and Working Beside Robots

A question to start with

A worker reaches into a packing area to clear a jammed tray. The robot arm is still moving. The worker's hand is in the path of the arm, and the arm does not know the hand is there. In another plant, a different worker does the same job, but the robot stops the moment the guard door opens, and the worker clears the jam safely. The difference is not luck. It is design: guards, sensors, stops, training, and a cleaning plan that were thought through before the line ever ran. This unit is about safety and hygiene around robots: how to keep people safe, how to keep food clean, and how to work beside a machine that is strong, fast, and blind to what it cannot sense.

Safety in a food plant has two sides that must both be held. Worker safety is about people: keeping them from being struck, trapped, cut, burnt, or hurt by a moving machine, by electricity, or by hot, cold, or wet work. Food safety is about the product: keeping microbes, chemicals, and foreign bodies out of food, and keeping the food at the right temperature and in the right condition. A robot affects both sides. It can remove people from a dangerous job, and it can also create a new danger if it is not guarded. It can cut hand touches on food, and it can also carry dirt into a clean area if it is not cleaned.

The rule that runs through this unit is simple: safety and hygiene are designed in before the robot arrives, not added after the first accident or the first failed hygiene check. A guard fitted after a near miss is better than no guard, but a guard planned from the start is cheaper, stronger, and easier to work with.

Questions to think about

  • A robot stops when a guard door opens. Why is that stop not enough, on its own, to make the work safe?
  • A robot is washed down every day. What could still make it a hygiene risk, even if it looks clean?
  • If a worker and a robot share a bench, who decides what the robot may do, and how is that decision written down?
Think about the jammed tray before you read on

The stop on the door is not enough on its own because the arm may still be moving when the door opens, or the worker may reach in through a gap, or the robot may restart while the worker is still inside. Safety needs the guard, the stop, a way to lock the robot off during cleaning and fixing, training, and a rule that nobody enters the guarded area until the robot is stopped and locked. A robot can look clean and still carry microbes in a joint, a fold, or a scratch where water sits, so cleaning must reach those places and be checked. When a worker and a robot share a bench, the risk check decides what the robot may do, and that decision is written in the risk assessment and in the safe working steps for that job.

Hazards around a robot

A hazard is something that can cause harm. Around a robot, the main hazards are easy to name, and naming them is the first step to controlling them. Impact is the arm, the tool, or the load striking a person. Trapping is a person caught between the arm and a fixed surface, such as a wall, a table, or another machine. Cutting is a sharp tool, a blade, or a sharp edge on a gripper or a pack. Burns are from hot food, hot surfaces, steam, or a heater. Cold is from a freezer or a chilled room. Electricity is from the robot, its controller, and its cables. Slips and falls are from water, oil, or food on the floor after washing. Noise is from the line, the air jets, and the machines.

Each hazard has a place and a time. Impact and trapping happen where the arm moves, and when the line is running. Cutting happens where a blade or a sharp tool is fitted. Burns happen near a heater, a cooker, or hot food. Slips happen after washing, when the floor is wet. A risk assessment looks at each hazard, asks who could be hurt and how, and decides what control will prevent it. The assessment is written down, and it is reviewed when the product, the tool, or the layout changes.

Quick self-check

Question: Why name the hazard, the person, and the time, rather than only saying "robots are dangerous"?

Because a general warning cannot be fixed. Naming the hazard, who could be hurt, and when it could happen points to a specific control: a guard for the moving arm, a lock for cleaning, a non-slip floor after washing. Without that detail, the fix is a guess.

Question: A robot works in a freezer. What hazards does the cold add, besides the moving arm?

Cold can numb hands, slow movement, and make a worker less able to react. It can also make surfaces slippery with ice, and it can make some materials brittle. The cold needs its own controls: warm clothing, short work periods, and a floor that is kept free of ice.

Guards, fences, sensors, and stops

Guards are the physical barriers that keep people out of the space where a robot moves. A fence with a locked gate, a light curtain that stops the robot when a beam is broken, a pressure mat that stops the robot when someone steps on it, and a guard door with a switch that stops the robot when it opens are all guards. They are chosen by the hazard and by the space: a fence suits a large fast robot, a light curtain suits a place where packs must pass through but people must not, a pressure mat suits a floor area where a person should not stand while the robot runs.

A stop is a device that halts the robot. An emergency stop button is a large red button that anyone can press to stop the robot at once. It must be easy to reach, easy to see, and it must stay pressed until it is reset by a person, not restart by itself. A safety stop is triggered by a guard: when the door opens, the light curtain is broken, or the mat is stepped on, the robot stops. A controlled stop brings the arm to a halt in a set way, so that it does not drop its load. The stop must be tested on a schedule, because a stop that has never been tested may not work when it is needed.

Lockout is the rule that a robot is locked off before anyone enters its space to clean, fix, or clear a jam. A lock is fitted to the power switch, the key is held by the person inside, and the robot cannot be restarted until that person has left and removed the lock. This rule protects against the most common serious accident: a robot restarted by someone outside while a worker is still inside.

Quick self-check

Question: Why must an emergency stop stay pressed until a person resets it?

Because a stop that resets itself could restart the robot while the danger is still there. A person must check the area, make it safe, and then reset the stop, so the restart is a deliberate act.

Question: A guard door has a switch that stops the robot. Why is a lock also needed for cleaning?

Because the switch only stops the robot while the door is open. If someone closes the door, or restarts the robot from outside, the robot could move while the cleaner is still inside. A lock on the power, held by the person inside, prevents any restart until that person has left.

Collaborative work: sharing space safely

Collaborative robots, introduced in Unit 2, are designed to work in the same space as people, but sharing space is not safe by name alone. The risk check must cover the whole job: the robot, the tool, the load, the speed, and the food. A cobot with a sharp cutter is still a cutting hazard. A cobot carrying a hot tray is still a burn hazard. A cobot carrying a heavy box can still strike a person hard enough to hurt.

Safe collaborative work uses limits. Speed is limited, so that contact, if it happens, is slow. Force is limited, so that contact, if it happens, is soft. The tool is chosen to be safe to touch: no sharp edge, no hot surface, no pinch point. The work is arranged so that the person and the robot are not reaching into the same space at the same time: the robot works on one side, the person works on the other, or they take turns. Sensors that detect contact and stop the robot are tested, and the test is recorded.

Training is part of collaborative safety. A worker who shares a bench with a cobot must know what the robot will do next, where it will move, when it will stop, and how to stop it. A worker who is surprised by a robot will step back, drop a tray, or reach into the wrong place. A short training session, with the robot moving slowly at first, lets the worker learn its pattern before the line runs at full speed.

Layers of protection around a robot

1. Guard: fence, door, light curtain, or mat keeps people out
↓
2. Stop: safety stop and emergency stop halt the robot
↓
3. Lockout: lock on the power, key held by the person inside
↓
4. Training: everyone knows the pattern, the stops, and the safe steps
↓
5. Check: stops, guards, and sensors tested on a schedule and recorded

Quick self-check

Question: A cobot is safe to touch. Can it carry a hot tray of food?

It can carry it, but the hot tray is still a burn hazard for a person who touches the tray, even if the arm itself is safe to touch. The risk check must cover the tray and the food, not only the arm, and the work must be arranged so the person does not touch the hot tray.

Hygiene: food contact, materials, and design

A robot that touches food, or that works above open food, must be designed for hygiene. The materials matter: food contact parts should be smooth, non-absorbent, resistant to corrosion, and easy to clean. Stainless steel is common for food contact parts, because it is smooth, strong, and can be washed. Plastics used for food contact must be food grade, which means they are made for contact with food and do not release harmful substances. Painted surfaces, rough welds, and open screw heads are risks, because they can flake, trap dirt, or hold water.

The design matters as much as the material. A food-grade robot has smooth covers, sealed joints, no horizontal ledges where water can sit, and no hollow parts where food and water can collect. Cables are covered or run inside the arm, so they do not hang over open food. The gripper is designed so that it can be taken off, washed, and put back without tools that are hard to clean. A robot that is hard to clean will not be cleaned well, however good the cleaning plan is on paper.

Zoning is the idea of keeping dirty and clean areas apart. Raw material areas, where soil, crates, and raw meat or fish are handled, are kept apart from ready-to-eat areas, where cooked or packed food is handled. A robot that moves between zones can carry contamination from one to the other, so a robot is placed in one zone, and food moves between zones through a controlled transfer, not by the robot crossing the line. People, tools, and cleaning equipment are also zoned: a brush used in a raw area is not used in a clean area.

Quick self-check

Question: Why is a horizontal ledge on a robot a hygiene problem?

Because water and food can sit on it, and microbes can grow in the standing water and food. A ledge that cannot drain will stay wet after washing, and the next batch of food can be contaminated from it.

Question: A gripper is used for raw chicken and then for cooked chicken. What is the risk?

Microbes from the raw chicken can be carried on the gripper to the cooked chicken, which will not be cooked again to kill them. The gripper must be cleaned, or a separate gripper must be used, and the two areas must be kept apart.

Cleaning the robot and the line

Cleaning a robot is a planned job, not a quick wipe. The robot is stopped and locked out, the gripper is removed if it can be, loose food is removed, the surfaces are washed with a cleaning solution, rinsed, and dried, and the gripper is cleaned and checked before it is put back. The cleaning steps are written down, with the solution, the time, the temperature, and the person who did the work. The record is kept, so that a later hygiene check can see that cleaning was done.

Water is both a cleaning tool and a risk. High-pressure water can drive dirt and microbes into joints, bearings, and electrical parts, and it can splash dirty water onto clean surfaces. For that reason, food robots are rated for washing: the rating shows how much water, and from which direction, the robot can take. A robot that is not rated for high-pressure washing must be cleaned with a gentler method, and a robot that is washed with a pressure jet it was not built for will fail early and may become a hygiene risk.

Drying matters too. A robot that is left wet will hold water in joints and folds, and microbes grow in standing water. After washing, the robot and its area should be allowed to drain and dry, and the floor should be squeegeed so that water does not pool. A cleaning check, by eye and by touch, looks for food, water, and damage: a cracked cover, a torn seal, or a worn gripper that can trap dirt must be reported and fixed.

Working beside robots: roles, training, and daily habits

Working beside a robot is a skill, and it is learned. The operator learns the normal pattern: what the robot does, in what order, and at what speed. The operator learns the safe steps: how to start, how to stop, how to clear a jam, how to clean, and how to call for help. The operator learns the signs of a fault: a new noise, a slower move, a gripper that drops pieces, a sensor that alarms often. Reporting a small change early is better than waiting for a stop, because a small change is often the first sign of wear.

Daily habits keep the work safe. Before starting, the operator checks that guards are in place, that the emergency stop works, and that the area is clear. During the run, the operator watches the flow and the count, and does not reach into the guarded area while the robot is running. At the end of the run, the operator follows the cleaning steps, records the cleaning, and reports any fault or damage. These habits are written as safe working steps for each job, and new workers are trained on them before they work alone.

Visitors and new workers need the same care. A visitor who does not know the line may walk into a guarded area, or reach for a pack on a moving belt. Visitors are told where they may stand, what they may touch, and who to ask. A plant that treats visitors as part of safety, not as an exception to it, avoids the accident that happens to the person who was never told.

Quick self-check

Question: An operator hears a new noise from a robot but the line is still running. What should the operator do?

Report it at once, and note when it started and what the robot was doing. A new noise is often the first sign of wear or a loose part, and catching it early can prevent a stop or a failure later. The operator should not wait for the line to stop on its own.

When something goes wrong: stops, faults, and first aid

Even a well-guarded line will stop, and the response to a stop must be planned. A jam is cleared only after the robot is stopped and locked out. A fault is read from the controller screen, and the fault code is written down with the time and the action taken. A repeated fault is reported, because a fault that repeats is a sign of a deeper problem: a worn gripper, a dirty sensor, or a program that is wrong for the product.

Injury, if it happens, needs a first aid plan: who gives first aid, where the kit is, who calls for help, and how the accident is reported. An accident report is not blame: it is a record that lets the plant find the cause and fix it, so the same accident does not happen again. A plant that hides accidents will repeat them. A plant that records them, and acts on them, gets safer.

A near miss, where nobody was hurt but someone could have been, is also recorded. A near miss is a free lesson: it shows a gap in the guard, a step that was skipped, or a training need, without the cost of an injury. The team that reports near misses is the team that prevents the next accident.

Try it yourself

Walk through a kitchen or a food shop in your mind, or visit one, and list five hazards you can see: a hot surface, a wet floor, a sharp knife, a heavy box, a moving belt. For each hazard, write who could be hurt, how, and one control that would prevent it. Then pick one hazard and write the safe working steps for it in three lines: what to do before starting, what to do while working, and what to do if something goes wrong. Compare your steps with a classmate: are your controls practical, and would a new worker understand them?

Key ideas to carry forward

  • Worker safety and food safety must both be held: a robot can remove people from danger and can also create a new danger.
  • Safety and hygiene are designed in before the robot arrives, not added after the first accident.
  • Hazards around a robot include impact, trapping, cutting, burns, cold, electricity, slips, and noise.
  • Guards keep people out of the robot's space: fences, locked gates, light curtains, pressure mats, and guard doors.
  • Emergency stops must stay pressed until a person resets them; safety stops are triggered by guards.
  • Lockout locks the power off, with the key held by the person inside, before anyone enters the guarded area.
  • A collaborative robot is safe only for the whole job, including the tool, the load, and the food, with limits on speed and force.
  • Food contact parts must be smooth, washable, and food grade, with no ledges or hollows where water and food can sit.
  • Zoning keeps raw and ready-to-eat areas apart, and a robot should not carry contamination between zones.
  • Cleaning is a planned, recorded job, with the right water pressure for the robot's rating, and drying after washing.
  • Daily habits, training, and reporting small changes and near misses keep the line safe.
  • Accidents and near misses are recorded so the cause can be found and fixed.

Glossary

Emergency stop

A large button that stops a robot at once and stays pressed until a person resets it. Food tech example: an operator presses the emergency stop when a tray jams and a hand is near the moving arm.

Food grade

Made for contact with food, without releasing harmful substances. Food tech example: a food grade plastic gripper can touch bread without risk to the food.

Guard

A physical barrier that keeps people out of a dangerous space. Food tech example: a fence around a palletising robot is a guard.

Hazard

Something that can cause harm to a person or to food. Food tech example: a moving robot arm is a hazard for impact, and raw chicken is a hazard for contamination.

Hygiene

Practices that keep food clean and free from contamination. Food tech example: washing hands, cleaning tools on schedule, and keeping raw and ready-to-eat food apart are hygiene practices.

Light curtain

A set of light beams across an opening that stops a machine when a beam is broken. Food tech example: a light curtain across the entrance to a robot cell stops the robot when a person steps through.

Lockout

Locking a machine off, with the key held by the person working on it, so it cannot be restarted. Food tech example: a cleaner locks out a robot before entering its cell to wash it.

Near miss

An event where nobody was hurt but someone could have been. Food tech example: a worker steps into a robot cell just as the arm stops, with no injury, is a near miss that must be reported.

Risk assessment

A written check that names hazards, who could be hurt, and what controls will prevent harm. Food tech example: a risk assessment for a packing robot names impact, trapping, and cutting, and lists the guard and the lockout as controls.

Risk check

The process of looking at a job, naming its hazards, and deciding what controls are needed. Food tech example: a risk check for a cobot covers the arm, the tool, the load, and the food.

Safe working steps

Written steps for a job that tell a worker what to do before starting, while working, and if something goes wrong. Food tech example: the safe working steps for clearing a jam say to stop, lock out, clear, and then restart.

Zoning

Keeping dirty and clean areas apart in a food plant. Food tech example: raw meat and ready-to-eat food are kept in separate zones, with separate tools and separate cleaning equipment.

References

  • Irudayaraj, J., Food Processing Operations Modeling, Design, and Analysis, 2nd edition, 2002.
  • Singh, R. P., Computer Applications in Food Technology, 1996.

Unit 7: Costs, Payback and Choosing the Right Automation

Note

Unit 7: Costs, Payback and Choosing the Right Automation

A question to start with

A small food plant packs jam in jars. The manager is offered a robot that will fill, cap, and label the jars. The price is high, and the manager must decide. The robot will pack faster than the team does now. It will also need power, cleaning, spare parts, a new gripper when the jar size changes, and training for the team. It will save the cost of some hand work, but it will not save the cost of the fruit, the sugar, the jars, or the energy to cook the jam. If the robot breaks down for a week, the plant must still pack jam. Is the robot a good buy? The answer is not in the price tag alone. It is in the full cost, the full saving, and the risk if the plan is wrong. This unit is about that decision: what automation costs, what it saves, how payback is worked out, and how to choose the right automation for a real plant.

A decision about automation is a decision about money, people, food, and time, all together. Money is the most visible part, but it is not the only part. A robot that saves money but damages soft fruit, or that needs a skill the plant does not have, or that stops the line for a long time when it fails, is not a good buy even if the numbers on paper look good. A good decision counts the costs that are easy to see and the costs that are easy to miss, counts the savings that are real and the savings that are only hoped for, and tests the plan with the real food before the money is spent.

Questions to think about

  • A robot costs a large sum to buy, but it works for years. How should that cost be spread across the years it will work?
  • If a robot packs faster but needs an hour of cleaning and setting up each day, is it faster over a whole day, and how would you prove it?
  • A supplier says a robot will pay for itself in a year. What would you ask to see before you believe that claim?
Think about the jam plant before you read on

The purchase cost should be spread across the years the robot will work, together with the running costs for each year: power, cleaning, parts, and training. A robot that is fast for an hour but needs an hour of cleaning and setting up each day may be slower over a whole day than a slower machine that needs little setting up, so the proof is a timed day, not a timed minute. Before believing a payback claim, ask to see the full list of costs and savings, the hours the line will actually run, the price of the labour saved, and a trial with the real jam, the real jars, and the real team.

What automation costs: the full list

The cost of automation starts with the purchase, but it does not end there. The purchase cost covers the robot, its controller, its tool, and its software. It may not cover the extras that make it work: the guard, the fence, the conveyor, the feeder, the camera, the tray supply, the installation, and the changes to the floor, the power, and the water supply. These extras can be a large part of the total, and a quote that leaves them out will look cheap and then grow.

Running costs follow every year. Power runs the robot, the conveyors, the lights, and the air supply. Cleaning uses water, cleaning solutions, and time. Spare parts wear out: grippers, suction cups, belts, filters, and seals all need replacing on schedule. Maintenance is the planned work that keeps the machine running: checking, lubricating, testing stops, and replacing worn parts before they fail. Training is the time spent teaching the team to run, clean, fix, and change over the machine, and teaching new workers when the team changes. Downtime is the cost of the line standing still: lost production, wasted food that was in the line when it stopped, and overtime to catch up.

There are also costs that are easy to miss. Changeover time is a cost, because the line is not packing while it is being changed. A new product may need a new gripper, a new tray, or a new program, and each of those costs time and money. A robot that needs a specialist to fix it, who must travel to the plant, will be stopped longer than a robot the team can fix. A robot that needs a part that takes weeks to arrive will be stopped longer than a robot whose parts are held in stock. These costs belong in the decision, not in a surprise after the purchase.

Quick self-check

Question: Why is the purchase price alone a poor guide to the cost of a robot?

Because the purchase price leaves out the guard, the conveyor, the installation, the power, the cleaning, the parts, the training, and the downtime. The full cost over the life of the machine is the number that matters, and it is larger than the price tag.

Question: A gripper wears out every few months. Is that a purchase cost or a running cost?

It is a running cost, because it repeats over the life of the machine. The decision should include the cost of each replacement and the time spent fitting it, for every year the robot will run.

What automation saves: the full list

Savings come from the same places as the pressures in Unit 1: speed, consistency, hygiene, safety, and labour. A faster line makes more packs in the same time, so the cost of each pack falls, provided the extra packs can be sold and the raw material is available. A more even line wastes less food and less pack material, because fewer packs are rejected for low weight, bad seals, or missing pieces. A cleaner line has fewer touches on open food, which can cut waste and rework. A safer line has fewer injuries, less strain, and less time lost to absence. A line that needs fewer people on each shift saves labour cost, but only if those people are moved to other useful work or if the plant truly needs fewer hours.

Savings must be counted honestly. Labour saved is only a saving if the hours are really saved: if the same people are still paid for the same hours, but now stand and watch, the saving is not real. Speed gained is only a saving if the line is fed, the packs are sold, and the rest of the line can keep up. Waste cut is only a saving if the waste is measured before and after, so the cut can be proved. A saving that is guessed, or that depends on the line running at a speed it never reaches, will not appear in the accounts.

There are also savings that are hard to put a number on, but that still matter. A line that is easier to clean is more likely to be cleaned well. A line that records its work gives a record that helps with traceability and with customer checks. A team that learns to run a robot gains a skill that helps the plant in other ways. These gains should be noted in the decision, even if they are not put into the payback sum, so that the decision is not made on money alone.

Quick self-check

Question: A robot replaces two workers on a shift, but both workers are moved to other jobs in the same plant. Is the labour cost saved?

Not as a cash saving for the plant, because the plant still pays both workers. The gain is that two people are now free for other work, which may be worth more, but the payback sum should not count their wages as saved unless their hours are truly cut or they leave and are not replaced.

Question: Why must waste be measured before and after automation, not only after?

Because without a before measure, nobody knows whether waste fell, rose, or stayed the same. A line that claims to cut waste must show the before and after counts, with the same product and the same measure, to prove the cut is real.

Payback: a plain way to work it out

Payback is the time it takes for the savings from a machine to equal its cost. The plain way to work it out is to add up the full cost, add up the saving each year, and divide the cost by the yearly saving. The result is the number of years before the machine has paid for itself. A short payback means the machine pays quickly. A long payback means the plant waits a long time, and the risk is higher, because the product, the volume, or the team may change before the machine has paid.

An example in plain numbers shows the idea, without claiming these numbers fit any real plant. If the full cost is 100 units of money, and the saving each year is 25 units, the payback is four years: 100 divided by 25. If the saving each year is 50 units, the payback is two years. The numbers in a real decision must come from the plant: the real cost, the real hours, the real wage, the real waste, and the real volume. A payback worked out from guessed numbers is a guess, however neat the arithmetic looks.

Payback has limits. It does not show what happens after the machine has paid for itself, when it may keep saving for years. It does not show the value of the machine at the end of its life, or the cost of replacing it. It does not show risk: a machine with a short payback but a high chance of failing, or a product with a short life, may be riskier than a machine with a longer payback and a steady product. For that reason, payback is a guide, not a verdict. It is used with the other checks in this unit: the trial, the risk, and the fit with the plant.

Working out payback, step by step

1. Add up the full cost: purchase, extras, installation, training
↓
2. Add up the running cost each year: power, cleaning, parts, maintenance
↓
3. Add up the saving each year: labour, waste, rework, and extra packs sold
↓
4. Take the saving each year minus the running cost each year
↓
5. Divide the full cost by that yearly net saving: the answer is the payback in years

Quick self-check

Question: The full cost is 100 units and the net saving each year is 20 units. What is the payback?

Five years: 100 divided by 20. The machine must run for five years at that saving before it has paid for itself.

Question: Why is payback a guide, not a verdict?

Because it does not show what happens after payback, the value of the machine at the end of its life, or the risk that the product, the volume, or the team will change. A short payback with high risk may be worse than a longer payback with a steady product.

Choosing the right automation: a step-by-step choice

Choosing automation is a decision in steps, and skipping a step is how plants buy the wrong machine. The first step is to name the problem: what is slow, uneven, unsafe, unclean, or wasteful, and how is it measured now? A problem that is not measured cannot be proved to be fixed. The second step is to name the job: what exact step will the machine do, and what will stay with people? The third step is to name the food: its size, shape, softness, stickiness, temperature, and variation, and how it arrives. The fourth step is to look at the simple fixes first: better layout, better guides, better training, better tools, and better cleaning, because a simple fix may solve the problem without a robot.

The fifth step is to compare options: a robot arm, a simpler machine, a vision system, a better conveyor, or a change in the pack. Each option is judged by the same measures: cost, saving, payback, fit with the food, fit with the team, safety, hygiene, and changeover time. The sixth step is a trial with the real food, at real speed, in real conditions, with the real team watching. The trial counts good packs, damaged packs, missed picks, rejects, stops, and changeover time. The seventh step is the decision, written down with the costs, the savings, the payback, the trial results, and the risks, so that the decision can be checked later.

Starting small is often the right choice. One robot at the end of the line, stacking boxes, is a small step that teaches the team, proves the supplier, and shows whether the plant can run and clean a robot well. A whole line automated at once is a large step that leaves no room to learn. A plant that starts small, learns, and then adds the next step is more likely to end with a line that works than a plant that tries to automate everything in one purchase.

Quick self-check

Question: Why name and measure the problem before looking at machines?

Because a machine that is bought before the problem is measured may solve a different problem, or no problem at all. A measured problem, such as a count of rejected packs or a time lost to a slow step, gives a clear test: did the machine fix the measured problem?

Question: Why try the simple fixes first, before buying a robot?

Because a better guide, a better layout, or better training may fix the problem at low cost, and may be needed even if a robot is bought later. A robot fitted onto a bad layout will still be a bad line, only faster.

Risk, flexibility, and the life of the machine

Every automation decision carries risk, and the risk should be named, not hidden. Product risk is that the product changes, the volume falls, or the pack size changes, and the robot no longer fits. Supplier risk is that the supplier is slow to help, or that parts take a long time to arrive. Skill risk is that the team does not learn to run the machine, or that the one person who can fix it leaves. Food risk is that the food varies more than the trial showed, and the gripper or the camera fails on the softest or stickiest batch. Downtime risk is that the line stops for a long time when the robot fails, and there is no hand method to fall back on.

Flexibility is the answer to product risk. A robot that can be reprogrammed for a new pack size, a gripper that can be changed quickly, and a program that is saved and named clearly all make the machine useful for more products and for longer. A machine that only fits one product, in one size, is a risk if that product changes. The choice should ask: how long will it take to change this machine to the next product, and what new parts will that change need?

The life of the machine also matters. A robot that is well maintained, cleaned, and checked can work for many years. A robot that is not maintained will fail early, and its payback will never arrive. The plan should include the maintenance schedule, the spare parts to hold, the training for new workers, and the review date when the machine will be checked again: is it still paying, is it still safe, is it still clean, and does it still fit the product? A machine that is never reviewed can go on costing money long after it has stopped saving it.

Choosing automation, step by step

Name and measure the problem
↓
Name the exact job and the food
↓
Try the simple fixes first
↓
Compare options by cost, saving, fit, safety, hygiene, and changeover
↓
Trial with the real food at real speed, and count the results
↓
Decide in writing, with costs, savings, payback, trial results, and risks

The people in the decision

Automation changes jobs, and the people who do those jobs should be part of the decision. The operator knows where the line jams, where the food breaks, and where the time is lost. The setter knows how long a changeover really takes. The maintenance worker knows which parts wear out and which machines are hard to clean. The quality checker knows which faults reach the buyer. A decision made without these people will miss the faults they see every day, and the machine will be blamed for a problem that the team could have named in advance.

The decision should also say what will happen to the people whose jobs change. Will they move to checking, setting, cleaning, and fixing? Will they be trained for those jobs, and when? Will new jobs be created, and who will do them? A plant that answers these questions before the machine arrives keeps trust and keeps skill. A plant that does not answer them will lose both, and the machine will be run by a team that does not want it there.

This subject ends where it began: with the food, the people, and the line. A robot is a tool. A good tool, chosen for the right job, fitted to the real food, run by a trained team, kept safe and clean, and counted honestly, makes food better, work safer, and a plant stronger. A tool chosen for the wrong job, or run without training, cleaning, and counting, does none of those things. The skill of a food technologist is not to buy the newest machine, but to choose the right machine, for the right step, for the right food, and to prove that it works.

Try it yourself

Pick a food plant job you know, such as packing jam, stacking boxes, or sorting fruit. Write the decision in one page: the problem and how it is measured now, the exact job the machine would do, the food and how it arrives, the simple fixes you would try first, the full list of costs, the full list of savings, and the payback worked out with plain numbers you state as your own working example. Then list three risks and, for each risk, one control that would lower it. Bring your page to a classmate and ask them to challenge one number: can you defend it, or does it need a trial to prove it?

Key ideas to carry forward

  • The full cost includes purchase, extras, installation, power, cleaning, parts, maintenance, training, changeover, and downtime.
  • Savings come from speed, consistency, hygiene, safety, and labour, but only savings that are real and measured count.
  • Labour saved is only a saving if the hours are truly saved or moved to useful work.
  • Payback is the full cost divided by the net saving each year, and it is a guide, not a verdict.
  • Choose in steps: measure the problem, name the job and the food, try simple fixes first, compare options, trial with real food, and decide in writing.
  • Start small, learn, and then add the next step, rather than automating a whole line at once.
  • Risk includes product change, supplier delay, skill loss, food variation, and long downtime; flexibility and quick changeover lower product risk.
  • A machine needs a maintenance schedule, spare parts, training for new workers, and a review date.
  • The people who run the line know where it fails, and they should be part of the decision and trained for the jobs that change.

Glossary

Downtime

Time when a line or a machine is stopped and not producing. Food tech example: a line stopped for a robot fault loses production for every minute it stands still.

Flexibility

How easily a machine can be changed for a new product or pack size. Food tech example: a robot with a quick-change gripper and a saved program is flexible for a new tray size.

Full cost

All the costs of a machine over its life: purchase, extras, installation, running, maintenance, training, and downtime. Food tech example: the full cost of a packing robot includes the guard, the conveyor, the cleaning, and the spare grippers.

Maintenance

Planned work that keeps a machine running well, such as checking, lubricating, and replacing worn parts before they fail. Food tech example: a monthly check of a robot's stops and grippers is maintenance.

Net saving

The saving each year minus the running cost each year. Food tech example: if a robot saves 25 units a year and costs 5 units a year to run, the net saving is 20 units a year.

Payback

The time it takes for the savings from a machine to equal its cost. Food tech example: a machine that costs 100 units and saves 25 units a year has a payback of four years.

Rework

Work done again to fix a pack or a product that failed a check. Food tech example: packs with a failed seal that are opened, refilled, and resealed are rework.

Running cost

The cost of running a machine each year: power, cleaning, parts, and maintenance. Food tech example: power, cleaning solution, and replacement suction cups are running costs for a picking robot.

Spare part

A part held in stock so a worn or broken part can be replaced without waiting for delivery. Food tech example: spare suction cups are held in stock so a worn cup can be changed at once.

Trial

A test run of a machine with the real product, at real speed, in real conditions, before a purchase is decided. Food tech example: a trial runs the plant's own jam jars through a filling robot to count good and damaged packs.

Changeover time

The time spent changing a line from one product to another, during which the line is not producing. Food tech example: changing a jam line from one jar size to another takes time to change the gripper, the program, and the label.

Waste

Food or pack material that is lost or rejected and cannot be sold. Food tech example: biscuits broken by a gripper and packs rejected for low weight are waste.

References

  • Irudayaraj, J., Food Processing Operations Modeling, Design, and Analysis, 2nd edition, 2002.
  • Singh, R. P., Computer Applications in Food Technology, 1996.