Don’t Automate a Process You Don’t Control

Oct 9

Don’t Automate a Process You Don’t Control

Knowledge Publication 008 Knowledge Series 002 – Better Decisions Build Better Bakeries

Why successful industrial bakery automation begins with a stable and understood process.

Automation can make a controlled process faster and more repeatable. It can also make an uncontrolled process fail more consistently.

Executive Summary

Automation can improve capacity, consistency, safety and labour efficiency. But automation does not automatically create control. If a process is unstable, poorly understood or dependent on constant operator correction, adding automation may simply transfer the problem into a more expensive system. The automated equipment is then blamed for poor performance. More sensors are added. More programming is requested. More technical support is needed. But the real problem may still be the process itself. Before automating, we need to understand how the process should work, where the variation comes from, what operators are currently correcting and what happens when conditions move outside the acceptable range. A process does not need to be perfect before it can be automated. But it does need to be understood and controlled.

Automation Is Not the Same as Control

Automation and process control are related, but they are not the same thing. Automation performs an action without requiring the same level of manual intervention. Process control keeps the outcome within an acceptable range. You can automate a process that is not under control.

The machine will still operate.
The conveyor will still move.
The robot will still pick.
The dosing system will still add ingredients.

But that does not mean the correct outcome is being achieved. If the inputs vary, the method is unclear or the acceptable operating range has never been defined, automation may simply repeat the wrong process more efficiently.

Operators Often Hide the Variation

This is something that is easily missed. A manual process may look stable because experienced operators are constantly correcting it.

They adjust the speed.
They change the timing.
They straighten products on a conveyor.
They separate products that are touching.
They compensate for differences in dough condition.
They remove products before they create a blockage.
They know when something does not look right.
Many of these corrections are never recorded.
They have simply become part of how the operation works.

Management sees a functioning process. But the process may only function because operators are continuously absorbing the variation. When the manual task is automated, that flexibility disappears. The automation now receives the real process exactly as it is. And suddenly the hidden variation becomes visible.

Missed products.
Poor positioning.
Frequent stops.
Rejects.
Jams.
Reduced speed. The automation did not necessarily create the problem. It exposed it.

A Practical Bakery Example

Imagine a bakery wants to automate the loading of products into packaging. The manual operation requires several people, so a robotic loading system appears to be a logical investment. During normal production, however, the products do not always arrive in the same condition.

Spacing changes.
Orientation changes.
Some products touch each other.
Product temperature varies.
Occasionally products stick to the belt.

An experienced operator sees this and corrects it. The operator separates products, changes their orientation and removes anything that could disturb the packing process. A robot will not automatically understand all those situations.

We can add cameras.
We can add more sensors.
We can improve the programming.

And perhaps some of that will be necessary. But first we should ask why the products are arriving inconsistently.

Is the upstream process stable?
Is the transfer between conveyors correct?
Is the cooling process under control?
Is the belt speed properly coordinated?
Are the products leaving the oven with consistent dimensions?
Is the product strong enough to be handled automatically at that point?

If these questions are not answered, we may spend a lot of money teaching the automation to compensate for problems that should have been solved upstream.

Start With the Required Outcome

Before discussing equipment, define what the process is expected to achieve.

What should the output look like?
Which variation is acceptable?
Which variation is not acceptable?
What production rate is required?
How should the process behave during startup, normal production, changeovers and shutdown?
What happens when a product or condition moves outside the accepted range?

Without clear answers, it is difficult to design reliable automation. A supplier can automate the movement. But the operation still needs to define the required result.

Understand the Operating Window

Every process has an operating window.

Product temperature.
Dough consistency.
Piece weight.
Spacing.
Position.
Line speed.
Cooling time.
Moisture.
Dimensions.

The automation needs to work within the actual range of conditions found in production.

Not only at the average.

A system designed around one ideal product, one ideal speed and one perfect operating condition may perform very well during the acceptance test.

Reality starts when production changes.
Different products are introduced.
Raw materials vary.
The line starts and stops.
Belts become dirty.
Temperatures change.
Changeovers take place.
Operators work around a disturbance.

Good automation needs to be designed for real operating conditions.

But we can only do that if we understand those conditions first.

Do Not Design Around the Average

An average can hide the problem. An average product may be 100 millimetres wide. But perhaps the actual range is 92 to 108 millimetres. The average spacing may be 150 millimetres. But perhaps products sometimes arrive with 50 millimetres between them and sometimes with 250. The average line speed may look stable. But perhaps short interruptions happen every few minutes. Automation does not experience the average. It experiences every individual product and every disturbance. That means we need to understand the variation, not only the average value. The system must either handle that variation or the process must be improved until the variation is reduced.

Decide What the Automation Should Handle

Not every variation can or should be removed. Some variation is part of the product or process. The automation may need to recognise it and respond correctly. But that should be a conscious decision.

Which conditions should the automation accept?
Which conditions should trigger an adjustment?
Which products should be rejected?
When should the system stop?
How will production recover?
When is operator intervention required?

These decisions should be made before commissioning. Otherwise they will be discovered during production, one stoppage at a time.

A Process Does Not Need to Be Perfect

Saying that a process should be controlled does not mean it needs to be perfect. No industrial bakery process operates without variation. The objective is not to remove every disturbance. The objective is to understand the important variation and keep it within a range the automation can manage. That means we should know:

the required outcome
the acceptable operating range
the most important sources of variation
which disturbances the automation can handle
which conditions require intervention
how the process returns to normal operation

Once those things are clear, automation becomes much easier to design.

And much more likely to succeed.

Automation Can Lock In a Poor Method

There is another risk. Once a process has been automated, changing the operating method may become more difficult. The mechanical design has been completed. The controls have been programmed. The guarding has been installed. The line interfaces have been established. The operating sequence is now built into the system. If the original method was poor, we have made it more permanent.

That is why automation should not begin with:

“Which machine should we buy?”

It should begin with:

“Is this the correct process, and is it ready to be automated?”

Improve the method first. Then automate the improved method.

What Needs to Change

Before automating a process, ask:
What outcome are we trying to achieve?
Is the current process stable?
Which variations are present?
What are operators correcting today?
Are those corrections visible and understood?
What operating range must the automation handle?
Which conditions should trigger adjustment, rejection or stopping?
How will the process recover after a disturbance?
Are we automating the correct method?
Would process improvement remove some of the need for complex automation?

These questions should come before equipment selection. Not after commissioning.

Closing

Automation is an important part of modern bakery operations.

It can improve safety.
It can reduce repetitive work.
It can increase capacity.
It can improve consistency.
And it can create valuable operational information.

But automation does not replace process understanding.

If operators are constantly compensating for an unstable process, removing them without understanding those corrections will not create control.

It will expose the instability.
Understand the process.
Define the operating window.
Reduce unnecessary variation.
Decide which disturbances the automation must handle.
Then automate.

Do not automate a process you do not control.

Bakery Expert Perspective

Good automation begins before equipment is selected. It begins by understanding the process, the product and the variation that exists in real production. The objective is not to create a system that works only when every condition is perfect. The objective is to create an operation that remains stable when normal variation occurs. Automation should strengthen process control. It should not be expected to replace it.

Key Takeaways

Automation and process control are not the same thing.
Experienced operators often hide process variation through constant correction.
Automation may expose problems that were already present.
Define the required outcome before selecting equipment.
Understand the full operating range, not only average conditions.
Decide which variations the automation should handle.
Define what happens during disturbances and how production recovers.
A process does not need to be perfect, but it does need to be understood.
Improve a poor operating method before making it permanent through automation.
Reliable automation begins with process control.

Knowledge Series 002 – Better Decisions Build Better Bakeries

This publication forms part of Knowledge Series 002, Better Decisions Build Better Bakeries, exploring the operational thinking behind successful industrial bakery and food-manufacturing operations.

Previous publication:

You Will Never Have All the Information.

Next publication:

The Best Machine Does Not Always Create the Best Line.

Willem Jongens