Waste takes many forms in manufacturing.
Slightly too much raw material is used. A product’s dimensions, shape or weight begin to deviate from the target. A process runs with incorrect settings. Packaging fails, or a finished product is rejected during quality inspection.
A single deviation may seem insignificant. But when the same issue occurs thousands or millions of times, even minor process variation can become a significant cost.
Automation can help reduce waste at many stages of production. However, simply adding more automation does not solve the problem.
First, we need to understand where waste occurs, what causes it and how early a deviation can be detected.
Many manufacturers already know fairly accurately how many products are rejected.
That is a good starting point. But knowing that 2.3% of the day’s output was rejected does not tell us what needs to change in the process.
The key is to link finished-product quality data to information about the conditions under which each product was manufactured.
What settings were used? What were the temperatures, pressures, operating speeds and dosing quantities? Had there just been a material changeover? Had any measured value started to drift before the first rejected product appeared?
When a rejected product can be linked to the process data recorded during its manufacture, we can begin to identify the causes of waste.
Final quality inspection determines whether a finished product is acceptable.
To reduce waste, we also need to ask whether we could detect earlier that the process is drifting away from its intended operating conditions.
Not all deviations occur suddenly. Product weight may gradually increase, temperature may rise, the quantity dispensed may decrease, or the product’s shape may slowly change.
Individual readings may remain within the alarm limits. Yet historical measurement data may reveal a change in process behaviour.
This is why monitoring upper and lower limits alone is not always enough. Sometimes, the trend matters more than any individual reading.
The earlier a change is detected, the sooner corrective action can be taken.
Machine vision has a wide range of applications. For example, we have used 3D machine vision to measure the shape and height of sandwich biscuits. A 3D image can be used to measure product height, dimensions, shape, volume and surface profile.
In this application, the biscuits were packed in groups of 12. If even one biscuit was too tall and prevented the pack from closing properly, all 12 biscuits were rejected: one that was too tall and 11 that were perfectly acceptable.
Detecting the excessively tall biscuit would have made it possible to stop the defective pack from moving further down the line. However, detection alone would not have addressed the underlying cause.
Based on the measurement data, the process was modified so that the solid filling was trimmed to a lower height before the top biscuit was placed on it. This kept the finished biscuit within the height required for the pack to close properly.
Once a single excessively tall biscuit no longer caused 11 perfectly good biscuits to be rejected as well, the savings were substantial.
Detecting a defective product helps prevent it from reaching the customer. Reducing waste also requires us to understand and address why the defect occurs.
Reducing waste does not always require a new sensor, a machine vision system or a major investment in automation.
Sometimes, the solution lies in how the existing production line is operated.
In wood processing and mining, for example, raw material quality can vary considerably. The same operating speed may not work equally well under all conditions. When material is more difficult to process, higher speeds can increase material losses or compromise the quality of the output.
In such cases, it is worth calculating the effect of slowing down that process stage when material quality requires it.
How much would throughput decrease? How much raw material or finished product would be saved? What would the financial impact be?
If the savings from reduced waste outweigh the financial impact of lower throughput, and the additional capacity is not currently needed, running more slowly can be a highly cost-effective solution.
Later, as production volumes grow and more capacity is needed, a larger investment can be made in that process stage. Making that investment years earlier may not be justified simply because a more technically efficient solutions.
Not all waste ends up in the bin.
If a product has a specified minimum weight and the manufacturing process has substantial variability, it may be necessary to produce it at an average weight slightly above the intended target. This provides a margin to help prevent individual products from falling below the permitted minimum.
Every product may therefore meet specifications while still containing slightly more raw material than necessary. This excess is known as product giveaway.
If process variability can be reduced, it may be possible to bring the average weight closer to the target without increasing the number of underweight products.
For an individual product, the difference may be only a few grams. Across a large production volume, the impact can be considerable.
This type of waste does not appear in the reject rate.
Once sufficient process and product quality data are available, the next question is how to use them.
If a product characteristic starts to change, can that change be linked to temperature, dosing, operating speed, raw material properties or machine settings?
If the relationship is understood well enough, corrective action can sometimes be automated through closed-loop control. Measurement feedback is used to adjust the process towards the target, and subsequent measurements verify the effect of the adjustment.
It is not always necessary to go that far. Simply enabling an operator to detect a deviation earlier and understand where to intervene can provide significant value.
The objective remains the same: to reduce the occurrence of defects as well as detect them sooner.
Waste reduction may involve additional measurements, software modifications, changes to production methods, machine vision, automatic process control or more advanced data analytics.
The technology should be selected once the problem is understood.
If waste can be reduced by changing a single setpoint, a more complex system is unnecessary. If several factors affect the process simultaneously, more measurement data and analysis may be needed.
The financial cost of waste should also be quantified. How much is lost each year in raw materials, machine time, energy and labour? What would a one-percentage-point reduction in the reject rate, for example, be worth?
Only then can we assess how much it makes sense to invest in solving the problem.
Sometimes, a small software modification is enough. Sometimes, machine vision or an additional measurement is needed. And sometimes, the calculation shows that an investment is not justified.
Automation should reduce waste where doing so is technically feasible and economically worthwhile.
Reducing waste does not necessarily begin with purchasing a new automation solution.
At Esys, we start by working with the customer to identify where waste occurs, what causes it and at which stage of the process action can be taken to reduce it.
Based on that understanding, we assess the most appropriate solution. This may involve a small modification to the existing automation or production method, more effective use of process data, an additional measurement, machine vision or automatic process control.
Our aim is to find a solution that reduces waste effectively and delivers a sound financial return.
If you would like to identify opportunities to reduce waste in your production, we can review your process with you and determine where to focus the first improvements.
