Image source: timoelliott.com
An unfortunate reality of modern process control and performance improvement is that many still see factory automation as a threat to their jobs, rather than an improvement opportunity to help them by removing the drudgery of every day repetitive operations which frankly process operators never really liked doing and were never particularly suited for.
When I first started my career, several decades ago, in the development of performance improvement systems, very few, if any of us, could envisage where this could take us today.
At that time Quality Control, as it was often called, was standard procedure. This was largely based upon production making things as quickly as possible, often incentivised to do so, whilst quality inspection teams picked out the defects in their noble efforts to ensure that only conforming products reached the end user. In principle, this was considered normal and was virtually universally accepted as standard practice. The problem was that it was flawed, in that defective products, often reached the end user and consequential customer complaints were, for many industries, substantial. The quality control departments, rather than production, often bore the brunt of such complaints.
However, some years earlier visionaries, such as Edwards Deming and Walter Shewart developed Statistical Process Control. The theory was that SPC, as it became widely known, was not based upon product inspection, but upon statistical SPC was to determine process capability, set realistic control limits, to take account of normal process variation, and also predict the extent of failures (out of spec components) to be expected from any repetitive production process. Sadly, this new technology was not widely accepted due to the fact that, during the post-war years, Western managed companies had full order books and were simply ‘too busy’ to take on a new ‘unproven’ technique when inspection had, for decades, been the accepted norm. But, as has already been mentioned, manual inspection, however diligently carried out, was flawed largely because the ability of human beings to remain attentive for extended periods was limited (even 10 minutes at a time often proved ineffective) and production lines were getting faster, making inspection even more difficult and unreliable.
Enter the Japanese, whose production, post-war was ‘on its knees’ and the quality of goods produced was, to say the least, unreliable.
So Deming and Co. took their vision to Japan, where the Japanese eagerly embraced the new techniques, developed them further and, in almost every area of business, beat us at our own game. Automobiles, televisions, cameras, etc., all began to develop impressive reputations for quality and reliability, that the West simply envied.
Inspection gradually, and often grudgingly, during the next few decades, became discredited and the old inspection-based Quality Control has now been largely replaced by Quality Assurance, where the Quality Department set the standards and acceptable control limits, together with rigorous auditing to ensure that the Standard Operating Procedures are adhered to.
Lean Six Sigma has developed from the early beginnings of SPC. Much of the credit for the development of Lean goes to Toyota and for Six Sigma to Motorola. It was only a matter of time before Lean (improving efficiency) and Six Sigma (reducing variation and defects) would combine to form Lean Six Sigma, as we know it today, but it is still not a perfect science.
Anybody who has run a Six Sigma programme will quickly appreciate that Six Sigma is always based upon reducing variation in an effort to ensure that only conforming product is sent to the customer. However, even a Six Sigma process (literally meaning six standard deviations), would still be statistically able to deliver three DPMO (defects per million opportunities). If you happen to be one of the three in a million that receives the defect, then you will obviously not be very happy. Some processes, unfortunately, with significant inherent variation, cannot achieve Six Sigma, perhaps only three or four Sigma, showing 66,000 DPMO and 6,000 DPMO respectively.
Where this variation cannot be reduced, it could mean that the level of defects produced and the risk of customer complaints could be unacceptably high. A return, therefore, to inspection could make sense, but definitely not to manual inspection, as that has already been discredited.
This is where AI or automated intelligence can come to the rescue through, in one of its simplest forms, the installation of automated on-line vision inspection to eliminate defects. Dependent upon the nature of the defects to be rejected, this can work extremely well and will certainly achieve a much higher level of success than manual inspection for defects could ever achieve, but again, it is not a panacea.
One area where this can get tricky is in the reading and confirmation of alphanumeric characters on packaging. For this to be as reliable as possible, the alphanumeric characters must conform to a standard known as OCRB, which is generally only of sufficient quality to be read with a high degree of success when the code is printed by laser coders or by thermal contact printers. Inkjet printers, though very common in food and drink manufacturing processes, do not generally provide a sufficient print quality to be consistently read accurately by camera technology, despite what some camera manufacturers might claim and despite the fact that inkjet printing can still be read by the human eye.
However, AI is improving all the time and cameras and the software behind them is no exception. So we may well find in future that on-line vision technology can also reliably read virtually anything which is readable to the naked eye.
Though we also manufacture vision systems amongst our factory floor information management systems, we refuse to make exaggerated claims for them and believe that, with current technology, the best solutions are still created from a combination of automation, where possible, and human interaction and the decision-making process that follows.
Roy Green, Harford Control Ltd.