Manufacturing with vision
29 June 2009
As margins are squeezed manufacturers are under increasing pressure to ensure they minimise waste. Anna Mitchell recently caught up with UKIVA director, Don Braggins to explore the use of machine vision in statistical process control to achieve these aims.
The UK Industrial Vision Association (UKIVA) was formed in 1992 in response to uncertainty in the industrial machine vision business. Don Braggins, a consultant specialising in processing and analysis was drafted in to help run the trade association and said the organisation aims to demonstrate to industry the various expertise available in the machine vision field.
Inspection remains the biggest single application for vision and Braggins says using vision in statistical process control can provide huge cost and quality benefits for manufacturers. ‘The idea of statistical process control is that every now and then during a manufacturing process you take a sample and you measure it,’ says Braggins. ‘In the old days this would occur every quarter of an hour or so, but the principle remains the same today. You take very careful measurements of the product and plot them. This allows you to follow trends, such as a product being consistently larger than it should be, and allows you to take corrective action.’
Braggins illustrates his point with an example of a factory making turned parts on an automatic lathe. ‘As the tool wears down very slightly, the diameter of the turned part gets slightly larger. So, you set a point. If several successive measurements are at that point you take action and move the tool forward a little so it doesn’t make such large diameters. The basic idea is take corrective action before something gets out of tolerance and your production will always be in tolerance,’ he explains.
‘In the old days you would be measuring with a micrometer or similar tool. You would need a quarter of an hour, or even half an hour, in order to take one accurate measurement,’ he continued. ‘Vision can give you a measurement on every single piece. It may not be quite as accurate as using a micrometer but because you are doing it so many times you’ve got a much better average and an average gives you a better answer than one single measurement.’
Statistical process control involves taking measurements of products and plotting them
Once you’re collecting that kind of data, according to Braggins, you might as well use it to control your process and take corrective action. ‘Typically the data is a dimension,’ he elaborates but it doesn’t have to be. For example, imagine printing a label and after while it starts to get a bit fuzzy. That indicates it’s time to clean the screen printer.’
Braggins argues that using a vision system to throw out anything that is rubbish is not particularly helpful to a process. ‘The whole point is to control a process so you don’t make rubbish in the first place. It is much more economic to use it in a role where you take action to keep things in control rather than wait until your system is producing something that is not acceptable – by that point you’ve wasted money’.
One of the barriers to widespread adoption of machine vision in statistical process control is simply a lack of knowledge. ‘In pre-automation days,’ recalls Braggins. ‘It was received wisdom that you shouldn’t make measurements too frequently otherwise you’d get into the kind of loop where you’re correcting too early. Then you over-correct so you’ve got to correct the other way. You get into an unstable situation. But with vision, you’re averaging over many, many measurements in series and you will be able to spot a trend. Instead of plotting a line through a sparse set of measured points you’ve got the entire line there for you.
A worker measures a pin with a dial caliper
‘You know exactly when something happened,’ he continues. ‘For example, in the food industry, if you load a new ingredient into a hopper and when it starts to come through you realise something has changed then you can investigate whether the new material was exactly what it was meant to be.’
Braggins illustrates his point, recalling a manufacturer of O-Rings that was having problems with the quality of his product. ‘O-rings are used in baking systems and they’re safety critical so they must be right,’ he starts. ‘The manufacturer in question found they could relate when they were making o-rings that were not satisfactory to the outside temperature. That seemed a bit odd but then it was discovered that when the factory, which didn’t have air conditioning, got a bit hot they opened the doors. The resulting draft cooled down the elastic compound used to make the o-rings far quicker than it should have been.’
Asked whether there were any industries that were well versed with the use of vision in statistical process control, Braggins answers: ‘The car industry has a long tradition of insisting that its suppliers have SPC systems in place.’ He continued, ’They inspect their suppliers’ factories rather than the product coming out of them. To become accredited as a supplier to the primary auto industry you will need to have your statistical process control system inspected and accredited. As far as I know they don’t insist on machine vision but this is obviously the best way to manage most processes.’
After demonstrating how useful machine vision systems can be in industry, Braggins concludes that the sector in the UK is largely impeded by a lack of skilled engineers in the field. He is concerned the country is being left behind by other rapidly developing nations and says: ‘There’s an awful lot of vision training going on in Chinese Universities. I heard a talk recently by a professor in Beijing, he claimed that he had more than 100 post graduate PhD students working on vision in his automation institute.’ It seems the UK will have to keep on its toes as it faces increasing pressure in this field from the rest of the world.
Contact Details and Archive...
Most Viewed Articles...