The automotive industry is a very different place from that of Henry Ford’s first production line of the 1910s. Now, with rising wage-costs in Europe and North America, outsourcing to developing countries, particularly India, has become increasingly fashionable, and any one car will normally be manufactured from components produced in more than one continent, not just more than one factory.
With this comes the obvious problem of traceability. In such a high-value industry, manufacturers need to spot faults as soon as they emerge, and this means being able to track a component’s path throughout the production process and across the globe.
Machine vision is providing the solution to this, in addition to many other high-level processes of vehicle production. It is almost impossible to imagine a time when machine vision wasn’t a staple of the industry, providing automation and detailed measurement on the production line.
Absolute Vision has tackled the problem with 2D data matrices – a kind of barcode composed of dots rather than lines – that provide a faster, and more reliable, method of storing data about the component than traditional alphanumeric codes.
The dots store data in a digital rather than analogue format, so that unlike linear barcodes, their size holds no information. It is much easier for the camera to read the matrices even in bad light and low contrast, which can be a frequent problem for shiny automotive components.
The dots also have a far higher density than other methods, meaning the parts can be individually serialised, rather than given a group batch number. This provides even greater detail on when, and where, difficulties occurred.
Applications of the technology can however be more sophisticated than simple tracking, with additional data such as the performance trajectory of engine parts attached to the serial number and stored in a global database. Then, during assembly, often on the other side of the world, the performance curve can be programmed into the engine, to give much finer control over pollutant emissions.
Absolute Vision's 2D data matrices store more information and are more reliable than traditional bar codes
‘With the increased risk of counterfeit components, the ability to know whether an item is genuine is becoming increasingly important as parts are made all over the world,’ says Charlie Plain-Jones, European sales manager of Absolute Vision. At least three high-profile automotive manufacturers, including Perkins, have already installed the systems.
Even within one factory, this kind of traceability is necessary, with many manufacturers, such as Cabi in Italy, producing components for several different high-profile clients. Confusing the different components could be disastrous.
To ensure this doesn’t happen, Cabi uses a two-tier system. Cognex’s Checker product detects a unique graphical symbol to identify the client and then the more sophisticated In-Sight camera is used to read a 2D code that provides a unique serial number for the individual part. Often, it is more than just a part’s journey across the globe that needs to be tracked. Many factories will make tiny adjustments to production procedures which can cause difficulties further down the line if they are not noted. Rhex Edwards, the director of sales and marketing at Perceptron, once had a client in Detroit whose processes had altered over the years so that they produced car body sides just 2mm narrower than the CAD designs, making them too small to fit with the rest of the vehicle.
It is a common concern in today’s automotive industry, where many similar vehicles are built around a common platform of the same components. If the physical process is not first monitored, redesigns for new models could be incompatible with the existing platform.
Perceptron’s products use laser triangulation to produce a 3D profile of the different components, with an accuracy of up to 100μm. ‘It can collect very high rates of data and build a totally 3D image. It will give all the information about the part, including any drifts or variation in the measurement’ says Edwards. This can then be compared to the original CAD designs to map unplanned changes that have occurred during manufacturing.
A Perceptron scanner captures 3D information about a component at a rate of 458,000 points per second.
‘It is very common these days to outsource parts worldwide,’ says Edwards. ‘Now you can digitally share the information about the parts. You can literally ask craftsmen to make the design in the model shop, and it will be exactly the same in either the design centre or the assembly point.’
Laser triangulation is not the only method available to provide a 3D image of automotive components. Solving3D’s Promt.stereo system uses MVTec’s Halcon software library to compare the images from two cameras, giving information about the size and position of features such as drill holes, which can then be compared back with the CAD drawings in a similar way.
Sometimes, the 3D images provide the CAD drawings themselves if the originals are not available. It is called reverse engineering, and is used to make components for classic cars from the 1920s to the 1940s that are no longer available – a surprisingly large market at the moment.
Its value is not limited to resuscitating 1926 Bentleys. Computer models, which would previously use ready-made CAD data to predict the performance of a car and the effects of a car crash, now use the scanned data of prototypes as part of their models. It helps to save even more money and time by reducing the number of wrecked cars necessary to guarantee the greatest safety.
Sometimes the car itself is scanned and fed into the models, so the program can help to provide greater insight into the milliseconds it takes for a crash to occur. Previously, high-speed photography was necessary for painstaking reconstruction.
Perceptron’s scanners are lightweight and portable, weighing as little as 340g, so they can be taken wherever they are needed on the factory floor. They have found a place with many leading manufacturers, including BMW, Mercedes, and Jaguar. However, Perceptron’s Edwards claims that it is the power of the software that has really allowed these applications, even for relatively uneducated users: ‘Twenty years ago you’d have needed a PhD to process the data. Now, a laptop and three to five days training allow an operator to set up and maintain the equipment.’ Accuracy, too, has improved. Just five years ago, 4,000 points per second would have been optimistic; now, more than 458,000 are possible, providing a much clearer and more meaningful image.
If these innovations are anything to go by, the future bodes well for both the automotive and machine vision industry. It is a long way from the early Model T Fords, each of which took the careful, personal attention of three engineers on its journey to the road from the draftsman’s board.
Managing the many high-temperature, molten processes has always been a problem for automotive manufacturers, but it is now easier with the new high-resolution IR cameras currently available. Renault recently opted for Flir’s ThermaCam S65 to view the injection moulding of plastic fenders in its Modus, Clio, Mégane and Espace models.
To avoid weaknesses, the temperature must be evenly distributed and the cooling must be uniform during the moulding. The camera gave a detailed view of the processes involved, particular at the far reaches of the fender, and allowed the engineers to adapt the flow regulation and the cooling cycle to the complex geometry of the piece.
The cameras are also used to test the quality of imported components. ‘Before the company steps into a huge purchase and delivery contract for automobile parts, we examine the supplier’s product and manufacturing process as thoroughly as possible’, says Jean-Baptiste Blumenfeld, validation technician and infrared camera expert at the Renault’s Technocentre. ‘An infrared camera is perfect to assess the welding quality of plastic parts and components.’
