Assembly line upgrade for defect elimination
120,000 automotive parts manufactured by Meister are checked each day using Cognex machine vision
Cognex Corporation (NASDAQ: CGNX), the major global supplier of vision systems, sensors and bar code readers, presents an example of an application using industrial vision in precision mechanics. Intended for automobile equipment manufacturers, the parts manufactured by Meister France are subjected to very strict appearance checks with the aim of completely eliminating defective parts before delivery to customers. An ambitious gamble when you are producing almost 40 million parts per year. A gamble which has paid off thanks to Cognex In-Sight® industrial vision sensors.
The Meister Group is a Belgian industrial group mainly supplying the automobile market. The company has factories in Belgium, France and the Czech Republic specialising in the mass production of cut steel parts.
The challenges for these modern production units, which use specialist precision lathes, is to manufacture relatively complex parts in a few seconds, and to guarantee the conformity of the parts on delivery, without ignoring the essential and continual search for gains in productivity.
In France, the Meister factory is located at Scionzier, in Haute-Savoie. It manufactures electric valve parts for automobile equipment manufacturers specialising in ABS braking systems. Nearly twenty-four multi-spindle lathes produce 120,000 parts each day, representing an annual production of 35 to 40 million parts.
The demand for quality is the main problem on which all of its efforts are concentrated: it has to try and avoid delivery of defective parts to customers. In a sector where the smallest incident on an assembly line can bring about exhaustive investigations and can lead to complicated and costly procedures for the subcontractor, the search for Zero Faults is the only acceptable way forward.
However, the manufacturing techniques used and the demands of mass production do not allow such an objective to be reached directly from the use of machines. Checking and sorting has to be carried out in order to remove defective parts, with the faults consisting of missing components, metal shavings, loose components, damage from vibration or knocks…
Checks were previously carried out by the naked eye by operators which limited the number of defective parts to around 1 in a 1000. This was still too many, and so studies had to be carried out in order to reduce this rate as a value lower than 100 ppm was aimed for. It was also necessary to work on reducing the impact of manpower costs on the cost price of the parts.
Considering the automation of these checks by vision tools was a natural approach for Meister’s technicians, who already had experience of industrial vision systems for a dimension checking application. A seminar organised by the Alpsitec company also allowed them to find out about the performance and capacities of the In-Sight vision sensors, made by Cognex, the world leader in industrial vision.
Alpsitec is an approved partner system integrator of Cognex. The company was called upon to carry out a demonstration directly on the production line so as to verify that the Cognex cameras were capable of “seeing” the faults they had to detect. After this first feasibility test, a prototype was brought up to date and assessed over a month. The simplicity of the use of the In-Sight sensors was a decisive factor in the choice of the system. The Terms and Conditions were drawn up and two test systems were ordered.
The checking system – in fact two independent test benches – was installed at the end of the production line in order to carry out a final check of the parts just before they were packed. All the parts produced were sent to this control point, therefore 100% of the production is checked.
The parts are put into their packaging – the mesh – by a robot. Once the packaging is completed, the robot picks it up and places it on the test surface. Then the robot takes hold of the Cognex In-Sight 1000 vision sensor linked to a lighting system and passes it along the mesh, over the parts. It’s important to remember that the vision sensor must inspect each part in order to detect any of the four types of faults to be removed: presence of metal shavings, missing components, loose components, and damage from knocks or vibration. The sensor sends information on the checks carried out to the robot’s control centre. The robot puts the vision sensor down and takes the defective parts and deposits them into a chute – one chute for each type of fault – which then carries them to a hopper. Then the system continues its operation.
One of the test benches is fitted with two In-Sight 1000 sensors and operates at a rate of 6000 parts per hour. The other system comprises a single sensor and works at a rate of 4000 parts per hour. Both systems worked as dual sorters during the first few months of the operation.
The important part of the work of updating the application consisted of identifying the various faults which the checking systems had to recognise and to “teach” them to the vision sensors. This procedure is essential for optimising the efficiency of the checking system.
The rate of faulty parts delivered to customers has rapidly dropped to 40 per million. The power of the processor algorithms of the In-Sight sensor and the finer analysis of the faults to be removed should allow this rate to be brought down even more to below 20ppm.
Jean-Marc Sermet, Technical Director of Meister France, has supervised this project from beginning to end. He is very pleased with Cognex products and Alpsitec’s service, the combination has provided him with the necessary skills and experience in setting up vision solutions in industry. “Above all, we are engineers specialising in precision metal cutting”, stated Mr. Sermet. “Alpsitec’s contribution has allowed us to make rapid progress on this project and to benefit from efficient and reassuring support.”
Alpsitec has also trained a technician who has been able to rapidly take charge of setting the vision sensors’ parameters. So Meister is able to input data by itself for new types of faults to be learned and for modifying the parameters in relation to the 15 different types of parts to be inspected.
“We were concerned that these test systems would slow down production rates. We have noted with satisfaction that the implementation of these industrial vision solutions do not have a negative effect on production”, added Jean-Marc Sermet. “The solution used appears to be particularly stable, and the operators do not have any need to intervene.”
The biggest reward is that there has been a real return on investment in less than six months. Customer relations have been strengthened from the significant improvement in quality. Meister is currently looking at other applications for industrial vision on its production lines.