QUAR

QUAR – Artificial Intelligence for Industry 4.0

Most engine blocks leaving the factory gates today were produced by fully automated processes. Now the FORWISS Institute at the University of Passau is developing an intelligent system to automate the monitoring of some of these production processes: With the help of machine learning, the system should be able to make accurate predictions about the wear condition of processing machines.

The production of aluminium engine blocks works without human intervention: after casting, the parts are decored and processed fully automatically by robots. One plant consists of 13 processing stations, each of which carries out a specific task.

In pre-decoring, for example, two pneumatic hammers strike a component simultaneously to loosen the sand of the inner sand cores. 'These hammers always operate at their own stress capacity and that of the manufactured component', said Dr Erich Fuchs, General Manager of the Institute of Software Systems in Technical Applications of Computer Science (FORWISS Passau). Precision is paramount. 'If the hammers are just five millimetres off, it could damage the engine block.'

There is another risk: If one of the machines in the process series fails, the entire plant with all its 13 stations comes to a complete standstill. As a number of engine blocks are processed simultaneously at any given time, such outages are potentially very cost-intensive.

Intelligent system is trained on specific signals

This is where the QUAR project comes into play (the acronym stands for the German project title for 'predictive maintenance and quality assurance in machining'): The researchers aim to use artificial intelligence to develop a fully automated monitoring system. Using machine learning, the system will be able to make accurate predictions when, for example, the hammers will cease to work reliably or fail altogether.

To this end, the researchers identify signals that could indicate such failures. Although mechanical changes in pneumatic hammers are gradual and difficult to observe, they do, however, lead to deviations in the processes. The researchers are trying to make these measurable with the help of vibration and other sensors and by tracking fluctuations in electric power consumption, among other things.

The collected data are fed into the intelligent system, which determines the optimal times for the replacement of critical components, and thereby minimising unplanned downtimes due to machine failure. With this intelligent monitoring and maintenance system, the QUAR project contributes a building block to the digitalisation of the entire production process.

Project partners and funding

Professor Tomas Sauer (Chair of Digital Image Processing) heads the project jointly with FORWISS General Manager Dr Erich Fuchs. The Institute deals with the theoretical part, i.e. the selection and implementation of suitable learning algorithms and their mathematical modelling. The industrial project partner is R. Scheuchl GmbH, a company based in Ortenburg, Germany, which as a manufacturer of special-purpose machine tools is responsible for the complete assembly and test operation of the parts processing plant.

The project is financed using funds from the Informations- und Kommunikationstechnik ('Information and Communication Technology') research and development programme of the Bavarian State Ministry of Economic Affairs and Media, Energy and Technology.