In the car industry, like in other modern industries, the innovation cycles have become increasingly shorter. Driven by the objectives of environmental protection laws, hazardous manufacturing materials must be replaced by more adequate new materials. Equally important is the reduction of cost while the highest possible quality standard is being maintained. In many branches, new materials such as glass mat reinforced thermoplastics (GMT) are currently introduced and increasingly more used for manufacturing products, and thereby replacing steel and metal constructions.A GMT is a composite consisting of two components, namely a thermoplastic reinforced by glass fiber. An example is the manufacturing of car seats. The high security standards and other requirements (e.g. concerning wear and tear) can now be satisfied by using GMTs. For example, the rear part of a bucket seat for a car can now be manufactured with GMT, instead of more costly metal constructions.
Figure 1 shows the production process with GMTs. It consists of a preparation phase, a pressing phase and a finishing phase. In the preparation phase the raw material is put on a conveyer belt that moves it through the tunnel kiln, where it is heated. In order to avoid an undesired cooling, the material is then immediately put into the hydraulic press, where the geometry of the car seat is pressed before it is cooled off so that its form is maintained. During the finishing phase unwanted bumps must be removed.
The pressing of the material depends on a number of parameters with complex interrelationships. The temperature of the material influences the volume per unit time which is responsible that the material reaches every part of the pressing form. As soon as the material is put into the press, the press is closed with a speed of about 800 mm/s. As soon as the press reaches the material the speed is reduced to a value between 5 and 15 mm/s. After the press is closed a constant pressing force is exerted on the material for some duration. After that, the material is left in the press for some time to cool off. The duration of cooling depends on the temperature of the material and the tool, the tool geometry, the topology of the cooling capillaries of the tool etc.
In product and production planning, "system development" and "parameter optimization" are distinguished as two separate phases, which can also be called primary and secondary engineering [\protect\citeauthoryearKrottmaier1991]. In the primary phase, a prototype of the product and the corresponding manufacturing process is developed. In some previous research it was already shown how machine learning techniques can be applied for supporting the primary engineering phase [\protect\citeauthoryearReinartz and Schmalhofer1994]. More specifically, it was shown how an explanation based abstraction method [\protect\citeauthoryearSchmalhofer and Tschaitschian1993] can be used for abstracting planning schemata from success cases of the real world [\protect\citeauthoryearSchmalhofer et al.in press]. In the secondary phase, appropiate parameters must be found for the respective primary design. In this paper, we are solely concerned with this secondary design phase. In particular, we propose a knowledge base and knowledge evolution techniques for documenting and maintaining all available information and knowledge. This knowledge concerns the various parameters and how they determine the desired characteristics of the product.