Intelligent Manufacturing Systems (IMSs) are expected to solve, within certain limits, unprecedented, unforeseen problems on the basis even of incomplete and imprecise information [34]. According to its definition, the intelligence has two main characteristics: the capability to receive and store the knowledge and the capability to apply the knowledge.

The information technology involved in the production systems is to provide the intelligence of the IMSs, consequently, it is expected to be able to realize the above mentioned characteristics.

To acquire and to store the production related knowledge, artificial neural networks (ANNs) can be used as production models because they can handle strong non-linearites, large number of parameters, missing information. Based on their inherent learning capabilities, ANNs can adapt themselves to changes in the production environment and can be used also in case if there is no exact knowledge about the relationships among the various parameters of manufacturing.

In this work the stages of the "classical strategy of ANN applications" are identified and some related problems are pointed out: problems of reusing the knowledge; modeling of non-invertable dependencies; solving different assignments arising in different stages and levels of production and production planning; modeling of connected processes, e.g. production lines.

The assignment dependent model building strategy is the reason for the above mentioned problems. Having recognised this reason a new method was realised to build up the ANN model independently of the given or possible assignments. It has some advantages:

• It solves the problem of modeling of non-invertable dependencies.
• It considers the accuracy requirements already during the model building stage.
• It finds all the dependencies among the process parameters.
• It builds up the ANN model.

Because the ANN model was built up assignment independently, and incorporates all the dependencies among the parameters it can be called: as the "general model" of the given process.

A new search method was elaborated to solve various process-related assignments based on the general model of the given process. The repeated running of this search algorithm results in different solutions of the given assignments. Tasks arising in different levels and stages of production and production planning can be solved using this tool; consequently, this method is applicable also from the viewpoint of knowledge reuse.

The optimization tool, "ProcessManager" has been also worked out, which:

• is able to model connected processes, e.g. production lines,
• is able to handle different process models,
• solves various optimizations task with various constraints,
• can be used to determine compromises among different targets of the optimization,
• can be applied in different stages of production and production planning.

All of the methods have been realized in the software packages of "Neureca2", "FunctionManager" and "ProcessManager". All of the tests connected with these new methods were performed using the software "TestManager" and "PatternMaker". The usability of these methods and tools has been shown through different tests and applications.