THE USE OF A HYBRID MODEL OF THE EXPERT SYSTEM FOR ASSESSING THE POTENTIALITY OF MANUFACTURING THE ASSUMED QUANTITY OF WIRE HARNESSES

Background: Control plays the main role in ensuring the stability of production processes, while digital models of processes and methods of artificial intelligence are used more and more commonly in it. Production of highly diversified items in small lots at low inventory levels is characterised by a much lower stability as compared with large-lot manufacturing. Additionally, innovations created for items or processes result in disturbances to current work. Although this turbulence is usually momentary, it may lead to a loss of function or manufacturing stability, which in turn translates into financial losses, as well as   losing customers. This paper presents the potential of using simulation models and artificial neural network models to assess the stability of a reorganized production system.

Methods: The problem analysed in the paper is that of merging a simulation model with an ANN model by designing a hybrid model. A direct connection of both types of models is not possible due to their various structures, specificity, and different purposes, as well as the various types of input and output data. Therefore, the idea of merging these two types of models through an expert knowledge base and fuzzy inference was proposed. The results from the simulation model and the ANN model were used to gather the knowledge on the production system being analysed. It has been proposed that the output from the simulation model provided knowledge of the risk level, while the output from the ANN model provided knowledge of process stability.

Results: The paper presents the idea of projecting a hybrid model of the expert system in order to assess the stability of a reorganized production system. A model of a hybrid expert system was developed to assess the potential of executing the assumed production plans. The level of risk and the level of stability determined by the simulation model and the ANN model are entered into the system. The output from the expert model is the value of the variable determining the potential of achieving the goal. In the construction of the model, fuzzy inference was used, which uses linguistic variables and is characterized by a knowledge system in the form of fuzzy rules "if ... then ...". For both the independent variable and for the dependent variable, a set of membership functions representing accepted linguistic variables was proposed, and then decision rules were determined.

The idea of merging simulation models with ANN models was tested on a practical example in production system that manufactures products for dishwashers.

Conclusions: The potentiality to execute production plans depending on the level of risk and the level of stability of the production system is too complicated to be modelled mathematically, but based on the analysis of data from the simulation and ANN models, it is possible to obtain information concerning the relations between corresponding input and output values.