Metoda technologii grupowej bazująca na sieciach neuronalnych

The production system that uses typical technologies is exported by large assortment of produced elements and quick changing market conditions. The idea of this system is manufacturing elements that are similar in their constructional forms and manufacturing processes using identical methods. Decreasing of production costs can be obtained as a result of application of the group technology. The method of information coding and methods of elements' clustering are basic problems in group technology. In the literature there are no references for method of machine elements description according to both design and technological features. In connection with above the following goal of the work is to develop the method of research the design and technological similarity using the Kohonen net as a clustering tool. The Kohonen net is one of the neural nets, belonging to group of nets that work without supervising and can analyse similarities in the input data. Basing on the assumptions presented above, the goal of the work is determined. The main task that have been solved in this work is the method of description the design and technological parameters and the method of clustering that uses the Kohonen net. The structure and algorithm of a two-layer net have been elaborated. This net has been used for clustering machine elements. Also the classical algorithm of the Kohonen net has been changed to apply the method of "neural gas" and new learning rate function. These changing let for more effective clustering of machine elements. The original method of description the technological and design parameters is also proposed. This method bases on application of the utility function for information coding about machine tools and manufacturing processes that arc used in chosen manufacturing plant. Experiments and results show that the elaborated method let cluster machine elements. Analysis of experiments results present that this method is a very effective for computer aiding of similarities analyse of machine elements.