A Study on a Learning Model for Setup of Grinding Parameters. Experimental Verification.
Moriaki Sakakura, Ichiro INASAKI · TRANSACTIONS OF THE JAPAN SOCIETY OF MECHANICAL ENGINEERS Series C · 1995
A learning model for the setup of grinding parameters had been developed by the authors using genetic algolithms and the fuzzy reasoning algorithm. Although the effectiveness of the model has been proven through a wide range of computer simulations, further practical investigations are required for industrial applications. In this respect, the performance of the model is investigated experimentally in this study. Since the performance of any learning model is affected by learning data, particular attention is paid to the influence of learning data. Ouality and quantity of the data obtained from the grinding process is discussed. The experimental results prove that the number of learning data has an influence on the rules obtained. Insufficient data cannot generate sufficient rules. On the other hand, an excessive amount of data degrades quality of the rules, which could be attributed to confliction of learning data.