Determination of lapping machine settings for hypoid gears using artificial intelligence

The present study deals with a computer determination system of lapping machine settings for hypoid gears, to which neural net is applied. A policy for selecting teacher signals, which are used for training the system, was determined to achieve high system determination performance. However, it was difficult to realize high determination performance for some of setting parameters; e.g., a pinion cone motion and a set-over. The present paper describes a new policy for selecting teacher signals to improve the determination performance. Furthermore, a new method for applying a neural net to the system is also proposed. In the method, multiple neural nets are used to enable the other setting parameters to be considered for determining a machine setting parameter. As a result, relatively high determination performance is achieved even for the pinion cone motion. This means that the developed system could be a substitute for experienced operators.