VHDL implementation of a neural diagnosis system: application to induction machine fault detection

The purpose of this paper is to present an implementation method using a neural network dedicated to diagnostic. The example given is related to induction machine diagnosis but we explain how the methodology can apply to various classification systems. The proposed implementation is based on VHDL language in order to have a flexible solution in term of implementation, the prototype being supposed to be realized into FPGA architecture. We describe the VHDL modelling of the network, and how it can be used for the "training" of the network and the results obtained. We demonstrate the accuracy of the model and its reliability.