Artificial Neural Networks with Machine Learning Design for a Polyphasic Encoder

Sergio Alvarez-Rodríguez, Francisco G. Peña-Lecona

Artificial neural networks are a powerful tool for managing data that are difficult to process and interpret. This article presents the design and implementation of backpropagated multilayer artificial neural networks, structured with a vector input, hidden layers, and an output node, for information processing generated by an optical encoder based on the polarization of light. A machine learning technique is proposed to train the neural networks such that the system can predict with remarkable accuracy the angular position in which the rotating element of the neuro-encoder is located based on information provided by light’s phase-shifting arrangements. The proposed neural designs show excellent performance in small angular intervals, and a methodology was proposed to avoid losing this remarkable characteristic in measurements from 0 to 180° and even up to 360°. The neuro-encoder was implemented in the simulation stage to obtain performance results, where the main evaluation metric employed to assess the performance is the total error. This proposal can be useful to improve the capabilities of resolvers or other polyphasic sensors used to obtain outstanding precision and accurate data, even when working under hard and noisy industrial conditions.