A High-Accuracy Two-Stage Deep Learning-Based Resolver to Digital Converter

Abstract

High-efficiency Permanent Magnet Synchronous Motors (PMSMs) are widely used in industrial applications. Subsequently, resolvers as position sensors experience increasing usage in the drive of PMSMs. Measuring the rotor’s position using a resolver needs a costly Resolver to Digital Converter (RDC). Therefore, in this paper, a software-based, low-cost RDC is presented against the conventional Angle Tracking Observers (ATOs). The proposed RDC is a high-accuracy two-stage Deep Neural Network (DNN)-based one. In this regard, an overview of the resolvers’ function and the general methodology to DNN training is given. Then, the case study is done on the gathered data from the set of the resolver hybrid reference model and the vector control of PMSM. Eventually, the trained DNN is examined under different speeds and architectures against the conventional methods to confirm the investigations.