Neural State Estimator for Complex Mechanical Part of Electrical Drive: Neural Network Size and Performance of State Estimation

Abstract

This paper presents the results of simulation research of an off-line-trained, feedforward neural-network-based state estimator. The investigated system is the mechanical part of an electrical drive characterised by elastic coupling with a working machine, modelled as a dual-mass system. The aim of the research was to find a set of neural network structures giving useful and repeatable results of the estimation. The mechanical resonance frequency of the system has been adopted at the level of 9.3-10.3 Hz. The selected state variables of the mechanical system are load, speed and stiffness torque of the shaft.

Virtual Signal Calculation Using Radial Neural Model Applied in a State Controller of a Two-Mass System

“…Intelligent structures with neural networks are widely investigated in the modern literature [31][32][33][34][35]. Neural networks can be utilized in electric drives for multiple purposes, such as state variable estimation [36], motor condition monitoring and diagnosis [37], enabling control with damaged sensors (often referred to as Fault-Tolerant Control) [38], and adaptive control [39]. To achieve adaptive qualities, training of the network must be performed online, i.e., during the operation of the drive.…”