Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Physics-guided neural networks for inversion-based feedforward control applied to hybrid stepper motors

Daiwei Fan, Max Bolderman, Sjirk Koekebakker, Hans Butler, Mircea Lazar

Abstract

Rotary motors, such as hybrid stepper motors (HSMs), are widely used in industries varying from printing applications to robotics. The increasing need for productivity and efficiency without increasing the manufacturing costs calls for innovative control design. Feedforward control is typically used in tracking control problems, where the desired reference is known in advance. In most applications, this is the case for HSMs, which need to track a periodic angular velocity and angular position reference. Performance achieved by feedforward control is limited by the accuracy of the available model describing the inverse system dynamics. In this work, we develop a physics-guided neural network (PGNN) feedforward controller for HSMs, which can learn the effect of parasitic forces from data and compensate for it, resulting in improved accuracy. Indeed, experimental results on an HSM used in printing industry show that the PGNN outperforms conventional benchmarks in terms of the mean-absolute tracking error.

Physics-guided neural networks for inversion-based feedforward control applied to hybrid stepper motors

Abstract

Rotary motors, such as hybrid stepper motors (HSMs), are widely used in industries varying from printing applications to robotics. The increasing need for productivity and efficiency without increasing the manufacturing costs calls for innovative control design. Feedforward control is typically used in tracking control problems, where the desired reference is known in advance. In most applications, this is the case for HSMs, which need to track a periodic angular velocity and angular position reference. Performance achieved by feedforward control is limited by the accuracy of the available model describing the inverse system dynamics. In this work, we develop a physics-guided neural network (PGNN) feedforward controller for HSMs, which can learn the effect of parasitic forces from data and compensate for it, resulting in improved accuracy. Indeed, experimental results on an HSM used in printing industry show that the PGNN outperforms conventional benchmarks in terms of the mean-absolute tracking error.
Paper Structure (9 sections, 25 equations, 7 figures)

This paper contains 9 sections, 25 equations, 7 figures.

Table of Contents

  1. INTRODUCTION
  2. PRELIMINARIES
  3. First--principle modeling of an HSM
  4. Field--oriented control architecture of an HSM
  5. PROBLEM STATEMENT
  6. FEEDFORWARD CONTROL OF HSMS USING PHYSICS--GUIDED NEURAL NETWORKS
  7. EXPERIMENTAL VALIDATION
  8. CONCLUSIONS
  9. ACKNOWLEDGEMENTS

Figures (7)

  • Figure 1: FOC architecture including the HSM, the current control with $dq$--transform and the position feedback--feedforward control setup.
  • Figure 2: Reference (top window), feedforward signal (middle window), and the resulting tracking error (bottom window) for the feedforward controllers using the physical model \ref{['eq:FeedforwardPhysicalModel']} and the NN \ref{['eq:FeedforwardNeuralNetwork']} on a simulation example.
  • Figure 3: Schematic overview of the physics--guided neural network.
  • Figure 4: Example for imposing physical knowledge via $T(\cdot)$, i.e., improved extrapolation capabilities when training a NN with $T(y) = \textup{mod}(y)$ compared to $T(y) = y$ on a limited data set (top window), and the reduction of the required amount of neurons $n_1$ to achieve an approximation of similar quality (bottom window).
  • Figure 5: HSM FL57STH51--2804A by FULLING MOTOR with encoder.
  • ...and 2 more figures

Theorems & Definitions (4)

  • Remark II.1
  • Remark II.2
  • Remark III.1
  • Remark IV.1