Application of OPTIMUS in motor control system optimization technology

1. Problem Statement

Application Overview

The design of motor control systems is a multidisciplinary problem involving mechanical, electrical, and control systems. OPTIMUS can explore the design space and automatically improve the design without user intervention through its powerful optimization algorithms. This application case demonstrates how OPTIMUS can be integrated with Simulink to optimize the output torque ripple and switching loss of a motor by modifying (a) the pole width of the motor and (b) the excitation signal.

Design Problem

The control system model is built in Simulink. The electromagnetic field corresponding to different pole widths of the motor has been calculated in ANSYS, and the coil inductance is imported into the Simulink model as a result. The inverter circuit is modeled in PSpice, and the circuit module is converted into a Simulink-compatible module through the SLPS interface to complete the integration.

Software tools used

• OPTIMUS and its Matlab/Simulink interface
• Simulink
• PSpice (SLPS interface)
• ANSYSSimulation

process and OPTIMUS workflowThe OPTIMUS graphical user interface integrates the simulation programs, their workflows, and input and output files. Through the OPTIMUS and Matlab/Simulink interface, OPTIMUS easily parameterizes the simulation input files and parses the required output parameters from the output files (Figure 1).

Figure 1 – OPTIMUS workflow for motor control system simulation

2. Design parameters and methods

The motor in the case is a switched reluctance motor (SRM) with 3-phase excitation, 6-pole stator and 4-pole rotor. The control system simulation is established in Simulink, the motor electromagnetic field is calculated in ANSYS, and the inverter circuit is modeled in PSpice. (Figure 2)

Figure 2 – Simulink model of the motor control system

Select the design parameters.

The three design parameters are the motor pole width, the starting angle of the excitation signal, and the width of the excitation signal. The optimization goal is to minimize the ripple of the motor output torque. The constraint is that the motor speed is greater than 1000RPM.

The design of experiments (DOE) method and the response surface model (RSM) are used to explore the design space. In this case, the Latin hypercube method with 100 sample points is applied. On this basis, a least square response surface based on Taylor polynomials is established to fit the experimental design sample points.

Design optimization

OPTIMUS In this case, the adaptive evolution (SAE) genetic algorithm is applied to find the minimum motor output torque ripple by solving on the response surface, while also satisfying that the motor speed is not less than 1000 RPM. The optimal solution found on the response surface is used as the starting point in the local optimization process of the simulation workflow solution. In this way, through the strategic combination of several optimization algorithms and different solution methods, the global optimal design can be found in the end, while shortening the time of the optimization process.

Figure 3 - Contribution plot showing that motor pole width and excitation signal start angle are the design parameters that have the greatest impact on output torque ripple

3. Results

The Latin Hypercube Experimental Design method was run to establish the sample of the response surface. Figure 4 shows that the motor pole width and the excitation signal width are important design parameters that have a great influence on the output torque ripple. This response surface model is an approximation of the simulation model. During the optimization process, if the simulation model needs to be solved continuously and in large quantities, a considerable amount of calculation will be required. The proper use of the response surface model can effectively reduce the amount of calculation and improve the efficiency of the optimization process. The quality of the response surface model (and its reliability for the optimization process) can be confirmed by the regression coefficients obtained during the establishment process.

Figure 4 – OPTIMUS response surface showing the relationship between output torque ripple and selected input parameters

OPIMUS found the design with the smallest motor output torque ripple and satisfied the speed constraint (Figure 5). Compared with the initial design, the optimal design effectively reduced the motor output torque ripple by 13.8% (Figure 6).

Figure 5 – Convergence of the optimization objective function: Minimizing output torque ripple

Figure 6 – Output torque and speed of the motor before and after optimization

4. Conclusion

OPTIMUS successfully automated the Simulink simulation and found the optimal pole width, starting angle and width of the excitation signal, which effectively reduced the output torque ripple of the motor and ensured that the motor speed was always higher than the specified speed.