Automatic optimization of real systems

Content

Automatic optimization of real systems.. 1

Virtual part of the proposed optimization structure. 2

Automatic optimization of real systems

INTRODUCTION

Based on MATLAB Simulink, a structure for automatic optimization of real systems under workload has been developed. Only the optimized parameters of the real system, for example, the parameters of the PID controller, are transferred to the virtual Simulink environment. The connection of the virtual environment with the rest of the real system is provided via the COM port. This structure eliminates errors related to the inadequacy of models and allows full use of the powerful mathematical apparatus of the integrated MATLAB environment.

Parameter optimization structures

The development of regulators and optimization of their parameters can be performed at the level

· controllers of real control systems (Figure 1),

· models of control systems (Figure 2)

· real system with a regulator model (Figure 3)

Figure 1. Control system circuit.

Figure 2. Model of the control system.

Figure 3. Model of a regulator in the structure of a real control system.

Of the three options under consideration, development and optimization at the controller level (Figure 1) is hampered by limited specialized development resources and is characterized by the highest costs.

Minimization of costs is ensured by the development of control systems using modeling (Figure 2). For these purposes, there are many specialized programs with powerful mathematical tools. The disadvantage of this option is the possible deviation of the model parameters (inadequacy of the model) from the parameters of real systems, which leads to deterioration of the optimization results.

In the variant (Figure 3), parameters, such as the PID controller, are optimized in a virtual environment with constant operating parameters of the object. The error associated with the “inadequacy” of the model is completely eliminated, while the minimum costs increase only for the construction of a communication channel between the controller and the controller model using standard means.

Virtual part of the proposed optimization structure

An example of the optimization structure of 4 parameters of the PID controller of a real system, taken out into a virtual environment, is shown in Figure 4. The environment consists of Simulink blocks and includes

· generator of a given (step) effect

· error signal calculator

· discrete PID controller with output signal limitation

· channel for transmitting the PID controller signal to the COM port.

· channel for receiving position feedback signal from COM port

· means of optimization and display of variables.

Communication between the virtual environment (Figure 4) and the real system is carried out by bytes via the COM port at a speed of 115200 bits/s.

Figure 4. Structure of the virtual part.

Optimization criteria and parameter change ranges are specified in the blocks of the Simulink à Simulink Design Optimization library section. In the presented version (Figure 4), the controller parameters are configured to ensure the required response of the real system to a step effect.

The current values of the 4 optimized parameters of the discrete controller are displayed in real time (green blocks). The set value, the reaction of the real system and the real impact (controller output) are accumulated and displayed on the plotter (crimson block).

The real-time clock cycles are set by the real controller. It also provides the connection between the real and virtual parts: it receives the output signal of the virtual controller and transmits to Simulink the readings of the real sensor measuring the system's response.

Figure 5. Examples of displaying in Simulink a setpoint (yellow line), a real system response (blue line), and a controller output (red line).

After optimization, the regulator parameters obtained in Simulink are written into the controller. The real control system is ready for independent operation.

If the system parameters change, leading to deterioration of its performance, it is necessary to repeat the optimization in the same shared environment.