How to fix the graph step response system (overshoot/peak) in Simulink

I already trained a Neural Network on MATLAB, after that I want to simulate that into my control design in Simulink. So, I use gensim to generate the NN block to the simulink. After I run the model in Simulink, it shows a bad results in a graph (blue line). How to overcome this overshoot things? Is there anything wrong with my control design?

Answer 1

[Piyush Dubey]

Hi Agnes,

The control design of your Neural Network can be modified or normalized to restrict the output within a given range. Alternatively, there are ways by which the scope block can be tweaked to incorporate the overshot range encountered while running your model.

These are as follows:

1. Tune controller parameters: This step involves adjusting the gains of the controller to optimize the system's performance. You can use various methods like trial and error, Ziegler-Nichols method, or model-based tuning techniques.
2. Using anti-windup mechanism: Anti-windup mechanisms are designed to limit the integrator output when the actuator saturates or reaches its limits. This prevents the accumulation of excessive control action during saturation, reducing overshoot.
3. Apply filtering: Filters can help reduce the impact of rapid changes in the system, which can lead to overshoot. Low-pass filters are commonly used to eliminate high-frequency noise.

Please refer to the following documentation to get insights on how scope can be configured to adapt to variety of visualizations: https://www.mathworks.com/help/dsp/ug/configure-time-scope.html

Hope this helps.