Model Predictive Control of Two Mass Resonant Mechanical System Optimized by Neural Network

More complex systems are introduced in industry nowadays. Expected speed response has more difficulties to be achieved especially during transients. The final product deteriorates quality using traditional speed controllers incapable of controlling complex systems. Therefore, speed control has to be improved by applying advanced control algorithm. Accordingly, the paper presents implementation of Model Predictive Control algorithm, estimation of difficult-to-measure or unmeasurable quantities, and prediction of system parameters in complex operating and environmental conditions using Neural Network. Model Predictive Control algorithm is recognized as solution for control of complex systems due to possibility of predicting system behavior and keeping speed response error at minimum required value. Neural Network adapts static, unmeasurable system parameters when working conditions change. To reduce computational complexity, control algorithm and Neural Network can be implemented on separate processor units. Mathematical model is developed in Matlab Simulink. Validation of Neural Network parameter estimation is performed with random data set which are not in Neural Network training data set.