Development of An Auxiliary Offline Controller Based on Neural Network for a Shaking Table With An Online Pid Control System

Abstract eng:
The shaking table is a powerful tool in dynamic testing in which a high signal reproduction fidelity is considered as a distinguishing feature of the system. However, a precise control of the platform motions to ensure desired accelerations is still a challenging target because of the interference of the different mechanical, hydraulic and electronic parts as well as the test specimen particularly their strong inherent nonlinearities. This study addresses the improvement of acceleration signal matching on a shaking table using an additional online controller based on a recurrent neural network algorithm. The latter is developed to provide an additional correction to the command signal produced by the actual shaking table control system, in order to reduce the signal distortion between the reproduced signal and the desired one. The shaking table is simulated by its transfer function implemented in Matlab/Simulink, including a PID online controller. The training of the Neural Network is performed offline. The response of the transfer function of the system acceleration is used as input to the NN. The desired signal in acceleration represents the NN target. After training, the NN is implemented in Matlab/Simulink model in order to provide an online correction to the output signal, in the same time as the PID. To evaluate the performance of the proposed control scheme i.e a PID-NN controller, a correlation coefficient R has been calculated between the target signal and the system output. Results show that the NN controller improves the signal matching in acceleration.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket