Self-learning simulations on grouting pressure control by the artificial neural networks for a dynamic system

For the security of dam foundation, the control of grouting pressure is one of the most important issues. In order to avoid the dangerous pressure fluctuation and to improve the control precision, a feedback proportional-integral-derivative control method is proposed in this work. Because the grouting pressure is influenced by many factors such as grouting flow, grouts density and geological conditions, such a proportional-integral-derivative methodology should be tuned. To this end, back-propagation artificial neural networks were employed to model the grouting control process and sensitivity analysis algorithm. Furthermore, to obtain the optimal parameters, an iteration algorithm was adopted in each sampling interval time through the discrete Lyapunov function of the tracking error. The simulation results showed that self-learning tuning was robust and effective, which was meaningful for the realization of the automatic control device in the grouting process.