Research On Neural Network Parameters Adaptive-Control For Deep-sea Mining Heave Motion Compensation

In this paper, the study background was deep-sea mining of polymetallic nodules, while the lifting pipe's heave compensation of mining ship as the research object, and then this paper set up three kinds of physical simulation model of light-load, medium-load, heavy-load heave compensation experimental system's hydraulic system. It attempted to put the neural network adaptive PID controller as control strategy into the system. Through the simulation and experimental, it obtained that the physical simulation model was correct and the designed controller was feasible.Firstly, mathematical modeling was established for the 4/3-way electro-hydraulic proportional valve-control-cylinder. The paper simulated proportional valve's dead zone characteristic based on SIMSCAPE. We knew that the proportional valve's dead zone had a significant impact on the hydraulic system's dynamic performance. The harmful aspect of the proportional valve's dead zone to hydraulic system was eliminated by voltage compensation. After that, three kinds of physical simulation model of light-load, medium-load, heavy-load heave compensation experimental system were established by SIMSCAPE.Secondly, all the heave compensation system had much negative factors such as saturation, dead zone nonlinearity, hysteresis, time-varying parameters. In order to get satisfactory accuracy, enhance system's robustness, and improve theirimmunity and adaptive capacity. The author proposed two kinds of neural network adaptive PID controller including single neuron adaptive PID and BP neural network adaptive PID, which were realized through Simulink blocks and S-function.Thirdly, three kinds of physical simulation model were simulated respectively. The author analysed comprehensively flow, pressure, displacement, force and others of the system. Simulation data will be theoretical reference to experimental research and engineering applications. In order to test all the controller's adaptive capacity including the conventional PID controller and both of neural network adaptive PID controllers, contrastive simulation were carried out by changing the sinusoidal frequency of disturbance signal. Simulation result showed that the neural network adaptive PID controller had the advantages of better robustness, stronger anti-interference, higher precision and adaptive capacity.Fourthly, the author built light-load heave compensation experimental platform and used the xPC Target to finish the simulation experiment of light-load heave compensation. The sinusoidal position tracking comparative experiment about the conventional PID controller and single neuron adaptive PID controller and step dynamic response comparative experiment by changing the system oil pressure were finished. Experimental results showed that the single neuron adaptive PID controller was better than that of the conventional PID controller in interference and adaptive capacity. Light-load heave compensation system adopted single neuron adaptive PID controller as position feedback controller, the system had stronger robustness and adaptive ability and enhanced its interference ability.Finally, it could conclude through simulation and experiment:(1)Three kinds of physical simulation models of heave compensation experimental system created by SIMSCAPE could simulate flow, pressure, displacement, force and others of the actual system, and had stronger physical meaning, more convenient parameters modification and portability.(2)Composed of the neural network and PID called neural network adaptive PID controller having adjustable PID parameters, simulation and experiment results showed that the neural network adaptive PID controller had higher compensation accuracy, stronger robustness and adaptive ability.