Study On Adaptive PID Control Of Ship Course Based On Neural Networks

In the process of the voyage, load change, the speed change and the wave interferencewill cause serious nonlinear control problem. To overcome these problems, The controllerhaving stronger adaptive ability and better robustness was needed. At present, most ofcontroller designs still rely on a certain model, they can’t satisfy control requirements whenvoyage environment is bad or some changes appear. Therefore, Many researchers at home andabroad have made unremitting efforts on how to improve the control performance of shipcourse, and made some achievements. In this development tendency, Considering that neuralnetwork has the ability of approximate any nonlinear function, this paper combined neuralnetwork with PID control and fuzzy theory robustness, then used them in ship course controland designed a fuzzy PID controller based on BP neural network for ship course control. Realexample simulation in the case of no wave interference, wave interference and the modelparameter perturbations were done, and simulation results showed that this controller cansolve the nonlinear control problem well.Firstly, the historical background of this issue and its significance, the developingsituation of ship course control and neural network control were introduced. Then themathematical model of ship course motion, steering engine, and environment interferencewere elaborated, which mainly analysised the six degrees of freedom model ship and deducedthe first order nonlinear KT model, and gave the steering gear servo system simulation figureand wave spectrum analysis and the waves of disturbing torque was simulated. Secondly, thebasic model of neural network and its way to work were introduced, then analyzed BPnetwork and its learning arithmetic and the stability of neural network briefly, explained theneural network stability and convergence, laid the foundation for neural network adaptivefeasibility. Thirdly, introduced self-adaptive PID control method based on BP neural networkand designed prediction model according to the need of the control method. The convergencerate of the control system was slow and sometimes reached a local minimum, in order toavoid the disadvantages, added a fuzzy module to the control system. The result of thesimulation showed that the convergence rate has increased, so this method is feasible. At last,the conjugate gradient algorithm was analyzed, since its conjugate characteristics was largelydetermined by negative gradient algorithm, so it had not reached the global optimal faults.Therefore, this paper put forward the improvement conjugate gradient algorithm. Sincenetwork convergence speeded up and attained global optimum along the conjugate direction when the complexity was unchanged in improved conjugate gradient method. This paper usedconjugate gradient BP algorithm instead of the normal BP algorithm in design and added thefuzzy module into the control system. Then the fuzzy PID control system based on improvedBP neural network was finished designing. Through real example simulation we can see thatthis method had better robustness in ship course control, so this control method is feasible.