Temperature Control For Hydraulic Source Based On Neural Network

Oil system is an important part of a test-bed mechanism.It provide reliable quality and stable hydraulic source for Simulator of the test-bed mechanism,to meet the load-driven simulator to achieve the corresponding movement characteristics. So the test-bed request for high performance about oil temperature. So must control the scale of temperature to ensure its accuracy.Based on analyzing working-parameters and characteristics of the hydraulic system, build control system for hydraulic source. The construction of temperature control system is composed of selection of cooling device and design of cooling circuit. It determine all parameters of a removable plate heat exchanger on the basis of calculation of the system's largest heating power in different conditions. It choose three-percentage-flow valve to adjust cooling water continuously, by regulating the ratio of three spool valves to regulate the cooling water flowing through opening of plate heat exchanger and changing its heat-exchanging ability to achieve the purpose of temperature control.By analyzing, it know the temperature control for hydraulic system is a big-delay, nonlinear, time-varying and uncertain system in essence, and it is difficult to use precise mathematical model to describe it, which bring higher demand on temperature control. Refer at research and application on the temperature control about large-delay system home and abroad, First, it use PID algorithm to study the system and carry out the experiment under certain conditions. The results basically match with simulation that system response very slowly for a long time and temperature changes largely.It finally design a single neuron PID control algorithms, which combine neural networks and PID control. Training on the neurons self-learning, it change the connection-weight between neurons according to the impact on output performance of system when the object changes parameters.Thus its steady state correspond to the PID controller parameters under the control of optimization. Unter the simulation analysis and experimental verification, it shows that single neurons PID control has stronger adapting ability and higher control precision.In order to interprete nonlinear dynamic process of the temperature control system better and predict or estimate changing trends of the system, it use a neural-network predictive control algorithms. It design the neural network predictive control system using SIMULINK dynamic simulation tools and related objects, through training of the neural network , geting the satisfied results of training. Dynamic visual simulation find that the system will not only tracking performance better, but have a stronger robust to the model mismatch.