Optimization Design Of Magnetorheological Brake Based On RBF Network Agent Model

Magnetorheological fluid(MRF)is a new type of smart material which is developed in recent years.When it is in no extra magnetic field,the rheological properties are similar to ordinary Newtonian fluid properties;When it is outside to strengthen under the action of magnetic field,the magnetic rheological effect occurs that the apparent viscosity is orders of magnitude increase in milliseconds;And the yield stress and viscosity can be reversible changes with the magnetic field.The magnetic rheological brake is a new type of linear actuator.There is a contradiction between the optimization goal of the brake torque and the brake weight in the multi-objective optimization design of the magnetic rheological brake,it requires not only the braking torque meeting the minimum value of the torque requirements,but also reducing the weight of brake.For this multi-objective optimization problem,the RBF network model has a good solution,but the RBF network itself can not be selected for the measurement data,so it must be through a large number of sample data to be able to accurately fit the multi-dimensional hyper surface.In combination with a mathematics method and statistical method,the response surface method adopts a finite number of experiment design and regression fitting to describe the relationship between the input variables and the output response of complex systems.So the response surface method and the RBF network method are combined,which not only can guarantee the accuracy of the computation,but also can improve the efficiency of operation,so it has significant meaning about the quick response of complex engineering design.The approximate model method can make the black box problem of design variables and the optimization of objective function,approximate representation by implicit function,then uses the optimization algorithm to improve the optimization effect on the basis of approximate model.The approximate model method mainly includes two parts of the experiment design and the model establishment of the two part.The Latin hypercube sampling method is widely used in experiment design,but there are some problems that the sampling time and sampling quality can not be well balanced at the same time.So in this paper,a heuristic method of Latin hypercube sampling is proposed,which will only randomly generate the first sample point while the rest of the sampling points are obtained by the heuristic method.Based on this method,a variety of response surface models are constructed and tested.The results show that the proposed method is not only a short sampling time,but also the sampling points are distributed evenly,which is helpful to construct the high precision and high efficiency response surface approximation model.In this paper,a Latin hypercube sampling method is used to construct the RBF networksurrogate model of the source function.The optimal solution set of the geometric parameters of the magnetic rheological brake is obtained by using the NSGA-II multi-objective genetic algorithm.The results show that the method proposed in this paper can better solve the problem of multi objective optimization design of magnetic rheological brake,and it also can be used for the optimization design of other complex mechanical and electrical products.