Linear Motor Structure Optimization Design Based On Multiple Support Vector Machine

Permanent magnet synchronous linear motor is widely used in modern machine tools, transportation, logistics, transportation equipment and information devices areas. Because of the characteristic of large thrust,simple structure and high efficiency, linear motor become a new type of drive technology. With the development of linear motor technology, a large number of research scholars domestic and overseas have developed many practical products and equipments, which include the initial linear motor conveyer driven, coal machines, crane, maglev train, etc. Chinese scholars began to research the linear motor in the1970s, and have made considerable achievements since then. They have developed a variety of industrial equipments based on the linear motor technology,such as curtain machine, fried tea machine, conveyor, automatic device and so on.Because of the end effect and the cogging effect, linear motor thrust fluctuation is one of important problems which affect the performance of the linear motor. To control the linear motor thrust fluctuation is of great significance to promote the application areas of the linear motor. This paper present a method to optimize the motor structure parameters for control the thrust ripple thoroughly. The multiple support vector machines was used to establish the non-linear relation model of motor structural parameters with cogging force harmonic characteristics, the non-linear model was then repeatedly used to optimize the motor structure parameters by decision-tree method. The simulation and experiment of a double secondary permanent magnet synchronous linear motor was used to verify the possibility of this method by calculate the thrust fluctuation rate of the optimized linear motor.The main research content is as follows:Firstly, the finite element model of the linear motor was established to calculate the motor sample data space. To explore the relation between structure parameters and output parameters of the linear motor, the main structure parameters were chose to be the input parameters of the finite element model,which affect the output parameters deeply, especially the thrust fluctuation. Based on the finite element model,the orthogonal experiment and random sample were used to obtain the sample data space.Secondly,in order to solve the rapid computing problems of parameter optimization for the linear motor, a nonparametric modeling method was proposed based on the multiple output support vector machine. The SVM model of the permanent magnet synchronous linear motor was established by the kernel function combination of the high/low-order polynomial kernel function where the penalty parameter was optimized by the cross validation algorithm. The computational efficiency and precision of the motor model can meet the large-scale computing requirements in the motor structure optimization area. The multiple support vector machine model of the linear motor can provide high precision multidimensional input data and output data for the optimization process.Thirdly,the decision tree model of the linear motor can be used to gradually narrow the motor structure parameters scopes,and meet the requirements of structure parameter values ultimately. The optimal motor’s3D FEM model has been established to calculate the optimal motor thrust and thrust fluctuation.The result of the experiments above shows the possibility of this method. A multi-SVM motor model can be used to provide data support and establish a mapping from a motor structure parameter input space to a motor output parameter space. A decision tree algorithm has been used to find the optimal motor parameters. The motor with the optimal structure parameters can provides the thrust steady, and the thrust fluctuation can be decreased in a limited range. The method has provide a new way to solve the thrust fluctuation problem. The thrust can meet the motor design standards and reach the steady value in a short time.