Application Of Improved Particle Swarm Optimization In System Identification Of Voice Coil Motor

Voice coil motor is a special form of a direct drive motor. The systems are non-linear in nature and their dynamic equations have several unknown parameters. Therefore, system identification is a prerequisite to analysis of a dynamic system. However, in the system identification methods, One of the most promising algorithms for solving global optimization problems is the Adaptive Particle Swarm Optimization. The APSO is proposed for handling non-linear constraint functions with boundary limits of variables to find the best values for the unknown parameters of voice coil motor system within a set of search space. In this algorithm, the APSO guarantees fast speed convergence and accurate solutions regardless of the initial conditions of parameters. In this paper, the experimental platform is LED wire bonders. The voice coil motor control horn fast movement in the Z-axis direction. Through analyzing the voice coil motor and its’ closed-loop system. Several tests are carried out and using improved APSO to identify the voice coil motor system. Then, with Matlab/simulink software, simulating the identified model of the voice coil motor to validate the accuracy of selected unknown parameters. Simulation display can be a good fit between the simulation and experimental results. The main work of this paper has the following aspects:1. A brief review of the development status of the voice coil motor. Introducing the domestic and foreign of the methods used in the parameter identification. Focus on the comparison and selection of a voice coil motor parameter identification method. this paper makes a more detailed description in particle swarm algorithm’s inventions, research and applications.2. based on the voice coil motor in LED wire bonders experimental platform, paint its PRO/E three-dimensional model diagram to introduce its structural characteristics. Analysis of the structure of the voice coil motor model. Build a voice coil motor mathematical model. Analysis of the voice coil motor control in LED wire bonders experiment platform for "PID+velocity loop, position loop double-loop control system". Lay the foundation for the following closed-loop experiments, using improved particle swarm optimization algorithm for identification parameters and the models simulation.3. Analysis and describe the functional role of TURBO PMAC control card in LED bonders experimental platform specifically and its corresponding data acquisition software PEWIN32PR-O2. Make a detailed explanation of the open-loop experiments. experimental data obtained plotted in Figure. Plot experimental data in FIG Make a detailed explanation of the sampling frequency of the closed-loop experiments. TURBO PMAC experimental procedures TURBO PMAC experimental procedures. Design eight groups of different parameters Experimental program and plot experimental data which obtained in closed-loop experiment in Fig. Also designed two specific PID parameter values and got the experiment values.4. Describe the particle swarm optimization algorithm, And explain the reasons of premature convergence for PSO. In view of this shortcoming for PSO, this paper proposes a improved particle swarm optimization. Make a detailed explanation of the improved method. Therefore, the PSO algorithm can avoid falling into local optimum and it can find the global optimum solution easily and efficiently5. on the basis of the eight groups of the closed-loop experimental data. Use improved adaptive particle swarm optimization algorithm processing data, identify parameters. In the eight groups of the recognized value, the smallest identification error is the fifth group. Use the fifth group to be simulation target. Simulate the other seven groups of experimental data. Experiment and obtain experimental data. Still use the fifth group to be simulation target. Simulate the results of these two sets of PID parameters.Making the eight groups of data with the PID parameters into two groups for cross-examination. square-error and relative maximum absolute error were calculated separately. verifying the reliability of the model parameter identification.