Particle Swarm Optimization Algorithm And Its Application In Electromechanical Equipment

With the development of high automatization, intellectuality and reliability, the modern electromechanical equipments are becoming more and more complicated in both mechanical and control systems. Therefore, the higher standards are brought up in the equipment designing, the status monitoring and the system controlling, etc, where many of problems can be regarded as the optimization ones and be solved by the socalled optimization algorithms. The Particle Swarm Optimization (PSO) algorithm, as a population-based and heuristic optimization technique, has been successfully applied to optimize a variety of optimization problems in engineering practice because of the simple and rapid convergence speed, and easy implemention. In this thesis, an improved PSO (DDPSO) algorithm is proposed under the theory analysis of the PSO algorithm, and then its applications in the design, the status monitoring and the control field of electromechanical equipments are discussed too. The main works in this thesis are listed as follows.(1) The general mathematical model of optimization problems and the PSO algorithm are introduced firstly,including the research background and present status, the basic principle and main characteristics, and the influence of parameter on the convergence performance of the basic PSO algorithm.(2) To improve performance and avoid trapping to local optimum, a modified PSO algorithm (DDPSO) is proposed based on Dimension Information Share and Dynamic Cognition in the animal cognition and the decision-making process. One dimension of the particle is selected randomly and copied to other dimensions when updating the velocity of the particle. Meanwhile, DDPSO simulates the animal cognition in every running, and the different cognitive stages are represented by the respective velocity formula, in order to adapt for different optimization problems. Therefore, it is an immense improvement in model structure of PSO and more accordant to cognitive process, social nature and swarm intelligence of animal. The effectiveness and practicability are demonstrated by the simulation results in testing with benchmark functions, training of back-propagation neural network (BPNN) and optimizing parameters in the chaos systems.(3) BPNN becomes more and more popular in fault diagnosis, especially in the field of gear-boxes. Unfortunately, the back-propagation algorithm is a gradient-based method, where some inherent problems are frequently encountered when using this algorithm, such as slow convergence speed in training, easiness to trap into a local minimum, etc. The DDPSO algorithm is used to train BP neural network in fault diagnosis of the gear box. Each particle position vector is made up of all of the weights and thresholds in BPNN, and the minimum sum of mean square error (MSE) is treated as the fitness of DDPSO. The experimental results verified that DDPSO, as the intelligent method, can escape from local minimum and has faster convergence than back-propagation, PSO-TVIWD, PSO-TVACD and PSO-DV in training BPNN. DDPSO can achieve quite high accuracy rate of recognition and provide a new way in fault diagnosis of the complicated non-line mechanical system.(4) Based on the testing system for protective effect evaluation of the thermal protection jacket and field data, the mathematical model to adjust the thermal protection jacket is established. As an essentially a typical combinational optimization problem, a discrete Binary Particle Swarm Optimization (BPSO) algorithm is developed to solve this problem. Furthermore, with discussing BPSO in real-time control system, the rectificating error method is proposed to overcome the shorcomings of BPSO. Experimental result shows that the presented algorithm can not only achieve the automatic adjustment of the thermal radiation and automation of the testing process, but also meet the requirement of rapidity, stablity and accuracy for the control system.(5) In the traditional design of electrical control system, the distribution cabinet is usually made manually by professional engineers, who expert a great influence on the quality of the job depends mainly on the experience of engineers, such as high costs, long development cycles and frequent errors. Aimed to the questions above, a new method is proposed to do this work automatically based on PSO. Firstly, the mathematical model of the distribution cabinet is established. Secondly, the methods and steps of PSO to solve this problem are discussed. And finally, the proposed method and DDPSO are experimented in a real application of automatic welding equipment for wire mesh. Experiment results prove the proposed method is fast and effective, can accomplish the intelligent and automatic design of the distribution cabinet, and can also shorten the design cycle time and reduce the design error risks. This work provides a new way to achieve automatic distribution cabinet design for the complicated electrical control system, and it can be highly significant to raise the information and automation level in equipment manufacturing industries.