基于群智能辨识的轧辊偏心补偿方法研究

Study on Roll Eccentricity Compensation Method Based on Swarm Intelligence Identification

为消除轧辊偏心引起的板带厚度周期波动,设计了融合蚁群优化的粒子群(Particle Swarm Optimization Combined with Ant Colony Optimization,PSO-ACO)算法对轧辊偏心进行辨识和补偿。为避免粒子群算法陷入局部最优,PSO-ACO算法通过将伪随机比例规则引入粒子群种群最优值的确定中,以提高种群多样性;结合信息素浓度更新式改进粒子位置更新式,使粒子注重当前搜索信息,加快粒子搜索速度。仿真实验结果表明,在几种典型测试函数和离线轧辊偏心辨识求解上,融合蚁群优化的粒子群算法具有较高的求解精度和满意的求解速度;基于此算法的偏心辨识结果进行的主动轧辊偏心补偿,可明显减小偏心对板带厚度的影响,提高板带厚度精度。

To eliminate the periodic fluctuation of strip thickness caused by roll eccentricity,the particle swarm optimization combined with ant colony optimization(PSO-ACO)algorithm is designed to identify and compensate the roll eccentricity.In order to avoid the PSO algorithm falling into local optimum,pseudo-random-proportional rule is introduced to the determination of the population optimal value in PSO for improving the diversity of population.The update expression of particle position is improved in combination with the update expression of the pheromone concentration,which makes particles pay more attention to the current search information and accelerates the search speed of particles.The simulation experiment results show that the PSO-ACO algorithm has higher accuracy and satisfactory solution speed in the solution of several typical test-functions and off-line roll-eccentricity identification.Through active roll-eccentricity compensation based on the eccentric identification results of PSO-ACO algorithm,the influence of eccentricity on the strip thickness is reduced significantly and then the accuracy of the strip thickness is improved.