基于混合粒子群算法的沥青转运车螺旋搅料器模型及仿真

 为解决沥青转运车螺旋搅料器搅拌特性建模困难的问题，运用人工智能理论建立了以螺距、叶片半径为输入，沥青混合料的离析率为输出的神经网络模型。为克服BP算法与粒子群算法（PSO）的缺陷，将L-M算法与PSO算法相融合的混合粒子群算法PSOLM应用于该神经网络模型的学习算法中。为避免PSOLM算法在全局最优值附近的搜索过程变慢，采用一种从PSO搜索到L-M搜索的启发式算法。仿真试验结果表明，与BP算法、PSOBP算法相比，该算法不仅对螺旋搅料器模型的精度和建模的效率有显著的提高，而且改善了PSO算法的全局寻优能力，提高了算法的收敛速度和计算精度，避免了PSO算法早熟现象的出现，为螺旋搅料器搅拌特性的建模提供一条新的有效解决途径。

Modeling and Simulation of Spiral Stirrer for Asphalt Conveyor Based on Hybrid PSO Algorithm

To overcome the difficulty of modeling for agitation characteristics of spiral stirrer for asphalt conveyor, a Neural Network (NN) model was established, based on both pitch and radius of helical blade being used as the input vector, with the rate of mixture segregation used as the output vector of NN. To avoid the shortcomings of the Back Propagation (BP) algorithm and the Particle Swarm Optimization (PSO) algorithm, a hybrid PSO algorithm PSOLM was employed as the learning algorithm of NN, which was combined with Levenberg-Marquardt (L-M) algorithm. To avoid the slow search speed reated with the global optimum in the PSO-BP algorithm, a heuristic way was adopted to give a transition from the particle swarm search to L-M search. Simulation experiment was performed with model experiment for asphalt conveyor. The results of simulation and experiment show that the proposed approach not only has faster convergence and higher computational precision than the original PSO, but also can overcome the drawbacks of BP algorithm. It is also a new and effective approach for model building of spiral stirrer agitation characteristics.