基于改进PSO-LSSVM的浆体管道临界淤积流速预测

The prediction of critical deposition velocity in slurry pipeline based on improved PSO–LSSVM

针对现有各种类型的浆体输送管道临界淤积流速经验公式形式不一,适用面较窄,且实测值与计算值的误差很大的问题,采用了基于改进的粒子群算法与最小二乘支持向量机相结合的新方法对临界淤积流速展开预测。所改进的粒子群算法在采用异步变化的学习因子、二次型惯性权重递减策略的同时,融合了自然选择的思想。通过实验仿真,结果显示本文采用的算法所取得的效果相比传统的方法要更优。同时,和所选的具有代表性的经验公式相比有着更高的精度。

In view of the problems that existing various types of the empirical formula in critical deposition velocity of slurry pipeline are different, the application scope of formula is narrow, and the error between the measured value and the calculated value is very large, the present paper is aimed at combining an improved particle swarm optimization algorithm with least squares support vector machine algorithm as a new method to predict the critical deposition velocity. The improved particle swarm algorithm used a strategy of asynchronous changes in learning factor and quadratic decreasing inertia weight for nonlinear adjustment of the inertia weight. At the same time, this algorithm also introduced the idea of natural selection. The simulation results show that the prediction result obtained by this algorithm is superior to conventional methods. Meanwhile, it has higher precision when compared with the selected empirical formula.