基于支持向量机的离心泵磨损特性分析方法

Method for analyzing the wearing characteristics of centrifugal pump based on support vector machines

针对离心泵磨损机理错综复杂和高度非线性的特点,提出了基于最小二乘支持向量机的离心泵磨损特性分析方法,通过对算法的实现,建立了离心泵的磨损特性分析和几何参数优化的智能模型,模拟得到离心泵的磨损特性关系,分析了磨损随轮叶片几何参数的变化规律。与神经网络和普通支持向量机方法进行计算比较,结果表明,最小二乘支持向量机磨损预测模型得出的的平均相对误差只有0.005%,学习速度为12步,训练时间为1.1s,学习速度和预测精度得到了很大的改善。可为离心泵的磨损特性分析及其抗磨可靠性设计提供新的可行方法。

An intelligent method based on the least squares support vector machine is suggested for analyzing the wearing characteristics of centrifugal pump.The corresponding model is established which can be used to analyze the relationship between wearing and geometric parameters of the impeller in pump.The mean relative error of predicted wearing of a test pump using the model based on least square support vector machine method reaches 0.005%,which is much better than that predicted by the models based on neural network method and common support vector machine.