滤料过滤效率的数值预测研究

Filtration Efficiency Numerical Forecast of Filter Material

以耐高温滤料过滤效率的试验数据为基础数据,分别采用LS-SVM、RBF神经网络和BP神经网络3种方法,以空气温度、过滤风速、发尘浓度作为输入,过滤效率作为输出训练算法,预测不同测试条件下滤料的过滤效率,并对3种算法的预测结果进行了对比分析.结果表明,采用LS-SVM无论从预测精度还是运行时间上,预测效果都最为理想;LS-SVM的过滤效率预测模型与RBF、BP神经网络相比,平均绝对百分比误差最小,程序运行时间最短.

Based on the filtration efficiency experimental data of the high-temperature resistant filter material ,the filtration efficiency was forecasted in different test conditions using the method of IS-SVM, RBF neural network and BP neural network respectively. The input includes temperature ,filtration velocity and dust concentration. The forecast results of the three-intelligent arithmetic were contrasted and analyzed. It indicated that the LS-SVM arithmetic was the best compared with the RBF and BP neural network in precision and running time, because the LS- SVM arithmetic was the most time-saving method with the least mean absolute percentage error.