改进的遗传算法在汽车故障诊断中的应用

Application of Improved Genetic Algorithm in Automobile Fault Diagnosis

摘要:

本文提出了一种基于遗传算法的自适应模型来预测发动机部件的特征图。将数值模型和测试数据的主要性能参数的区别函数作为目标函数,并且考虑了元件特性图的耦合因子作为优化参数,自适应模型和测试数据之间的主要性能的参数和过程参数差异范围显示在0.05％内。同时,该部分的总温度和压力控制在1％以内。此外,故障诊断模型是通过小偏差方程方法实现的,其中实现了气路分析和症状测量参数,代表发动机性能参数的变化。它表明了症状参数的选择值对故障诊断误差影响很大,最佳选择值为阈值的1/3。故障诊断模型的症状参数与实际故障之间的变量值的相对误差可以控制在5％以内,因而可以正确评估故障类型,并且在故障诊断模型的所有执行中都不存在误诊。

Abstract:

An adaptive model is proposed based on genetic algorithm to predict the characteristic map of components. The difference functions, of the primary performance parameters between numerical model and test data, are taken as objective function. The coupled factors of component characteristics" map as optimized parameters are considered. The difference of the main performance parameters and process parameters between the adaptive model and the test data are shown to be within the range of 0.05%.Meanwhile, the section"s total temperature and pressure are controlled within 1%. Furthermore, a fault diagnosis model is developed by the small deviation equation method in which the gas path analysis is implemented and the symptom and measuring parameters represent engine performance parameters" variation. It shows that the selection and relatively variable value of symptom parameter have great effect on fault diagnosis error, and the best selection of value is 1/3 of threshold. The relative error of variable value between the symptom parameter of fault diagnosis model and the real fault can be found to be controlled within 5% and it can do the correct evaluation of fault type. And the fault diagnosis model has no misdiagnosis in all the performed conditions.