基于网格搜索和投票分类模型的喷油器故障诊断研究

Research on Fuel Injector Fault Diagnosis Based on Grid Searchand Voting Classification Model

为了提高高压共轨试验台对喷油器检修效率,提出一种基于网格搜索和投票分类模型的喷油器故障自动诊断方法。由于压电喷油器故障数据采集困难,使用AMESim软件模拟不同轨压和脉宽状态下压电喷油器可能出现的多种故障情况。随后,将采集到的1 760组数据使用由随机森林、支持向量机和GBM组成的投票分类模型进行训练,并使用网格搜索法优化各分类器的超参数。实验结果表明:该模型对压电喷油器的5种故障状态及正常状态诊断时的准确率、精确率、召回率和F1-score分别为98.86%、99.13%、98.56%、98.83%,表现出较高的准确性和稳定性。该方法能够快速高效地对喷油器故障情况进行定位。

In order to improve the efficiency of injector maintenance for high-pressure common rail test benches,an automated diagnosis method for injector faults was proposed based on grid search and voting classification models.Due to the difficulty in collecting fault data of piezoelectric injectors,AMESim software was used to simulate various fault conditions that may occurred in piezoelectric injectors under different rail pressures and pulse width states.Subsequently,the collected 1760 sets of data were trained using a voting classification model composed of random forest,support vector machine,and GBM,and the hyperparameters of each classifier were optimized using grid search method.The experimental results show that the accuracy,precision,recall,and F1-score of this model in diagnosing the 5 fault states and normal state of piezoelectric injectors are 98.86%,99.13%,98.56%and 98.83%,respectively,demonstrating high accuracy and stability.This method can be used to locate injector faults rapidly and efficiently.