基于组合分类器的MSF发动机综合故障诊断研究

Study on the Comprehensive Fault Diagnosis about MSF Engine Based on the Combining Classifier

摘要: 本文针对发动机综合性故障准确定位困难及瞬时转速和声压信息量大等特点，提出了基于投票表决方案的组合分类器决策层融合算法，指出了神经网络学习算法存在的不足，并对学习算法进行了改进，利用三重BP网络结合投票表决方案，提出了基于组合神经网络分类器的发动机综合故障MSF诊断方法。通过试验研究和结果对比，表明了该方法能可靠有效地提高发动机综合性故障诊断能力。

Abstract: Aiming at the engine characteristics of fault location trouble and large amount of information in instantaneous speed and pressure, the fusion algorithm of the decision level about the combining classifiers is proposed based on voting by ballot; the deficiency of neural network learning algo-rithm is pointed out; and the learning algorithm is improved; with voting scheme using three BP network, integrated fault diagnosis method of engine combined MSF is proposed based on neural network classifier. Through the experimental research and comparative analysis, the method is more reliable and effective in improving the comprehensive fault diagnosis capability of the engine.