内燃机故障声响信号时间序列模型诊断方法

TIME SERIES MODEL FOR FAILURE DIAGNOSIS OF INTERNAL COMBUSTION ENGINE BY SOUND SIGNALS

本文运用统计模式识别原理,提出了一种以内燃机声响信号的时间序列自回归系数及自回归倒谱系数作为模式的特征参数,对参考模式和待检模式用灰色系统的关联测度、Mahalanobis距离测度和Kullback-leiber信息量距离测度同时进行识别,并用层次分析法对识别结果进行综合,从而达到诊断故障目的的方法。通过实验验证,表明其是有效而可行的方法。

On the basis of the principle of statistical pattern recognition and bymeans of the time series of the sound signals of the internal combustion eng-ine, a new method for failure diagnosis of the internal combustion enginehas been developed. Autoregressive coefficients and autoregressive cepstralcoefficients are used as the feature vector of pattern. The relational measureof grey system, the measure of the Mahalanobis distance. and the measure ofdistance of the quantity of information of Kullback-Leiber are employed fordiscriminating the testing pattern to see to which reference pattern it belongs.The results of every discrimination are synthesized by AHP (analytic hierachyprocess). Then the goal of diagnosis is achieved. The method has been verif-ied in a process for diagnosing Diesel Engine-19% in laboratory. The theore-tical and experimental results have shown that the method is practical andeffective.