Study On Rotate Speed Estimation And Abnormal Sound Characteristics Extracting Based On The Engine Working Sound Analysis

The object matter of this paper is extracting characteristic parameters ofabnormal noise signal of engine. For acoustic signal of engine has muchabnormal state information and malfunction state information of engine, wecan find the trouble location and the source of trouble by extractingcharacteristic parameters of abnormal noise signal of engine accurately.Two parts are discussed in the paper. Firstly, we extracted many importantcharacteristic parameters through the short-time analysis of abnormal noisesignal of engine and introduce two methods of pitch estimation in time domain.Further we bringed forward a method of rotate speed estimation based on pitchestimation and we can estimate the rotate speed of engine through the pitch ofsignal and analyze the operating mode of engine. Secondly, linear predictionanalysis: extracted the linear predictor coefficient of abnormal noise signal ofengine by using durbin arithmetic, schur arithmetic and cholesky arithmeticrespectively and then analyzed the relative merits of these algorithm.Furthermore we can educe other characteristic parameters of abnormal noisesignal through the linear predictor coefficient of abnormal noise signal andbuild union eigenvector through all the characteristic parameters in order toextract the characteristics of abnormal noise signal inextenso.The main work of the paper is as follows:1. Analyzing the abnormal noise signal of engine in time domain.In the paper we discussed several processing methods of abnormal noisesignal in time domain and introduce emphatically some methods such asshort-time energy, short-time average amplitude, short-time average zeropassage frequency, short-time autocorrelation function and so on. The physicalmeaning of these methods is specific and it is easy to realize these methods andalso we can obtain many important characteristic parameters of abnormal noisesignal in time domain through the methods. We introduced the application oftime domain analysis for pitch estimation and analyze two methods of pitchestimation emphatically: the methods of pitch estimation based on short-timeautocorrelation function and based on short-time average amplitude differencefunction.2. Rotate speed estimation based on pitch estimationRotate speed is an important characteristic of engine and the malfunctionof engine will induce the anomaly of rotate speed. In the paper we discussedthe rotate speed estimation based on pitch estimation. Through analyzing thefundamental period of signal, which is gathered under different rotate speed,we can build the comparison form between fundamental period and rotatespeed through which we can judge the working state of engine.3. Linear prediction analysis for the abnormal noise signal of engineIn the paper we analyzed the abnormal noise signal of engine by linearprediction, extracted the linear predictor coefficient of abnormal noise signal ofengine and labor the relations and operation methods for predictor rank,predictor coefficient and model plus. We introduced four linear predictionmethods, which are durbin method, schur method, cholesky method andimproved cholesky method. Through analyzing and comparing the methods wecan demonstrate the validity of linear prediction analysis.4. Principal component analysisIn the paper we extracted linear predictor coefficient through linearprediction analysis and through the linear predictor coefficient we educe othercharacteristic parameters of abnormal noise signal such as part correlationcoefficient, lsp coefficient and cepstrum coefficient. Through them we can build unioneigenvector and process principal component analysis. Principal componentanalysis can reduce the dimension of vector and predigest operation.Anyway the method of rotate speed estimation based on pitch estimationwhich is discussed in the paper is innovative. We can not only estimate rotatespeed but also judge the working state of engine by using the method. At thesame time the linear prediction algorithm based on correlation method andcovariance method has extensive applicability, reliability, progress andintegrality. The effect of the method that using it to extract the characteristicparameters of abnormal noise signal of engine is satisfactory and the methodhas great theoretical meaning and use value.