Research On Rolling Bearing Performance Degradation Assessment Based On Local Linear Embedding And Support Vector Machine

Rolling bearing is the critical component in large rotating machinery,which is called the "industrial joint".The fault of rolling bearing will not only cause huge economic loss,but also lead to significant casualty accident.Therefore,early detection of rolling bearing fault is more important than dealing with the fault itself.It is particularly important how to effectively assess the running condition of the rolling bearing,in order to achieve early detection of faults,and formulate scientific maintenance program to prevent major accidents.From normal condition to complete failure condition in the whole life cycle,rolling bearing need go through a series of different performance degradation condition.Effective monitoring of these performance degradation conditions can effectively organize production and formulate scientific maintenance schedule in advance,which can realize the prevention of major accidents and maximize production efficiency.Rolling bearing performance degradation assessment is a kind of maintenance technology proposed on the basis of this idea.It focuses on the measurement of rolling bearing performance degradation degree over the entire life cycle,rather than too much attention on the diagnosis of fault category in a certain point.The paper is Studied according to such a train of thought as feature extraction,feature reduction,pattern recognition and performance degradation assessment.The main research contents include:1.Rolling hearing vibration signal has characteristics of non-stationarity,complexity of frequency components and so on.Wavelet packet decomposition is adopted to realize multi-resolution analysis of the signal and the time-frequency resolution of high-frequency portion of the signal can be improved,combined with the indexes of time domain and frequency domain,the feature of the rolling bearing can be extracted.2.Local linear embedding(LLE)algorithm is researched and adopted in the reduction of the extracted high-dimensional features for removing the redundancy among them.And the similarity measurement of high-dimensional space in the algorithm is improved.The improved algorithm is applied to the feature reduction,and its superiority can be proved by the experimental comparison.3.The theory of support vector machine(SVM)is researched and adopted in the recognition of rolling bearing different performance degradation degree.Rolling bearing performance degradation assessment index is established.And the performance degradation assessment curve is given combined with the actual data.At last the performance degradation assessment model is given.