Application Of Improved HHT Algorithm With Grey Model In Fault Diagnosis

Modern manufacturing industry is developing with features of large-scale integrated, information intelligence and automatic procedure. To improve the production quality and production accuracy, research of fault diagnosis should be enforced in machine automation field. HHT(Hilbert-Huang Transform) is one of the algorithms which can recover all the initial features of the original signal for fault detection. The project focuses on the research of FPC(Flexible Circuit Board) manufacturing and adopts the Grey Prediction Model with HHT method for analysis of the monitored signal. Based on HHT, it can find out the possible fault position and evaluate the manufacturing situation. The paper focuses on:1. Propose a second order Grey Model(GM(2,1)) improved with HHT algorithm: GM(2,1)-HHT. Based on the second order Grey Function, this method adopts firstly the GM(2,1) for boundary extension. Secondly, the extended signal could be transformed with HHT algorithm to observe Hilbert frequency chart. It eliminates the influence of boundary problem with HHT algorithm and is capable of forecasting non-monotone fluctuation fault signal.2. Propose a Verhulst Prediction Model improved with HHT algorithm: Verhulst-HHT. Based on the Verhulst function, this method adopts firstly the Verhulst Model for boundary extension. Secondly, the extended signal could be transformed with HHT algorithm to observe Hilbert frequency chart. It eliminates the influence of boundary problem with HHT algorithm and is capable of forecasting saturation type fault signal.3. To compare the prediction performance with the proposed two improved model, this paper conducts simulation prediction experiment with the improved HHT algorithm based on GM(1,1): GM(1,1)-HHT. It can conclude from result of the experiment that GM(1,1)-HHT is capable of processing the fault signal in exponential type; GM(1,1)-HHT is capable of processing the fault signal in non-monotone type; and Verhulst-HHT can decrease forecast error for predicting saturation signal.