Mine Hoist Fault Diagnosis Based On Data Field Model

Mine hoist fault diagnosis plays a very important role in safe production. Thereare many methods for mine hoist fault diagnosis; Fuzzy C-mean algorithm andspectral clustering algorithm are two typical methods. As random central uncertainty,traditional fuzzy C-mean method is easy to fall into local optimum. Traditionalspectral clustering method needs to determine the cluster number artificially, and it issensitive to the initial cluster centers and weak of robustness. As a kind ofmathematical model of non-contact interaction among data objects, Data field modelcan reveal the clustering characteristic of data objects. In this thesis, data field modelis introduced into the fuzzy c-means clustering method and spectral clustering methodto avoid the shortage and improve the performance of fault diagnosis.This thesis proposes the clustering algorithm based on data field model and fuzzyC-mean algorithm. Using the potential value of fault points in theconstructed fault field, the proposed algorithm find out noise points and kick out.Obtain the initial cluster centers by the natural aggregation behavior of fault points inthe fault field to guide the fuzzy c-means clustering and optimize clusteringperformance. The experimental results on the Aggregation dataset show that the fuzzyC-means algorithm based on data field model is effective to avoid the shortage of thetraditional fuzzy C-means method and improve the performance of fault diagnosis.This thesis proposes the clustering algorithm based on data field model andspectral clustering algorithm. The algorithm adjusts the similarity matrix appropriately,with the help of data field model to obtain the k value and the initial cluster centersthat spectral clustering algorithm needs. Finally, cluster the adjusted eigenvectorsusing the K-means algorithm. The experimental results on the Iris dataset, Winedataset, Zoo dataset show that the spectral clustering optimization algorithm based ondata field model is effective to improve the stability and accuracy of spectralclustering algorithm.Two improved clustering algorithm are applied to mine hoist fault diagnosis. Thetest results on the real electrical machinery dataset and real hoist bearing dataset showthat they can efficiently enhance the performance of mine fault diagnosis.