Research On Bearing Fault Diagnosis Method Based On Improved Association Rule Algorithm

With the development of industrial informatization,various mechanical equipments are monitored and managed in real time through information technology,and a large number of historical data are stored.How to use data mining technology to intelligently analyze these data,thus providing scientific basis for fault diagnosis and equipment maintenance,is an urgent problem to be solved.Bearing is an important component of mechanical equipment that is indispensable and prone to failure.Bearing failure will affect other components,and then affect the working condition of the whole system,resulting in the performance reduction of the whole mechanical equipment,and even failure.Therefore,it is very important to diagnose bearing failure in time.By analyzing the bearing vibration signals,the corresponding time domain and frequency domain signal features can be extracted,and these features can be used to construct the feature dataset required for data mining.Data dimensionality reduction is performed on the feature dataset to reduce data redundancy,followed by the selection of a suitable method to process the feature set data.Traditional data mining methods and improvement methods are mostly serial algorithms,but the processing power of a single computer is very limited and not suitable for large-scale and multi-dimensional data mining.On this basis,a bearing fault diagnosis method combining clustering algorithm and association rule algorithm is proposed,which can effectively reduce the classification errors and the number of repetitions by clustering algorithm and evaluate the clustering effect using clustering evaluation index.By clustering analysis of rolling bearing data,the characteristic data can be divided into intervals according to the clustering effect,which can lay the foundation for further fault diagnosis and data mining.The spectral clustering(SC)algorithm is mainly used to process the data in the feature set and compare with the K-means algorithm to verify the efficiency of the spectral clustering algorithm and to prepare for the subsequent association rule mining.If only the support and confidence in association rules are used,it is not possible to filter out all the strong association rules,resulting in too many rules being generated,making the matching time too long.To solve this problem,boosting degree can be introduced to optimize the judgment framework of strong association rules to filter out association rules with stronger relevance and reduce the matching time of fault diagnosis.Meanwhile,by mining and analyzing rolling bearing data,the SC-Apriori algorithm is proposed to mine the association relationship between different characteristic data of rolling bearings,which can be used to determine their fault types based on the matching rate and facilitate timely response.Specific experimental analysis using the bearing failure dataset publicly released on the website of the Bearing Data Center of Case Western Reserve University,USA,verifies the efficiency of the SC-Apriori algorithm with higher accuracy and lower computational overhead compared to the K-Apriori algorithm and the K-FP-Growth algorithm.The Apriori algorithm is a commonly used data mining algorithm,but it generates too many candidate item sets,which leads to excessive time consumption and limits its use in practical applications.To solve this problem,a new algorithm,SC-FP-Growth algorithm,is proposed,which combines spectral clustering with FP-Growth algorithm,effectively reducing the number of candidate item sets and decreasing the time complexity of the algorithm.Through experimental validation,it is found that the difference between SC-Apriori algorithm and SCFP-Growth algorithm in matching rate is not significant,but SC-Apriori algorithm still outperforms SC-FP-Growth algorithm in terms of time overhead.Therefore,this experiment again verifies the superiority of the SC-Apriori algorithm.Although the SC-FP-Growth algorithm can reduce the number of candidate item sets,it still cannot completely replace the Apriori algorithm,and a suitable algorithm needs to be selected according to the specific situation in practical applications.