Research On Equipment Fault Diagnosis Method Based On Incremental Ensemble Learning

With the deep integration of information technology and industrial manufacturing,a large amount of state data generated in the process of equipment operation has been retained,which makes the effective identification and prediction of equipment fault using large data analysis method become the mainstream in the field of fault diagnosis.However,due to the change of equipment operation status with time in the production process,the original fault type is repaired,and the new fault type is generated,the traditional machine learning method can not meet the actual demand of real-time change of equipment operation status in the production process.Applying incremental learning method to fault diagnosis field can continuously learn new knowledge from incremental data,while preserving most of the previously learned knowledge,so as to improve accuracy,save time and space costs,and reduce the amount of calculation.However,the incremental state data in the operation of equipment has the characteristics of mass,imbalanced,high noise and strong causal correlation.If not processed,the accuracy of fault identification will be seriously affected.In view of the above problems,according to the characteristics of equipment operation status data and referring to a large number of domestic and foreign literatures,this study proposes a dynamic weight ensemble learning model based on incremental information.The innovation of this model is mainly embodied in the following three aspects:(1)Aiming at the imbalance characteristics of incremental state data,a dynamic weighted resampling algorithm is proposed,which dynamically adjusts the sample weight and realizes real-time processing of imbalance incremental data.(2)Considering that the original fault is repaired with the generation of new fault modes over time,an ensemble learning-base classifier elimination mechanism is proposed to dynamically adjust the model,abandoning the base classifier with low classification accuracy and not suitable for the current running state,so as to adapt to the current running state of the equipment and effectively improve the reliability of the classification model.(3)In view of the strong causal correlation between equipment faults,this model proposes an effective information screening strategy,filters more valuable new features and trains a new base classifier to update the original ensemble classification model,realizes incremental information fusion,and solves the problem of fault pattern classification for incremental state data.Experiments show that the research method in this paper can effectively solve the problem of fault diagnosis of incremental imbalanced equipment operation status data.Incremental learning is used to realize the incremental information of the effective information in the new data with the original model and the dynamic updating and adjustment of the classification model.The real-time state data generated in the process of equipment operation are processed efficiently,and the imbalanced data processor is used.The method eliminates the negative influence of data imbalance on diagnostic accuracy,realizes real-time processing of imbalance incremental samples,significantly improves the diagnostic accuracy of the model,saves time and space costs,and meets the actual needs of industrial production.