Research On Equipment Lifecycle Management And Fault Prediction

With the development of the Industrial Internet of Things and the advent of the era of big data,equipment has increased in complexity and intelligence,and the management requirements of enterprises on equipment have gradually increased.Traditional equipment maintenance strategies are based on passive maintenance methods such as fault-triggered emergency repair and excessive maintenance which cannot meet modern maintenance needs.In this context,product lifecycle management(PLM)and failure prediction and health management(PHM)technologies have emerged.Through the collection,storage,statistical analysis of the data generated during the entire process of equipment operation and maintenance until decommissioning and scrapping,and with the help of advanced algorithms,timely prediction of the remaining useful life(RUL)of the equipment avoid serious production accidents in time.Condition-based maintenance provides technical support and plays an important role in modern industry.Based on the data-driven PHM technology,this thesis combines deep learning-related algorithms to study the life cycle management and remaining life of mechanical equipment under complex conditions.Aiming at the multi condition attributes of the complex service equipment system,the CanopyKmeans algorithm is proposed to identify the condition attributes of the equipment,so as to reduce the interference of condition changes on RUL prediction.A direct prediction method of RUL based on hybrid neural network model is studied.Convolutional neural networks(CNN)is used to automatically extract the deep features of complex working conditions to avoid the problem that traditional feature extraction methods rely too much on expert experience,long short-term memory(LSTM)is used to extract time-series dependence characteristics and predict equipment RUL.Data enhancement method is used to improve the overall learning performance of the hybrid model.Compared with the traditional K-means clustering algorithm,the proposed Canopy-Kmeans algorithm is more accurate in selecting the initial clustering centroid,has better clustering effect for complex working conditions.Based on CNN-LSTM hybrid neural network model,the RUL direct prediction method studied in this paper can mine the deep features of time series data,reduce the dependence on domain expert knowledge to a certain extent,and make the model more universal.Finally,based on PLM theory and PHM technology,using network development technology,a set of visualization platform of equipment life cycle management and fault prediction is built to realize the data collection and statistical analysis of equipment life cycle.In this paper,the verification experiments are carried out on C-MAPSS public data set,the results show that the complex condition recognition model of this paper has better robustness to the changes of equipment working conditions.Compared with the traditional shallow model,the hybrid neural network CNN-LSTM prediction model achieves higher prediction accuracy.At present,the system has been put into use in a certain unit in Shanghai.The system has been running stably and has completed the life-cycle information management of more than 20000 sets of equipment.The system fault early warning is timely and accurate,and the visualization management of equipment real-time performance improves the internal digital management level,which has good application value.