Research On Fault Diagnosis And Fault Trend Prediction Of Centrifugal Blowers Based On Multi-source Information Fusion

With the continuous progress of technology and manufacturing,modern industrial equipment is developing towards the direction of large-scale,automation,and information,and its structure is becoming more and more complex,and the working environment is harsh,the frequency of failure has also increased significantly.The centrifugal blower is an important part of rotating machinery,widely used in buliding,chemical,energy and other industries.Because of its complex structure,many types of fault,and long maintenance period,it is of great significance to realize the intelligent fault diagnosis and trend prediction of centrifugal blowers.With the development of sensors and big data technology,the operation of mechanical equipment contains a large amount of monitoring data.Mining fault information from multi-source sensor data can provide an effective way to realize the diagnosis and trend prediction of key equipment.This article utilizes the special key demonstration project of Chongqing Technology Innovation and Application Development,and has carried out research on fault diagnosis and trend prediction methods and systems of centrifugal blowers based on multi-source information fusion.Firstly,the composition of the multi-source sensor information of the centrifugal blowers is systematically analyzed,and a multi-level information fusion framework combining data-level and feature-level is constructed according to the characteristics of the multi-source information.The fault diagnosis model of centrifugal blowers based on adaptive deep convolutional neural network is established on the basis of multi-source information.10-fold cross-validation technology and two model evaluation metrics(Accuracy,F1-score)are used to evaluate the diagnosis model.Secondly,sparse auto-encoder is used to realize the feature fusion of multi-source information of centrifugal blowers,and the health index of fault trend prediction of the centrifugal blowers is constructed by calculating the difference between the network reconstruction value and the input value.The fault trend prediction model of centrifugal blowers based on long short-term memory network is established on the basis of multi-source information.Four model evaluation metrics(RMSE,MAE,MAPE,R~2)are used to evaluate the model.Thirdly,the corresponding remote operation and maintenance system for turbine equipment is developed.The overall framework and workflow of the system are proposed,the application environment and configuration of the system are described.The implementation process and functional interface of the software function modules such as the system’s running status module,fault diagnosis module,fault trend prediction module and historical query are introduced in detail.Finally,combined with application cases,the proposed method of fault diagnosis and trend prediction of centrifugal blowers based on multi-source information fusion is analyzed,and compared with different single signal,diagnosis and trend prediction methods to verify the feasibility and effectiveness of the proposed method.