Development Of An Online Monitoring System For The Safe Operation Of Large Induced Draft Fans

As the auxiliary equipment of industrial boiler,the safe and reliable operation of large induced draft fan is directly related to the stability of boiler operation.Because the induced draft fan works in the extreme environment of high speed and high load for a long time,it is easy to break down and even cause economic losses or casualties.So the on-line monitoring of large induced draft fan is of great significance.However,due to the complex production environment and high upgrade cost,the traditional condition monitoring of induced draft fan is carried out by manual monitoring and threshold alarm,which is difficult to achieve unified management,and the fault alarm has a certain lag.Based on the actual project requirements of a heat source plant,combining cloud computing technology and deep learning method,this paper designs and develops a set of online monitoring system for the safe operation of large induced draft fan.The system can monitor the operating state of the induced draft fan equipment and provide functions such as data visualization,equipment fault identification and abnormal state alarm.The main work contents are as follows:(1)Taking the induced draft fan of a heat source factory as the research object,the state characteristic parameters of the induced draft fan are studied according to the actual production conditions of the industrial field,the types and characteristics of the induced draft fan fault are analyzed,and the problems existing in the fault diagnosis of the induced draft fan are analyzed,and then the traditional condition monitoring system is optimized and improved,which lays the foundation for the following research.(2)According to the weak fault feature signal of boiler induced draft fan in the early stage,traditional Variational Mode Decomposition(VMD)can not extract fault feature effectively due to improper parameter setting.An improved whale optimization algorithm combined with envelope entropy is proposed to determine the number of modes and penalty factor in VMD.Firstly,dual-parameter optimization is carried out on the Mode number K and penalty factor in VMD by whale optimization algorithm.The signals to be decomposed are decomposed through the VMD after parameter optimization,and K Intrinsic mode functions(IMF)are obtained.Kurtodegree value and correlation coefficient are used as evaluation criteria for component selection.Extract effective IMF component reconstruction signals.Finally,the effectiveness of the proposed method in fault feature extraction of induced draft fan weak feature signal is verified by comparing various optimization algorithms with experimental data.(3)Aiming at the problem of low fault recognition rate and low recognition speed in the process of fault diagnosis,a fault diagnosis method of VMD and convolutional neural network based on parameter optimization is proposed,and a convolutional neural network model is established.Secondly,it introduces the division of experimental data.There are 720 sample data of 6 types of faults.Each sample is trained and tested by using the features extracted from VMD of parameter optimization as the input of convolutional neural network model,so as to realize the fault diagnosis of induced draft fan.Finally,the original data collected from the industrial field are used to verify the diagnostic classification capability of the proposed method,and compared with BP network,support vector machine and autoencoder.The results show that the proposed method has faster convergence and diagnostic accuracy,which verifies the superiority of the proposed method.(4)The online monitoring system for the safe operation of large induced draft fan is designed and developed.First,according to the overall requirements of the system,complete the overall architecture of the online monitoring system,including the selection of sensors and acquisition schemes,complete the design of network transmission packet format,and achieve lossless compression of original data at the edge gateway.The fault diagnosis platform is constructed by means of multi-language mixed programming,and the C language reconstruction of signal noise reduction,spectrum analysis and fault diagnosis algorithms is realized.The software of fault diagnosis platform is written by C# language.Finally,the system is connected to the induced draft fan system of a boiler in a heat source factory for the whole test.The test results show that each function runs stably,can realize the accurate signal acquisition and fault diagnosis,and meet the requirements of industrial production safety and practicability.