Research On Fan Fault Monitoring And Diagnosis System Based On Improved Neural Network

Fan is a device that can improve gas pressure through mechanical energy and transmits gas.It is widely used in various fields in human production and life.When the fan fails,it will directly affect the industrial production and daily life,resulting economic losses and even casualties.With the development of artificial intelligence and Internet of things technology,the real-time monitoring system of fan with early warning and rapid diagnosis can effectively reduce the loss.In the actual investigation,it is found that even the top domestic wind turbine enterprises have not yet put the wind turbine monitoring and diagnosis system into practical or still in the time domain signal acquisition,but not the stage of analysis and diagnosis.Therefore,it is necessary to improve the existing neural network fault identification rate under real conditions and develop a set of practical fault monitoring and diagnosis system for fan faults.In this paper,an improved neural network is obtained by combining the appropriate neural network with the corresponding optimization algorithm for the PYHL-14A axial fan used in subway ventilation.Using MATLAB build a fault diagnosis system for fan,this includes signal feature extraction,pattern training,fault identification and self-learning.In addition,the fan fault research platform is established to collect the fault signals and establish the corresponding fault database.The main work and achievements of this paper are as follows:1.Through a large number of literature research and the comparative analysis of BP neural network,SVM support vector machine and ELM extreme learning machine,selecting the most suitable fault classification neural network for this system.In addition,the genetic algorithm(GA),particle swarm optimization(PSO)and cuckoo algorithm(CS)are compared and analyzed,and the most excellent algorithm is selected for the improvement of neural network structure.Through the analysis and research,the fan fault identification system based on ELM-CS improved neural network is proposed for the first time,and its effect is improved by 2%compared with the neural network without optimization.The recognition rate of rotor misalignment fault and pedestal looseness fault was increased by 4.2%and 4.1%respectively.2.The PYHL-14A fan fault diagnosis system is established,and the overall architecture and fault diagnosis system of the system are introduced.In addition,the modules of the system are introduced in detail,and the fault diagnosis and diagnosis of the fan fault diagnosis system are demonstrated.Among them,the system has self-learning function so that the system can carry out the related parameters through the sensor to collect abundant fault database,and based on this self-learning and upgrading the recognition system,so as to make the system more suitable for application in engineering practice.Among them,the system has self-learning function so that the system can carry out the related parameters through the sensor to collect abundant fault database,and based on this self-learning and upgrading the recognition system,so as to make the system more suitable for application in engineering practice.3.A fault research platform for PYHL-14A fans is established to measure the experimental data.This paper describes the experimental platform of the hardware and the function for several fault types of fault simulation,fault signals for the establishment of the database and get the training data and forecasting data acquisition.In this paper,the PYHL-14A fan experimental platform is built for the first time,and the fault simulation experiment is carried out.The obtained data is used for training and verification of neural network.The fan fault identification system established in this paper can improve the feasibility,recognition accuracy and stability compared with the traditional system,and has certain practical value.In addition,the system established in this paper plays a key role in the future related research.