Research On Fault Diagnosis System Of Air Compressor Based On LSTM

With the advancement of social industrialization,air conditioning system is more and more widely used in industrial production.To ensure the accuracy and stability of workshop temperature has become the key condition for each large factory to effectively avoid potential safety hazards and efficient production.Whether the air compressor,as the most key equipment in the air conditioning system,can operate normally in the production process of the workshop is the key to the whole air conditioning system Unified safety and even the production efficiency of the workshop play a very important role.In this paper,the air compressor is taken as the research object.The aim is to study the influence of various operating parameters of air compressor on air compressor failure.A prediction and monitoring model of air compressor operating parameters based on variable long and short time memory network(v-lstm)is proposed.The relevant parameters are determined by Pearson correlation coefficient,and then the abnormal failure is determined by 3 ? criterion according to the collected data Barrier threshold.Finally,the simulation experiment is carried out according to the actual data of the factory to judge the validity of the variant LSTM model proposed in this paper.The experimental results show that our variant long and short-term memory network model has higher efficiency and accuracy for the fault prediction and monitoring of air compressor.Finally,a small fault mechanism reasoning expert system is established through the monitoring results for fault diagnosis.The main contents of this paper are as follows:First of all,this paper comprehensively summarizes the working principle,mechanical structure,fault type and fault mechanism of air compressor,introduces the source of this topic,the current situation at home and abroad,research content,research significance and in detail,mainly studies and analyzes the current situation of fault diagnosis system,the current situation of deep learning algorithm and the research status of long-term and short-term memory network technology According to the analysis of the characteristics of the basic data,the suitable deep learning algorithm is determined.Secondly,the paper introduces the shortcomings of the cyclic neural network in data prediction,studies and analyzes the basic principles and training methods of the traditional long-term and short-term memory network,and puts forward a lightweight algorithm model v-lstm in the traditional LSTM algorithm.Because there are many parameters in the traditional LSTM algorithm,a lot of training is necessary to make the model converge.In the face of data scarcity and small lightweight scenes,the effect is not obvious.The v-lstm neural network unit proposed in this paper only uses the input gate and forgetting gate,and obtains the output result by simply adding(without introducing new parameters,directly adding the corresponding positions of matrix elements),which greatly optimizes the parameter scale and makes the network model training convergence faster Speed.Then we use PLC to collect data in real time for model training,use the trained model to monitor and predict the equipment parameters in real time,and then use our improved network model v-lstm to compare with the traditional LSTM,BP neural network and SVM,which proves that the improved long short memory network algorithm model proposed in this paper should be more effective.At the same time,in order to make our v-lstm achievebetter training effect,that is to achieve faster convergence and better model robustness,we add L2 norm to the optimization objective,and the experiment shows that the improved optimization objective can make the model achieve better effect.Moreover,this paper studies and analyzes the basic working principle of air compressor,determines the possible failure phenomena and causes,and develops a lightweight expert system for air compressor fault reasoning.Once there is abnormal data in the process of model prediction and monitoring,the expert system will automatically give the failure point,the failure situation,the cause of the failure,and finally give the corresponding treatment method.Finally,this paper develops a web-based air conditioning fault diagnosis system using C#and MATLAB.The main functions of the system include: the front-end page real-time monitoring air conditioning system parameters data,including real-time working conditions,working condition judgment,working condition reasoning;the background control mainly includes: air compressor exhaust parameters prediction and monitoring,knowledge base and case base add management.The system is efficient and convenient to use,and can detect and diagnose faults quickly.