The Reliability Data Analysis And Fault Diagnosis Of Reciprocating Compressor Model

With the development of technology and industry,reciprocating compressors are often used in petroleum,natural gas and chemical industries as an important mechanical equipment.Reciprocating compressor is characterized by more parts and more complex structure,the level of its safety and reliability will directly affect the actual production of economic benefits.Research on its safety and reliability has always been highly regarded by researchers.In the past,scholars used the traditional methods to simulate the thermal performance and mechanical performance for reliability research.This article is based on the current data application ideas for reliability research.There are many types of data involved in the study,and reciprocating compressor parts related to more information,Without uniform management,it is easy to cause data omission.Based on the theory of database and combined with the principle of practicability,we developed a professional database management system integrating functions of data entry,inquiry,modification and authority setting.It realizes the visual operation of various research data of reciprocating compressor,which has the advantages of convenience and quickness.Reciprocating compressor research data,there is a type of life-type fault data,Studying it can give us a reliability index,which means the average time between failures.In this paper,according to the small sample data of reciprocating compressor parts,a priori distribution model of parameters is established by Bayesian theorem,and the posterior distribution of parameter is solved by using Markov chain-Monte-Carlo method in depth learning.Finally,through the joint use of algorithm programs and Python language.The mean time between failures of components under Weibull distribution was 491.64 h,which verified the effectiveness of the model.The fault characteristic value data,as another kind of fault data of the reciprocating compressor,can realize the fault pattern recognition and classification for its research.In this paper,the BP neural network and the fuzzy adaptive neural network are improved,and the corresponding fault diagnosis models are established,which are respectively applied to the simple fault eigenvalue data and the complex fault eigenvalue data.The average diagnostic accuracy of the former is 90.48%.The latter has good classification effect when ?=0.7,?=0.55 and ?=0.85.The overall idea of this article is to start with the data characteristics of the research object.A corresponding algorithm model was established to quantitatively study the reliability of the reciprocating compressor.And,the established reliability data analysis model and fault diagnosis model can be popularized and applied.