Life Prediction Model And Reliability Analysis Of Sucker Rods,pipes And Pumps In Mechanical Production Well

In recent years,with the rapid development of computer technology,new sciences such as data mining,machine learning and artificial intelligence have been gradually applied in various industries.New models based on large volumes of data have been explored using modern techniques for parameter solving and multi-domain applications of implementation model have been widely developed.Nearly six decades have been passed since the development of Daqing Oilfield in 1960.Due to a series of objective reasons such as crustal movement and equipment aging,the equipment abrasion rate is on the rise,resulting in the frequency of failures.A large amount of information on oilfield equipment stored in the relevant oilfield departments needs to be utilized and studied.It is imperative to construct intelligent oil fields with data.Based on the effective data collected from the oilfield in the past decades,this paper conducts the reliability analysis on important components of oilfield machinery.From the perspective of statistics,a general model has been established adopting a Burr-Ⅻ model commonly used in industry and Generalized Half Normal(GHN)distribution that are applied to failure problems caused by fatigue damage and wearing of components.For the maximum likelihood estimation and Bayesian estimation of model parameters,the model parameters are solved by MH-MCMC(Metropolis-Hastings Monte Carlo Markov Chain)method,Particle Swarm Optimization(PSA),Genetic Algorithm(GA),Differential Evolution(DE)algorithm,etc.to verify the validity of the model via Monte Carlo simulation.Finally,the model and theories are applied in oilfield departments,acquiring the life distribution rules of oil sucker rods,pumping pipes and oil-well pumps to get the reliability analysis of the equipment.The highly matching between the theoretical results and the actual data illustrates the rationality of the model.The conclusion provides an important theoretical basis for the reliability analysis of oilfield components and makes a certain contribution to the engineering application.Based on the data collected from the site,the statistical analyses of failures of oil sucker rods,pumping pipes and oil-well pumps in one district of Daqing have been conducted and fault trees of them were established respectively.Lifetime analysis and reliability studies were performed using Generalized Half Normal(GHN)distribution model and the Burr-Ⅻ model.This paper mainly studies the following contents:Firstly,according to failure causes and invalidation forms of oil sucker rods,a fault tree is established.Based on the stepwise I interval setting sampling,the oil sucker rod’s life prediction model of 2pBurr-Ⅻ distribution is established.The Differential Evolution(DE)algorithm and Quasi-Newton method(QN)are used to determine the two-parameter maximum likelihood estimation value,which are evaluated by Monte Laro simulation.Based on the sucker rods’ data collected in the oilfield,the life prediction model of sucker rods under different working conditions will be calculated.At the same time,with the reliability analysis of sucker rods performed,it can provide a certain theoretical reference for the later maintenance of related departments,Considering that the data will increase year by year,the model will be extended to the 3pBurr-XII distribution model in order to improve the accuracy.Secondly,with failure causes and invalidation forms of pumping pipes studied,the fault tree has been established.Based on mixed sampling,the oil-pipe life prediction model of the Generalized Half Normal(GHN)distribution is established.Differential evolution(DE)algorithm and genetic algorithm(GA)are applied to determine the parameter estimates of the generalized half normal distribution model of pumping pipes and Monte Carlo simulation is used to compare statistical performances of Genetic Algorithm(GA)method and Differential Evolution(DE)method.The life cycle distribution model of the pumping pipe was calculated and the reliability analysis was carried out thanks to the employment of on-site data.Finally,the model was extended to propose the Double-truncated Generalized Half Normal(DTGHN)model.The MH-MCMC algorithm is used to realize the parameter solving and confidence interval of the distributed quantile fractile is obtained via using the percentile method of bootstrap.Finally,with failure causes and invalidation forms of oil-well pumps discussed,the fault tree has been established.The life prediction model of the oil-well pumps is proposed based on Bayesian procedure with a rebate warranty policy.The proposed method requires fewer assumptions than its competitors to search for optimal sampling plans when the lifetimes of products follow a 3pBurr-XIID.Newton-Raphson method,Particle Swarm Optimization(PSO)algorithm and Genetic Algorithm(GA)are used in the proposed Empirical Bayesian(EB)procedure to search for reliable estimates of the parameters in the Bayesian model,and the obtained estimates are denoted as QN-EB,PSO-EB and GA-EB,respectively.The performance is evaluated via using Monte Carlo simulations.Finally,in terms of the reliability analysis of the oil-well pumps,its optimal preventive cycle and strategies are obtained.The model was eventually extended to accelerated three-parameter Burr-Ⅻdistribution considering the needs of later research.The Particle Swarm Optimization(PSO)algorithm is used to obtain maximum likelihood estimates of 3pBurrⅫ distribution,denoted by PSO-MLEs.