Sealing Reliability Study For Pipe Joint

In aerospace and petrochemical field,sealing performance is critical to the operation of equipment or system.However,the sealing problem is complex and remains a challenge in engineering.Statistics show that most sealing failures occur in pipe joints.And the sealing performance of pipe joint is uncertain due to the variability of its parameters in practice.Therefore,it is necessary to study the sealing performance of pipe joint in view of reliability by considering all kinds of uncertainty.In this paper,the sealing reliability of pipe joint is studied.With the development of computer science,finite element analysis(FEA)has become a popular method to study the sealing performance of pipe joint.Therefore,the complex-FEA-based sensitivity analysis and computation of small failure probability which are all related to the sealing reliability of pipe point are studied.A case study of a pipe joint is also presented to demonstrate the application of the proposed method.The main works in this paper are as follows:Firstly,the study methods of pipe joint are discussed in detail and the mechanism of metal to metal seals is introduced.For a certain type of pipe joint,we study the simulation method and FEA model of sealing performance.The sealing criterion of the pipe joint is also built and the test envelope FEA is performed.Then,the most dangerous load point of the test envelope is obtained.Then a new algorithm combining Sobol method and Kriging metamodel is proposed to deal with sensitivity problem based on complex computer simulation in this paper,which is referred to as AK-Sobol(Adaptive-Kriging-based Sobol)algorithm.By using adaptive Kriging metamodel to predict response value instead of evaluating true response value,AK-Sobol algorithm can be more efficient to compute sensitive indices and reflects the relative importance of each input variable.The accuracy and efficiency of AK-Sobol is demonstrated by several numerical examples in this paper.Then,the AK-Sobol algorithm is used to perform sensitivity analysis of the pipe joint and the first and second order effect of each variable are obtained.Finally,to deal with small failure probability problem based on complex computer simulation,this paper also proposes a new reliability method combing the improved cross-entropy-based importance sampling(CE-IS)method and adaptive Kriging metamodel,which is referred to as AK-CE/IS~2(Adaptive-Kriging and improved Cross-Entropy-based Importance Sampling)method.The AK-CE/IS~2 method uses Gaussian mixture as the IS density.By resorting to Kriging metamodel with the active learning function,the IS density is optimized by minimizing the Kullback-Leibler cross entropy with the optimal IS density.This new method can obtain a near-optimal IS density in few rounds and the Kriging metamodel is also adaptively updated at the same time.With the near-optimal IS density,the Kriging metamodel continue to be updated to correctly classify the population generated from the near-optimal IS density and the failure probability can be obtained.After demonstrating the accuracy and efficiency of the new method by numerical examples,we use the new algorithm evaluate the failure probability of a pipe joint.