Prediction of pipeline failure probability with multiple failure modes and variable correlation
Aiming at the problem of competing multiple failure modes and dependent variables of corroded pipelines,and low efficiency of predicting the probability of small failures,a time-varying failure probability prediction method for corrosive pipelines based on SS method is proposed.Based on the limit state function of leakage and burst failure,considering the competition between leakage and burst failure,a prediction model for the pipeline competitive failure probability was constructed.In order to solve the above model,the Nataf transformation method was used to describe the correlation between the depth and length of the defects in the simulated samples,and a method to solve the failure probability of corroded pipelines was proposed by considering the competitiveness of failure modes and the correlation of variables.Monte Carlo Simulation(MCS)was used to verify the above method.Finally,the above method was applied to investigate the effect of the weak correlation(correlation coefficient 0-0.3)between defect depth and length on the probability of failure within 15 years of pipeline service.The results show that when the failure probability is greater than 10-6,the calculation results of this method are basically consistent with those of MCS,and the calculation efficiency is higher than that of MCS,and the minimal probability events with the probability level of 10-12 can be predicted.Within 15 years of pipeline service,the greater the correlation coefficient of defect variables is,the greater the probability of pipeline burst failure is,and the earlier it reaches the threshold year of burst failure probability.As the correlation coefficient increases,the smaller the probability of leakage failure is,and the later it reaches the threshold year of leakage failure probability.At the later stage of the prediction,the effect of the correlation coefficient increase on the probability of the two kinds of failures is weakened.
NETL
NSTL
万方数据
