基于数据融合和非线性维纳过程的埋地管道退化过程预测

Degradation prediction of buried pipelines based on data fusion and nonlinear Wiener process

管道在多种因素的作用下会不断发生壁厚减薄,为反映性能退化过程的非线性、个体差异性和测量随机性,选取影响管道外腐蚀的主要控制因素进行数据融合,结合非线性维纳过程和加速退化轨道模型建立腐蚀管道剩余寿命的概率密度函数和可靠度函数,并采用蒙特卡洛马尔科夫链方法进行参数抽样和参数估计,最后利用卡尔曼滤波对模型参数进行更新,实现退化过程的实时预测。通过实例进行分析验证,结果显示该模型不仅降低了外腐蚀因素之间的相关性,同时避免了线性维纳过程的缺陷,客观反映了不同检测时间下的管道剩余寿命,与二元逆高斯模型和线性维纳模型相比剩余寿命的预测误差更小,预测精度有所提高。

In order to reflect the nonlinearity,individual difference and measurement randomness of the performance degradation process,the main controlling factors affecting the external corrosion of the pipeline were selected by kernel principal component analysis,and the probability function of the remaining life of the corroded pipeline was established by combining the nonlinear Wiener process and the accelerated degradation track model.Monte Carlo Markov chain method is used for parameter sampling and parameter estimation.Finally,Kalman filter is used to update parameters in real time.The results show that the proposed model not only reduces the correlation between external corrosion factors,but also avoids the defects of the linear Wiener process,and objectively reflects the remaining pipeline life under different detection times.Compared with the binary inverse Gaussian model and the linear WP model,the results show smaller errors and improved prediction accuracy.The results provide a theoretical basis for improving the integrity management level.