为预防因热氧老化异常引发的燃气聚乙烯（PE）管道泄漏事故，提出管道热氧老化程度的超声波智能检测方法。提取PE管的线性和非线性超声检测特征参数，并分析其与管道热氧老化程度的关系；构建并优化BP神经网络，将超声检测特征向量和老化时间作为网络输入和输出；扩展检测特征参数并用于网络训练，训练能够评估PE管老化程度的神经网络模型。研究结果表明：由声速、声衰减、非线性系数可构成用于智能评估老化程度的超声检测特征向量，经优化并训练后的神经网络可评估PE管老化时间，老化时间评估值的平均相对误差低于6.9%。

In order to prevent gas polyethylene (PE) pipeline leakage accident caused by abnormal thermal oxidative aging, an ultrasonic intelligent detection method for thermal oxidative aging degree of pipeline was proposed. The linear and nonlinear ultrasonic testing characteristic parameters of PE pipe wereextracted, and the relationship between the characteristic parameters and the aging degree of PE pipe was analyzed. The BP neural network was constructed and optimized, and the ultrasonic testing characteristic vector and aging time were respectively used as the input and output of the network. The testing characteristic parameters were expanded and used for network training, and the neural network model that could evaluate the aging degree of PE pipe was obtained. The results show that the acoustic velocity, acoustic attenuation and nonlinear coefficient can be used to form the characteristic vector of ultrasonic testing for intelligent evaluation of aging degree. The optimized and trained neural network can be used to evaluate the aging time of PE pipe, and the average relative error of the evaluation value of aging time is less than 6.9%.