为了克服ＢＰ算法收敛速度慢的问题，提出了一种基于混合学习规则的ＢＰ算法，并采用模归一化方法，成功地定量组织了故障的学习样本，建立了能够定量分析发动机气路部件故障的人工神经网络（ＢＰＮ）。通过分析测量系统随机误差的影响和实际试车数据的效验结果，表明该网络具有较强的推广能力及适应性，能基本满足故障定量诊断的要求，并具有较好的工程实用性。 In order to overcome the problem of slow convergence of BP algorithm, this paper proposes a BP algorithm based on hybrid learning rules, and uses the method of modulo normalization to quantitatively organize the learning samples of faults, and builds a BP algorithm that can quantitatively analyze the faults of engine air line components Artificial neural network (BPN). By analyzing the influence of the random error of measurement system and the actual test data, it shows that the network has strong promotion ability and adaptability, can basically meet the requirements of quantitative diagnosis of faults, and has good engineering practicability.