汽车变速箱疲劳寿命试验中可能发生的故障种类很多。又由于汽车变速箱变速挡位多，多对齿轮同时啮合，结构紧凑，部件之间相互耦合，因而监测诊断用的振动加速度信号与故障之间的映射关系非常复杂。特征与故障之间并非简单的一一对应关系。以信号处理技术为手段的常规诊断技术已显示出故障诊断能力较弱的不足。ＡｒｔｉｆｉｃｉａｌＮｅｕｒａｌＮｅｔｗｏｒｋ（简称ＡＮＮ，人工神经网络）以其非线性大规模并行分布式处理、联想、记忆、高度的容错性、自适应、自组织及自学习能力和极强的非线性映射能力，在模式识别、信号处理与图像处理、控制理论、故障诊断等领域得到了广泛应用。本文详细介绍了多层前向网络（ＭｕｌｔｉｌａｙｅｒｆｅｅｄｆｏｒｗａｒｄＡＮＮ）在汽车变速箱状态监测和故障诊断中的应用情况，给出了基于ＡＮＮ的寿命预测结果，并与传统方法得到的相应结果进行了比较。并就汽车变速箱状态监测和故障诊断中的特征选择、特征提取、诊断模型和寿命预测模型的建立等问题进行了讨论。 There are many kinds of faults that may occur in the fatigue test of automobile transmission. Because of the many gearbox transmission gears, many pairs of gears meshing at the same time, the structure is compact and the components are coupled with each other. Therefore, the mapping relation between the vibration acceleration signal for monitoring and diagnosis and the fault is very complicated. Between features and failures is not a simple one-to-one correspondence. Conventional diagnostic techniques that use signal processing techniques have shown a weakness in fault diagnosis. Artificial Neural Network (referred to as ANN, artificial neural network) with its large-scale non-linear parallel distributed processing, associative, memory, a high degree of fault tolerance, adaptive, self-organizing and self-learning capabilities and very strong nonlinear mapping capabilities, , Signal processing and image processing, control theory, fault diagnosis and other fields have been widely used. This paper introduces the application of multi-layer forward network (Multilayer feedforwardANN) in condition monitoring and fault diagnosis of automobile transmission in detail, and presents the life prediction results based on ANN, and compares them with the corresponding results obtained by traditional methods. The problems of feature selection, feature extraction, diagnosis model and life prediction model in status monitoring and fault diagnosis of automobile transmission are discussed.