论文部分内容阅读
目前已经使用振动测量和神经网络方法开发出了齿轮传动的故障诊断技术。实施振动测量采用背对背结构的齿轮试验装置 ,在施加负荷和按齿轮节圆线速度条件下运转 ,这些都是工业齿轮应用的标准型式。齿轮的普通故障模式体现在试验齿轮上。本文所研究的运行齿轮有 :正常齿轮、齿尖断裂的齿轮、磨损的齿轮、径向跳动超差的齿轮、综合导程误差超差的齿轮和润滑有限制的齿轮。描述齿轮传动运行状态的特征参数有振动、电机功率、油温和转速。振动测量使用加速度计和声发射传感器。自组织映射 (SelfOrganizingMap ,SOM)是一种著名的神经网络方法 ,它是使用上述这些特征参数进行训练和试验。所有经过训练的网络都是有效的 ,其成功率可达到 95 %以上。神经网络能够识别运行状态 ,甚至可以达到 10 0 %的成功率。
Vibration measurement and neural network methods have been used to develop fault diagnosis techniques for gear drives. Implementing Vibration Measurements A gear test device with back-to-back configuration, operating under load and at the pitch line speed, is the standard type of industrial gear application. The normal mode of gear failure is reflected in the test gear. The running gear studied in this paper are normal gear, gear with broken tooth tip, worn gear, gear with over runout, gear with integrated lead error tolerance and gear with limited lubrication. The characteristic parameters describing the running state of the gear drive are vibration, motor power, oil temperature and speed. Vibration measurements use accelerometers and acoustic emission sensors. Self-Organizing Map (SOM) is a well-known neural network method that is trained and tested using these characteristic parameters. All trained networks are effective, with a success rate of up to 95%. Neural networks can recognize the state of operation and can even achieve a success rate of 100%.