论文部分内容阅读
因为Spiking神经网络(Spiking neural networks,SNNs)能同时传递时空信息,SNNs包含优于传统神经网络的许多特性,因而更适用于动态时序信号的分析。碰撞和受阻是机械臂在靠近抓取位置时常见的两种故障。为区别此两种故障状态与正常工作状态,提出一种基于SNNs的新型机械臂故障诊断方法。讨论所提出的SNNs故障诊断方法的体系结构,比较了当SNNs故障诊断方法选用不同Spiking神经网络拓扑结构和不同参数时的诊断结果。试验结果表明所提出的基于Spiking神经网络的机械臂故障诊断方法是有效的。该方法有助于机械臂故障的正确诊断,并且对平稳安全的生产具有重要意义。
Because Spiking neural networks (SNNs) can simultaneously transmit spatio-temporal information, SNNs contain many features superior to traditional neural networks and are therefore more suitable for the analysis of dynamic timing signals. Collisions and obstructions are two common faults that a robotic arm may have when approaching a gripping position. In order to distinguish the two fault states and the normal working state, a novel SNNs-based fault diagnosis method of manipulator is proposed. The proposed SNNs fault diagnosis system architecture is compared and the diagnostic results are compared when using different Spiking neural network topologies and different parameters for SNNs fault diagnosis. The experimental results show that the proposed method of robot fault diagnosis based on Spiking neural network is effective. This method is helpful for the correct diagnosis of robotic failure, and is of great significance to the smooth and safe production.