基于粗糙集和RBF神经网络的工业机器人故障诊断

Fault Diagnosis of Industrial Robot Based on Rough Set and RBF Neural Network

工业机器人作为制造业高速发展不可或缺的重要元件之一,其典型故障的突然发生导致停机,将直接影响到整机寿命周期和工厂生产效益,因此,针对工业机器人典型故障的诊断具有重要的意义。提出一种基于粗糙集和RBF神经网络结合的故障诊断方法:运用粗糙集理论对典型故障关键位置采集的数据进行处理,对不确定的数据进行表达;在保持现有数据信息分类功能不变的特性下,得到最小的表达方式;通过探究数据间的相应依赖关系得到难以表达的模式;通过分析数据信息得到相关规则。依据粗糙集的最小条件属性集对RBF神经网络进行设计,完成网络测试、分类,提升故障分析和处理的能力,切实解决了工业机器人典型故障原因多样且数据相互交织不确定性的问题,从而实现了挖掘隐含特征和规律,进行准确的故障诊断。该方法可有效减少冗余数据信息处理和网络训练时间及计算量,大幅提升工业机器人典型故障诊断准确率。

As one of the indispensable components for the rapid development of the manufacturing industry, the sudden occurrence of typical faults of indusrial robots leads to downtime, which directly affects the life cycle of the machine and the production efficiency of the factory. Therefore, it is of great significance to diagnose typical faults of industrial robots. A fault diagnosis method based on rough set and RBF neural network is proposed. The method uses rough set theory to process the data collected from typical fault key locations and express the uncertain data.Under the characteristic of keeping the existing data information classification function unchanged, the minimum expression is obtained. The pattern which is difficult to express is obtained by exploring the corresponding dependency relationship between the data. Relevant rules are obtained by analyzing data information. According to the minimum conditional attribute set of rough set, the RBF neural network is designed to complete the network testing, classification and improve the ability of fault analysis and processing. It effectively solves the problem that the typical fault causes of industrial robots are diverse and the data are intertwined with uncertainty, thus realizing the mining of hidden features and rules and accurate fault diagnosis.This method can effectively reduce the time and computation of redundant data information processing and network training, and greatly improve the accuracy of typical fault diagnosis of industrial robots.