A robot anomaly detection device and detection method

Abstract

The invention discloses a robot anomaly detection device and detection method, comprising a detection module, a data processing module, an optical neural network module and an anomaly judgment module. The information is transmitted to the data processing module; the data processing module is used to preprocess a plurality of current state information, generate a predetermined number of modulation signals, and output them to the optical neural network module; the optical neural network module modulates the optical signal according to the modulation signal , and after performing optical calculation on the optical signal, the calculation result is converted into a feedback electrical signal and output to the abnormality determination module; the abnormality determination module determines whether the robot is in an abnormal state according to the feedback electrical signal and predetermined determination criteria.

Description

technical field [0001] The invention relates to the field of robot fault detection, in particular to a robot abnormality detection device and detection method. Background technique [0002] At present, the realization of neural network through software algorithm is relatively mature, but this is extremely dependent on the configuration conditions of the computing unit itself, which consumes a lot of power and takes a long time to compute. Optical neural networks have attracted people's attention due to their ability to compute data at the speed of light. In the current field of robot anomaly detection, the identification of abnormal state is realized by processing and analyzing the data obtained by the detector based on various data analysis methods. [0003] Patent CN101200066B discloses, "Method and Control System for Monitoring the State of Industrial Robots", which uses the electrical processing unit to measure some characteristics of the robot (joint friction value, cl...

AI Technical Summary

Problems solved by technology

This invention requires multiple single-frequency lasers, which are costly and have complex optical path structures. In the case of many input sample dimensions, it is unrealistic to use multiple single-frequency lasers to build; the micro-ring is adjusted by the radius of the micro-ring and the coupling length. The filtering wavelength cannot realize timely adjustment of network parameters during neural network training

Images