Design And Implementation Of Safety Monitoring System For Exoskeleton Robot

In this paper,the lower extremity exoskeleton is mainly aimed at rehabilitation medical scene.There is a most important point in the medical rehabilitation scenario is to ensure the safety of patients in rehabilitation training.This paper studies the safety monitoring system for exoskeleton robot.For the safety problems,this paper will focus on the possible failures in the operation of the exoskeleton robot.An end-to-end fault diagnosis model that can automatically extract features and fuse data is designed to realize state monitoring during the running process of the exoskeleton,judge whether the current running state of the machine is normal or not,and locate the fault location if there is a problem.The main work of this paper is as follows:(1)Firstly,the software architecture of the security monitoring system is designed.The monitoring function is studied,and the data interaction method between the exoskeleton training data and the system platform is designed,which provides a platform for data acquisition and calculation for the following research.On this basis,the implementation process of the status monitoring function is designed,and the problems existing in the actual use of exoskeleton are analyzed,and the problems that the safety monitoring system needs to address are determined: motion control system fault analysis,and finally the fault data set is collected and constructed.(2)To improve the diagnosis accuracy of the traditional model,it is necessary to analyze the problems of sensor data fusion and feature extraction.The multi-sensor data fusion method for the exoskeleton is explored,the application methods of three traditional data fusion schemes on the exoskeleton fault data are analyzed,and the experimental results are analyzed.Conclusion: The improvement of diagnosis accuracy of traditional fault diagnosis model is extremely dependent on the selection of artificial feature extraction and data fusion scheme,but the manual operation is relatively complex,and there will be redundant and invalid in features,resulting in the waste of computing resources.(3)To improve the existing problems in the traditional model: 1.The traditional model only focuses on the data at the current moment,but the occurrence of exoskeleton fault is temporal,and the study is only carried out at a single moment,and the performance of the monitoring data will be similar to that of other states,leading to low diagnostic accuracy;2.The improvement of diagnostic accuracy of traditional schemes relies heavily on the selection of artificial feature extraction and data fusion schemes,and its operation process is complex,with redundant and invalid features.This paper introduces a method based on convolution and temporal neural network.The one-dimensional convolutional neural network is used to extract the numerical features of the sensor and the temporal correlation features are extracted by the temporal neural network.The final designed model can automatically perform feature extraction and data fusion,simplify the process of data fusion and feature extraction,save the complexity of manual operation,and realize end-toend mixed fault diagnosis.Experimental results show that the proposed method achieves the highest diagnostic accuracy with the simplest input data.