Research On Safety Technology Of Industrial Robot Based On Invisible Protective Fence

With the rapid development of intelligent manufacturing,industrial robots have been widely used at home and abroad,and their safety issues have gradually come to the fore.Under actual working conditions,enhancing the safety protection capability of industrial robots is an important guarantee to ensure the safety of personnel and reduce the probability of accidents.At present,the safety protection measures of industrial robots are mainly based on physical fence,safety grating and safety carpet,but these protection methods lack versatility and flexibility,and do not give industrial robots complete safety protection capability.To ensure personnel safety,this paper constructs an invisible safety fence around the industrial robot based on 2D Li DAR,and develops a complete safety protection system,which empowers the industrial robot with good human sensing ability and risk assessment capability.The main research contents are as follows:Firstly,the design requirements of the industrial robot safety protection system are proposed according to the shortage of various safety protection measures.After comparing and analyzing the current common sensing sensors and considering the actual needs of the subject,2D LIDAR is finally selected to build the invisible protection fence for industrial robots,and the overall design of the protection system is completed based on this,and the whole system is divided into several parts: human body sensing in the robot monitoring area,target tracking and overall risk assessment.Secondly,research was conducted on human perception algorithms in the monitoring area of industrial robots.Due to the unordered and sparse nature of point clouds output by laser radar,a model of point cloud and image mutual conversion was proposed to address this issue.The point cloud data was visualized through pre-processing such as filtering,segmentation,and enhancement.YOLOX was then applied to detect point cloud images,and improved using an attention mechanism CA and an adaptive feature fusion strategy ASFF,enhancing the focus on leg features and the ability to fuse features,and improving the detection performance of YOLOX.Then,in order to solve the problem of system instability due to missed detection by the perception algorithm,a Kalman filter-based human tracking algorithm is proposed to track and predict the human targets appearing in the monitoring area of the robot in real time until the targets leave the monitoring range of the robot.In addition,under actual working conditions,multiple workers may appear in the monitoring area of the robot.To avoid tracking confusion,this paper solves the matching problem of multiple targets by using the nearest neighbor association algorithm and the Hungarian matching algorithm.Next,the risk assessment model for industrial robots is studied.In light of the limitations of the current risk assessment models in practical applications,a novel risk assessment model was developed.According to the different sources of risk,the risk indicators of industrial robots are divided into static risk indicators and dynamic risk indicators.The off-line assessment model of the static risk of industrial robots is constructed by using the linear weighted model based on the risk matrix method,and the static risk coefficient of industrial robots is calculated considering the nonlinear spatial distribution.Then the dynamic risk index of the industrial robot is obtained through the fuzzy comprehensive evaluation model,and the final risk level is obtained through compound calculation with the static risk coefficient of the industrial robot.Finally,the experimental platform of the industrial robot safety protection system was built,and the deployment of the algorithm and the development of the upper computer software were completed by using QT and other tools.The experimental results show that the industrial robot safety protection system developed in this paper can meet the initial design requirements and protect the workers who enter the robot monitoring area,which has certain feasibility.