Research on the Risk of Radar Antenna Array Maintenance Operations in Real Working Conditions Based on Intelligent Evaluation Tools

The assessment of worker posture load risk, as they interact with their work environment, is pivotal in evaluating biomechanical overload and preventing work-related musculoskeletal disorders (WMSDs). This paper investigates the application of artificial intelligence (AI) algorithms for posture risk assessment in authentic work settings. Workers' actions were recorded using smartphones, and key posture frames were extracted through pose recognition algorithms from these video samples. The key frames were digitally skeletalized, and their pose load was assessed using the Rapid Upper Limb Assessment (RULA) method. A total of 9431 key frames were extracted from a 32-min video of a single worker performing a task, after conducting two repetitive experiments. The findings demonstrate the efficacy of the method in recognizing key postures within actual work scenarios. Achieving a joint recognition accuracy of 84.6%. Moreover, 89.1% of the assessed tasks in radar antenna array maintenance resulted in upper arm scores exceeding 3 points, indicating a significant risk. This research lays a robust foundation for establishing a human-centric Industry 5.0 system, facilitating the online, real-time assessment of human posture risks.