Research On The Methods Of Data-Driven Assembly Line Work Fatigue Evaluation And Work Optimization

The ultimate service object of manufacturing system is human,and as the most dynamic and energetic role in manufacturing system,human always dominate the whole manufacturing system.Industry 5.0 offers an industrial vision beyond efficiency and productivity,with a focus on sustainable development and human-centered flexible industry,advocating that human security,access and happiness in manufacturing systems should receive more attention,which makes human-centeredness a core feature and inevitable trend in the development of current manufacturing systems.However,in recent years,the frequent occurrence of production safety accidents caused by work fatigue not only has brought great impact on the life safety and health of workers,but also caused huge economic losses to manufacturing enterprises.In this case,safe and efficient production has become a common goal pursued by manufacturing companies and workers.Identifying and predicting operational fatigue is the main means to scientifically prevent production accidents and improve production efficiency.Work fatigue is a state of mental,physical and emotional exhaustion caused by the longterm continuous labor of the worker,concentrating in both physiological and psychological aspects,which will have a negative impact on the work efficiency,work quality and safety and health of workers.In this paper,we study work fatigue based on the physiological and psychological characteristics of workers in a typical assembly line work.Firstly,we analyzed the characteristics of assembly line operations,the causes and the external performance of job fatigue in detail,and selected the heart rate data and facial expression data of operators to conduct research on work fatigue.Then,a LSTM neural network model was constructed to extract 10 features of heart rate,and a regression tree integration algorithm was constructed to extract 3 facial features including average blink duration,blink frequency and yawn frequency,and these features were fused.Next,a DMLPNN operational fatigue evaluation model was established with fusion features as input and fatigue state as output.Affected by COVID-19,the proposed method and model were verified by experiments of the simulation of PCB assembly operation,and 450 sets of valid data from 12 healthy subjects were collected to train and evaluate the model.The accuracy of the trained model was 86.67%.Finally,the trained model was used to evaluate the work fatigue of the PCB assembly worker,the existing work was designed and optimized based on the fatigue evaluation results.The changes of workers’fatigue state and production efficiency before and after optimization were compared,and the results showed that the proposed method in this paper can effectively alleviate the fatigue of the workers and improve the production efficiency.