Evaluation And Prediction Of Upper Limb Muscle Fatigue In Static Overhand Work

In human led maintenance and repair,overhand work is relatively common in fields such as construction machinery assembly,aircraft and automobile maintenance,radar maintenance,etc.Upper limb musculoskeletal disorders(MSDs)resulting from this condition have received increasing attention.Aiming at the problem of upper limb muscle fatigue in static overhand work,this thesis studied the fatigue characteristics of upper limb based on surface electromyography(sEMG)through experimental design,data collection and processing,and realized the evaluation and prediction of upper limb muscle fatigue in static overhand work.The main research contents are as follows:(1)The mechanism of sEMG generation was studied,the sEMG collection method of upper limb muscle fatigue for static overhand work was analyzed,and the corresponding relationship between upper limb movement and muscle was discussed.According to the working conditions of overhand,sEMG collection experiment of upper limbs in static overhand work was designed.sEMG equipment was used to collect sEMG of 12 subjects in static work at the height of 0cm,5cm and 10 cm,and 75 groups of data were collected in total.(2)For sEMG noise reduction processing,notch filtering,bandpass filtering,wavelet analysis and multi-scale principal component analysis were studied,and the advantages and disadvantages of multi-scale principal component analysis and wavelet denoising were compared.The results show that the noise reduction effect based on multi-scale principal component analysis was better than wavelet threshold.The time domain,frequency domain and nonlinear feature extraction methods of sEMG were studied,13 features were selected for muscle fatigue evaluation,and the performance of 13 fatigue evaluation algorithms were evaluated and compared in terms of SVR index,goodness of fit and algorithm complexity.(3)The active muscle group of the upper hand during static operation was analyzed through the time-domain and frequency-domain characteristics,and the main power generating muscle and fatigue state characteristics were determined.Fatigue stages were divided by Borg CR-10 subjective fatigue scale and time domain and frequency domain characteristic changed experiments.Multi-classification support vector machine,BP neural network and C4.5-based decision tree algorithm were used to classify muscle fatigue states.The feasibility of machine learning algorithm to classify the fatigue state of upper limb muscle in static work was verified.The results showed that the multi-classification support vector machine had the best effect,and the classification accuracy of muscle fatigue in static overhand work reached 97.78%.(4)Aiming at the prediction of upper limb muscle fatigue in static overhand operation,a Convolutional Neural Networks(CNN)-Long short-term memory(LSTM)hybrid method for muscle fatigue prediction was used.CNN data feature extraction capability was used to extract the advanced features of the input EMG feature value,and then input the LSTM model with memory ability for prediction.The average absolute percentage error was 6.42% for the prediction of EMG features 30 s in advance.Compared with the traditional CNN and LSTM models,CNN-LSTM had the characteristics of high prediction accuracy,good stability.