Intelligent System Identification Applied to the Biomechanical Response of the Human Trunk during Sudden Loading

EMG;sudden loading;system identification;low back pain;trunk;biomechanics

Lawrence, Brad Michael ; Carolyn Sommerich, Committee Member,David Kaber, Committee Member,Gary Mirka, Committee Chair,Gregory Buckner, Committee Co-Chair,Lawrence, Brad Michael ; Carolyn Sommerich ; Committee Member ; David Kaber ; Committee Member ; Gary Mirka ; Committee Chair ; Gregory Buckner ; Committee Co-Chair

University:North Carolina State University

关键词: EMG;    sudden loading;    system identification;    low back pain;    trunk;    biomechanics;   

Others  :  https://repository.lib.ncsu.edu/bitstream/handle/1840.16/5208/etd.pdf?sequence=1&isAllowed=y

美国|英语

来源: null

【 摘 要 】

Current techniques in biomechanics are not sufficient for modeling sudden loads.Modeling techniques used in intelligent system identification, where self-adapting basis functions identify time-varying model parameters, may provide a superior method for describing human trunk dynamics during sudden loading.By deriving time-varying system dynamics, more insight is gained into how the trunk responds to sudden loads.Six males were subjected to sudden loads with varying conditions of expectation, fatigue, and training.Electromyographical and trunk motion data were recorded, and processed using a novel system identification algorithm to yield dependent measures of peak and average trunk stiffness, peak muscular torque, work, and impulse.This system identification model yielded valid, accurate, and robust results with minor experimental cost.The model calculated physiologically valid stiffness magnitudes ranging from an average stiffness mean of 532 Nm/rad (standard deviation = 306 Nm/rad) to a peak stiffness mean of 1314 Nm/rad (standard deviation = 814 Nm/rad).The model root mean square accurately predicted empirical angular displacement at a value of 0.1˚.Across varying model inputs, dependent variable output magnitudes differed by less than 5%.On average, the model required 4000 iterations per trial (less than 5 minutes per trial) to converge and obtain dependent variable output.Model output indicated that expectation significantly increased peak and average stiffness (p<0.0001) by 70% and 113% respectively, and significantly decreased peak torque, work, and impulse magnitudes (p<0.0001) by 36%, 63%, and 56% respectively; training significantly decreased peak torque and work magnitudes (p<0.05) by 25% and 36% respectively; the interaction of expected loads with training significantly increased peak and average stiffness (p<0.05) by 17% and 20% respectively, and decreased impulse magnitude (p<0.05) by 53%; and the interaction of unexpected loads with fatigue significantly decreased work and impulse magnitudes (p<0.05) by 23% and 17% respectively.These results led to workplace recommendations that could potentially reduce low back pain.This intelligent system identification approach is easy to implement and has potential application to address a multitude of research questions in the field of biomechanics.