The Application Of Neural Network Control In The Implementation Of Active Musculature

Traffic accidents is an issue of national concern,and the head and neck are themost likely body parts to be injured in car accidents.Considering its magnitude,gravity and the consequent negative impacts on economy,public health and the generalwelfare of the people,it’s important to focus on the research on improving the modelsand the methods which have been put in research in this field. Implementing activemusculature in human head and neck model has become the mainstream of research.In order to study on the implementation of active response in human models,aSimMechanics model based on a finite element human head and neck model was built.It consists of a head, vertebrae, the soft tissues known to have a more significanteffect on the motivation of head and neck during the impact, such as ligaments andmuscles. Muscles with passive properties was modeled with Force ElementBody,while the active properties of each muscle were added to the actuator via theoutput of neural network(NN) control loop performed in Simulink toolbox,usingHill’s methodology to calculate the force that active muscles produce,which is alsothe output of the whole control loop. The differences in joint angles are the feedbacksignals from the SimMechanics model,controller is implemented to output theactivation level of each muscle,and the BP neural network (BPNN) is to enable theactivation level to be time-variant according to the changes in the system.In this study, the said built SimMechanics head and neck model(SimM model)with active muscle properties is simulated and compared to sled tests performed onhuman volunteers at Naval Biodynamics Laboratory by Ewing et al(or the well-knownNBDL tests,1976),to validate the implementation of active muscles.According to the simulation results,the following conclusions were made.Comparing the result with the data utilized from NBDL tests,good agreement wasachieved between the two data for the time period between70ms and200ms;a goodcorrelation was also shown in the simulation.Thus it is believed that the approach ofSimMechanics model combined with neural network control is useful and reliable inthe implementation of active musculature.