Predicting the seat transmissibility of a seat-occupant system exposed to the whole-body vibration with combined artificial neural network and genetic algorithm

Inter-subject variability and seat conditions are complex and may affect the dynamic responses of the occupant-seat system to vibration. This research was aimed to clarify the contributions associated with the occupant and the seat to seat transmissibilities and thus developed an optimized artificial neural network model with the genetic algorithm to represent the cross-axis coupling and nonlinearity of seat transmissibilities with various seat conditions. Different vibration magnitudes, backrest inclinations, cushion thicknesses, frequencies and the mass of 12 subjects were set as the input parameters to predict the vertical in-line and horizontal cross-axis transmissibilities. For the predictive performance metrics (RMSE and R2), the mean values (0.118 and 0.889) were obtained for both seat transmissibilities within the testing data sets from BP-ANN models, and those with GA-BP-ANN models were optimized with 0.072 and 0.947, respectively. The seat transmissibility predicted from the model exhibited resonance behavior similar to that observed in the whole-body vibration test. With the optimization of the genetic algorithm, GA-BP-ANN models can provide enhanced predictions of the cross-axis coupling and nonlinearity of seat transmissibilities when compared to BP-ANN models.