The ability to model the spine's response to mechanical shock and vibration is an important step in assessing the health hazards of repeated impacts to vehicle passengers. Current linear models, such as the Dynamic Response Index (DRI) and the British Standard 6841 filter (BS 6841), perform poorly when the input consists of large-magnitude shocks typical of those experienced by personnel in military vehicles. In this study a recurrent neural network (RNN) was developed which models the spinal acceleration response of the seated passenger at the L4 vertebra to vertical accelerations applied at the seat. A RNN is a universal nonlinear approximator that can, in theory, model any system with memory if trained with a representative set of measured input-output data. The seat-spine system was modelled as a network with four inputs and one output. The backpropagation algorithm was used to train the network by adjusting network parameters to minimize the square of the prediction error. The inputs to the network were delayed values of the inputs and outputs. The trained network significantly outperformed the two linear models examined for predicting the z-axis acceleration at the L4 vertebra.
Modelling the dynamic response of the human spine to shock and vibration using a recurrent neural network
Modellierung des dynamischen Antwortverhaltens des menschlichen Rückgrats auf Stöße und Schwingungen mit Hilfe eines neuronalen Netzes
Heavy Vehicle Systems ; 4 , 2-4 ; 145-165
1997
21 Seiten, 13 Bilder, 1 Tabelle, 34 Quellen
Article (Journal)
English
Automotive engineering | 1997
|Investigation of the Effect of Whole-Body Vibration on the Human Body by Spine Modelling
SAE Technical Papers | 2005
|Modelling of a traffic cell based on a recurrent-neural network
Automotive engineering | 2011
|Automotive engineering | 2010
|