Identification of Spring-Force Factors of Suspension Systems Using Progressive Neural Network on a Validated Computer Model

The spring-force factor is a multiple scalar that amplifies (or reduces) the spring force of a road-wheel suspension. The setting of spring-force factors for individual suspensions can change the stiffness property of an entire suspension system of a tracked vehicle. In engineering practice, these factors may not be accurately known. In this article, a novel progressive Neural Network (NN) technique is suggested to determine the suspension factors based on the dynamic response (displacement) of the road-wheels. A rope-length sensor can easily measure the displacement. The NN model is established and trained off-line using initial training data including a set of assumed spring-force factors and the displacements of road-wheels computed from a validated model built by the tracked vehicle toolkits (ATV) of ADAMS (a multi-body dynamics simulation tool). The trained NN model can inversely determine the suspension factors by feeding in the displacements of the road-wheels. The identified factors are then used to update the factors in the ATV model from which the new displacements of the road-wheels can be generated. A progressive retraining process of the NN model is continuously conducted until the errors between the calculated displacements and the measured ones decrease to an acceptable level. The procedures are examined for the determination of suspension factors of an in-service tracked vehicle. It is found that the present technique is very robust for the determination of the suspension factors of the tracked vehicle.