An Optimisation Technique for Vehicle Suspensions with the Vibration Dose Value as Optimisation Parameter 922142
A modified Nelder & Mead (N&M) algorithm (Simpleks algorithm) is used, with the Vibration Dose Value (VDV) as optimisation parameter, to optimise the springs and dampers, functioning in unison with defined suspension bump stops, of the suspension system of a vehicle.
The teqnique for determination of VDV's was origionally developed for use during physical testing of vehicles where it was used as a single number to provide an objective assesment of the severity of the vehicle ride as experienced by the driver and passengers of a vehicle. However, the VDV has proved to be ideally suited as optimisation parameter during the development of new vehicles and during investigations into existing vehicles and vehicle upgradings as regards the optimisation of the springs and dampers of their suspension systems.
The choice of the N&M algorithm above other optimisation methods for performing the optimisation was based on the saving in computer simulation runs, computer time and resulting costs necessary to determine an optimum suspension configuration in conjunction with the unique definition of the optimum spring and damper characteristics. This indicates a vast improvement above the previously used method which made use contour lines of equal VDV values fitted through a matrix of VDV's determined from 3-D vehicle simulations with varying spring and damper combinations. The tendency of the contour lines then indicate the location of the optimum spring and damper values.
For the testing and qualification of the modified N&M algorithm as optimisation method for springs and dampers as well as for accellerated development of the control program of the modified algorithm it was coupled to a three dimensional interpolation subroutine where it was tested on an existing proven matrix of VDV's. The results indicate that the correct optimum value is singularly defined and that the the optimum value is obtained in approximately one third of the time needed for the matrix an contour method.