Performance measures and control laws for active and semi-active suspensions

Storey, I 2012, Performance measures and control laws for active and semi-active suspensions, Doctor of Philosophy (PhD), Aerospace, Mechanical & Manufacturing Engineering, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Storey.pdf Thesis Click to show the corresponding preview/stream application/pdf;... 6.94MB
Title Performance measures and control laws for active and semi-active suspensions
Author(s) Storey, I
Year 2012
Abstract This thesis concentrates on two competing performance requirements of general suspension systems: "smoothness" and tracking. The focus of the thesis is on real-time feedback controls which can be applied in microprocessors with relatively limited capacity.

Evolutionary algorithms (EAs) are used as a tool in the investigation of a wide range of control algorithms. Jerk (the rate of change of acceleration) is used as the basis of the suspension comfort performance measure, and a nonlinear cost function is applied to tracking, which targets the travel limits of the suspension (termed the "rattlespace"). Tracking measures currently in use generally fail to explicitly refer to the working space width. This matter is analysed, showing that driver slowdown is a complicating factor.

The test rig of the physical experiment is of the semi-active type. High performing semi-active controls are generally based on active controls. Thus active controls are also investigated in this thesis.

By stiffening the suspension as it moves away from equilibrium it can be made to combine softness over smooth roads with the capacity to react to large bumps when needed. Electronic control produces a much greater range of possible responses than is possible with just rubber or neoprene bump stops.

Electronic, real-time control can attempt to target a smooth chassis trajectory within the possible future limits of rattlespace. Two general methods are proposed and analysed: one that adjusts the suspension stiffening according to the current road state, and another that targets edge trajectories within the possible future movements of the rattlespace. Some of these controls performed very well. With further investigation, they may be developed into extremely high performance controls, especially because of their high adaptability to varying conditions.
The problem of avoiding collisions with rattlespace limits is related to the problem of avoiding overshoot of a limit distance. It becomes apparent that the residual acceleration at the point of closest approach needs to be limited, otherwise instability results.

This led to the search for controls that attain rest without overshooting the final rest position. It was found that the minimum jerk needed for a general minimum-time control that does not overshoot zero displacement is always the control with just one intermediate switch of control, instead of two switches that are generally needed.

This was proven to be optimal, and because of its optimality it works consistently when applied as a closed-loop, real-time optimal control. This control deals with the most difficult part of the trajectory: the final, "docking" manoeuvre. The control proved to be robust in physical experiments and it may itself have a number of applications.

Some heuristics have been developed here to account for stochastic movement of the rattlespace edges in suspension controls, and these have proven quite successful in numerical experiments.

Semi-active suspensions have a limit on the forces they can apply (the passivity constraint), but clipped versions are known to produce uncomfortable jerk. One method developed in this thesis produces a vast improvement in semi-active controls in the numerical experiments.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Aerospace, Mechanical & Manufacturing Engineering
Keyword(s) Multiobjective
Evolutionary Algorithms
Version Filter Type
Access Statistics: 1310 Abstract Views, 3925 File Downloads  -  Detailed Statistics
Created: Mon, 29 Oct 2012, 09:11:34 EST by Brett Fenton
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us