Major drawbacks of transportation by motor vehicles are crashes and the consequences thereof like injuries and fatalities. The safety belt and the airbag, often referred to as the restraint system, have been introduced to reduce the number and severity of injuries. The restraint system should behave differently for different crashes and/or different occupants. State-of-the-art belt and airbag systems are “adaptive”, meaning that they have a limited set of modes of operation to adapt to different occupants and crashes. Examples of these modes of operation are different deformation characteristics of the load limiter in the belt system and different points of triggering the inflators of the airbag. Design of such modes of operation focus on achieving a satisfactorily low risk of injury for classes of occupants and crashes. Examples of measures for the risk of injury are the maximum chest acceleration, the maximum chest deflection and the maximum head acceleration. Appropriate modes of operation are typically obtained by minimization of the risk of injuries, using complex nonlinear models to simulate a vehicle and occupant, subjected to a crash test. Such an approach is time-consuming and the obtained modes are a compromis. In this thesis, an innovative view on (the design of) restraint systems is elaborated. The idea is to add sensors and actuators in order to allow feedback control of the restraint system. The airbag and/or the belt are manipulated during the crash to force one or more occupant variables, representing the risk of injuries, to follow an a priori defined reference signal. This reference signal represents the lowest possible risk of injuries. This view on restraint systems can be seen as a starting point for the development of future restraint systems, and as a basis for an effective design expedient for modes of operation of real world restraint system components. The concept of active restraint systems has been elaborated using the numerical model of a mid-size male dummy as the “driver” of a mid-size passenger car, subjected to the US-NCAP frontal crash test. To manipulate the airbag, the size of the vent in the airbag and the mass flow into the airbag have been chosen. To manipulate the belt, the force in the belt section near the load limiter, has been chosen. The chosen occupant variables are the chest acceleration, the head acceleration and the chest deflection. Reference signals are pragmatically determined. Controllers to manipulate the airbag or the belt, cannot be designed using the available numerical model, since that is nonlinear and far too complex. Therefore, linear timeinvariant (LTI) control design models are derived to approximate the relevant dynamic behavior of the restraint system, the dummy and their interactions. These control design models are obtained with the approximate realization method, using the responses of the occupant variables to stepwise perturbations in the manipulated variables of the restraint system. Low order feedback controllers are designed using “loopshaping” techniques, aiming at a stable closed loop system with satisfactory performance. Finally, the controllers are implemented and evaluated in the closed loop system with the complex nonlinear model. In comparison with the original restraint system, control of the chest acceleration by manipulation of the belt force can reduce the risk of chest injuries by 60%. Control of the head acceleration by manipulation of the vent size reduces the risk of head injuries by 50%. Appropriate simultaneous control of the chest and the head acceleration reduces the risk of injury to the chest and head by 50% or more. The modelling and the control design strategy have also been applied successfully to arrive at a controller for the chest deflection bymanipulation of the belt force. In addition, the strategies have been applied successfully to arrive at controllers for the chest and the head acceleration for the case of a small female dummy as the “driver”. It has become clear that the dynamic behavior of the belt and/or the airbag, interacting with the dummy can be considered linear, at least for control design purposes. Besides that, low order feedback controllers are effective to enforce the desired behavior of the complex nonlinear model. Furthermore, it turned out that the control design problem for simultaneous control of the head acceleration and the chest acceleration by manipulation of the vent size and the belt force can be treated as a decoupled control design problem. The modelling and control design strategy have shown to be effective and effi- cient. Insight into appropriate modes of operation for adaptive restraint systems can be obtained from results of closed loop simulations. (Author/publisher)
Samenvatting