Seat belt control : from modeling to experiment

Abstract

In the last decades, vehicle safety has improved considerably. For example, major improvements have been made in the area of the structural crashworthiness of the vehicle, various driver assistance systems have been developed, and enhancements can be found in the restraint systems, the final line of defense in occupant protection. Despite this increase of vehicle safety measures, many fatalities still occur in road transportation. Regarding the unavoidable crashes, a significant amount can be attributed to the fact that the seat belt system does not perform optimally. No crash event or occupant is identical, yet conventional seat belts are – in general – not able to adjust their characteristics accordingly. The system is therefore optimal for only a limited number of crash scenarios and occupant types. With the current sensor and processor technology, it may be possible to develop a seat belt that continuously adapts to the actual crash and occupant conditions. Such a device is referred to as a Continuous Restraint Control (CRC) system, and the work presented in this thesis contributes to the development of this type of systems. The main idea of seat belt control is to add sensors and actuators to the seat belt system. The force in the seat belt is prescribed by the actuator during the crash, such that the risk of injuries are minimized given the current impact severity and occupant size and position. This concept poses several technological challenges, which are in this thesis divided into four research topics. Although many sensor technologies exist nowadays, so far no methods have been proposed to measure the occupant injury responses in real-time. These responses are essential when deciding on the optimal belt force. In this thesis, a solution has been presented for the problem of real-time estimation of (thoracic) injuries and occupant position during a crash. An estimation is performed based on modelbased filtering of a small number of readily available and cheap sensors. Simulation results with a crash victim model indicate that the injury responses can be estimated with sufficient accuracy for control purposes, but that the estimation heavily depends on the accuracy of the model used in the filter. A numerical controller uses these estimated injury responses to compute the optimal seat belt force. In this computation, it has to be taken into account that the occupant position is constrained during the crash by the available space in the vehicle, since contact with the interior may result in serious injury. The controller therefore has to predict the future occupant motion, using a prediction of the future crash behavior, a choice for the future seat belt force, and a model of the vehicle-occupant-belt system. Given the type of control problem, a Model Predictive Control (MPC) approach is used to develop the controller. Simulation results with crash victim models indicate that using this controller lead to a significant injury risk reduction for the thorax, given that an ideal belt actuator is available. The injury estimator, the prediction and control algorithm proposed in the foregoing are designed with simple mathematical models of occupant, seat belt and vehicle interior. It is therefore recognized that such accurate, manageable models are essential in the development of CRC systems. In this thesis, models of various complexities have been constructed that represent three types of widely used crash test dummies. These models are validated against both numerical as experimental data. The conclusion of this validation is that in frontal crashes, the neck and thoracic injury criteria can well be described by linear (time-invariant) models. However, when the models are to be used in the design of a belt control system, more attention has to be given to the modeling of the chest and seat belt. The severity and duration of a typical impact require a seat belt actuator with challenging specifications. For example, it has to deliver very high forces over a large stroke, it must have a high bandwidth, and must be small enough to be fitted in a vehicle post. These devices do not yet exist. In this thesis, a semi-active belt actuator concept is presented. It is based on a pressure-controlled hydraulic valve, which regulates the belt force through an hydraulic cylinder. The actuator is designed and constructed at the TU/e, and evaluated experimentally. Moreover, a moving sled setup has been developed which allows testing the actuator under impact conditions. Experimental results show that the belt actuator meets the requirements, except for the maximum force. The actuator can therefore at this point be used to prescribe belt forces in a safety belt in low-speed impacts.