|Abstract or Summary
- Vehicle dynamics evaluation is divided in two domains: ride quality and handling. The evaluation of characteristics related to both of these domains requires the knowledge of the vibrations developed by the vehicle. Vibrations can be described by acceleration, velocity or displacement (linear or angular). The inertial properties of large vehicles imply low frequency vibration characteristics.
Estimation of the ride quality involves primarily acceleration. Currently, sensors allow the measurement of near zero frequency acceleration with high precision. Therefore, they are suitable and satisfactory for ride quality evaluation. Standards for passenger acceleration exposure tolerance have been established. However, in the vehicle handling domain, the measurement of displacement is very important, since it defines the motion of the vehicle body and its response to the driver's actions. Obtaining the displacement by double integration of the acceleration is, by its simplicity, attractive. However, commercial accelerometers always exhibit considerable zero drift, mostly caused by low frequency thermal phenomena. This drift is amplified by double integration, which results in strong distortion of the obtained displacements. Therefore, for vehicle motion measurement, the signal obtained from acceleration requires further processing.
The purpose of this research is to evaluate vehicle dynamics using commercial DC accelerometers. A method for the recovery of vehicle displacements from an acceleration record is proposed. It combines a double integration and a model-based estimation method.
The output signal of the accelerometer is decomposed in three parts: (1) the actual acceleration, (2) the sensor drift, and (3) the remainder of error and random noise. The output signal of the accelerometer is numerically integrated and converted to displacement units. The estimation procedure requires a model structure describing the general form of displacement (obtained by the double integration of the sensor output) in response to the excitation of the vehicle.
Therefore, the estimation procedure employs a parameterized model of the; (1) anticipated displacement, and (2) displacement signal drift. The vehicle displacement model is derived by taking into account first principles that govern the vehicle dynamics. Parameters to be estimated have a physical interpretation (for example the natural frequency and the damping ratio of the vehicle considered as a single degree of freedom system). The structure of drift model is of empirical nature. The corresponding parameters to be estimated can not, a priori, be translated into obvious physical characteristics of the sensor. Based on this model structure, (1)+(2), and the double integration of the accelerometer output, the parameters of the model are estimated using the Levenberg-Marquardt method.