Event Based Sampling with Application to Spectral Estimation

Sammanfattning: In this thesis event based sampling, estimation of a single resonance frequency and the application indirect tire pressure monitoring system are studied.Event based sampling is an alternative to traditional uniform or equidistant sampling. In event based sampled systems, signals are sampled only when certain pre-defined events take place. Event based sampling occurs naturally in many contexts such as motion systems where angles and positions are measured. In this thesis special attention is paid to rotating axles where the position is measured with a rotational speed sensor. Due to production tolerances such a sensor has a nonideal performance. This causes a periodic error in the event based sampled signal which in many cases cannot be neglected. Two different methods for elimitating the sensor errors are proposed.To be able to use standard tools for analyzing the signal from the rotational speed sensor, the signal is interpolated to a uniformly sampled signal. A desirable feature of the interpolation is to incorporate a lowpass filter to avoid aliasing. We investigate and compare three different methods, all based on a non-parametric regression.A survey of different methods to estimate a single resonance frequency in the rotational speed signal is presented including aspects of estimation accuracy and computational complexity.As an application of event based sampling and spectral estimation, a Tire Pressure Monitoring System (TPMS) is designed and analyzed. The TPMS is based on event based sampled measurements from a wheel speed sensor. The tire is modeled as a spring-damper system. The idea is to monitor the resonance frequency caused by the spring-damper system. This frequency is correlated with the inflation pressure, and by monitoring the resonance frequency, it is possible to detect inflation pressure changes. The system is implemented and tested in a real vehicle with promising results.