PROGNOSTICS OF AUTOMOTIVE SENSORS: TOOLS AND CASE STUDY

摘要

Nowadays, a diversity of sensors are mounted in the vehicle to capture data, which are critical input to control units to make modern cars more reliable, safer, less environmentally damaging and offer greater performance than ever before. Nevertheless, sensors themselves degrade, just as any other dynamic system. A degrading sensor may cause decline in vehicle performance, shut down, or even a fatal accident. Therefore, it is essential to track and monitor the health conditions of the sensors. To predict the degradation of sensor before failure, based on a comprehensive literature review, a confidence value (CV) based prognostic decision model is proposed, which is built on the direct output signals of sensors. Features are extracted from the signals and artificial intelligent algorithms are used to assign a confidence value, and identify the degradation level. To illustrate the feasibility and efficiency of the decision model, a case study with wheel speed sensor is performed. Also a comprehensive patent search for wheel speed sensor diagnosis is conducted. Features related to pulse duty factor and amplitude of the signal are extracted and they are fed into the statistic pattern recognition (SPR) model to calculate the CV value, which determines the current health status of the monitored sensor.