Use of generalised Pareto models to set on-board diagnostic failure thresholds
暂无分享,去创建一个
Vehicles sold in the US and Europe have to be equipped with a Diagnostics, called OnBoard Diagnostics (OBD), which monitor the performance of various elements of the emission control system. The driver is informed as to any failures by the use of a Check Engine Light on the dashboard of the vehicle and then should return the vehicle to the dealership for rectification. The vehicle manufacturer's aim is to ensure that the Check Engine Light is only illuminated for legitimate failures. For the calibration of an On Board Diagnostics there needs to be sufficient separation between the response of a good sensor and that of a failed sensor the setting of this threshold should be based upon a statistical model of the data so that the predicted failures rate can then be determined. In practice generating a statistical model which is appropriate for the setting of fault thresholds can prove difficult. In this paper this issue is resolved by making use of an extreme statistics method, Generalised Pareto Distribution model to allow a threshold for an Airflow Sensor diagnostic to be set.