Innovative Model-Free Onboard Diagnostics for Diesel Particulate Filter

Abstract

Recent legislations require very low soot emissions downstream of the particulate filter in diesel vehicles. It will be difficult to meet the new more stringent OBD requirements with standard diagnostic methods based on differential sensors. The use of inexpensive and reliable soot sensors has become the focus of several academic and industrial works over the past decade. In this context, several diagnostic strategies have been developed to detect DPF malfunction based on the soot sensor loading time. This work proposes an advanced online diagnostic method based on soot sensor signal projection. The proposed method is model-free and exclusively uses soot sensor signal without the need for subsystem models or to estimate engine-out soot emissions. It provides a comprehensive and efficient filter monitoring scheme with light calibration efforts. The proposed diagnostic algorithm has been tested on an experimentally validated simulation platform. 2D signatures are generated from soot sensor signal for nominal and faulty configurations. Gaussian dispersions on soot estimator (30%) and sensor model (15%) have been considered. Based on a statistical analysis, a relevant threshold is defined satisfying a compromise between non-detection and false alarm rates. The selected threshold is then used for online DPF diagnostic using NEDC cycle. The obtained results are promising and clearly show the performance of the proposed method in terms of non-detection and false alarm rates. The resulting diagnostic scheme can be easily integrated in the ECU for onboard DPF monitoring.