Soot Sensor Elimination with DPF Substrate Failure Monitoring

The automobile industry is going through one of the most challenging times, with increased competition in the market which is enforcing competitive prices of the products along with meeting the stringent emission norms. One such requirement for BS6 phase 2 emission norms is monitoring for partial failure of the component if the tailpipe emissions are higher than the OBD limits. Recently PM (soot) sensor is employed for partial failure monitoring of DPF in diesel passenger cars.. PM sensor detects soot leakage in case of DPF substrate failure. There is a cost factor along with extensive calibration efforts which are needed to ensure sensor works flawlessly. This paper deals with the development of an algorithm with which robust detection of DPF substrate failure is achieved without addition of any sensor in the aftertreatment system. In order to achieve this, a thermodynamic model of DPF substate was created using empirical relations between parameters like exhaust flow rate, exhaust gas temperature and soot mass content. The modeling was done in both empty (no soot) and filled (threshold soot content) DPF substrate conditions. There were two methodologies, namely integration method and normalization method. In integration method the pressure drop across DPF substate in actual running condition is cumulatively summed up and compared with the integrated modelled value achieved from the thermodynamic model. If the ratio between modelled and actual crosses the threshold value, a DPF substrate failure flag is raised. In normalization method, actual pressure drop across DPF is corrected using empirical thermodynamic relations for exhaust temperature and soot mass content. The corrected pressure drop is normalized for the exhaust flow rate and then averaged based on release conditions. If the averaged values is lower than the threshold, a DPF substrate failure flag is raised. DPF substrate failure is intimated only when both the methodologies raise the failure flag. The algorithm was tested with actual failed DPF sample and robust detection was observed (more than 90% detection accuracy) and no misdetection.