Model-Based Fault Diagnosis of a NOx Aftertreatment System

The Lean NOx Trap (LNT) is an aftertreatment device used to attain a reduction in nitrogen oxide emissions for Diesel and lean burn engines. The LNT is typically used as a storage device, capturing NOx during lean engine operation. The trap can be regenerated by controlling the exhaust air-fuel ratio to create a rich gas mixture. Under rich conditions, the stored NOx is released and catalytically converted. This way, tailpipe emissions can be significantly reduced by properly modulating the lean (storage) and rich (regeneration) periods. To maintain the LNT operate with high conversion efficiency, an optimized control of the regeneration scheduling is required. In addition, LNT systems require fault diagnostic schemes to detect and isolate faults, typically related to sulfur and thermal damages. This paper deals with the design and validation through simulation of a model-based fault diagnosis scheme for a LNT system. The mathematical model of the subsystem, based on the physics of the processes involved, consist of timevarying nonlinear ODEs. The proposed diagnostic approach is based on the generation of residuals using system models and through comparison of the predicted and measured value of selected variables, including the AFR, catalyst output temperature and the NOx concentration at the output of the LNT. The paper focuses on detection and isolation of controller faults and LNT parametric faults related to sulfur and thermal damage. The model utilized in the diagnostic scheme, which includes sulfur poisoning and thermal deactivation processes, has been experimentally validated from data collected on a Diesel LNT system and integrated with a quasi-steady engine and vehicle simulator to estimate tailpipe emissions during standard driving cycles.