The use of three-way catalysts in automotive exhaust after-treatment systems is essential for reducing tail-pipe emissions to the levels demanded by environmental legislation. A three-way automotive catalyst reduces automotive emissions by oxidizing both the unburned hydrocarbons and carbon monoxide and by reducing the nitrogen oxides contained in the engine exhaust gas. In order to ensure that the components of a catalytic automotive emission control system continue to operate properly with age, effective and robust on-board system monitoring is essential. Because emissions from a small fraction of malfunctioning vehicles are believed to account for a large fraction of the total automotive-related emissions, on-board monitoring has the potential to significantly reduce hydrocarbon and nitrogen oxide air pollutants. This observation, along with legislation that mandates on-board diagnostic (OBD) systems to monitor the health and performance of the catalyst system, has led to interest in monitoring strategies that are able to reliably indicate when the catalytic emission control system is no longer meeting specification or when a fault is present in the system.

Legislated targets aimed at improving air quality through the reduction of automotive emissions specify both the emission levels that must be achieved by new vehicles as well as the deviation from these levels that must be detected by on-board diagnostic (OBD) systems. In the United States, SULEV (Super Ultra Low Emission Vehicle) emission standards specify stringent hydrocarbon and nitrogen oxide emission levels over an 120K mile vehicle lifetime. PZEV (Partial Zero Emissions Vehicle) emission standards specify further reductions in emission levels over an 150K mile vehicle lifetime. Meeting these requirements is particularly challenging due to the lack of cost-effective and reliable automotive emissions sensors. This limitation has led to catalyst control and monitoring approaches that consider the oxygen storage capacity of the catalyst rather than the conversion efficiency. Although the catalyst oxygen storage capacity can not be measured directly, it can be inferred using pre- and post-catalyst exhaust gas oxygen (EGO) sensors with a suitable model of the catalyst. These sensors are available in modern engine air fuel ratio (AFR) and emission control systems making this approach to control and monitoring attractive.

The technology development to address the related requirements for emission control and diagnostic monitoring have largely been carried out independently. Although there are benefits to a partition of the control and diagnostic functions arising from the reduction in coordination effort during development and mutual failure modes during operation, there are also significant disadvantages associated with this separation of function. It is more difficult to decouple the aging response of the catalyst from its dynamic behavior without the use of intrusive tests or severely limiting the conditions under which diagnosis can be performed. The determination and calibration of these conditions is also a significant task. Finally, the operation of the diagnostic can require suspension of the control strategy together with an associated emissions penalty. Because both catalyst control and diagnostics are ultimately based on some model of the catalyst behavior, exploiting this commonality can reduce the calibration burden and improve the overall catalyst system performance.

A reasonable next step in the progression of catalyst control and diagnostic system development is the integration of model-based approaches to both the control and diagnostic functions using the same dynamic model of the catalyst behavior. We present a model-based catalyst control and monitoring strategy in which the catalyst is modeled as a nonlinear integrator with a time-varying integral gain. Because this gain reflects the catalyst oxygen storage capacity and, therefore, the health of the catalyst system, it can also be used in the determination of a diagnostic metric in addition to providing information on the catalyst behavior for the controller. Although changes in this gain are an indicator of catalyst system health, there are also normal variations in this value as a result of different driving conditions with a healthy catalyst. A critical requirement for the success of this integrated strategy is the ability to detect changes that are a reflection of improper catalyst system operation and changes that are a result of normal variation with a healthy catalyst.

Various nonparametric statistical tests based on test metrics computed from a moving horizon of the time-varying gain are used to perform this discrimination between normal and abnormal catalyst operation. These test metrics are compared to determine their ability to detect changes in the catalyst system performance using a number of differently aged and intentionally deactivated catalysts. Experimental vehicle tests carried out using a 4.6L ULEV II gasoline engine operated over several EPA federal test procedure (FTP) drive cycles with each of the different catalysts are used as the basis for comparison. Preliminary results indicate that it is possible to perform very accurate discrimination between catalyst operation, even near the OBD detection threshold, using statistical classification techniques based on the cumulative distribution function of the catalyst gain.