Model-Based Fault Detection for a Gasoline Engine Three-Way Catalyst

In this paper we develop a non-intrusive modelbased catalyst monitoring algorithm that operates by determining the magnitude of the catalyst's frequency-domain transfer function in a specific range of frequencies corresponding to a small-amplitude square-wave input command. This method requires the calibration of two parameters within a simple catalyst model consisting of a system gain, a time delay, and an integrator. This method for catalyst monitoring offers several advantages over current methods. Specifically, it is robust to catalyst downstream oxygen-sensor location, does not require any new hardware, and operates non-intrusively in a semi-continuous manner. Thus, the method does not require restrictive steady-state entry conditions or specific events such as deceleration fuel cut-off to operate. Results obtained over a statistically significant number of varied drive cycles demonstrate the ability of the catalyst monitor to accurately detect the presence of a degraded catalyst.