NO_x Sensor Aging Evaluation, 03-R8153

Principal Investigator
Theodore M. Kostek

Inclusive Dates:  06/01/10 – 09/01/10

Background - The diesel engine industry has recently widely adopted use of a sensor to detect NO_x in exhaust. The tailpipe emissions of NO_x are highly regulated, and this sensor provides important information. By measuring the NO_x in the exhaust, the engine system can decide how much diesel exhaust fluid to inject into the exhaust. The NO_x and diesel exhaust fluid mix and flow together over a special catalytic converter — a selective catalytic converter — that eliminates the NO_x. Not only is the NO_x sensor vital for the operation of the selective catalytic converter, the NO_x sensor is also vital to detect failed or broken systems. Beginning in 2013, the on-board diagnostics requirements become significantly more complex. Achieving good control and good diagnostics requires a deep understanding of the behavior of the NO_x sensor both in the fresh condition and in the aged condition.

Approach - The purpose of this project is to develop methods to fully characterize NO_x sensors. Various data have appeared in the public literature, some from SwRI. This project advances the state of the art by specifically addressing nonlinear behavior and cross-sensitivity. The regulated species "NO_x" is composed of NO and NO₂, and the response of the sensor to these two compounds is not necessarily the same. Because the sensor operates on a diffusion principle, the sensor reading depends on the pressure in the exhaust. These various dependencies may or may not be linear, and so the methodology should be capable of detecting any nonlinear and interaction terms. A client provided four sample sensors: two known to be within specification and two known to be out-of-specification. The goal of the project was to provide a full characterization of all four sensors. The methodology applied in this project was based on ideas from statistical design of experiments. One area within design of experiments is response surface analysis, which focuses on identifying nonlinear multi-input models. By adapting ideas from response surface analysis to the NO_x sensor, SwRI researchers were able to fit models to all four sensors. Furthermore, the terms in the models were statistically analyzed to see whether they were significant or spurious.

Accomplishments - The project succeeded in fully characterizing all four sensors. A separate nonlinear model was identified for each sensor, and the models showed excellent agreement with the sensors. The correlation coefficients between the actual sensor output and the model output were more than 0.98, and the maximum error was less than 5 percent. By studying the structure of the models, the behavior of the sensors can be more fully understood. Now that a statistically rigorous method has been developed to characterize the sensors, the natural next step is to apply the method to a large sample of sensors at various levels of aging. Such a NO_x sensor "library" would be useful to set limits on the ability to control and diagnose the selective catalytic converter. If the library is industry-wide, then it would also serve as a baseline for discussions with government regulators.

2010 Program Home