An Adaptive Estimation Approach for Integrating Real-World Operation Dynamics in Engine-Out NOx Emission Modeling of a Wheel Loader

Abstract

AbstractAccurately predicting engine-out nitrogen oxides (NOx) emissions on-board is crucial for effective emission control in heavy-duty engines. Real-world engine operating conditions, especially in non-road applications with frequent dynamic changes, can significantly affect NOx emission characteristics. However, these engine emission characteristics are conventionally measured on steady-state or regulated driving cycles, which may not fully reflect the emission levels under real-world operational dynamics. This highlights the necessity of integrating engine performance during transient operation into the NOx prediction model to enhance the accuracy of on-board predictions. This paper introduces a novel data-driven model to predict engine-out NOx emissions during the construction activities of a wheel loader. This paper begins by addressing discrepancies between steady-state map predictions and on-board NOx measurements. To bridge these gaps, the model identifies engine transient operating conditions by analyzing the time derivatives of engine speed and torque. The model structure integrates steady-state and transient emission maps, with the transient map being iteratively refined using the Kalman filter principle, thereby improving its accuracy and robustness in response to engine dynamics. The proposed method maintains a model structure that is easily implemented and similar to conventional steady-state emission maps, while also enabling online self-learning for model parameter updates. Model validation shows that the model has high prediction accuracy and the ability to differentiate between steady-state and transient engine working conditions during construction activities.