With emissions regulations becoming increasingly restrictive and the advent of real driving emissions limits, control of engine-out NO emissions remains an important research topic for diesel engines. Progress in experimental engine development and computational modelling has led to the generation of a large amount of high-fidelity emissions and in-cylinder data, making it attractive to use data-driven emissions prediction and control models. While pure data-driven methods have shown robustness in NO prediction during steady-state engine operation, deficiencies are found under transient operation and at engine conditions far outside the training range. Therefore, physics-based, mean value models that capture cyclic-level changes in in-cylinder thermo-chemical properties appear as an attractive option for transient NO emissions modelling. Previous experimental studies have highlighted the existence of a very strong correlation between peak cylinder pressure and cyclic NO emissions. In this study, a cyclic peak pressure-based semi-empirical NO prediction model is developed. The model is calibrated using high-speed NO and NO emissions measurements during transient engine operation and then tested under different transient operating conditions. The transient performance of the physical model is compared to that of a previously developed data-driven (artificial neural network) model, and is found to be superior, with a better dynamic response and low (<10%) errors. The results shown in this study are encouraging for the use of such models as virtual sensors for real-time emissions monitoring and as complimentary models for future physics-guided neural network development.

Models used to predict the instantaneous injected fuel mass are of varied interest in automotive applications, including for providing inputs to CFD calculations or for engine control. While multiple injection strategies are now commonly used in diesel engines, the overall approach may be susceptible to injection fusion, which is defined as two successive injections that are partly or totally coupled due to the short time interval between each event. In this work, a new model to predict the instantaneous mass flow rate from a diesel injector is proposed based on the analytical solution of a first-order linear dynamic system exposed to an impulsion. Experiments are also conducted to quantify the main injection characteristics of a solenoid indirect-action injector under different injection pressures, backpressures and injection durations, representing a total of 33 different conditions. From these results, a model is proposed and validated against experimental data using a single injection strategy. Then, the model is enhanced to predict split injection with and without injection fusion. Successful comparisons are realized between the model and the experiment. The model is then used to successfully simulate a piezoelectric injector experiencing different levels of fusion available in the literature so as to illustrate the universality of the proposed approach.

Gaseous fuels for heavy-duty internal combustion engines provide inherent advantages for reducing CO, particulate matter (PM), and NO emissions. Pilot-ignited direct-injected NG (PIDING) combustion uses a small pilot injection of diesel to ignite a late-cycle main direct injection of NG, resulting in significant reduction of unburned CH emissions relative to port-injected NG. Previous works have identified NG premixing as a critical parameter establishing indicated efficiency and emissions performance. To this end, a recent experimental investigation using a metal engine identified six general regimes of PIDING heat release and emissions behavior arising from variation of NG stratification through control of relative injection timing (RIT) of the NG with respect to the pilot diesel. The objective of the current work is to provide comprehensive description of in-cylinder fuel mixing of direct injected gaseous fuel and its impacts on combustion and pollutant formation processes for stratified PIDING combustion. In-cylinder imaging of OH*-chemiluminescence (OH*-CL) and PM (700 nm), and measurement of local concentration of fuel is considered for 11 different , representing 5 regimes of stratified PIDING combustion (performed with MPa and ). The magnitude and cyclic variability of premixed fuel concentration near the bowl wall provides direct experimental validation of thermodynamic metrics ( , , ) that describe the fuel-air mixture state of all 5 regimes of PIDING combustion. The local fuel concentration develops non-monotonically and is a function of RIT. High indicated efficiency and low CH emissions previously observed for stratified-premixed PIDING combustion in previous (non-optical) investigations are due to: (i) very rapid reaction zone growth ( m/s) and (ii) more distributed early reaction zones when overlapping pilot and NG injections cause partial pilot quenching. These results connect and extend the findings of previous investigations and guide the future strategic implementation of NG stratification for improved combustion and emissions performance.

Natural gas (NG) is an attractive fuel for heavy-duty internal combustion engines because of its potential for reduced CO, particulate, and NO emissions and lower cost of ownership. Pilot-ignited direct-injected NG (PIDING) combustion uses a small pilot injection of diesel to ignite a main direct injection of NG. Recent studies have demonstrated that increased NG premixing is a viable strategy to increase PIDING indicated efficiency and further reduce particulate and CO emissions while maintaining low CH emissions. However, it is unclear how the combustion strategies relate to one another, or where they fit within the continuum of NG stratification. The objective of this work is to present a systematic evaluation of pilot combustion, NG combustion, and emissions behavior of stratified-premixed PIDING combustion modes that span from fully-premixed to non-premixed conditions. A sweep of the relative injection timing, , of NG and pilot diesel was performed in a heavy-duty PIDING engine with  = 140-220 bar,  = 0.47-0.71, and a constant NG energy fraction of 94%. Apparent heat release rate and emissions analyses identified interactions between the pilot fuel and NG, and qualitatively characterized the impact of NG stratification on combustion and emissions. Changes in the resulted in six distinct PIDING combustion regimes, for all considered injection pressures and equivalence ratios: (i) RIT-insensitive premixed, (ii) stratified-premixed (early-cycle injection), (iii) NG jet impingement transition, (iv) stratified-premixed (late-cycle injection), (v) variable premixed fraction, and (vi) minimally-premixed. Parametric definitions for the bounds of each regime of combustion were valid for the wide range of and investigated, and are expected to be relevant for other PIDING engines, as previously identified regimes agree with those identified here. This conceptual framework encompasses and validates the findings of previous stratified PIDING investigations, including optimal ranges of operation that provide significantly increased efficiency and lower emissions of incomplete combustion products.

The internal combustion engine faces increasing societal and governmental pressure to improve both efficiency and engine out emissions. Currently, research has moved from traditional combustion methods to new highly efficient combustion strategies such as Homogeneous Charge Compression Ignition (HCCI). However, predicting the exact value of engine out emissions using conventional physics-based or data-driven models is still a challenge for engine researchers due to the complexity the of combustion and emission formation. Research has focused on using Artificial Neural Networks (ANN) for this problem but ANN's require large training datasets for acceptable accuracy. This work addresses this problem by presenting the development of a simple model for predicting the steady-state emissions of a single cylinder HCCI engine which is created using an metaheuristic optimization based Support Vector Machine (SVM). The selection of input variables to the SVM model is explored using five different feature sets, considering up to seven engine inputs. The best results are achieved with a model combining linear and squared inputs as well as cross correlations and their squares totaling 26 features. In this case the model fit represented by values were between 0.72 and 0.95. The best model fits were achieved for CO and CO, while HC and NO models have reduced model performance. Linear and non-linear SVM models were then compared to an ANN model. This comparison showed that SVM based models were more robust to changes in feature selection and better able to avoid local minimums compared to the ANN models leading to a more consistent model prediction when limited training data is available. The proposed machine learning based HCCI emission models and the feature selection approach provide insight into optimizing the model accuracy while minimizing the computational costs.