**1. Introduction**

Modern internal combustion engines (ICEs) are designed with the aim to reduce the pollutant and CO2 emissions, while delivering the required torque performance, complying with the binding legislations for vehicle homologation [1]. Referring to spark ignition (SI) engines, car manufacturers are facing the challenging path towards the low-emissions vehicles through the development of innovative, and sometimes very complex, engine architectures. In the present-day scenario, various technical solutions, characterized by a different cost/effectiveness compromise, have been successfully implemented in SI engines for the control and the abatement of noxious species at the exhaust. The most common technology still consists in the adoption of the three-way catalyst (TWC) along the exhaust line. TWC device requires a close-to-stoichiometric air/fuel (A/F) mixture to guarantee a high efficiency, with significant performance losses at the engine cold start operations. To overcome this issue, a greater interest is also devoted to the emerging techniques to

Tornatore, C.; Bozza, F.; Valentino, G. Effect of Cylinder-by-Cylinder Variation on Performance and Gaseous Emissions of a PFI Spark Ignition Engine: Experimental and 1D Numerical Study. *Appl. Sci.* **2021**, *11*, 6035. https://doi.org/10.3390/ app11136035

Academic Editor: Ramin Rahmani

Received: 29 April 2021 Accepted: 26 June 2021 Published: 29 June 2021

**Citation:** Teodosio, L.; Marchitto, L.;

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

limit the in-cylinder production of pollutant emissions for SI engines, such as the adoption of innovative combustion modes moving towards the Low-Temperature Combustion (LTC) concept: the homogeneous charge compression ignition (HCCI), the spark-assisted compression ignition (SACI) and the turbulent jet injection [2]. Turbulent jet injection (TJI) demonstrates to be a promising technique to reduce the exhaust emissions of SI engines [3], especially in the case of an active pre-chamber thanks to the ultra-lean combustion; on the other side, HCCI and SACI combustion modes allow significant improvements in NOx emissions, while some penalties on the HC and CO are obtained [4].

In addition, well-established technologies, such as the employment of cooled EGR [5], direct or port water injections [6] and alternative fuels, including ethanol/gasoline blends [7] or methanol/gasoline blends [8], allow certain benefits on the main exhaust emissions.

In addition to the above-discussed techniques to improve pollutant species, a particular attention has to be devoted to the control of cylinder-by-cylinder variation, since it could lead to a combustion deterioration with consequent increase in the exhaust emissions, especially of HC and CO species.

Different factors may play a role on the onset of cylinder-by-cylinder variation in SI engines. Indeed, the increasing complexity of engine subsystems, the high number of mechanical components, the manufacturing tolerance, the components aging can be considered as examples inducing the cylinder-by-cylinder variation and leading to a worsening in both efficiency and exhaust emissions. Of course, this aspect cannot be overlooked and should be taken into account during the engine development phase. An individual control and optimization of combustion for cylinders in a multi-cylinder engine can further contribute to suppress the cylinder-by-cylinder variation and to improve emissions. To this aim, both experimental and modeling approaches are employed. Experimental and numerical methods are increasingly combined to merge their relative advantages and to offer the availability of validated numerical tools capable to reproduce the behavior of both engine and single cylinders under various operating conditions.

Despite the relevant effects of cylinder-by-cylinder variation on SI engine performance and emissions, few technical papers are available in the literature. In addition, a reduced attention is devoted to understanding the causes that originates this phenomenon. As an example, Czarnigowski [9] investigated the effect of cylinder-by-cylinder variation on indicated mean effective pressure (IMEP) in a radial nine-piston engine, founding differences in single cylinder IMEP up to 40%. Zhou et al. [10] conducted an experimental and numerical study on a four-cylinder spark ignition engine to evaluate the influence of cylinder-by-cylinder variation on the performance. They attribute this phenomenon to the non-uniformity of gas exchange between cylinders and estimated a relative deviation of individual cylinder IMEP larger than 30%. Recently, Xu et al. [11] proposed a combustion variation control strategy, optimizing the thermal efficiency of a lean burn spark ignition engine by means of the reduction in the cylinder-by-cylinder variability. They reduced the combustion variation up to 28% with a maximum and an average increase in the brake thermal efficiency of 0.32% and 0.13%, respectively.

Referring to the cylinder-out emissions, Einewall et al. [12] carried out individual cylinder measurements of emissions and pressure cycles on a six-cylinder lean burn natural gas engine. Mixture quality variations between cylinders were confirmed by the analysis of heat release and A/F ratio. Emission measurements in each cylinder were performed only at high/medium loads and low speeds. No clear trend between A/F ratio and HC emissions of single cylinders was observed, while lower NOx emissions were detected with the air/fuel mixture leaning.

Concerning the numerical approach, a number of methodologies were adopted by worldwide researchers to explore the potentials of the models in reproducing both engine and cylinder performance. In the case of exhaust emissions, regression methods, Artificial Neural Network (ANN) models, Adaptive Neuro-Fuzzy Systems (ANFIS) and various phenomenological sub-models integrated into the fluid-dynamic codes are used. As an example, Sayin et al. [13] utilized an ANN approach to predict the overall performance, HC

and CO emissions at the exhaust of an SI engine fueled with gasoline at different octane numbers. It was observed that the ANN model was able to reproduce the engine behavior at different speed/load points with very low root mean square errors. Zschutschke et al. [14] coupled a 1D code to a detailed chemical kinetics solver for the estimation of the engine-out NOx emissions of a direct injection spark ignition engine. The model was validated in a single operating point against the experimental findings and the outcomes of a developed 3D CFD model. Then, it was extended for the prediction of NOx emissions in the entire engine operating map, denoting a good agreemen<sup>t</sup> with the experimental data.

In a previous work [15], the authors studied a turbocharged gasoline engine through a 1D code to predict the combustion and the emissions (HC, CO and NO) of a single cylinder under a limited set of operating points. A certain inaccuracy was found in reproducing the experimental trend of HC emission at part load and increasing the EGR content.

Liu et al. [16] analyzed the combustion process and the emissions of an original compression-ignition (CI) engine converted to an SI natural gas (NG) engine using 3D Gequation based RANS simulations. According to a unique set of model tuning parameters, 3D model was able to qualitatively predict the effect of NG composition on emissions over a reduced range of operating conditions.

In the light of the above-discussed literature works, it emerges a lack of combined numerical/experimental studies on the main pollutant emissions of SI engines over different operating conditions (variation in speed/load point and EGR rate), also including the effects of cylinder-by-cylinder variation.

The main topic of the present paper is represented by the combined experimental and 1D numerical analyses of a small turbocharged PFI spark ignition engine in order to provide fast and accurate predictions for individual cylinder-out emissions at different operations and with an apparent cylinder-by-cylinder variation. A previous dedicated experimental study on the examined engine has shown a certain difference in the injected fuel quantity by the port-injectors, mainly ascribed to the fuel rail geometry [15].

In this work, in a first stage, the original engine test bench was modified with the aim to measure both the cylinder-out exhaust emissions and the overall emissions at the engine exhaust, just upstream of the TWC. An extensive experimental campaign was carried out: at each operating condition, the individual cylinder behavior was characterized both in terms of performance, combustion evolution and stability, knock occurrence and exhaust emissions.

Tests were performed at various speed/load points of the engine domain, including operations under different external EGR rates to accurately explore the emission variations for the engine and cylinders. In particular, part-load points typical of the engine Worldwide harmonized Light-duty vehicles Test Cycle (WLTC) were investigated under stoichiometric A/F ratio and increased residual contents.

The experimental outcomes were employed to validate a 1D model of the entire engine, developed within the GT-PowerTM code. Refined sub-models of turbulence, combustion, heat transfer and pollutant emissions were utilized and integrated within the commercial code. The propagation of the cylinder-out noxious species within the exhaust system up to the TWC was also considered. The main innovative aspect of present work is represented by the adopted modeling approach which allows to easily forecast the effects of cylinderby-cylinder variation on both combustion and exhaust emissions of a SI engine and in a wide range of operating conditions. An additional novelty of work, compared to the ones reported in the current literature, is represented by the first-attempt prediction of the single cylinder emission characteristics. Once validated, the model is applied to reproduce the improvements in terms of Indicated Specific Fuel Consumption (ISFC) and pollutants resulting from the suppression of A/F ratio imbalance between engine cylinders.

Summarizing, the proposed numerical procedure can be considered a valuable tool to control the cylinder-by-cylinder variation with the aim to optimize the individual cylinder behavior for improved engine stability, fuel economy and pollutant emissions.
