Search Results

The combustion process in spark-ignition engines can vary considerably cycle by cycle, which may result in unstable engine operation. The phenomena amplify in natural gas (NG) spark-ignition (SI) engines due to the lower NG laminar flame speed compared to gasoline, and more so under lean burn conditions. The main goal of this study was to investigate the main sources and the characteristics of the cycle-by-cycle variation in heavy-duty compression ignition (CI) engines converted to NG SI operation. The experiments were conducted in a single-cylinder optically-accessible CI engine with a flat bowl-in piston that was converted to NG SI. The engine was operated at medium load under lean operating conditions, using pure methane as a natural gas surrogate. The CI to SI conversion was made through the addition of a low-pressure NG injector in the intake manifold and of a NG spark plug in place of the diesel injector.

With continuously increasing concern for the emissions from two-stroke engines including regulated hydrocarbon (HC) and oxides of nitrogen (NOx) emissions, non-road engines are implementing proven technologies from the on-road market. For example, four stroke diesel generators now include additional internal exhaust gas recirculation (EGR) via an intake/exhaust valve passage. EGR can offer benefits of reduced HC, NOx, and may even improve combustion stability and fuel efficiency. In addition, there is particular interest in use of natural gas as fuel for home power generation. This paper examines exhaust throttling applied to the Helmholtz resonator of a two-stroke, port injected, natural gas engine. The 34 cc engine was air cooled and operated at wide-open throttle (WOT) conditions at an engine speed of 5400 RPM with fueling adjusted to achieve maximum brake torque. Exhaust throttling served as a method to decrease the effective diameter of the outlet of the convergent cone.

Machine learning algorithms are effective tools to reduce the number of engine dynamometer tests during internal combustion engine development and/or optimization. This paper provides a case study of using such a statistical algorithm to characterize the heat transfer from the combustion chamber to the environment during combustion and during the entire engine cycle. The data for building the machine learning model came from a single cylinder compression ignition engine (13.3 compression ratio) that was converted to natural-gas port fuel injection spark-ignition operation. Engine dynamometer tests investigated several spark timings, equivalence ratios, and engine speeds, which were also used as model inputs. While building the model it was found that adding the intake pressure as another model input improved model efficiency.

This paper examines the energy pathways of a 29cc air-cooled two-stroke engine operating on natural gas with different exhaust geometries. The engine was operated at wide-open-throttle at a constant speed of 5400 RPM with ignition adjusted to yield maximum brake torque while the fueling was adjusted to examine both rich and lean combustion. The exhaust configurations examined included an off-the-shelf (OTS) model and two other custom models designed on Helmholtz resonance theory. The custom designs included both single and multi-cone features. Out of the three exhaust systems tested, the model with maximum trapping efficiency showed a higher overall efficiency due to lower fuel short-circuiting and heat transfer. The heat transfer rate was shown to be 10% lower on the new designs relative to OTS model.

Machine learning models were shown to provide faster results but with similar accuracy to multidimensional computational fluid dynamics or in-depth experiments. This study used a support-vector machine (SVM), a set of related supervised learning methods, to predict the dynamic performance (i.e., engine power and torque) of a heavy-duty natural gas spark ignition engine. The single-cylinder four-stroke test engine was fueled with methane. The engine was operated at different spark timings, mixture equivalence ratios, and engine speeds to provide the data for training and testing the proposed SVM. The results indicated that the performance and accuracy of the built regression model were satisfactory, with correlation coefficient quantities all larger than 0.95 and root-mean-square errors close to zero for both training and validation datasets.