Effects of artificial neural networks characterization on prediction of diesel engine emissions

Tehranian, Azadeh

doi:10.33915/etd.1298

Cited by 7 publications

(10 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The learning mechanism of an ANN imitates the learning mechanism of the brain. Artificial neural networks have been used to control and/or to predict in many different applications including engine and emissions research [86][87][88][89]. Because neural networks were showed to be a powerful tool to predict nonlinear relationships, an ANN architecture was chosen to predict engine efficiency in this study.…”

Section: Artificial Neural Network (Ann) Modelmentioning

confidence: 99%

Linear engine development for series hybrid electric vehicles

Tóth-Nagy¹

View full text Add to dashboard Cite

Linear Engine Development for Series Hybrid Electric Vehicles by Csaba Tóth-Nagy This dissertation argues that diminishing oil reserves, concern over global climate change, and desire to improve ambient air quality all demand the development of environment-friendly personal transportation. In certain applications, series hybrid electric vehicles offer an attractive solution to reducing fuel consumption and emissions. Furthermore, linear engines are emerging as a powerplant suited to series HEV applications. In this dissertation, a linear engine/alternator was considered as the auxiliary power unit of a range extender series hybrid electric vehicle. A prototype linear engine/alternator was developed, constructed and tested at West Virginia University. The engine was a 2-stroke, 2-cylinder, dual piston, direct injection, diesel engine. Experiment on the engine was performed to study its behavior. The study variables included mass of the translator, amount of fuel injected, injection timing, load, and stroke with operating frequency and mechanical efficiency as the basis of comparison. The linear engine was analyzed in detail and a simple simulation model was constructed to compare the trends of simulation with the experimental data and to expand on the area where the experimental data were lacking. The simulation was based on a simple and analytical model, rather than a detailed and intensely numerical one. The experimental and theoretical data showed similar trends. Increasing translator mass decreased the operating frequency and increased compression ratio. Larger mass and increased compression ratio improved the ability of the engine to sustain operation and the engine was able to idle on less fuel injected into the cylinder. Increasing the stroke length caused the operating frequency to drop. Increasing fueling or decreasing the load resulted in increased operating frequency. This projects the possibility of using the operating frequency as an input for feedback control of the engine. Injection timing was varied to investigate two different modes of engine operation experimentally. The two modes were direct injection compression ignition (DICI) and "pseudo" homogeneously charged compression ignition (PHCCI). Simulation was performed to include HCCI operation in the study. The study showed that the HCCI operation resulted in higher peak cylinder pressure than that of DICI operation. A combined genetic algorithm-artificial neural network predictor model was used along with the simulation model to find the combination of engine parameters that yielded the highest engine efficiency. The predictor-simulator model suggested the most efficient combination of engine parameters. I thank Dr. Nigel N. Clark for his guidance during my postgraduate studies. Not only he did provide invaluable input to my research, he also showed the way with his integrity, philosophy, and attitude. I also thank Dr.

show abstract

Section: Artificial Neural Network (Ann) Modelmentioning

confidence: 99%

Linear engine development for series hybrid electric vehicles

Tóth-Nagy¹

View full text Add to dashboard Cite

show abstract

“…This control strategy utilizes the equivalent consumption minimization strategy that is based on equivalent fuel flow rate to minimize the overall FC over a given driving cycle [39,40]. For example, the equivalent fuel rate can be given by:…”

Section: Real Time Control Strategymentioning

confidence: 99%

Impact of Intelligent Transportation Systems on Parallel Hybrid Electric Heavy Duty Vehicles

Al-Samari¹

View full text Add to dashboard Cite

Impact of Intelligent Transportation Systems on Parallel Hybrid Electric Heavy Duty Vehicles Ahmed Al-Samari A hybrid electric vehicle uses multiple sources of energy that can be independently or all together used to propel the wheels. In the presented work, the vehicle propulsion controller (VPC) for a parallel heavy duty hybrid electric vehicle (HEV) model has been modified to manage the alternative power source in advance based on the forthcoming traffic information. The goal is to prepare the powertrain for the next power event by making more energy storage capacity to capture free energy via regenerative braking or store more energy for expected need. The method of preparation will be by managing the battery state of charge (SOC), which is a metal hydride battery for this study, to take advantage of opportunistic regeneration. Autonomie software was used to simulate parallel HEV models. The results revealed that the proposed looking-ahead control strategy for a class 8 parallel hybrid heavy duty vehicle with an engine power of 410 kW had a substantial contribution in preparing the system for forthcoming power demand. The looking-ahead strategy employed in this study improved fuel economy from 0.5% on flat terrain to about 3% on mountain terrain. Moreover, a looking-ahead strategy can contribute significantly to maintaining adequate power for the vehicle on different terrain types. The engine power can be downsized (with looking-ahead strategy) therefore improving fuel economy up to 13% while maintaining adequate power over different terrain types. The battery energy capacity can be downsized (with looking-ahead strategy) by half while maintaining nearly the same benefits (i.e. fuel economy and adequate power) compared to the hybridization system without looking-ahead strategy. Since different routes types (i.e. flat, hilly and mountain terrains) were used to investigate the impact of the lookingahead strategy on heavy duty parallel HEV, these results can generally be applied to many terrain and traffic situations.

show abstract

“…WVU has conducted prior modeling, using artificial neural networks [21], where a network is trained on one or more test cycles and is then used to predict emissions on another. However, this is an involved process which would restrict its use by any local authority trying to predict emissions from a link, and would not be used by a vehicle operator seeking to compare two vehicles that were tested for emissions on different cycles.…”

Section: Relevence To Inventory Modelsmentioning

confidence: 99%

“…Average NO X on Transient mode and Cruise mode combined and averaged versus UDDS y = 0.43x + 0.68 R 2 lbs UDDS PM (g/mile) 56,000 lbs Average Cruise+Transient PM (g/mile)Linear(1973( -1994( and 2000( -2003 Linear(1973( -2003 Average PM on Transient mode and Cruise mode combined and averaged versus UDDS4.5.5. CORRELATION OF AC5080 VERSUS OTHER CYCLESThe AC5080 was used to predict emissions from the Cruise mode, Transient mode and UDDS, as shown inFigure 17toFigure 22.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Cycles and weight effects on emissions and development of predictive emissions models for heavy duty trucks

Vora¹

View full text Add to dashboard Cite

When heavy-duty truck emissions are expressed in distance-specific units (such as g/mile), the values may depend strongly on the nature of the test cycle. Prior studies have compared emissions gained using different cycles and have proposed techniques for translating emissions factor rates between cycles. First, emissions data from the 5-mode CARB HHDDT Schedule, UDDS, and AC5080 were reviewed, with reference to each other. NO X and PM emissions were the two components of emissions reviewed. A heavy-duty chassis dynamometer was used for emissions characterization along with a full scale dilution tunnel. The vehicle test weights were simulated at 30,000 lbs, 56,000 lbs, and 66,000 lbs. For each vehicle, average data from one cycle have been compared with average data for a different cycle. When the Cruise mode and Transient mode of the HHDDT schedule were compared, it was evident that injection timing strategies affected the average Cruise mode emissions of NO X. As a result, there was substantial data scatter when mode-averaged Cruise emissions were plotted against mode-averaged Transient emissions. Moreover, the relationship between Cruise and Transient NO X was not the same for the various test weights. Correlations for PM varied widely in goodness of fit to the data. This is because PM may increase substantially due to elemental carbon "puff" which occurs when the turbocharger has not reached full boost pressure at the onset of an increase in engine load. The AC5080, originally developed for inspection and maintenance applications, showed reasonable correlation with the UDDS, although a best fit line still caused the AC5080 to mis-predict 16 out of 25 UDDS NO X values by over 20%. It was concluded that information is needed on both the transient and steady high speed emissions characteristics of a vehicle before an emissions factor can be estimated for a road link. Next, two-dimensional correlations were used to predict the emissions rate on one cycle from the rates of two other cycles. The vehicle test weights were simulated at 56,000 lbs. The multidimensional analysis using two cycles yielded better predictive correlations for the emissions than single cycle correlations. The UDDS yielded the highest distance-specific emissions and it showed similar emissions as the combination of the Cruise mode and Transient mode of the HHDDT. Weight, like transients, significantly effects the emissions of a vehicle. Its effect is different for different species. The emission of NO X , CO, CO 2 , HC and PM were analyzed with respect to weight. There were three types of vehicles analyzed: HHDDT, MHDDT and MHDGT. These Kuntal A. Vora Cycles and Weight Effects on Emissions and Development of Predictive Emissions Models for Heavy Duty Trucks v 2. Acknowledgments

show abstract

Effects of artificial neural networks characterization on prediction of diesel engine emissions

Cited by 7 publications

References 43 publications

Linear engine development for series hybrid electric vehicles

Linear engine development for series hybrid electric vehicles

Impact of Intelligent Transportation Systems on Parallel Hybrid Electric Heavy Duty Vehicles

Cycles and weight effects on emissions and development of predictive emissions models for heavy duty trucks

Contact Info

Product

Resources

About