Study On Performance Optimization In The Entire Load Range Of An Atkinson Cycle Engine Based On Artificial Neural Network And Genetic Algorithm

The energy crisis and environment protection have become a global problem,and thus conventional internal combustion engine (ICE) face huge challenge. Studyon highly efficient, energy and environment-friendly engine technology is the onlyway for the sustainable development of the conventional ICE industry. This thesisresearches on an Atkinson cycle engine with bigger geometrical compression ratio(GCR) comparing to conventional Otto cycle engines. In this situation, the Atkinsoncycle engine can realize bigger expansion ratio, thus more heat energy is converted tomechanical energy and higher thermal efficiency is resulted. When this kind ofAtkinson cycle engine is used in a hybrid car, it can mainly work in the fuel efficientarea and greatly reduce the engine fuel consumption.The knock tendency for the Atkinson cycle engine increases due to the big GCR.Thus late intake valve closure (LIVC) operation is necessary to decrease the effectivecompression ratio in order to avoid the knock. Bigger the GCR is, more LIVCoperation is needed. However, at WOT (widely open throttling) operating condition,the LIVC operation would reduce the engine effective displacement. When designingthe Atkinson cycle engine, we need to determine an optimum GCR to maximize thefuel economy while maintaining enough WOT torque. Furthermore, the Atkinsoncycle engine mainly works in the medium to high load operating area, and thus thefuel economy improvement for part loads is more useful for the fiiel consumptionreduction of the hybrid car.The Atkinson cycle engine addressed in this thesis has many design andoperative parameters, and these parameters are highly interrelated with each other andinteractively influence the engine performances. In order to decrease cost and enhancedesign and optimization efficiency, this thesis investigates on the methodology thatoptimizes the engine performances in the entire load range of the Atkinson cycleengine based on artificial neural network (ANN) and genetic algorithm (GA).Additionally, thermodynamic cycle analysis is also conducted to provide theoreticalguidance for the design and optimization. Therefore, the main study aspects for thisthesis include: (1)In order to research on the effect characteristics of the important design andoperating parameters on the cycle thermal efficiency of the Atkinson cycle engine, theifnite-time thermodynamics was used to establish the irreversible cycle model for theAtkinson, Otto and Miller cycle heat engine, respectively. The following conclusionsfrom the numerical computation results could be obtained: there is an optimumcompression ratio and peak cylinder pressure that maximizes the cycle efficiency ofthe Atkinson cycle heat engine; in some range, as the expansion ratio increases, thepeak cylinder pressure increases and the exhaust temperature decreases, thusimproving the cycle thermal efficiency; as equivalence ratio increases，the cyclethermal efficiency decreases; Miller cycle heat engine is closer to the real Atkinsoncycle engine while the cycle thermal efficiency of the Atkinson cycle heat engine isthe cycle efficiency improvement extreme of the real Atkinson cycle engine.(2)The Atkinson cycle engine was developed based on a1.8L Otto cycle enginewith GCR of10.6. We first established the WOT GT-Power simulation model for thebaseline Otto cycle engine. Corresponding experimental data were measured andcollected to precisely calibrate the GT-Power model. The maximal error between thesimulated and experimental values for the WOT torque and BSFC in the whole enginespeed range is2%and2.9%, respectively. Calibrate the relevant parameters inGT-Power knock model to make the knock index equal to200represent the real slightknock.(3)The GCR, intake and exhaust valve timing, spark angle and air-fuel ratiointeractively influence the knock intensity, and final engine performance of theAtkinson engine. This increases the difficulty for determining the optimum GCR andoptimizing the operative parameters. In order to more efficiently and preciselyconduct the optimization for the design and operative parameters, the artificial neuralnetwork (ANN) method was used to establish the optimization models for the targetAtkinson cycle engine. The ANN models were trained and tested using the datacollected from the GT-Power computations. Then, the GCR and the operatingparameters were optimized by combining the ANN models and genetic algorithm(GA). Atfer optimized by the GA, the optimum GCR was determined as12.5. Theexperimental results for the prototype Atkinson cycle engine indicate that, in thespeed range below4400rpm, the predicted operating parameters such as the sparkangle by the ANN models are very close to the experimental ones. Furthermore, themaximal error between the ANN prediction and the experimental for the WOT torqueand BSFC is only2.2%and2.53%, respectively. (4)In order to optimize the part load operating parameters aimed at maximizingthe fuel economy, the GT-Power simulation models at a series of representativespeed-load points covering the entire speed-load operating range of the Atkinsoncycle engine were established. These GT-Power models were all precisely calibratedusing the experimental data of the prototype Atkinson cycle engine. Experimentalresults show that, in the entire part load area the GT-Power models have highprediction accuracy with the maximum BSFC prediction error of8.5%. Then, partload operating parameters for the Atkinson cycle engine were optimized by the GAthrough MATLAB/GT-Power coupling to maximize the fuel economy merit.(5)According to the GA optimization results for the operative parameters in theentire load range, we proposed a novel torque-based load control strategy that baseson combination of the IVC (Intake Valve Closure, IVC) timing and ETC operation.This strategy could be described as: in the medium to high load range, LIVC+ETCoperation is used to control the engine output; and in the low load range, only theETC is adopted to control the engine load while advancing the IVC timing to improveengine cycle thermal efficiency.The experiments in a test-bed indicate that, atfer optimized by the GA andexperimental calibrations, fuel economy in the entire load range for the Atkinsoncycle engine is obviously optimized. For the WOT operating conditions below4400rpm, the maximal torque reduction for the Atkinson cycle engine is less than6%comparing to the baseline Otto cycle engine while the maximal fuel economyimprovement for the Atkinson cycle engine is13%at2400rpm. For the WOToperating conditions above4400rpm, in order to maintain enough engine rate power,less LIVC operation is used, but the spark angle delaying and rich mixture have to beadopted to avoid the knock. Therefore, the fuel economy for the Atkinson cycleengine deteriorates with the average fuel consumption increasing as3%.Atfer the part load operating parameters of the Atkinson cycle engine wereoptimized by GA, part load experimental fuel economy are obviously optimized withthe maximal improvement of7.67%. Comparing to the baseline Otto cycle engine, theoperating area for the lowest fuel consumption of the Atkinson cycle engine isobviously larger, which is very useful for reducing the fiiel consumption of the hybridcar using the Atkinson cycle engine.