Experimental Optimization of a Heavy-Duty Diesel Engine Using Automated Genetic Algorithms 2002-01-0960
A micro-genetic algorithm (μGA) optimization method was applied to a heavy-duty, direct-injected diesel engine via an automated test bed system. The goal of this application was to demonstrate the feasibility and advantages of automated optimization experiments. With the genetic algorithm, no user input was required other than the factors of interest and their allowable ranges. This means that once the routine was initiated, it was essentially run undisturbed until a preset objective level was reached or a preset number of generations had been run. The automated μGA was successfully demonstrated at all points of the six-mode transient cycle simulation, excluding idle.
To accomplish the automated experiments, an automated testing system was developed around a Caterpillar single-cylinder diesel engine. An exhaust gas recirculation (EGR) pumping system was installed along with analog and/or serial communication to and from the pump drive, the gaseous emissions analyzers, the intake and exhaust pressure controllers, the electronic unit injector (EUI) fuel injection system, and the AVL dynamic particulate analyzer that was used for online soot measurement. Customized software was also developed to run the optimization routine and interface with the μGA code, the laboratory devices, and the engine.
For comparison of optimization methods, a response surface method (RSM) was also performed at the high speed (1737 rev/min), medium load (57% max.) Mode 5. Interestingly, at this mode both the RSM and the μGA located optima that had similar parameter values. In addition, at these optimum points, the engine was able to meet the 2002/2004 regulated emissions levels using the standard EUI and single injections.