In this study a modified artificial bee colony algorithm and the cooperative-swarm variant of particle swarm optimization were applied to minimize diesel engine emissions and fuel consumption in the laboratory at medium load conditions. Tests were conducted using No. 2 diesel fuel in a four-cylinder, production diesel engine with series turbochargers and a high-pressure exhaust gas recirculation loop. Emissions were recorded at steady-state conditions and input into custom scripts in Matlab.

Both triple-injection strategies, consisting of a pilot-main-post injection scheme, and quadruple-injection strategies, using two pilots, were investigated for a high exhaust gas recirculation rate of 38%. A two-factor design of experiments study was also completed to examine the individual and interaction effects of six variables when using three injections. The modified artificial bee colony algorithm achieved 40% reductions in soot and nitric oxide emissions within 176 engine runs using a triple injection schedule with six variables. The cooperative-particle swarm method optimized an eight variable, quadruple injection schedule in only 84 engine tests. Cooperative-particle swarm algorithm was unable to find a similar optimum to artificial bee colony in triple injection experiments and appeared to stagnate.

A longer burn time was observed with the quadruple injections which also displayed decreased maximum cylinder pressures, maximum cylinder pressure rise rates, and fuel consumption results. Triple injections were able to achieve lower nitric oxide emissions. Optimized triple and quadruple injection schedules called for similar centers of combustion early in the expansion stroke resulting in similar hydrocarbon, soot, and carbon monoxide emissions. Results of the design of experiments testing illustrated the strong effect of main injection timing and fuel pressure on all aspects of the objective function. Limited effects were observed from interaction terms, except in the case of carbon monoxide and hydrocarbons.