A robust optimization of an automobile valvetrain is presented where the variation of engine performances due to the component dimensional variations is minimized subject to the constraints on mean engine performances. The dimensional variations of valvetrain components are statistically characterized based on the measurements of the actual components. Monte Carlo simulation is used on a neural network model built from an integrated high fidelity valvetrain-engine model, to obtain the mean and standard deviation of horsepower, torque and fuel consumption. Assuming the component production cost is inversely proportional to the coefficient of variation of its dimensions, a multi-objective optimization problem minimizing the variation in engine performances and the total production cost of components is solved by a multi-objective genetic algorithm (MOGA). The comparisons using the newly developed Pareto front quality index (PFQI) indicate that MOGA generates the Pareto fronts of substantially higher quality, than SQP with varying weights on the objectives. The current design of the valvetrain is compared with two alternative designs on the obtained Pareto front, which suggested potential improvements.

This content is only available via PDF.
You do not currently have access to this content.