Engine Calibration With Surrogate-Assisted Bilevel Evolutionary Algorithm.

Area of Science:

• Automotive Engineering
• Optimization Algorithms
• Computational Intelligence

Background:

• Engine calibration involves complex, costly black-box optimization problems with objective space constraints.
• Existing surrogate-assisted evolutionary algorithms often neglect variable sensitivity analysis, leading to inefficient optimization.
• Understanding variable impact is crucial for effective engine calibration.

Purpose of the Study:

• To propose a novel surrogate-assisted bilevel evolutionary algorithm for real-world engine calibration.
• To address the limitations of existing methods by incorporating variable sensitivity analysis.
• To enhance efficiency in constraint handling and optimize fuel consumption.

Main Methods:

• A surrogate-assisted bilevel evolutionary algorithm was developed.
• Principal Component Analysis (PCA) was used for variable sensitivity analysis and classification into lower-level and upper-level variables.
• An ordinal-regression-based surrogate model estimated solution feasibility.

Main Results:

• The proposed algorithm effectively handles constraints in engine calibration.
• It achieved a smaller fuel consumption value compared to state-of-the-art methods.
• Variable sensitivity analysis directed optimization efforts towards more impactful parameters.

Conclusions:

• The surrogate-assisted bilevel evolutionary algorithm offers an efficient approach to engine calibration.
• Incorporating variable sensitivity analysis significantly improves optimization performance.
• The method demonstrates superior constraint handling and fuel economy optimization for gasoline engines.