New Math Trick Could Make Jet Engine Design Faster and Cheaper
Speedier design could mean more fuel-efficient planes and lower power bills for everyone.
Designing a jet engine blade is a headache of choices. How curved should it be? How thick at the tip? How many blades in the row? These are the "dimensions," meaning separate knobs you can turn, and a modern design can have 30, 60, even 90 of them. Testing each combination with realistic airflow software takes hours on powerful computers (a "black box" problem, because you can run the test but never see the hidden rules behind it). Engineers only get a limited number of tries before the budget runs out, so picking the right designs to test matters enormously.
The new method, DA-EGO, attacks this by chopping the huge puzzle into smaller pieces and solving each one with a fast stand-in model that approximates the slow simulation. The best answers from each piece get merged into one promising candidate. Crucially, the pieces are not fixed. Each round, the algorithm checks which knobs affect each other (like realizing salt and baking time interact in a recipe) and redraws the groups, while also narrowing where it searches next.
Tested on 21 standard benchmark problems at 30, 60 and 90 dimensions with a strict budget of 1,500 evaluations, DA-EGO performed well on problems where the knobs act mostly independently. Honesty matters here: on one notoriously tangled test called shifted Rosenbrock, a rival method, GSGA, did better at 60 and 90 dimensions. The real proof came on genuine engineering: a transonic rotor blade with 28 variables and a compressor stage with 60 variables, both improved.
The payoff for the rest of us is indirect but real. Engines that are designed faster and better tend to burn less fuel, which shows up in airline ticket prices, electricity bills and emissions over the following years. The catch is patience. This is a specialist tool published in an engineering journal, not a product you can use, and its wins depend on the specific problem. The savings arrive slowly, through manufacturers, not overnight.
- A new method called DA-EGO splits huge design puzzles (30 to 90 adjustable features) into smaller chunks instead of tackling them all at once.
- It matched or beat rivals on 21 test problems and improved real hardware designs, including a jet engine rotor blade and a compressor stage.
- Fewer expensive computer simulations means lower design costs, and the resulting parts could be more fuel-efficient, affecting flight and power prices later.
Why It Matters
Faster, cheaper engine design could mean more fuel-efficient jets and power plants, trimming fuel bills and emissions over time.