Software tool will help engineers design jet engines

Purdue University researchers have created a software tool that is more than 100 times faster than other programs used by engineers to improve jet engine designs

The software analyzes engine models and quickly extracts information that indicates whether the design is mechanically sound, said Mario Rotea, a professor in Purdue’s School of Aeronautics and Astronautics.

Considering the complex inner workings of a jet engine, software aimed at predicting how well a new design will function can be cumbersome and time-consuming. Jet engines house numerous rotating disks containing blades. The mechanical properties of these blades are difficult to predict because they change as they wear and because no two blades are perfectly identical: they emerge from manufacturing with minute variations in geometric shape and mechanical properties.

“But even tiny variations can lead to drastic changes in vibration levels, compromising engine performance and reliability,” Rotea said.

Conventional software aimed at evaluating the mechanical properties of blades can take weeks or longer to predict how well the “bladed disks” will work.

However, time is money in industry, and for efficiency’s sake engineers cannot afford to wait weeks for a program to crunch data.

“If it takes a month to give you the answer, that’s not very practical,” Rotea said. “What we developed was a technique that is much more intelligent.”

Rotea presented new findings about the software tool in July during the 38th Joint Propulsion Conference and Exhibit in Indianapolis and also during the 15th World Congress on Automatic Control in Barcelona, Spain.

Engine designers use computer models to test designs before actually building an engine. The models predict how the multitude of critical engine parts will react to factors such as wear or damage and manufacturing variations.

“That’s because it’s less expensive to use models,” Rotea said. “Industry has the ability to develop good models to analyze the vibratory responses of bladed disks. But these models are very complicated. They contain lots of unpredictable parameters, and what was lacking, in my opinion, was a good method to analyze those models to help them extract the numbers they need to say, ’This is a good design or this is a bad design.’”

The software he developed with former graduate student Fernando D’Amato is based on an “optimization algorithm,” which is a step-by-step procedure for solving a mathematical problem. This algorithm calculates the worst-case vibration level of the blades due to variations in mechanical properties.

A model of the bladed disk and the range of the possible variations are required to run this algorithm. Although it is difficult to predict exactly which variations a specific blade will have, engineers know what the range is.

“Some parameters change during the life of the engine,” Rotea said. “For example, you have blades that get nicked or they wear and they change mechanical properties.”

“How do you incorporate that into the model? You cannot predict all those variations that the engine will see in the field. But if you know the ranges for these parameter variations, you can determine the worst-case effect the parameter changes will have in the blade stress and vibration levels without actually searching through all possibilities.

“What we did was to develop an optimization algorithm that calculates the things they want much more efficiently,” he said. “We use optimization not to do the design but to actually predict the worst-case behavior over a known range of parameters.”

The algorithm analyzes one blade, or a small group of blades, and deduces the worst-case vibration level of any blade in the disk.

The time it takes to calculate the worst-case vibration level grows with the number of parameters that must be considered. The optimization algorithm that was developed by Rotea and D’Amato is about one and a half times faster for each parameter, compared with more conventional software tools.

“The more parameters there are, the more time is saved, Rotea said. “That means bigger problems are solved even more efficiently than smaller ones. If there are 60 parameters, the time savings is very large.”

He estimates that, for the average job, the tool is more than 100 times faster than other tools on the market.

The work has been funded by the National Science Foundation and private industry.

Writer: Emil Venere, (765) 494-4709, venere@purdue.edu

Source: Mario Rotea, (765) 494-6212, rotea@ecn.purdue.edu

Purdue News Service: (765) 494-2096; purduenews@purdue.edu