Artificial Intelligence Augmented Compressor Design

Key objectives and aims of the research
To develop a procedure for fast automated data generation using computational fluid dynamics (CFD) to enable quick exploration of the compressor design space.
To use the procedure to generate data and validate the use of machine learning for extracting low order physical understanding in the design of axial compressors which can then be used to produce low order models for the design of compressors.
To formally evaluate the parameters that should be used in low order models in the preliminary design of compressors
To use these findings to develop a more general research approach using machine learning which speeds up the research and development process
The approach
A large data-set will be generated using computational simulations for a simple compressor design problem.
Human expertise and data analysis techniques will then be used to abstract interesting features from the flow-field and further statistical techniques then used to quantify which of these features may be important to the performance of axial compressors.
These results will be compared with current understanding of the simple problem to validate the ability of machine learning technique to identify new physical understanding.
These techniques will then be applied to a more complicated problem to try and learn new physics
The novel Physical sciences/engineering methodology that will be carried out during the research
The use of machine learning to gain new physical understanding and so augment the design process.
Making use of CFD to explore the design space for physical understanding. Although the absolute results from the CFD are not accurate enough for a predictive model the trends in CFD provide information about the underlying physics.