Metamodel-based design optimization in industrial turbomachinery

Fans and Blowers community is experiencing, during those years, an incredible push in rethinking design approaches and strategies. The change in regulations on minimum efficiency grades and market requirements on even more customized products demand a changing in the way design in fan technology is perceived. In this context, even if synthetic approaches for fan design and analysis are still valuable tools, they need to be flanked by metamodels in order to overcome the limitations and criticism introduced by empirical relationships developed in the past for specific applications. In addition, by replacing computation-intensive functions with approximate surrogate models, it is possible to adopt advanced and nested optimization methods, such as those based on Evolutionary Algorithms, drastically improving the overall optimization computational time. Surrogate-based Optimizations based on Evolutionary Algorithm should become common practice in design optimization because of their capability of find optima in the design space, thanks to their intrinsic balance between exploitation and exploration. This work proposes methods for interweave elements of metamodeling techniques and multi-objective optimization problems with the synthetic approaches classically developed by the turbomachinery community. The entire Thesis can be ideally divided into two parts; the first gives a brief survey on the classical fan design and analysis approaches and reports two synthetic in-house codes for axial fan performance prediction. The second part present the state-of-the-art in metamodeling and optimization techniques, underlining the role of metamodeling in supporting design optimization and focusing in the more reliable and accurate framework for multi-objective optimization in fans engineering design.