Dr Linwei He

AI-in-the-loop design optimization of building structures

Building sector accounted for near 40% of energy and process-related CO2 emissions in recent years. To ensure a sustainable built environment, it is vitally important to design buildings that are structurally efficient. Structural layout/topology optimization methods are powerful means of designing highly efficient building structures to tackle the climate crisis. Since the optimization methods are often based on mathematical derivations (e.g., gradient), they require clearly defined design inputs to function correctly. However, real-world scenarios are complicated, so building designs are subject to numerous factors (e.g., aesthetics) that are challenging to incorporate in well-posed mathematical formulations. Moreover, designers may wish to make frequent adjustments in the design loop, since it is not always feasible to foresee all design requirements beforehand. To address this, this project aims to develop a so-called AI-in-the-loop framework, taking advantage of the fast-growing AI technology to aid building optimization. In addition to mathematical optimization algorithms, the framework will integrate an AI-guided intervention step, which adjusts the building designs by, e.g., evolving structural forms and adding/removing design requirements. In this project, the mathematical elements will ensure highly efficient structural designs with significantly reduced CO2 emissions, while the AI elements will help eradicate the practical barriers to stimulate more widespread adoption of innovative buildings in industry. There is an opportunity in this project to develop software tools for industrial users from world-leading UK companies.

Concurrent design optimization of 3D printed truss-like structures and manufacturing configurations

Additive manufacturing (AM, or ‘3D’ printing) has developed rapidly in recent years and has the potential to enable fabrication of structurally optimized components that have hitherto been impossible to manufacture. There is a growing interest in 3D printed truss-like structures, not only because they permit light-weight mechanical components for aerospace and automotive applications, but also because they enable innovative building forms in civil and architectural engineering applications. To optimize efficient 3D printed structures, common AM issues (such as the overhang limitations) should be taken into account in the design process. Therefore, a concurrent design optimization approach is proposed, which simultaneously optimizes structural layouts/topologies and manufacturing configurations during the AM process. The aim of this project is to develop concurrent optimization algorithms that can generate (near-)optimum truss-like structures fabricated via modern AM processes, such as the direct energy deposition (DED) and wire arc additive manufacturing (WAAM) approaches. Preliminary research has been conducted and a holistic optimization method was proposed, which, however, focused primarily on 2D problems. This project will therefore extend the methods by optimizing 3D structures, taken from either aerospace/automotive engineering or civil/architectural engineering applications. There is also an opportunity in this project to fabricate physical models and conduct laboratory tests, so as to validate the design methods.

If you're interested in one of these projects, or would like to self-propose a project, please contact Linwei at the above email. Current PhD opportunities in the department can be viewed here.