Multi-fidelity probabilistic optimisation of composite structures under thermomechanical loading using Gaussian processes

A multi-fidelity probabilistic optimisation method for the design of composite structures subjected to thermomechanical loading is introduced in this work for the first time. The proposed multi-fidelity approach offers considerable computation efficiency as well as sufficient accuracy, enabling probabilistic optimisation to include more design variables in the early design phase. This approach incorporates both nonlinear information fusion algorithms and multi-level optimisation to achieve increased accuracy and computation time savings. In this optimisation process, a High-Fidelity Model (HFM) covers only a part of the entire design space with information collected uniformly while providing high-fidelity information of other design spaces sparsely without causing extra computational cost. Simultaneously, a Low-Fidelity Model (LFM) explores the whole design space to compensate lack of high-fidelity information. In this manner, the number of high-fidelity information to construct a multi-fidelity model is dramatically reduced. The Reliability-Based Design Optimisation (RBDO) demonstrated the proposed multi-fidelity method of a mono-stringer stiffened composite panel under thermomechanical loading using Gaussian Processes (GPs).