ABSTRACT The aerospace industry employs brazed ceramic–metal joints. The component's reliable performance requires an apt composite material combination. An inverse design‐based artificial intelligence (AI) model may be used for this purpose. This study investigates the stability, sensitivity, and performance of AI by implementing a multi‐output Autoencoder (AE). The model can assist in selecting the constituent or composite material properties (coefficient of thermal expansion/Young's modulus) for a given performance parameter/target (average Von Mises Stress VMS). Stability and sensitivity analysis are performed by running 500 Monte Carlo (MC) simulations with 5% Gaussian noise injection. A classical model, namely, the Genetic Algorithm (GA), is implemented for performance comparison of AI and classical method. The MC simulations showed that the AE model converges after 200 iterations, is stable up to 5% noise level, and predictions are sensitive to VMS and the effective coefficient of thermal expansion input parameters. For a given VMS of 91.07 MPa, Al 2 O 3 ceramic, InCuSiI–ABA braze, and Kovar metal are the composite materials selected from the AE model. The AE model's predictive performance is higher (3.84% average percentage error APE) than the classical GA model (22.65% APE). This study will assist the engineering industry by accelerating Virtual Prototyping for composite material selection.
Khod et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: