Deep learning classifier for life cycle optimization of steel–concrete composite bridges

Abstract

The ability to conduct life cycle analyses of complex structures is vitally important for environmental and social considerations. Incorporating the life cycle into structural design optimization results in extended computational durations, underscoring the need for an innovative solution. This paper introduces a methodology leveraging deep learning to hasten structural constraint computations in an optimization context, considering the structure’s life cycle. Using a composite bridge composed of concrete and steel as a case study, the research delves into hyperparameter fine-tuning to craft a robust model that accelerates calculations. The optimal deep learning model is then integrated with three metaheuristics: the Old Bachelor Acceptance with a Mutation Operator (OBAMO), the Cuckoo Search (CS), and the Sine Cosine Algorithms (SCA). Results indicate a potential 50-fold increase in computational speed using the deep learning model in certain scenarios. A comprehensive comparison reveals economic feasibility, environmental ramifications, and social life cycle assessments, with an augmented steel yield strength observed in optimal design solutions for both environmental and social objective functions, highlighting the benefits of meshing deep learning with civil engineering design optimization.