Data-driven multidisciplinary and multi-objective optimization with a cooperative constraint-handling mechanism

As multidisciplinary design optimization (MDO) problems grow in complexity, modern challenges increasingly involve multiple objectives and constraints, often requiring expensive black-box evaluations. These characteristics pose significant difficulties for existing MDO algorithms. This article develops a data-driven multidisciplinary and multi-objective optimization algorithm tailored to tackle such problems. The algorithm integrates data-driven multidisciplinary design (DDMD) and multi-objective optimization (DDMO). The DDMO incorporates a cooperative constraint-handling mechanism that analyses the correlation between objectives and constraints, simplifying MDO problems. A dynamic tasking optimization strategy identifies promising solutions and efficiently explores the design space. To enhance computational efficiency, DDMD employs a data-driven multidisciplinary feasible method to accelerate decoupling and improve system-solving efficiency. The algorithm's effectiveness is validated through comparisons with five state-of-the-art methods across constrained multi-objective problems, six MDO problems and seven MDO engineering problems. Its applicability is further demonstrated by solving an MDO problem involving unmanned underwater vehicles, integrating fluid dynamics and structural mechanics.