Designing Intelligent Mechanical Systems through Generative AI and Prompt Engineering

Abstract

Mechanical system design often fails when optimization methods are applied in conditions that change rapidly or involve nonlinear interactions. Techniques that perform well in static environments are unable to manage uncertainty or high-dimensional trade spaces, and the result is frequent delays in engineering progress and restricted exploration of alternatives. Generative modeling has been introduced to widen the design search process by creating multiple structural options and accelerating early-stage evaluation. While this approach can uncover solutions that conventional methods overlook, it also produces designs that lack physical feasibility when guidance is absent. The present study develops a framework that links generative modeling with prompt engineering to ensure that model outputs reflect engineering requirements. The framework contains three stages: data preparation, generative modeling, and structured prompt design. Prompts are used to state design objectives, apply boundary conditions, and reference prior information. Iterative adjustment of prompts provides feedback that directs the generation of valid and practical solutions. Evaluation of the framework indicates three principal contributions. Efficiency improves through the rapid generation of multiple alternatives and faster convergence to acceptable options. Adaptability is achieved by revising prompts to match new requirements without retraining the underlying models. Interpretability increases because each prompt establishes a direct and reviewable link between engineering inputs and system outputs. Challenges remain in data quality, careful construction of prompts, and the need for validation through physics-based models. The study offers a systematic approach that enhances reliable generative design and broadens opportunities for innovation in intelligent mechanical systems.