A Domain Adaptation of Large Language Models for Classifying Mechanical Assembly Components
Fatemeh Elhambakhsh, Daniele Grandi, Hyunwoong Ko
TL;DR
The paper tackles the scarcity of structured functional data in mechanical design by proposing a supervised domain-adaptation framework that fine-tunes a domain-general LLM (GPT-3.5 Turbo) on domain-specific data from the Oregon State Design Repository to classify functions of mechanical assembly parts. The method leverages labeled domain data $D_{T_{ ext{labeled}}}$, function class definitions $D_L$, and textual descriptions $D_{L_{ ext{dis}}}$ to guide learning and uses hyperparameter optimization over $E$ (epochs), $B$ (batch size), and $LR$ (learning rate multiplier) with a cross-entropy loss $L$. Evaluations on OSDR show that domain-adapted models outperform pre-trained GPT models, and when applied to the ABC CAD dataset, achieve about 99% accuracy within eight predefined function classes, with only ~1% misclassification. This work enhances early-phase design reasoning by generating high-quality, function-labeled data that semantically enriches the design representation and supports more effective exploration of conceptual designs.
Abstract
The conceptual design phase represents a critical early stage in the product development process, where designers generate potential solutions that meet predefined design specifications based on functional requirements. Functional modeling, a foundational aspect of this phase, enables designers to reason about product functions before specific structural details are determined. A widely adopted approach to functional modeling is the Function-Behavior-Structure (FBS) framework, which supports the transformation of functional intent into behavioral and structural descriptions. However, the effectiveness of function-based design is often hindered by the lack of well-structured and comprehensive functional data. This scarcity can negatively impact early design decision-making and hinder the development of accurate behavioral models. Recent advances in Large Language Models (LLMs), such as those based on GPT architectures, offer a promising avenue to address this gap. LLMs have demonstrated significant capabilities in language understanding and natural language processing (NLP), making them suitable for automated classification tasks. This study proposes a novel LLM-based domain adaptation (DA) framework using fine-tuning for the automated classification of mechanical assembly parts' functions. By fine-tuning LLMs on domain-specific datasets, the traditionally manual and subjective process of function annotation can be improved in both accuracy and consistency. A case study demonstrates fine-tuning GPT-3.5 Turbo on data from the Oregon State Design Repository (OSDR), and evaluation on the A Big CAD (ABC) dataset shows that the domain-adapted LLM can generate high-quality functional data, enhancing the semantic representation of mechanical parts and supporting more effective design exploration in early-phase engineering.