An LLMs-Based Environmental Adaptability Analysis Method for Product Reconfiguration

Abstract

A product's performance is subjected to the combined effects of multiple environmental factors during its service life. Conducting environmental adaptability analysis can provide critical information for product reconfiguration, thereby improving product performance and extending service life. Product repair records contain rich information about product failure caused by environmental factors, yet systematic methods for extracting and analyzing such information remain lacking. This paper proposes an environmental adaptability analysis method based on large language models (LLMs), constructing an agent-based work flow to realize the generation of product-level environmental adaptability reports from unstructured repair records. Specifically, Agent 1 transforms repair records into triplet consisting of object, symptom, and environment, completing the standardized structured expression of information, Agent 2 categorizes the environmental factors according to different products based on the defined environmental classification system (mechanical, electrical, natural, usage), Agent 3 conducts statistical and association analysis from two dimensions, namely overall environmental factor coverage rate and component environmental sensitivity to generate final analysis reports. Experimental results demonstrate that the workflow based on LLMs can effectively characterize the differentiated features of various product categories in terms of environmental adaptability and accurately identify weak components under the influence of different environmental factors.