Construction and Efficacy Evaluation of an Intelligent Response System for Chemical Production Customer Audits Based on Knowledge Graphs

Abstract

Chemical production customer audits face intractable challenges, including heterogeneous audit standards, inefficient manual responses, and inadequate handling of complex cross-standard issues. Traditional manual response models have struggled to meet the evolving demands of high-stakes supply chain audits. This study integrates 236 heterogeneous audit standards from 127 core customers—including industry leaders such as Contemporary Amperex Technology Co. Limited (CATL) and Tesla—to construct a ternary knowledge graph (TKG) centered on “process parameters-quality indicators-compliance clauses.” An NLP-driven intelligent response system was developed to enable rapid semantic understanding, precise knowledge retrieval, and standardized response generation for audit queries. Comprehensive validation, including laboratory testing and 12 months of industrial application, demonstrates that the system achieves a question matching accuracy of 91.3%, reduces response time from 48 hours (traditional manual) to 15 minutes, and supports 27 customer audits with a 100% pass rate. The complex issue resolution rate reaches 89.6%, significantly reducing enterprise audit costs and compliance risks. The proposed technical framework effectively addresses the core pain points of multi-customer heterogeneous standard integration and intelligent audit response, providing a replicable technical pathway for audit management in the chemical industry and offering practical insights for the application of knowledge graphs and NLP in industrial compliance scenarios.