Annotation

Annotation: The problem of maintaining the relevance of the electronic structure of a product in a dynamically changing environment of a manufacturing enterprise is considered: replacement of components, changes in specifications, the emergence of new suppliers and the launch of design notices of changes. A method is proposed for updating the electronic structure of a product based on events generated by an enterprise resource management system (ERP system) using the "generate and verify" approach based on large language models (LLM). The key element of the method is the "generate and verify" cycle: a large language model generates the proposed change in the form of an intermediate representation of the electronic structure of the product that satisfies a context-free grammar, after which the intermediate representation is verified for compliance with the constraints of referential integrity and data type tolerance. An algorithm for processing ERP events of three classes has been developed: structural, attributive, and referential. A software implementation in Python based on an eventoriented approach is described. The proposed algorithm is integrated into the architecture of the product's electronic structure synthesis system, expanding its capabilities at the stage of product maintenance in production and operation. The algorithm has been experimentally tested on a model database with scenarios corresponding to various classes of events generated by the enterprise resource management system (ERP system).