An agent construction method for alloy material reverse design

Abstract

The application relates to the technical field of material intelligent design, and particularly discloses an agent construction method for alloy material reverse design, which comprises the following steps: constructing a paper library; classifying the intention of user questioning to generate a search keyword; searching the paper library based on the search keyword, returning multiple papers, and screening a specified number of papers; structurally disassembling corresponding text segments associated with multi-modal key information of the screened papers in a thinking chain mode, aligning cross-literature knowledge, and outputting aggregated structured design knowledge; taking the aggregated structured design knowledge as a basis, taking context experience prior as a constraint, taking historical design cases as a reference, outputting multiple groups of differentiated candidate schemes through a large language model; and adjusting the candidate schemes according to user feedback information on the design results. The application can break away from the dependence on large-scale data labeling and complex model training, and output traceable schemes in a low-cost and high-efficiency mode.

Images