A gas turbine fault experience diagnosis method and system based on a large language model and a knowledge graph

Abstract

The application discloses a gas turbine fault experience diagnosis method based on a large language model and a knowledge graph, and comprises the following steps: S1, collecting gas turbine operation abnormal phenomena, training and constructing a vertical domain large model; S2, performing semantic analysis on the natural language description and generating a structured abnormal feature representation; S3, constructing a gas turbine fault knowledge graph for computer retrieval and reasoning; S4, inputting candidate fault modes and their associated entity relationships as external knowledge constraints into the large language model; S5, the large language model performs multi-step reasoning analysis under the knowledge constraint condition, and obtains an auxiliary diagnosis result; and S6, outputting the auxiliary diagnosis result in the form of natural language. The application organizes the gas turbine field expert experience in the form of a knowledge graph, and combines the natural language understanding and reasoning capability of the large language model, so that the accuracy, practicality and engineering interpretability of the gas turbine fault diagnosis are improved.

Images