A mechanism and data dual-driven industrial production line root cause diagnosis method and system

Abstract

This invention discloses a mechanism- and data-driven root cause diagnosis method and system for industrial production lines. The method includes: acquiring and preprocessing multi-source heterogeneous industrial data from the production line; performing anomaly detection on the preprocessed multi-source heterogeneous industrial data, identifying abnormal time windows, and extracting multimodal anomaly data; constructing a three-level hybrid diagnostic model and generating a preliminary root cause candidate set based on the model's diagnostic output for the multimodal anomaly data; performing source tracing analysis on the preliminary root cause candidate set and generating a structured natural language diagnostic report based on the source tracing analysis results. This invention significantly improves the accuracy and generalization ability of root cause diagnosis across a wide range of scenarios by constructing a three-level hybrid model architecture. Furthermore, through source tracing analysis and natural language extension, it provides a structured report supported by clear physical logic and data evidence, thereby improving operational decision-making efficiency, reducing reliance on scarce domain experts, and significantly reducing unplanned downtime.

Images