FlowExtract: Procedural Knowledge Extraction from Maintenance Flowcharts

2026-04-08

AI Summary

This study addresses the challenge of unlocking procedural knowledge embedded in static ISO 5807 standard flowcharts—commonly stored as PDFs or scanned images in manufacturing facilities—for use by modern intelligent systems. To this end, the authors propose FlowExtract, a structured knowledge extraction pipeline that decouples node detection from edge reconstruction. The approach leverages YOLOv8 for node detection and EasyOCR for text recognition, and introduces a novel integration of arrow orientation analysis with backward connection tracing to achieve high-precision edge relationship recovery. Experimental results on an industrial troubleshooting guide dataset demonstrate near-perfect node detection accuracy and significantly outperform vision-language model baselines in edge extraction. This work presents the first method capable of producing queryable, structured representations of procedural knowledge from industrial flowcharts.

Abstract

Maintenance procedures in manufacturing facilities are often documented as flowcharts in static PDFs or scanned images. They encode procedural knowledge essential for asset lifecycle management, yet inaccessible to modern operator support systems. Vision-language models, the dominant paradigm for image understanding, struggle to reconstruct connection topology from such diagrams. We present FlowExtract, a pipeline for extracting directed graphs from ISO 5807-standardized flowcharts. The system separates element detection from connectivity reconstruction, using YOLOv8 and EasyOCR for standard domain-aligned node detection and text extraction, combined with a novel edge detection method that analyzes arrowhead orientations and traces connecting lines backward to source nodes. Evaluated on industrial troubleshooting guides, FlowExtract achieves very high node detection and substantially outperforms vision-language model baselines on edge extraction, offering organizations a practical path toward queryable procedural knowledge representations. The implementation is available athttps://github.com/guille-gil/FlowExtract.

Problem

• procedural knowledge extraction
• maintenance flowcharts
• connection topology
• vision-language models
• ISO 5807

Innovation

• procedural knowledge extraction
• flowchart understanding
• directed graph reconstruction
• arrowhead-based edge detection
• YOLOv8 and EasyOCR integration