Failure Knowledge Extraction for Industrial Plant Deviation Response
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for extracting failure knowledge in industrial plants require manual curation of knowledge from diverse sources, leading to time consumption, human error, and loss of important information due to the cognitive demands of reconciling knowledge across different document types.
Innovation Solution
A system and method that automatically extracts failure knowledge from diverse information sources using optical character recognition, named entity recognition, and machine learning models to create a failure scenario knowledge repository, integrating process and equipment knowledge for real-time decision support.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If manual curation of knowledge from diverse sources is performed, then comprehensive failure knowledge can be extracted, but time consumption increases significantly
Solution Approach 1:
The patent replaces manual mechanical curation processes with automated computational systems including optical character recognition (OCR) for document scanning, natural language processing (NLP) for text analysis, and machine learning models for knowledge extraction. This substitution dramatically reduces time consumption while maintaining comprehensive knowledge extraction from diverse sources such as FMEA documents, FTA reports, HAZOP analyses, and incident reports.
Solution Approach 2:
The system enables self-service automated knowledge curation where the computational system independently processes diverse document sources, extracts failure knowledge, and structures it without requiring continuous human intervention. The automated system serves itself by performing OCR, text processing, knowledge extraction, and database population tasks autonomously, freeing operators from time-consuming manual curation work.
2Loss of information
If manual curation of knowledge from diverse sources is performed, then knowledge can be extracted, but human error occurs in the curated knowledge
Solution Approach 1:
The patent replaces human manual curation with automated computational processing systems that use OCR for accurate text recognition, NLP for systematic text analysis, and machine learning models for consistent knowledge extraction. These automated systems eliminate human errors such as transcription mistakes, inconsistent formatting, and subjective interpretation variations, thereby improving the reliability and accuracy of the curated failure knowledge.
Solution Approach 2:
The system incorporates feedback mechanisms where extracted knowledge is validated against multiple source documents and cross-referenced for consistency. The automated system provides feedback loops that verify extraction accuracy, reconcile conflicting information from different sources, and ensure the curated knowledge maintains high reliability before being stored in the database.
3Loss of information
If manual curation of knowledge from diverse sources is performed, then knowledge can be extracted, but cognitive demands lead to loss of important information
Solution Approach 1:
The patent replaces cognitively demanding manual analysis with automated computational systems that systematically process diverse document types including text reports, diagrams, and scanned documents. The system uses OCR to convert images to text, NLP to analyze unstructured text, and machine learning models to extract and structure failure knowledge, thereby eliminating cognitive overload while ensuring complete retention of important information from all source materials.
Solution Approach 2:
The system segments the complex task of knowledge curation into distinct automated processing stages: document ingestion and OCR conversion, text preprocessing and entity recognition, knowledge extraction using machine learning models, and structured storage in database. This segmentation of the complex cognitive task into manageable automated steps ensures comprehensive information retention without overwhelming cognitive demands.
4Loss of time
If automated extraction using OCR and machine learning is implemented, then time consumption is reduced, but the system complexity increases
Solution Approach 1:
The patent implements automated OCR and machine learning-based extraction systems that process diverse document sources simultaneously and continuously, reducing knowledge curation time from manual days to automated minutes or hours. Although this increases system complexity by introducing computational components, the automation enables parallel processing of multiple documents and eliminates sequential manual work, achieving net time savings despite the added technological complexity.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Failure analysis of industrial plants are stored in various types of documents associated with industrial plant. The documents are used by operators of plant to address any deviation that is active in plant. The operator generally has prior knowledge of relevant processes, equipment and sensors described in a deviation scenario in these documents. However, a system that is envisaged to aid operator in real-time does not have this information readily available as this knowledge is spread across documents. Currently available systems manually curate failure knowledge thereby making the process time consuming and prone to human errors. Present disclosure provides method and system for performing extracting and collating failure knowledge from diverse sources in industrial plant. The system automatically extracts failure knowledge present in text documents using trained models and links it with process, equipment, sensor relationships that are present in piping and instrumentation diagram to create failure scenario knowledge repository.