Text Mining System for Maintenance Record Classification

Solution Overview

Problem

Existing solutions for classifying maintenance records in the airline industry, such as those used in maintenance cost guarantee programs, rely on machine learning algorithms that require significant time and computing resources, making them inefficient for real-time classification.

Innovation Solution

A system that normalizes maintenance event information by converting engineering jargon into dictionary form, uses a weighted scoring system based on uniqueness scores to classify maintenance records without the need for machine learning, by accessing a database of authoritative engineering information indexed by classes, and labels the records accordingly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning algorithms are used to classify maintenance records, then classification accuracy can be improved, but time consumption and computing resource requirements increase significantly

Engineering Contradiction:

Improveclassification accuracyVSAvoidtime consumption

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts and utilizes structured classification codes (ATA codes) that are already embedded within the unstructured maintenance record text. By identifying and extracting these pre-existing classification markers, the system avoids the need for complex machine learning training while achieving accurate classification. The extraction process focuses on locating specific code patterns within the text rather than analyzing entire document semantics.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system creates a simplified representation of the maintenance record by copying and structuring the embedded ATA codes into a machine-readable format. This copying approach transforms unstructured text containing classification codes into a structured dataset that can be processed efficiently without requiring intensive computational resources for analysis.

Inventive Principle:

Principle #26Copying

2Measurement precision

If machine learning algorithms are used to classify maintenance records, then classification accuracy can be improved, but computing resource requirements increase significantly

Engineering Contradiction:

Improveclassification accuracyVSAvoidcomputing resource requirements

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent extracts and utilizes structured classification codes (ATA codes) that are already embedded within the unstructured maintenance record text. By identifying and extracting these pre-existing classification markers, the system avoids the need for complex machine learning training while achieving accurate classification. The extraction process focuses on locating specific code patterns within the text rather than analyzing entire document semantics.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system creates a simplified representation of the maintenance record by copying and structuring the embedded ATA codes into a machine-readable format. This copying approach transforms unstructured text containing classification codes into a structured dataset that can be processed efficiently without requiring intensive computational resources for analysis.

Inventive Principle:

Principle #26Copying

3Productivity

If traditional text analysis methods are used, then processing speed can be improved, but classification accuracy decreases

Engineering Contradiction:

Improveprocessing speedVSAvoidclassification accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent introduces structured ATA codes as an intermediary element between the unstructured maintenance record text and the classification system. These codes serve as explicit markers that bridge the gap between free-text descriptions and standardized classification categories, enabling both rapid processing and accurate classification without requiring complex analytical algorithms.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11954132B2Text mining electronic documents to classify a maintenance record for a machine

Publication Date: 2024.04.09 THE BOEING CO

AI summary

A method is provided for classifying a maintenance record for a machine. The method includes receiving the maintenance record including maintenance event information expressed as a collection of terms of interest, and normalizing the terms of interest to their non-abbreviated, dictionary form. The method includes accessing a database of authoritative engineering information from electronic documents, expressed textually and indexed by class of a plurality of classes. The method includes weighted scoring the authoritative engineering information that uses one or more of the terms of interest represented in the weighted scoring by uniqueness scores. Matching authoritative engineering information is determined for the maintenance event information based on the weighted scoring, and the maintenance record is labeled with the class to which the matching authoritative engineering information is indexed.