AI Decision Support for System Malfunction Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing systems for analyzing dysfunctions in complex systems, such as trains, are difficult to maintain due to the complexity of the equipment and the impossibility of stopping traffic outside of reduced time slots, leading to challenges in determining the correct corrective actions.

Innovation Solution

A process for analyzing system dysfunctions that includes an initialization phase where maintenance information is received to form a database and decision trees are obtained, and an operating phase where maintenance agents perform operations based on decision trees, send information for database updates, and apply learning techniques to optimize decision trees.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If traditional documentation methods (paper or scanned electronic form) are used for maintenance procedures, then information is available for reference, but maintenance agents cannot find appropriate corrective actions when breakdowns differ from documented cases

Engineering Contradiction:
Improveavailability of maintenance informationVSAvoidease of finding corrective action
Core Design Contradiction:
Loss of informationVSEase of operation

Solution Approach 1:

The patent replaces traditional static documentation systems (paper or scanned forms) with an intelligent software-based decision support system that uses artificial intelligence and machine learning to dynamically generate and update maintenance procedures, transforming the mechanical lookup process into an intelligent adaptive system

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service by allowing maintenance agents to independently query the AI-based decision support tool, which automatically analyzes the breakdown situation and provides appropriate corrective actions without requiring external expert intervention or manual search through documentation

Inventive Principle:
Principle #25Self-service

2Quantity of substance

If complex systems like trains are maintained with detailed documentation, then comprehensive maintenance information is available, but the complexity of equipment makes determination of breakdown and corrective action difficult

Engineering Contradiction:
Improveamount of maintenance informationVSAvoidcomplexity of equipment
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent introduces an AI-based decision support system as an intermediary between the complex equipment and the maintenance agent. This intermediary processes the complexity by analyzing equipment data, comparing it against learned patterns from historical maintenance information, and presenting simplified diagnostic conclusions and corrective actions to the maintenance agent

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system transforms the unmanageable complexity of equipment parameters into structured, analyzable data by changing how maintenance information is represented - from static documentation to dynamic AI-processed data that adapts to specific breakdown situations, enabling efficient analysis despite equipment complexity

Inventive Principle:
Principle #35Parameter changes

3Stability of the object's composition

If maintenance procedures are documented in advance, then standardized processes are available, but the procedures become outdated when new breakdown patterns emerge

Engineering Contradiction:
Improvestability of maintenance proceduresVSAvoidadaptability to new breakdown patterns
Core Design Contradiction:
Stability of the object's compositionVSAdaptability or versatility

Solution Approach 1:

The patent transforms static, fixed maintenance procedures into dynamic procedures that continuously adapt. The AI-based decision support system learns from new maintenance data and breakdown patterns, automatically updating its knowledge base and decision logic to reflect current equipment conditions and emerging failure modes while maintaining procedural stability through systematic learning

Inventive Principle:
Principle #15Dynamics

4Reliability

If more maintenance information is collected and stored, then better decision support is available, but the time and resources required to manage the information increase

Engineering Contradiction:
Improvequality of maintenance decisionsVSAvoidtime to manage information
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system performs automated information management through self-service mechanisms - the AI automatically processes, analyzes, and updates maintenance information without requiring manual curation. Maintenance agents simply query the system with breakdown data, and the AI handles the complex tasks of information retrieval, analysis, and procedure generation, eliminating manual information management time

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces manual information management processes with automated AI-based processing. Instead of maintenance agents manually searching, filtering, and analyzing maintenance information, the AI system automatically processes the information using machine learning algorithms, significantly reducing the time and human resources required to manage maintenance data

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP3853784B1Method for analysing malfunctions of a system and associated devices
Publication Date: 2025.05.14 HITACHI RAIL GTS FRANCE SAS
  • EP3853784B1 patent drawingFigure 1
  • EP3853784B1 patent drawingFigure 2~3

AI summary

The invention relates to a method for analysing malfunctions of a system, comprising an initialisation phase and an operating phase comprising the following steps: - providing a maintenance agent of a set of decision trees and a database, performing by the agent of a sequence of maintenance operations on the system according to a part of the decision trees, - sending information relating to the sequence of maintenance operations performed, - updating the databases using the information sent, and - updating the set of selected decision trees by applying a learning technique applied to the information sent, the update being carried out by a data processing unit of a computer platform (10), the operating phase being repeated for a plurality of maintenance operation sequences.