Manufacturing Problem Inference via Snapshot Similarity

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for detecting emerging problems in product manufacturing are inadequate due to their global nature, reliance on manual filtering, sensitivity to data semantics shifts, requirement for labeled data, and inability to detect issues in non-textual data formats, leading to false negatives and false positives, especially in complex manufacturing processes.

Innovation Solution

A computer-implemented method that analyzes a time-ordered set of snapshots using similarity thresholds and neural network-based similarity signatures to identify trends in product manufacturing, allowing detection of emerging problems without prior knowledge of their type or instrumentation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If trend detection methods are used to detect global changes in events, then the method can identify overall patterns, but events representative of emerging problems correspond to too small a portion of global volume and are difficult to distinguish

Engineering Contradiction:
Improvedetection precisionVSAvoidevent volume
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent segments the event data by grouping events according to their similarity to a reference event, creating distinct clusters that separate emerging problems from global trends. This allows precise detection of small subsets of relevant events without being overwhelmed by global volume.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary similarity measure that acts as a mediator between the reference event and candidate events. This similarity metric enables precise identification of emerging problems by comparing event characteristics without requiring manual filtering or predefined categories.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If weak signal detection based on explicit semantic analysis is used, then event categories can be extracted, but exploring all possible combinations of filters has exponential cost and requires manual filtering

Engineering Contradiction:
Improveproblem detection accuracyVSAvoidfiltering complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent enables the system to automatically identify emerging problems through similarity-based clustering without requiring manual filter configuration. The method self-adjusts by comparing events to a reference event and automatically grouping similar events, eliminating the need for exponential filter exploration.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent changes the approach from discrete filter combinations to a continuous similarity measure. By using similarity thresholds rather than binary filter combinations, the system reduces complexity from exponential to polynomial time while maintaining detection precision.

Inventive Principle:
Principle #35Parameter changes

3Quantity of substance

If sentiment analysis is used to detect quality issues, then user sentiment can be extracted, but sentiment is a generic signal that cannot discriminate inherently different non-quality events and is not actionable

Engineering Contradiction:
Improvesignal coverageVSAvoidproblem discrimination accuracy
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

Instead of starting with generic sentiment labels and trying to classify events, the patent inverts the approach by starting with specific event characteristics and measuring similarity to a reference event. This allows precise discrimination of different problem types while maintaining comprehensive signal coverage.

Inventive Principle:
Principle #13The other way round (Inversion)

Solution Approach 2:

The patent replaces the mechanical sentiment analysis system with a similarity-based comparison system. This substitution enables precise problem discrimination by comparing event attributes directly rather than relying on generic sentiment labels that cannot distinguish between different types of issues.

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

4Measurement precision

If anomaly detection is used to monitor data for outliers, then isolated occurrences can be detected, but these methods are less relevant for detecting recurrent issues from root causes in design or manufacturing

Engineering Contradiction:
Improveanomaly detection accuracyVSAvoidproblem type coverage
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal detection method that works for both isolated anomalies and recurrent issues by using similarity-based clustering. The method can detect emerging problems regardless of whether they are single occurrences or repeated patterns, making it adaptable to various problem types from design or manufacturing root causes.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4290322B1Inference of emerging problems in product manufacturing
Publication Date: 2026.04.01 DASSAULT SYSTEMES SA
  • EP4290322B1 patent drawingFigure 1
  • EP4290322B1 patent drawingFigure 2
  • EP4290322B1 patent drawingFigure 3

AI summary

The disclosure notably relates to a computer-implemented method for inferring an emerging problem in product manufacturing. The method comprises obtaining a time-ordered set comprising one or more snapshots of a product and one or more similarity thresholds. The method also comprises obtaining at least one recent snapshot, the at least one recent snapshot being time-ordered after at least one snapshot of the time-ordered set. The method also comprises retrieving a subset of one or more snapshots from the time-ordered set, the one or more snapshots being time-ordered before the at least one recent snapshot and satisfying, with respect to the at least one recent snapshot, a similarity above at least one of the one or more similarity thresholds. The method also comprises determining a trend from the retrieved subset and a baseline, the trend being a time distribution of the snapshots of the retrieved subset with respect to the baseline.