Automated Defect Detection Interface with Human Feedback Loop

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current automated inspection techniques for images captured by borescopes, such as those used in aircraft engine blade inspection, are prone to errors due to human inattention and struggle to detect defects outside pre-defined classes, limiting their effectiveness in identifying all types of blade damage.

Innovation Solution

An automated defect detection system that uses an image capture device to transmit data to a monitoring and analysis site for automated analysis, employing Robust Principal Component Analysis (PCA) and a classifier to identify defects, with human inspector feedback refining the system's accuracy and storing results in a database for future reference and training.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If automated inspection techniques categorize defects into pre-defined classes, then common defects can be detected, but defects outside those classes are not detected

Engineering Contradiction:
Improvedefect detection accuracyVSAvoiddefect class coverage
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system incorporates feedback loops where automated analysis results are reviewed by human inspectors, and their corrections are fed back to refine the automated analysis algorithms. This continuous feedback mechanism enables the system to adapt to new defect types and improve detection accuracy over time without being limited by pre-defined classes.

Inventive Principle:
Principle #23Feedback

2Reliability

If human inspectors review images for defect detection, then defect interpretation can be performed, but errors result from human inattention

Engineering Contradiction:
Improvedefect detection accuracyVSAvoidinspection time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The inspection process is segmented into two stages: automated analysis for initial defect identification and filtering, followed by human inspector review only for potential defects. This segmentation reduces the time burden on human inspectors while maintaining high detection accuracy through the combination of automated efficiency and human expertise.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The automated analysis system acts as an intermediary between the raw images and human inspectors, pre-processing and filtering images to identify potential defects. This intermediary role eliminates the need for human inspectors to review every image individually, reducing time loss while maintaining reliability.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Productivity

If automated analysis is performed without human feedback, then processing speed increases, but accuracy improves through continuous training

Engineering Contradiction:
Improveinspection throughputVSAvoiddefect detection precision
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

Human inspector feedback on automated analysis results is systematically collected and used to retrain and refine the automated analysis algorithms. This feedback loop enables continuous improvement of detection precision while maintaining high productivity, as the system learns from human expertise without requiring manual review of every image.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP2776815B1System and method for improving automated defect detection user interface
Publication Date: 2019.11.27 UNITED TECH CORP
  • EP2776815B1 patent drawingFigure 1
  • EP2776815B1 patent drawingFigure 2

AI summary

A system and method for improving human-machine interface while performing automated defect detection is disclosed. The system and method may include an image capture device for capturing and transmitting data of an object, performing automated analysis of the data and reviewing results of the automated analysis by a human inspector and providing feedback. The system and method may further include refining the automated analysis of the data based upon the feedback of the human inspector.