Information Processing for Defect Analysis via Dimension Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data analysis methods struggle with analyzing large datasets containing multiple defect forms, especially when defects have slight variations or combinations, leading to inefficiencies and inaccurate defect identification.
Innovation Solution
An information processing apparatus that compresses the dimensions of multi-dimensional objective variables using techniques like PCA, clustering, and NMF, and calculates influence degrees using multitasking Lasso, enabling efficient and accurate defect analysis by reducing the number of dimensions and utilizing a user interface for parameter input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual analysis methods are used for defect identification, then analysis accuracy can be maintained for simple cases, but processing time becomes excessively long for enormous datasets
Solution Approach 1:
The patent replaces manual mechanical analysis with automated computer-based image processing and machine learning algorithms. The system uses automated defect detection algorithms to process images, substituting human manual inspection with computational methods that can analyze enormous datasets rapidly while maintaining consistent accuracy.
Solution Approach 2:
The patent transforms defect detection from qualitative manual assessment to quantitative automated measurement by extracting multiple features (area, perimeter, circularity, texture, etc.) from defect images. These numerical parameters enable systematic computer-based analysis that is both rapid and accurate, resolving the contradiction between speed and precision.
2Adaptability or versatility
If multiple reference images covering all defect forms are prepared, then comprehensive defect detection is possible, but the complexity of the analysis system increases significantly
Solution Approach 1:
The patent segments defect analysis into multiple independent feature extraction components (area, perimeter, circularity, texture, shape features) rather than relying on complete reference images. Each feature is extracted and analyzed separately, allowing the system to handle diverse defect forms without requiring comprehensive reference libraries for every possible defect type.
Solution Approach 2:
The patent creates a universal defect detection framework that uses multiple extracted features to characterize all defect types. Instead of requiring specific reference images for each defect form, the system uses a unified approach where combination of extracted features identifies different defect types, reducing system complexity while maintaining versatility.
3Productivity
If traditional dimensionality reduction methods are used, then computational load is reduced, but information loss occurs affecting analysis accuracy
Solution Approach 1:
The patent applies Principal Component Analysis (PCA) to transform the original feature space into a new coordinate system where the first few principal components capture the majority of variance in the data. This dimensional transformation reduces computational load while preserving essential defect characteristics, as the principal components maintain the most significant information from the original high-dimensional feature set.
Data Source
AI summary
An information processing apparatus has an objective variable acquirer configured to acquire a multi-dimensional objective variable, an objective variable dimension compressor configured to compress the number of dimensions of the objective variable, an explanatory variable acquirer configured to acquire an explanatory variable, and an influence degree calculator configured to set at least one of a basis characterizing the objective variable and a coefficient weighting the basis as a new objective variable and calculate an influence degree on the new objective variable by using the explanatory variable.