Multi-Layer SVM Anomaly Detection for Manufactured Parts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection systems for manufactured parts are inadequate as they often require user-determined parameters, are unsuitable for different parts, and struggle with limited training data and varying patterns, making it difficult to detect anomalies accurately.
Innovation Solution
An anomaly analysis apparatus using multi-layer support vector machines (SVMs) that objectively select important features through domain-specific filtering and ensemble optimization, employing bootstrapped cross-validation and large margin linear discriminant optimization to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional machine learning models are used with user-determined parameters, then the system can detect anomalies in manufactured parts, but the determination result is often unsuitable for different manufactured parts due to subjective parameter selection
Solution Approach 1:
The system employs automated feature selection algorithms that objectively identify important features without user intervention. The algorithm autonomously determines which features are most relevant for anomaly detection, eliminating the need for subjective user parameter determination and ensuring consistent performance across different manufactured parts.
Solution Approach 2:
The system dynamically adjusts feature selection parameters based on the specific characteristics of each manufactured part. By changing the parameters according to the particular application, the system maintains high detection accuracy across diverse part types while adapting to their unique features.
2Device complexity
If analysis is performed only for some portions or features of the manufactured part, then the model training is simplified, but the obtained model is unsuitable for other manufactured parts
Solution Approach 1:
The system creates a universal feature selection framework that identifies features applicable across multiple manufactured parts. By selecting features that have broad relevance rather than part-specific features, the trained model can be generalized to detect anomalies in different manufactured parts without requiring retraining.
Solution Approach 2:
The system performs preliminary feature selection and analysis on a diverse set of manufactured parts during the training phase. This preliminary action ensures that the selected features are representative and applicable to various part types, enabling the model to generalize effectively to unseen parts.
3Measurement precision
If domain-specific filtering and ensemble optimization are used to objectively select features, then feature selection accuracy is improved, but the computational complexity and processing time increase
Solution Approach 1:
The feature selection process is divided into multiple stages: initial feature filtering, intermediate selection, and final optimization. This segmentation allows the system to progressively refine feature selection without overwhelming computational burden at any single stage, balancing accuracy with processing time efficiency.
4Reliability
If bootstrapped cross-validation is used to minimize overfitting, then classification accuracy is improved, but the computational resources and training time required increase significantly
Solution Approach 1:
The system applies bootstrapped cross-validation selectively to the most critical features and model parameters rather than exhaustively validating all aspects. This partial application achieves sufficient accuracy improvement while limiting the additional computational resources required.
Data Source
AI summary
A classifier network has at least two distinct sets of refined data, wherein the first two sets of refined data are sets of numbers representing the features values data received from sensors or a manufactured part. Performing, via at least two distinct types of support vector machines using an associated feature selection process for each classifier independently in a first layer, anomaly detection on the manufactured part. Then, using the stored data including refined data of at least two different types of data transforms and performing, via at least a two distinct types of support vector machines in a second layer, an associated feature selection process for each classifier independently. Forming at least four distinct compound classifier types for anomaly detection on the part using the stored data or coefficients. The ensemble of second layer support vector machine outputs compare the results to determine the presence of an anomaly.