Triangulation-Based Anomaly Detection in 3D Printers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting anomalies in 3D printed objects, such as rule-based approaches and unsupervised machine learning models, are ineffective in real-time and often result in false positives, leading to material wastage and inefficiencies, as they fail to consider the cumulative effects of anomalies over time and lack causal explanations.
Innovation Solution
The triangulation-based anomaly detection system uses two or more machine learning models, such as an autoencoder-decoder and a time series model, to predict anomalies in real-time by cross-verifying predictions and identifying contributing factors, thereby reducing false positives and negatives and enhancing detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If rule-based approaches are used for anomaly detection, then detection speed is improved, but detection accuracy deteriorates due to false positives
Solution Approach 1:
The patent combines multiple machine learning models (autoencoder-decoder model and time series model) into a triangulation system that cross-verifies predictions. This merging of models resolves the contradiction by maintaining fast detection through automated processing while improving accuracy through cross-validation, reducing false positives that plague single-model or rule-based approaches.
Solution Approach 2:
The system implements feedback mechanisms where prediction results from multiple models are cross-verified and used to refine anomaly detection. The triangulation approach uses feedback from each model's predictions to validate and improve the overall detection accuracy, resolving the contradiction between speed and precision by allowing rapid automated processing with built-in validation.
2Measurement precision
If unsupervised machine learning models are used, then detection accuracy is improved, but false positives increase leading to material wastage
Solution Approach 1:
By merging multiple machine learning models in a triangulation system, the patent reduces false positives that cause material wastage. The cross-verification mechanism ensures that only anomalies consistently detected by multiple models trigger print job termination, thereby maintaining high detection accuracy while minimizing false alarms that would otherwise waste material.
Solution Approach 2:
The triangulation system acts as an intermediary layer between individual model predictions and final anomaly confirmation. This intermediary cross-verifies predictions before taking action, reducing false positives and preventing unnecessary material wastage while preserving the enhanced detection capability of machine learning models.
3Loss of information
If manual inspection by domain experts is performed, then causal explanation of anomalies is improved, but detection time increases causing lag
Solution Approach 1:
The patent replaces manual expert inspection with an automated triangulation system that uses multiple machine learning models to detect and explain anomalies. This substitution eliminates detection lag by automating the process while preserving causal explanation quality through the use of interpretable models and cross-verification that identifies root causes without requiring manual expert intervention.
4Device complexity
If single machine learning model is used, then model complexity is reduced, but prediction reliability deteriorates
Solution Approach 1:
The patent merges multiple machine learning models into a triangulation system where each model contributes to the overall prediction. This combination resolves the contradiction by maintaining relatively manageable individual model complexities while significantly improving prediction reliability through cross-validation and consensus-based anomaly detection.
Data Source
AI summary
Examples of systems for triangulation-based detection of anomaly in a print job performed by a three-dimensional printer are described herein. In an example, a data pertaining to a set of layers printed based on the print job of the 3D printer may be provided to two models to obtain respective predicted anomalies. Thereafter, the obtained predicted anomalies may be triangulated to detect an anomaly in the layer being printed by the 3D printer.