Additive Manufacturing Diagnosis Using Physics-Assisted ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current additive manufacturing systems face challenges in diagnosing failed builds and identifying performance issues, requiring significant time and human labor, and struggle to efficiently determine the root cause of failures.
Innovation Solution
The implementation of a physics-assisted machine learning model that generates physics features from raw data, uses classifiers to diagnose components, and determines health status, enabling rapid diagnosis and enhanced accuracy by considering wear and tear effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual diagnosis by experts is used, then diagnostic accuracy can be achieved, but time consumption and labor requirements increase significantly
Solution Approach 1:
The patent replaces manual expert diagnosis with an automated machine learning system that analyzes sensor data from additive manufacturing devices. The system uses trained models to automatically identify failure modes and diagnose issues, eliminating the need for human experts to manually examine and diagnose each failure case, thus reducing time consumption while maintaining diagnostic accuracy.
Solution Approach 2:
The patent introduces sensor data as an intermediary between the additive manufacturing device and the diagnosis process. Sensors continuously collect operational parameters (temperature, pressure, vibration, etc.) that serve as objective indicators of device health, allowing the machine learning system to diagnose issues based on quantitative data rather than relying on expert subjective assessment.
2Measurement precision
If comprehensive analysis of all parameters is performed to identify root cause, then diagnostic accuracy improves, but computational complexity and time increase
Solution Approach 1:
The patent performs preliminary actions by pre-training machine learning models using historical sensor data and known failure cases. During actual operation, the pre-trained models can quickly diagnose issues without requiring complex real-time computations. The system also pre-identifies which sensor parameters are most relevant for detecting specific failure modes, reducing the computational burden during diagnosis.
Solution Approach 2:
The patent segments the diagnostic process into multiple independent analysis modules, each responsible for detecting specific failure modes (e.g., electron beam issues, powder bed problems, build chamber anomalies). This segmentation allows the system to analyze different parameters in parallel using specialized models, reducing overall computational complexity compared to a single comprehensive analysis system.
Data Source
AI summary
A system for diagnosing an additive manufacturing device is provided. The system includes a first module configured to: obtain one or more parameters for a digital twin of a component of the additive manufacturing device based on raw data from the component of the additive manufacturing device; and generate physics features for the digital twin of the component of the additive manufacturing device based on the one or more parameters and one or more transfer functions, a second module configured to obtain one or more classifiers for classifying the component as a first condition or a second condition based on physics features; and a third module configured to: determine a health of the component based on the generated physics features of the first model and the one or more classifiers.