Melt Pool Radiation Prediction for Reliable AM Deviation Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting process deviations in melting processes, particularly in powder bed-based AM systems, face challenges in accurately determining radiation intensity and wavelength due to location-dependent emission variability, leading to insufficient accuracy in quality monitoring.
Innovation Solution
A method involving a machine learning model, specifically a neural network, is used to predict the radiation intensity and wavelength of the melt pool by providing historical power profiles and associated data as input variables, allowing for precise determination of target values for process glow, which improves detection accuracy by using a trained model to set a more reliable target value for process monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If statistical deviation methods are used to detect process anomalies, then the detection method is simple to implement, but the detection accuracy is insufficient due to high variance in the signal
Solution Approach 1:
The patent transforms the detection approach by changing from statistical deviation analysis to direct comparison with physics-based target values. The target values are calculated using fundamental physical parameters (laser power, scan speed, hatch spacing, material properties) through a deterministic model, replacing the statistical method that relied on signal variance. This parameter transformation enables accurate detection without being affected by the high variance inherent in optical signals from melt pools.
2Device complexity
If no target value for spatially dependent emission is available, then the measurement system is simpler, but the detection of process deviations becomes unreliable
Solution Approach 1:
The patent implements preliminary calculation of target values based on process parameters before actual measurement and comparison. The target values representing expected radiation intensity and wavelength are pre-computed using a physics-based model that incorporates laser power, scan speed, hatch spacing, and material properties. This preliminary determination of reference values enables reliable deviation detection by providing a deterministic baseline against which actual measurements can be compared, eliminating the need for complex spatial calibration while ensuring measurement reliability.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the accuracy of process deviation detection, reduces computational time, and allows for efficient mapping of nonlinearities, leading to improved reliability in identifying and responding to process abnormalities by narrowing the detection limits and providing a precise setpoint for process regulation.
Implementation Method 1
detection of process light (emission in the visible and near-infrared range) is increasingly being used for process monitoring
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In summary, the invention relates to a method for determining the radiation intensity and/or wavelength of a process glow (E10), a method for determining process deviations (DEV) of a melting process, and a method for controlling a melting process. To improve the determination of the radiation intensity and/or wavelength of a process glow in a melting process, the following steps are proposed: - Providing the power profile (P) for a segment (dtPOI) of the path (x) as an input (IN) for a machine learning model (MLM), in particular for a neural network, wherein the model (MLM) is equipped with historical and/or synthetic power profiles (PH) and associated historical ortrained on synthetic radiation intensities (E10H) and/or process luminescence wavelengths (E10), and - determining the radiation intensity and/or process luminescence wavelength (E10) as the output (OUT) of the model (MLM).