Additive Manufacturing Defect Detection From Spatial Process Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Additive manufacturing processes often result in geometric deviations and defects such as pores, cavities, cracks, and delaminations due to thermal effects, which affect the mechanical properties of the manufactured components, and existing methods struggle to accurately detect these defects in real-time during the manufacturing process.
Innovation Solution
A method utilizing machine learning algorithms trained on initial process data from n objects to identify defects by correlating defect coordinates with spatially resolved process data, enabling non-destructive detection of defects during or after production, and adjusting manufacturing parameters to prevent or repair defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If additive manufacturing processes are used to produce objects, then manufacturing flexibility and complexity are improved, but geometric deviations and defects occur due to thermal effects
Solution Approach 1:
The patent implements a feedback mechanism by capturing process data during additive manufacturing, training a machine learning model to predict defects, and using these predictions to adjust manufacturing parameters in real-time. This closed-loop system continuously improves geometric accuracy by learning from actual process outcomes and applying corrections to subsequent manufacturing operations.
Solution Approach 2:
The patent applies preliminary action by training a machine learning model with process data from initial manufacturing runs to predict defects before they occur in subsequent production. The model learns from historical process data and measurement results, enabling proactive defect prevention through parameter optimization before actual manufacturing defects manifest.
2Measurement precision
If traditional defect detection methods are used after manufacturing, then defect identification is achieved, but real-time detection during the manufacturing process is not possible
Solution Approach 1:
The system captures process data during manufacturing and uses a trained machine learning model to predict defects in real-time, providing immediate feedback that enables detection during the manufacturing process rather than only after completion. This allows for timely intervention and parameter adjustment.
Solution Approach 2:
The patent replaces traditional post-manufacturing inspection methods with a computational approach using machine learning algorithms that analyze process data in real-time. This substitution enables defect prediction during manufacturing by leveraging patterns learned from historical data, eliminating the time delay associated with separate inspection steps.
3Loss of time
If machine learning algorithms are trained on initial process data to detect defects, then real-time defect detection is enabled, but additional processing time and computational resources are required
Solution Approach 1:
The patent performs preliminary action by training the machine learning model in advance using process data from initial manufacturing runs. This pre-training phase enables the model to quickly predict defects during subsequent manufacturing operations without requiring complex real-time computations, thus reducing operational processing time while managing computational complexity through offline preparation.
4Loss of substance
If defects are detected early in the manufacturing process, then time and material are saved by rejecting defective objects early, but the capability to detect and prevent defects must be established first
Solution Approach 1:
The patent implements feedback by using measurement data from initial defects to train a machine learning model that predicts defects in real-time during manufacturing. This enables early detection and prevention, allowing for timely rejection or correction of defective objects before significant material waste occurs, while the system complexity is justified by the substantial material savings.
Data Source
Figure 1
Figure 2
AI summary
The invention relates to a computer-implemented method (100) for determining defects of an object produced using an additive manufacturing process, wherein volume elements or spatial coordinates of the object to be produced are processed during the additive manufacturing process, wherein the method (100) has the following steps of: determining (102) spatially resolved first process data relating to n objects, which are captured during an additive manufacturing process for producing the n objects, wherein the first process data define a process data coordinate system for each of the n objects, determining (104) measurement data relating to the n objects by means of imaging non-destructive or destructive methods after the n objects have been produced, wherein the measurement data define, for each of the n objects, an object representation in a measurement data coordinate system, determining (106) which coordinates of at least one section of the measurement data coordinate system are defect coordinates which are assigned to a defect in the object representation; correlating (108) the at least one section of the measurement data coordinate system comprising the determined defect coordinates with a corresponding section of the process data coordinate system in order to collect training data, training (110) an adaptive algorithm for determining defect coordinates in spatially resolved process data, which have been captured during an additive manufacturing process for producing an object, by means of the training data, determining (112) spatially resolved second process data which are captured during an additive manufacturing process for producing an object, and analysing (114) the second process data for defects by means of the adaptive algorithm. This provides a method (100) which allows potential defects of an object to be directly detected from process data which arise during additive production.