Volume Image Defect Detection for Additive Manufacturing Irregularities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting process irregularities in additive manufacturing are either destructive or difficult to interpret, especially when dealing with large and expensive objects, and non-destructive methods like CT scans provide unclear insights into manufacturing parameters.
Innovation Solution
A method and system that train a detection system using volume image data to identify process irregularities by comparing image data from a reference object with deliberately induced defects to other areas, allowing non-destructive detection of mechanical defects based on characteristic data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If destructive tests are performed to evaluate mechanical properties, then reliable quality assessment is achieved, but the tested objects are destroyed
Solution Approach 1:
The patent creates virtual copies of physical objects through 3D scanning to generate digital twins. These digital models allow for virtual testing and analysis without damaging the actual objects. The system captures geometric data, material properties, and process parameters to create comprehensive digital replicas that can be analyzed repeatedly without loss to the physical object.
Solution Approach 2:
The patent replaces physical destructive testing with computational analysis of digital models. Instead of applying mechanical loads to physical specimens until failure, the system uses software to simulate stress, strain, and failure modes on virtual copies, eliminating the need for physical destruction while maintaining assessment reliability.
2Reliability
If CT scanning is used for non-destructive inspection, then object integrity is maintained, but the images are difficult to interpret and provide unclear insights
Solution Approach 1:
The patent merges multiple data sources including 3D scan data, CT scan images, manufacturing process parameters, and material properties into a unified digital twin. This integration allows the system to correlate imaging data with process conditions and geometric context, making the inspection results much more interpretable and actionable.
Solution Approach 2:
The patent introduces software algorithms and processing intermediaries that automatically analyze and interpret raw CT scan images. The system uses image processing techniques to highlight defects, compare them against acceptance criteria, and provide clear interpretations rather than raw, difficult-to-interpret images.
3Measurement precision
If detailed analysis of manufactured objects is performed to detect process irregularities, then detection precision is improved, but the complexity of the inspection system increases
Solution Approach 1:
The patent performs preliminary actions by capturing manufacturing process data and creating digital twins during or immediately after the manufacturing process. This advance preparation organizes and structures data before inspection is needed, enabling detailed analysis without requiring complex real-time inspection systems.
Solution Approach 2:
The patent creates a multi-functional digital twin system that serves multiple purposes: geometric verification, material property storage, process parameter archiving, defect detection, and quality assessment. This single comprehensive digital model replaces what would otherwise require multiple separate inspection systems and procedures.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A method of training a detection system is able to acquire volume image data in an additively manufactured object for the detection of process irregularities, and comprises the steps of: a) receiving process irregularity data referring to a selected location within an additively manufactured reference object in which selected location a predefined process irregularity occurred during the additive manufacture of the object, b) acquiring volume image data of a volume of the reference object comprising at least the selected location by said detection system, c) identifying within the volume image data characteristic data which represent a difference between the volume image data of the selected location in comparison with the volume image data of at least one other location of the reference object and/or of a number of other additively manufactured objects in which no process irregularity has occurred and/or no process irregularity is suspected, d) assigning to the predefined process irregularity the characteristic data as a representative of the predefined process irregularity.