Abstract:Reliably manufacturing defect free products is still an open challenge for Laser Powder Bed Fusion processes. Particularly, pores that occur frequently have a negative impact on mechanical properties like fatigue performance. Therefore, an accurate localisation of pores is mandatory for quality assurance, but requires time-consuming post-processing steps like computer tomography scans. Although existing solutions using in-situ monitoring data can detect pore occurrence within a layer, they are limited in their localisation precision. Therefore, we propose a pore localisation approach that estimates their position within a single layer using a Gaussian kernel density estimation. This allows segmentation models to learn the correlation between in-situ monitoring data and the derived probability distribution of pore occurrence. Within our experiments, we compare the prediction performance of different segmentation models depending on machine parameter configuration and geometry features. From our results, we conclude that our approach allows a precise localisation of pores that requires minimal data preprocessing. Our research extends the literature by providing a foundation for more precise pore detection systems.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to accurately locate the pore positions in defects during the laser powder bed fusion (PBF - LB/M) process. Specifically, the goal of the paper is to achieve high - precision prediction of the pore positions within a single layer by improving the existing in - situ monitoring data processing methods, thereby enhancing product quality and the reliability of the production process. ### Problem Background Laser powder bed fusion (PBF - LB/M) is a widely - used additive manufacturing technology, but its repeatability and product quality consistency still face challenges. In particular, during the manufacturing process, the frequent occurrence of pores can have a negative impact on mechanical properties (such as fatigue strength and lifespan). Therefore, accurately locating pores is crucial for quality assurance. However, traditional detection methods (such as computed tomography) are time - consuming and costly, and while existing methods based on in - situ monitoring data can detect the presence of pores, they are lacking in positioning accuracy. ### Core Problems of the Paper The paper aims to improve the positioning accuracy of pores within a single layer by introducing a new pore - positioning method, using a segmentation model combined with Gaussian Kernel Density Estimation (KDE). Specifically, the paper attempts to solve the following problems: 1. **Improving Pore - Positioning Accuracy**: Existing methods can usually only simplify pores into an overall quantity (such as porosity value) and are unable to accurately position pores within a single layer. The paper proposes a new method that improves positioning accuracy by estimating the probability of pore occurrence at each pixel position instead of directly locating the specific position of the pores. 2. **Utilizing Multiple Sensor Data**: The paper combines camera data from two different bands, visible light and near - infrared light, to more comprehensively capture pore information. This method avoids the limitations of single - sensor data and reduces the need for complex data conversion. 3. **Reducing the Influence of Man - Induced Defects**: To ensure the generalization ability of the model, the paper avoids the practice of artificially inducing defects through specific parameter configurations or geometric designs, and instead uses data under actual production conditions for training and validation. ### Solutions The paper proposes a method based on Gaussian Kernel Density Estimation, transforming the pore - positioning problem into a probability - estimation problem. The specific steps are as follows: - **Data Pre - processing**: Obtain the true positions of pores through CT scans and use KDE to generate the pore probability distribution maps for each layer. - **Model Training**: Use different segmentation models (such as Unet, FPN, LinkNet, DeepLabV3, MAnet) to train the model so that it can learn the pore probability distribution from in - situ monitoring data. - **Performance Evaluation**: Verify the prediction performance of different model architectures under different geometric features and process parameters through experiments, and analyze the key factors affecting model performance. ### Conclusions The research results of the paper show that the proposed method can achieve relatively high pore - positioning accuracy within a single layer while reducing the need for complex data pre - processing. In addition, the paper also discusses the directions for future research, including improving label - generation methods, exploring other probability - distribution estimators, and further optimizing the model architecture. In summary, this paper solves the problem of insufficient pore - positioning accuracy in the PBF - LB/M process through an innovative method, providing new ideas and technical means for improving quality control in additive manufacturing.

Estimating Pore Location of PBF-LB/M Processes with Segmentation Models

Simultaneous Pore Detection and Morphological Features Extraction in Laser Powder Bed Fusion with Image Processing

High-Precision Real-Time Pores Detection in LPBF using Thermal Energy Density (TED) Signals

Automated porosity segmentation in laser powder bed fusion part using computed tomography: a validity study

Predicting Laser Powder Bed Fusion Defects Through In-Process Monitoring Data and Machine Learning

Attenuation-adjusted deep learning of pore defects in 2D radiographs of additive manufacturing powders

Predicting defects in laser powder bed fusion using in-situ thermal imaging data and machine learning

In-situ visual monitoring for multi-scale defects detection in laser powder bed fusion

A physics-informed machine learning model for porosity analysis in laser powder bed fusion additive manufacturing

Defect-based analysis of the laser powder bed fusion process using X-ray data

In-situ porosity prediction in metal powder bed fusion additive manufacturing using spectral emissions: a prior-guided machine learning approach

In-Situ Characterization of Pore Formation Dynamics in Pulsed Wave Laser Powder Bed Fusion

A Pore-Throat Segmentation Method Based on Local Hydraulic Resistance Equivalence for Pore-Network Modeling

Machine learning–aided real-time detection of keyhole pore generation in laser powder bed fusion

Analytical prediction of keyhole porosity in laser powder bed fusion

An Integrated Fuzzy Logic and Machine Learning Platform for Porosity Detection using Optical Tomography Imaging during Laser Powder Bed Fusion

Quantification of Interdependent Dynamics during Laser Additive Manufacturing Using X‐Ray Imaging Informed Multi‐Physics and Multiphase Simulation

Concurrent modeling of porosity and microstructure in multilayer three-dimensional simulations of powder-bed fusion additive manufacturing of INCONEL 718

A hybrid deep learning model of process-build interactions in additive manufacturing

Encoding Stability into Laser Powder Bed Fusion Monitoring Using Temporal Features and Pore Density Modelling

Formation mechanism of pore defects and surface ripples under different process parameters via laser powder bed fusion by numerical simulation and experimental verification