A deep neural network for classification of melt-pool images in metal additive manufacturing

2018 ◽  
Vol 31 (2) ◽  
pp. 375-386 ◽  
Author(s):  
Ohyung Kwon ◽  
Hyung Giun Kim ◽  
Min Ji Ham ◽  
Wonrae Kim ◽  
Gun-Hee Kim ◽  
...  
Keyword(s):  
Neural Network ◽  
Additive Manufacturing ◽  
Deep Neural Network ◽  
Metal Additive Manufacturing ◽  
Melt Pool ◽  
Metal Additive

Melt pool sensing and size analysis in laser powder-bed metal additive manufacturing

2018 ◽  
Vol 32 ◽  
pp. 744-753 ◽  
Author(s):  
Bo Cheng ◽  
James Lydon ◽  
Kenneth Cooper ◽  
Vernon Cole ◽  
Paul Northrop ◽  
...  
Keyword(s):  
Additive Manufacturing ◽  
Size Analysis ◽  
Powder Bed ◽  
Metal Additive Manufacturing ◽  
Melt Pool ◽  
Metal Additive

Effect of Time Spacing and Hatch Spacing on Thermal Material Properties and Melt Pool Geometry in Additive Manufacturing of S316L

2019 ◽  
Author(s):  
Elham Mirkoohi ◽  
Daniel E. Sievers ◽  
Steven Y. Liang
Keyword(s):  
Heat Transfer ◽  
Additive Manufacturing ◽  
Material Properties ◽  
Metal Additive Manufacturing ◽  
Melt Pool ◽  
Experimental Values ◽  
Melt Pool Size ◽  
Transfer Mechanisms ◽  
Hatch Spacing ◽  
Metal Additive

Abstract A physics-based analytical solution is proposed in order to investigate the effect of hatch spacing and time spacing (which is the time delay between two consecutive irradiations) on thermal material properties and melt pool geometry in metal additive manufacturing processes. A three-dimensional moving point heat source approach is used in order to predict the thermal behavior of the material in additive manufacturing process. The thermal material properties are considered to be temperature dependent since the existence of the steep temperature gradient has a substantial influence on the magnitude of the thermal conductivity and specific heat, and as a result, it has an influence on the heat transfer mechanisms. Moreover, the melting/solidification phase change is considered using the modified heat capacity since it has an influence on melt pool geometry. The proposed analytical model also considers the multi-layer aspect of metal additive manufacturing since the thermal interaction of the successive layers has an influence on heat transfer mechanisms. Temperature modeling in metal additive manufacturing is one of the most important predictions since the presence of the temperature gradient inside the build part affect the melt pool size and geometry, thermal stress, residual stress, and part distortion. In this paper, the effect of time spacing and hatch spacing on thermal material properties and melt pool geometry is investigated. Both factors are found statistically significant with regard to their influence on thermal material properties and melt pool geometry. The predicted melt pool size is compared to experimental values from independent reports. Good agreement is achieved between the proposed physics-based analytical model and experimental values.

scholarly journals Metal Additive Manufacturing Parts Inspection Using Convolutional Neural Network

2020 ◽  
Vol 10 (2) ◽  
pp. 545 ◽  
Author(s):  
Wenyuan Cui ◽  
Yunlu Zhang ◽  
Xinchang Zhang ◽  
Lan Li ◽  
Frank Liou
Keyword(s):  
Neural Network ◽  
Additive Manufacturing ◽  
Convolutional Neural Network ◽  
Data Augmentation ◽  
Quality Inspection ◽  
Weight Decay ◽  
Metal Additive Manufacturing ◽  
Low Efficiency ◽  
The Impact ◽  
Metal Additive

Metal additive manufacturing (AM) is gaining increasing attention from academia and industry due to its unique advantages compared to the traditional manufacturing process. Parts quality inspection is playing a crucial role in the AM industry, which can be adopted for product improvement. However, the traditional inspection process has relied on manual recognition, which could suffer from low efficiency and potential bias. This study presented a convolutional neural network (CNN) approach toward robust AM quality inspection, such as good quality, crack, gas porosity, and lack of fusion. To obtain the appropriate model, experiments were performed on a series of architectures. Moreover, data augmentation was adopted to deal with data scarcity. L2 regularization (weight decay) and dropout were applied to avoid overfitting. The impact of each strategy was evaluated. The final CNN model achieved an accuracy of 92.1%, and it took 8.01 milliseconds to recognize one image. The CNN model presented here can help in automatic defect recognition in the AM industry.

scholarly journals Uncertainty quantification and reduction in metal additive manufacturing

2020 ◽  
Vol 6 (1) ◽  
Author(s):  
Zhuo Wang ◽  
Chen Jiang ◽  
Pengwei Liu ◽  
Wenhua Yang ◽  
Ying Zhao ◽  
...  
Keyword(s):  
Quality Control ◽  
Additive Manufacturing ◽  
Computational Model ◽  
Uncertainty Quantification ◽  
High Throughput ◽  
Surrogate Model ◽  
Metal Additive Manufacturing ◽  
Uncertainty Sources ◽  
Melt Pool ◽  
Metal Additive

AbstractUncertainty quantification (UQ) in metal additive manufacturing (AM) has attracted tremendous interest in order to dramatically improve product reliability. Model-based UQ, which relies on the validity of a computational model, has been widely explored as a potential substitute for the time-consuming and expensive UQ solely based on experiments. However, its adoption in the practical AM process requires overcoming two main challenges: (1) the inaccurate knowledge of uncertainty sources and (2) the intrinsic uncertainty associated with the computational model. Here, we propose a data-driven framework to tackle these two challenges by combining high throughput physical/surrogate model simulations and the AM-Bench experimental data from the National Institute of Standards and Technology (NIST). We first construct a surrogate model, based on high throughput physical simulations, for predicting the three-dimensional (3D) melt pool geometry and its uncertainty with respect to AM parameters and uncertainty sources. We then employ a sequential Bayesian calibration method to perform experimental parameter calibration and model correction to significantly improve the validity of the 3D melt pool surrogate model. The application of the calibrated melt pool model to UQ of the porosity level, an important quality factor, of AM parts, demonstrates its potential use in AM quality control. The proposed UQ framework can be generally applicable to different AM processes, representing a significant advance toward physics-based quality control of AM products.

Convolutional neural network-based inspection of metal additive manufacturing parts

2019 ◽  
Vol 25 (3) ◽  
pp. 530-540 ◽  
Author(s):  
Binbin Zhang ◽  
Prakhar Jaiswal ◽  
Rahul Rai ◽  
Paul Guerrier ◽  
George Baggs
Keyword(s):  
Neural Network ◽  
Additive Manufacturing ◽  
Convolutional Neural Network ◽  
Classification Accuracy ◽  
Online Monitoring ◽  
Quality Inspection ◽  
Content Type ◽  
Metal Additive Manufacturing ◽  
Part Quality ◽  
Metal Additive

Purpose Part quality inspection is playing a critical role in the metal additive manufacturing (AM) industry. It produces a part quality analysis report which can be adopted to further improve the overall part quality. However, the part quality inspection process puts heavy reliance on the engineer’s background and experience. This manual process suffers from both low efficiency and potential errors and, therefore, cannot meet the requirement of real-time detection. The purpose of this paper is to look into a deep neural network, Convolutional Neural Network (CNN), towards a robust method for online monitoring of AM parts. Design/methodology/approach The proposed online monitoring method relies on a deep CNN that takes a real metal AM part’s images as inputs and the part quality categories as network outputs. The authors validate the efficacy of the proposed methodology by recognizing the “beautiful-weld” category from material CoCrMo top surface images. The images of “beautiful-weld” parts that show even hatch lines and appropriate overlaps indicate a good quality of an AM part. Findings The classification accuracy of the developed method using limited information of a small local block of an image is 82 per cent. The classification accuracy using the full image and the ensemble of model outputs is 100 per cent. Originality/value A real-world data set of high resolution images of ASTM F75 I CoCrMo-based three-dimensional printed parts (Top surface images with magnification 63×) annotated with categories labels. Development of a CNN-based classification model for the supervised learning task of recognizing a “beautiful-weld” AM parts. The classification accuracy using the full image and the ensemble of model outputs is 100 per cent.

Sign in / Sign up

Export Citation Format

Share Document