A method for melt pool state monitoring in laser-based direct energy deposition based on DenseNet

Yuan, Junlin; Liu, Hanru; Liu, Weiwei; Martínek, Radek; Peng, Shitong

doi:10.1016/j.measurement.2022.111146

Cited by 18 publications

(4 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Due to the amount of data required to effectively train a CNN [27], it is proposed to use the transfer learning technique [18]. In transfer learning, a CNN is pre-trained in a very large dataset of images uncorrelated to the images under study.…”

Section: Hybrid Machine Learning Modelmentioning

confidence: 99%

“…Recently, there has been some research inclined towards an establishment of melt pool monitoring through imaging [17,18] with the aid of artificial intelligence methods [7,15]. They can also be linked to the rapid growth and improvement of image processing techniques, especially those using modern ML techniques, such as Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNNs).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A hybrid machine learning model for in-process estimation of printing distance in laser Directed Energy Deposition

Ribeiro

Núñez²,

Venter³

et al. 2023

Preprint

View full text Add to dashboard Cite

There are several parameters that highly influence material quality and printed shape in laser Directed Energy Deposition (L-DED) operations. These parameters are usually defined for an optimal combination of energy input (laser power, scanning speed) and material feed rate, providing ideal bead geometry and layer height to the printing setup. However, during printing, layer height can vary. Such variation affects the upcoming layers by changing the printing distance, inducing printing to occur in defocus zone then cumulatively increasing shape deviation. In order to address such issue, this paper proposes a novel intelligent hybrid method for in-process estimating the printing distance ( $Z_s$ ) from melt pool images acquired during L-DED. The proposed hybrid method uses transfer learning to combine pre-trained Convolutional Neural Network (CNN) and Support Vector Regression (SVR) for an accurate yet computationally fast methodology. A dataset with $2,700$ melt pool images was generated from the deposition of lines, at $60$ different values of $Z_s$, and used for training. The best hybrid algorithm trained performed with a Mean Average Error (MAE) of $0.266$ , which indicates an average target error of $6.7%$ . The deployment of this algorithm in an application dataset allowed the printing distance to be estimated and the final part geometry to be inferred from the data. Thus, the present method can aid on-line feedback control on the Z-axis increment, to regulate layer height, improving 3D shape geometry in L-DED.

show abstract

Section: Hybrid Machine Learning Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A hybrid machine learning model for in-process estimation of printing distance in laser Directed Energy Deposition

Ribeiro

Núñez²,

Venter³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Their work examined the possibility of keyholing porosity and balling instabilities, and the in-situ signatures, which may indicate flaws, were identified by observed melt pools. Junlin Yuan et al [17] developed new machine learning algorithms and verified the accuracy by a multi-template in-situ experiment. Their algorithms achieve higher accuracy than a traditional convolutional neural network.…”

Section: Introductionmentioning

confidence: 99%

Numerical simulation on melt pool and solidification in the direct energy deposition process of GH3536 powder superalloy

Liu

Liu²,

Li³

et al. 2023

Preprint

View full text Add to dashboard Cite

In the direct energy deposition (DED) process, the highly energetic laser, rapid melting, and solidification processes lead to complex heat transfer and flow phenomena. A three-dimensional finite element model (FEM) is established to study the effect of process parameters on the melt pool and solidification quality during the DED process. The heat transfer, fluid flow, and solidification in the DED process of the GH3536 superalloy are studied. By investigating the effects of laser power, scanning speed, and feed rate on the morphology of melt pool and interlayer fusion, the appropriate input parameters for GH3536 are obtained. Temperature gradient and solidification rate obtained in transient thermal distribution are applied to predict the quality and morphology of the solidified structure at the cut-off point. Results show that high laser power and low scanning speed or feed rate will enlarge the melt pool. Well-solidified microstructure frequently appears in the middle of the parameter set; focusing on the enlargement of the melt pool is not the best strategy. The correlation between feed rate and laser power is not obvious. The minimum threshold for scanning speed is found at a given feed rate. When the scanning rate is below the threshold, abnormal morphology of the melt pool and irregular solidification structures will occur. The laser power and scanning speed range suitable for the GH3536 superalloy are summarized, and the undesirable and possibly fluctuating parameters are marked. The middle part of the parameter set is recommended for the feed rate.

show abstract

“…Many novel defect detection techniques have emerged with the development of artificial intelligence technology for WAAM process. Yuan et al (2022) proposed a method to detect defects in the melt pool based on the DenseNet39 model, which has an accuracy of 99.3%, but its detection speed is less than 1 FPS. Hossain et al proposed a method for detecting defects in the melt pool based on a machine learning approach, and the model was able to detect many types of defect, but it could not detect defects timely (Renh et al , 2021).…”

Section: Introductionmentioning

confidence: 99%

An in situ surface defect detection method based on improved you only look once algorithm for wire and arc additive manufacturing

Huang

et al. 2022

RPJ

View full text Add to dashboard Cite

Purpose Wire and arc additive manufacturing (WAAM) is a widely used advanced manufacturing technology. If the surface defects occurred during welding process cannot be detected and repaired in time, it will form the internal defects. To address this problem, this study aims to develop an in situ monitoring system for the welding process with a high-dynamic range imaging (HDR) melt pool camera. Design/methodology/approach An improved you only look once version 3 (YOLOv3) model was proposed for online surface defects detection and classification. In this paper, improvements were mainly made in the bounding box clustering algorithm, bounding box loss function, classification loss function and network structure. Findings The results showed that the improved model outperforms the Faster regions with convolutional neural network features, single shot multibox detector, RetinaNet and YOLOv3 models with mAP value of 98.0% and a recognition rate of 59 frames per second. And it was indicated that the improved YOLOv3 model satisfied the requirements of real-time monitoring well in both efficiency and accuracy. Originality/value Experimental results show that the improved YOLOv3 model can solve the problem of poor performance of traditional defect detection models and other deep learning models. And the proposed model can meet the requirements of WAAM quality monitoring.

show abstract

A method for melt pool state monitoring in laser-based direct energy deposition based on DenseNet

Cited by 18 publications

References 43 publications

A hybrid machine learning model for in-process estimation of printing distance in laser Directed Energy Deposition

A hybrid machine learning model for in-process estimation of printing distance in laser Directed Energy Deposition

Numerical simulation on melt pool and solidification in the direct energy deposition process of GH3536 powder superalloy

An in situ surface defect detection method based on improved you only look once algorithm for wire and arc additive manufacturing

Contact Info

Product

Resources

About