Monitoring of friction stir welding based on vision system coupled with Machine learning algorithm

Sudhagar, S.; Sakthivel, M.; Ganeshkumar, P.

doi:10.1016/j.measurement.2019.05.018

Cited by 64 publications

(19 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While the above efforts used deep learning with a CNN, the machine learning can also be done in simpler ways such as using a support vector machine [236]. Literature search uncovered three publications using such conventional machine learning methods in FSW [237][238][239] but none of them used deep learning approaches.…”

Section: Machine Learning the University Of Kentucky Weldingmentioning

confidence: 99%

Advanced Welding Manufacturing: A Brief Analysis and Review of Challenges and Solutions

Zhang

Yang

Wei

et al. 2020

Journal of Manufacturing Science and Engineering

View full text Add to dashboard Cite

Welding is a major manufacturing process that joins two or more pieces of materials together through heating/mixing them followed by cooling/solidification. The goal of welding manufacturing is to join materials together to meet service requirements at lowest costs. Advanced welding manufacturing is to use scientific methods to realize this goal. This paper views advanced welding manufacturing as a three step approach: (1) pre-design that selects process and joint design based on available processes (properties, capabilities, and costs); (2) design that uses models to predict the result from a given set of welding parameters and minimizes a cost function for optimizing the welding parameters; (3) real-time sensing and control that overcome the deviations of welding conditions from their nominal ones used in optimizing the welding parameters by adjusting the welding parameters based on such real-time sensing and feedback control. The paper analyzes how these three steps depend on process properties/capabilities, process innovations, predictive models, numerical models for fluid dynamics, numerical models for structures, real-time sensing, and dynamic control. The paper also identifies the challenges in obtaining ideal solutions and reviews/analyzes the existing efforts toward better solutions. Special attention and analysis have been given to (1) gas tungsten arc welding (GTAW) and gas metal arc welding (GMAW) as benchmark processes for penetration and materials filling; (2) keyhole plasma arc welding (PAW), keyhole-tungsten inert gas (K-TIG), and keyhole laser welding as improved/capable penetrative processes; (3) friction stir welding (FSW) ......

show abstract

Section: Machine Learning the University Of Kentucky Weldingmentioning

confidence: 99%

Advanced Welding Manufacturing: A Brief Analysis and Review of Challenges and Solutions

Zhang

Yang

Wei

et al. 2020

Journal of Manufacturing Science and Engineering

View full text Add to dashboard Cite

show abstract

“…So, the machine vision method has attracted researchers' focus. Reference [31] proposed a method to monitor the friction stir welding surface quality by surface image, which used Maximally Stable Extremal Regions method to detect the blob and express the flaw, at the detected weld joint. Liu et al [32] proposed an image analysis method to detect the defect at addictive manufacturing.…”

Section: B Product Quality Monitormentioning

confidence: 99%

A Product Quality Monitor Model With the Digital Twin Model and the Stacked Auto Encoder

et al. 2020

View full text Add to dashboard Cite

With the development of intelligent manufacturing and computer science, the system of equipment in the workshop has become more and more complex. In the intricate environment, the state of device changes constantly, which could affect the accuracy of methods since they cannot adapt the changing context. Recently, Digital Twin (DT) has received great focus among academic world and industrial world, which provides a new normal form for solving problems. In this paper, the structure of DT is discussed and a DT and Stacked Auto Encoder (SAE) Based Model is proposed to monitor the product quality. Based on the classical structure of DT, the digital model of DT is further divided into two parts, a task-achieved part and a self-update part. The former that comprises an encoder network that is a part of SAE and an Artificial Neural Network (ANN)-based classifier could check whether products are qualified. And a decoder network, another part of SAE, and a parameters-update rule make up the self-update part that could detect the accuracy of the task-achieved part and retrain the neural networks as the accuracy is poor. Furthermore, a new loss function is put forward as a training criterion in order to magnify the tiny difference between input data and result. In order to emulate the changing environment, the experimental data are collected at two different points in time. The data are then input to the proposed model and two other traditional methods to test the ability of accuracy and the adaptability for changing context. The comparisons show that the proposed method has got improvements, especially in where the effect of working environment is significant.

show abstract

“…[3][4][5][6][7] More recently, the in-line process monitoring and artificial intelligence (AI) algorithms are applied to this process but are mainly targeting specific material configuration and joint thicknesses. [8][9][10][11][12][13][14] Gibson et al gave a review about control architecture involving closed-loop position, force, torque or hybrid control modes, as well as in-process evaluation systems such as e-NDE V R and MonStir V R . The latter methods use process forces coupled to discrete Fourier transforms, neural network, or dimensional reduction techniques to correlate defect formation or mechanical properties.…”

Section: Introductionmentioning

confidence: 99%

“…3–7 More recently, the in-line process monitoring and artificial intelligence (AI) algorithms are applied to this process but are mainly targeting specific material configuration and joint thicknesses. 8–14…”

Section: Introductionmentioning

confidence: 99%

Machine learning models applied to friction stir welding defect index using multiple joint configurations and alloys

Nadeau

Thériault

Gagné

2020

Proceedings of the Institution of Mechanical Engineers, Part L:

View full text Add to dashboard Cite

Friction stir welding process has been studied extensively in the last decades since its early stage. Most of the research done so far is related to the process development including tool design, material weldability, post-weld mechanical behavior, and microstructural properties. More recently, in-line process monitoring and artificial intelligence algorithms are introduced into this process, but mainly to specific material configuration and joint thicknesses. This study will focus on the evaluation of different machine learning approaches including principle component analysis, K-nearest neighbor, multilayer perceptron, single vector machine, and random forest methods on a friction stir welding cell environment. The input variables provided from this cell environment are namely divided into two groups: one group refers to the application variables and the other group is related to the friction stir welding process variables. The application variables target the aluminum alloys, joint configuration, sheet thicknesses, initial mechanical properties, and their chemical composition. The friction stir welding process variables dictate the rotational speed, travel speed, forging force, longitudinal and transverse forces, torque, and specific energy. The output response to model from these machine learning algorithms is the defect index, which has been quantified using high-resolution immersed bath ultrasounds. This nondestructive evaluation technique has been described previously, which can detect defects ≥150 µm in thin sheets. The defect index has been classified into five classes, which is distinguished by the nature of defect, cold weld, or hot weld, as well as the width of the internal volumetric defect upon ultrasound C-scan result. The dataset, which is composed of around 500 various process conditions, has been generated over the last few years and the variables were taken exclusively in constant weld regime and in the force control mode using the output average values. This paper compares the best resulting machine learning methods applied on a friction stir welding cell basis, which is the K-nearest neighbor and multilayer perceptron algorithms. The K-nearest neighbor model reaches a deviation of 0.55 on the defect index in comparison with the experimental values, which is slightly better than the multilayer perceptron model, which obtains a score of 0.69. Over the initial 59 available model parameters, 10 and 15 of them were retained in the final algorithm using these techniques. The main predictors include the material thickness, base material ultimate tensile stress, rotational speed, travel speed, weld forces, and specific energy. The K-nearest neighbor model was able to provide a map of defect indices with regard to rotational speed and travel speed but was only possible when a higher density of data was found within the prediction area. A data density score was also included within the model to inform the end-user about the prediction reliability. The machine learning models are mainly about differentiating various cases rather than representing the physical phenomena as determined using the finite element analysis. That being said, in order to improve the prediction reliability as well as the machine learning models, the data twinning concept, which consists of generating simulated friction stir welding process conditions by finite element analysis, is briefly discussed.

show abstract

Monitoring of friction stir welding based on vision system coupled with Machine learning algorithm

Cited by 64 publications

References 41 publications

Advanced Welding Manufacturing: A Brief Analysis and Review of Challenges and Solutions

Advanced Welding Manufacturing: A Brief Analysis and Review of Challenges and Solutions

A Product Quality Monitor Model With the Digital Twin Model and the Stacked Auto Encoder

Machine learning models applied to friction stir welding defect index using multiple joint configurations and alloys

Contact Info

Product

Resources

About