Machine learning for metal additive manufacturing: Towards a physics-informed data-driven paradigm

Guo, Shenghan; Agarwal, Mohit; Cooper, Clayton; Tian, Qi; Gao, Robert X.; Grace, Weihong Guo; Guo, Y. B.

doi:10.1016/j.jmsy.2021.11.003

Cited by 142 publications

(23 citation statements)

References 124 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Oftentimes, common statistical analysis has been misidentified to ML, however, the former aims on the accurate description of observations while the latter is constructing algorithms whose focus lies on successful data classification and approximations in order to predict the outcome [ 14 ]. ML is widely recognized as an effective tool for research and applied purposes, in fields like additive manufacturing [ 15 , 16 , 17 , 18 ], materials science [ 14 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 ], autonomous driving [ 30 , 31 , 32 ], solar cells [ 33 , 34 , 35 , 36 , 37 , 38 , 39 ], chemistry [ 40 , 41 , 42 ], welding industry [ 43 ], solar radiation [ 44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 ] and many more.…”

Section: Introductionmentioning

confidence: 99%

Artificial Intelligence in Physical Sciences: Symbolic Regression Trends and Perspectives

Angelis

Sofos

Karakasidis

2023

Arch Computat Methods Eng

View full text Add to dashboard Cite

Symbolic regression (SR) is a machine learning-based regression method based on genetic programming principles that integrates techniques and processes from heterogeneous scientific fields and is capable of providing analytical equations purely from data. This remarkable characteristic diminishes the need to incorporate prior knowledge about the investigated system. SR can spot profound and elucidate ambiguous relations that can be generalizable, applicable, explainable and span over most scientific, technological, economical, and social principles. In this review, current state of the art is documented, technical and physical characteristics of SR are presented, the available programming techniques are investigated, fields of application are explored, and future perspectives are discussed.

show abstract

Section: Introductionmentioning

confidence: 99%

Artificial Intelligence in Physical Sciences: Symbolic Regression Trends and Perspectives

Angelis

Sofos

Karakasidis

2023

Arch Computat Methods Eng

View full text Add to dashboard Cite

show abstract

“…The key elements of PINNs include (1) preprocessing input data to extract hidden physical information and/or supplement it with production parameters and simulation data; (2) ensuring physical consistency between model inputs and outputs by penalizing any output that does not comply with physical principles using a loss function; (3) incorporating physical intuition into activation functions that lack physical meaning; (4) training the model to perform a physically meaningful analysis of the input information during model optimization, which differs from (3) as it fundamentally alters the data representation, while (3) only changes how the data is processed without modifying the data itself; and (5) evaluating the physical consistency of the model’s predictive logic, which focuses on explaining the behavior of the model. Using this approach, PINNs enhance model interpretability and reduce errors by relying on physical principles . However, the application of PINNs in the process industry can be limited as first principles of the process may be complicated.…”

Section: Introductionmentioning

confidence: 99%

“…Using this approach, PINNs enhance model interpretability and reduce errors by relying on physical principles. 25 However, the application of PINNs in the process industry can be limited as first principles of the process may be complicated. Thus, it is important to find ways to incorporate other forms of prior knowledge, such as flowcharts.…”

Section: Introductionmentioning

confidence: 99%

Topology-Guided Graph Learning for Process Fault Diagnosis

Jia

Liu

et al. 2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Faults in the process industry can be diagnosed using various data-driven methods, but the intrinsic relationships between inputs and outputs, particularly the physical consistency of model prediction logic, have received little attention. To address this issue, we propose a topology-guided graph learning fault diagnosis framework that combines the concept of graphs with process physics. Our framework focuses on knowledge embedding and explanation and includes several key components: a topology graph based on the flowchart, a self-attention mechanism to discover distinctive knowledge from data, graph convolution to capture variable relationships, graph pooling to coarsen graph data, and a gating mechanism to establish long-term dependencies. We also use a graph explainer to assess the physical consistency of the model's prediction logic. We demonstrate the feasibility of our method using the Tennessee Eastman process and show that it is not a black-box model but rather has natural advantages in terms of effectiveness and explanation.

show abstract

“…Besides, PINNs can be thought of an unsupervised strategy without the need of labeled data, 28,29 such as results from prior simulations or experiments. Owing to these advantages, PINNs or other variations have been applied to diverse problems involving fluid dynamics, [30][31][32][33][34] thermodynamics, [35][36][37] geophysics, 38,39 material sciences, [40][41][42] to name a few. More detail literature reviews of the application of PINNs can be referred to References [43][44][45][46].…”

Section: Introductionmentioning

confidence: 99%

A nonlocal energy‐informed neural network for isotropic elastic solids with cracks under thermomechanical loads

Zhou

2023

Numerical Meth Engineering

View full text Add to dashboard Cite

In this paper, a nonlocal energy‐informed neural network is proposed to characterize the deformation behaviors of elastic solids with cracks subjected to thermomechanical loads in the framework of ordinary state‐based peridynamics. Based on principle of virtual work, a nonlocal energy approach is developed to recast the solution to peridynamic equilibrium equation as a problem of minimizing the potential energy of the system, which automatically satisfies the traction‐free boundary conditions. Meanwhile, the energy representation of physical system can be treated as the loss function for machine learning methods. Therefore, a nonlocal energy‐informed neural network is constructed to approximate the solution of the system. A distinct advantage of the proposed neural network is that the strain energy is expressed in terms of spatial integration instead of spatial derivatives, which avoids the invalidation of automatic differentiation at the crack surfaces in original physics‐informed neural networks. To demonstrate the convergence and accuracy of the proposed neural network, a series of problems in solid and fracture mechanics are conducted, and compared with results from analytical solutions or classical numerical methods. Additionally, for elastic material containing initial cracks, displacement extrapolation method is encoded into the proposed neural network to evaluate the static stress intensity factor.

show abstract

Machine learning for metal additive manufacturing: Towards a physics-informed data-driven paradigm

Cited by 142 publications

References 124 publications

Artificial Intelligence in Physical Sciences: Symbolic Regression Trends and Perspectives

Artificial Intelligence in Physical Sciences: Symbolic Regression Trends and Perspectives

Topology-Guided Graph Learning for Process Fault Diagnosis

A nonlocal energy‐informed neural network for isotropic elastic solids with cracks under thermomechanical loads

Contact Info

Product

Resources

About