A Physics-Informed Convolutional Neural Network with Custom Loss Functions for Porosity Prediction in Laser Metal Deposition

Physics-informed machine learning is emerging through vast methodologies and in various applications. This paper discovers physics-based custom loss functions as an implementable solution to additive manufacturing (AM). Specifically, laser metal deposition (LMD) is an AM process where a laser beam melts deposited powder, and the dissolved particles fuse to produce metal components. Porosity, or small cavities that form in this printed structure, is generally considered one of the most destructive defects in metal AM. Traditionally, computer tomography scans measure porosity. While this is useful for understanding the nature of pore formation and its characteristics, purely physics-driven models lack real-time prediction ability. Meanwhile, a purely deep learning approach to porosity prediction leaves valuable physics knowledge behind. In this paper, a hybrid model that uses both empirical and simulated LMD data is created to show how various physics-informed loss functions impact the accuracy, precision, and recall of a baseline deep learning model for porosity prediction. In particular, some versions of the physics-informed model can improve the precision of the baseline deep learning-only model (albeit at the expense of overall accuracy).

Identification of a Spatio-Temporal Temperature Model for Laser Metal Deposition

Laser-based additive manufacturing enables the production of complex geometries viaayer-wise cladding. Laser metal deposition (LMD) uses a scanningaser source to fuse in situ deposited metal powderayer byayer. However, due to the excessive number of influential factors in the physical transformation of the metal powder and the highly dynamic temperature fields caused by the melt pool dynamics and phase transitions, the quality and repeatability of parts built by this process is still challenging. In order to analyze and/or predict the spatially varying and time dependent thermal behavior in LMD, extensive work has been done to develop predictive models usually by using finite element method (FEM). From a control-oriented perspective, simulations based on these models are computationally too expensive and are thus not suitable for real-time control applications. In this contribution, a spatio-temporal input–output model based on the heat equation is proposed. In contrast to other works, the parameters of the model are directly estimated from measurements of the LMD process acquired with an infrared (IR) camera during processing specimens using AISI 316 L stainless steel. In order to deal with noisy data, system identification techniques are used taking different disturbing noise into account. By doing so, spatio-temporal models are developed, enabling the prediction of the thermal behavior by means of the radiance measured by the IR camera in the range of the considered processing parameters. Furthermore, in the considered modeling framework, the computational effort for thermal prediction is reduced compared to FEM, thus enabling the use in real-time control applications.

Short Time Correlation Analysis of Melt Pool Behavior in Laser Metal Deposition Using Coaxial Optical Monitoring

The development and improvement of monitoring and process control systems is one of the important ways of advancing laser metal deposition (LMD). The control of hydrodynamic, heat and mass transfer processes in LMD is extremely important, since these processes directly affect the crystallization of the melt and, accordingly, the microstructural properties and the overall quality of the synthesized part. In this article, the data of coaxial video monitoring of the LMD process were used to assess the features of melt dynamics. The obtained images were used to calculate the time dependences of the characteristics of the melt pool (MP) (temperature, width, length and area), which were further processed using the short-time correlation (STC) method. This approach made it possible to reveal local features of the joint behavior of the MP characteristics, and to analyze the nature of the melt dynamics. It was found that the behavior of the melt in the LMD is characterized by the presence of many time periods (patterns), during which it retains a certain ordered character. The features of behavior that are important from the point of view of process control systems design are noted. The approach used for the analysis of melt dynamics based on STC distributions of MP characteristics, as well as the method for determining the moments of pattern termination through the calculation of the correlation power, can be used in processing the results of online LMD diagnostics, as well as in process control systems.

The control of the height and shape of the track in laser metal deposition by the QCW laser mode

The control of the track shape in laser metal deposition technology by the QCW laser mode has been studied. The different geometric characteristics of the tracks are shown to obtain at the same average laser power, depending on the selected laser power control mode. The difference in the temperature regimes of track formation is shown.