Simulation tools improve various aspects of the additive manufacturing process, however, they come with an undesirable computational time for real-world applications. Finite Element Method (FEM) that solves partial differential equations (PDE) presents promising capabilities in simple additive manufactured components as an expository problem. Yet, PDE-based solutions take significantly long CPU time due to a large number of time steps required to simulate an additively manufactured part. Parallel computing and GPU machines are not capable expeditors because of the nature of sequential solve at each time step. A typical alternative solution for reducing the computational cost is to model the deposition of several layers with the behavior of a characteristic single layer. This approach is an approximation that depends on the number of merged layers and the trade-off between the computational cost and accuracy. These types of modellings can help to simulate more significant size components but are not practical when deployed for parts with shape complexity.

With modern machine learning (ML) capabilities, a new shift towards integration of  FEM and ML has been introduced, where ML algorithms emulate the behavior of the time-consuming PDE-solver for real-time analysis of PDE in a given application. In this paper, we present a deep learning (DL) model that can substitute the thermal analysis of the additive manufacturing process. The substitute solver enables real-time prediction of the temperature distribution in the deposited layers. This informed-deep-learning solver can serve as the basis for other downstream calculations and control solutions, *e.g.*, geometry correction, distortion mitigation, and material property improvement.

The temperature results generated by FEM train a DL model to forecast the temperature distribution at and around every deposition point. The application of engineered features is necessary to achieve reliable thermal results independent of part geometry or deposition location on the piece. The training data is obtained by sampling the established physical model’s behavior over different temperatures, cooling rates, and part’s geometries. At each time-step, a vector of \[?\] dimensions consisted of the measurements is fed through the DL model. We use 19time steps for training and predict the 20th\] step of the time-series.  The temperature forecast uses the original mesh structure of the FEM definition to avoid any spatial interpolation error between PDE and ML temperature maps. The network architecture is composed of a Recurrent Neural Network (RNN) with two layers of LSTM, each with 64 cells to model the temporal sequence of 1742 deposition temperatures derived by PDEs. We train the model on 133 FEM-generated samples. The reported R2 value on validations data is 87%, while the Mean Absolute Error (MAE) is 0.3%. This paper compares the performance between the PDE and ML forecast for the thermal results. We showed DL models are promising for simulation of  the additive manufacturing process, and can be reliable alternatives for computationally-expensive FEM tools.


Real-Time Prediction of Temperature Distribution in Additive Manufacturing Processes Based on Machine Learning

Division / Sponsors: 
Aerospace, Bioengineering, Applied Mechanics, Design Engineering Division, Advanced Energy Systems, Manufacturing Engineering Division, Noise Control and Acoustics, Nondestructive Evaluation, Diagnosis, and Prognosis (NDPD), Nuclear Engineering Division, Process Industries Division, Fluids Engineering, Heat Transfer, Materials, Microelectromechanical Systems, Safety Engineering and Risk Analysis, Computers & Information in Engineering


Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2020/volume/84478

IMECE2020 - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.