Yifan Tang

Digital twin (DT) has been applied for monitoring, control, and decision-making of real-world engineering systems, but how to keep a DT valid as the physical system degrades over the lifecycle remains rarely explored. Instead of identifying degradation parameters from expensive aging experiments, this paper proposes a lifelong update method to capture degradation effects directly from system responses via continuously tuning DT models. The core idea is to represent the system degradation as temporal changes of DT model configurations through degradation stages. During the lifelong update process, an autoencoder compresses DT model parameters into latent features, and an LSTM learns the temporal trend of those features to predict latent features for future degradation stages. Based on the predicted latent features, the DT model configuration is reconstructed to predict the system responses affected by degradation. The proposed lifelong update method is evaluated with two real-world engineering datasets, e.g., the battery degradation dataset from Oxford University and the flight engine degradation dataset from NASA. Results demonstrate that the proposed update method successfully captures the degradation effects on system responses during the lifecycle. The one-tailed t-tests confirm that the proposed method statistically outperforms the conventional fine-tuning method in prediction accuracy and robustness at future unseen stages.

A digital twin (DT) is a model that mirrors a physical system and is continuously updated with real-time data from the physical system. Recent implementations of reduced-order-model-based DT (DT-ROM) have been applied in aerodynamics and structural health monitoring, where partial differential equations (PDEs) are utilized to update reduced bases and coefficients. However, these methods are not directly applicable when the PDEs of the system are unknown. This paper addresses the online update challenge for DT-ROM in scenarios lacking known PDEs of the system. To tackle the challenge, a systematic online update and application method is proposed. During the online update, the projection residual of online data on the reduced bases determines the necessity of updating reduced bases; the prediction residual of online data obtained by the current DT-ROM is used to decide whether to update the coefficient model. By sequentially evaluating both criteria, the method selectively incorporates essential online data for the online DT model update. During the online application, a criterion defined based on online data is adopted to determine whether the offline DT-ROM or the online one is applied to output final predictions. The capability of the proposed method is tested through three numerical and three engineering problems. Results indicate that the proposed online update method consistently reduces both projection and prediction residuals, thereby progressively enhancing the performance of the online DT-ROM on test data. Meanwhile, the online application method provides a prediction performance better than using offline DT-ROM only. Both demonstrate that the proposed work could be applied to online DT updates where the PDEs of the system are unknown.

Considering data insufficiency in metal additive manufacturing (AM), transfer learning (TL) has been adopted to extract knowledge from source domains (e.g., completed printings) to improve the modeling performance in target domains (e.g., new printings). Current applications use all accessible source data directly in TL with no regard to the similarity between source and target data. This paper proposes a systematic method to find appropriate subsets of source data based on similarities between the source and limited target datasets. Such similarity is characterized by the spatial and model distance metrics. A Pareto frontier-based source data selection method is developed, where the source data located on the Pareto frontier defined by two similarity distance metrics are selected iteratively. This method is integrated into an instance-based TL method (decision tree regression model) and a model-based TL method (fine-tuned artificial neural network). Both models are then tested on several regression tasks in metal AM. Comparison results demonstrate that (1) the source data selection method is general and supports integration with various TL methods and distance metrics, (2) compared with using all source data, the proposed method can find a subset of source data from the same domain with better TL performance in metal AM regression tasks involving different processes and machines, and (3) when multiple source domains exist, the source data selection method could find the subset from one source domain to obtain comparable or better TL performance than the model constructed using data from all source domains.

Various data-driven modeling methods have been developed to predict the thermal field in metal additive manufacturing (AM). The generalization capability of these models has been shown with simulation, but rarely tested with online physical printing. Instead, this paper aims to study a practical issue in metal AM, i.e., how to predict the thermal field of yet-to-print parts online when only a few sensors are available. This work proposes an online thermal field prediction method using mapping and reconstruction, which could be integrated into a metal AM process for online performance control. Based on the similarity of temperature curves (curve segments of a temperature profile of one point), the thermal field mapping applies an artificial neural network to estimate the temperature curves of points on the yet-to-print layer from measured temperatures of certain points on the previously printed layer. With measured/predicted temperature profiles of several points on the same layer, the thermal field reconstruction proposes a reduced order model (ROM) to construct the temperature profiles of all points on the same layer, which could be used to build the temperature field of the entire layer. The training of ROM is performed with an extreme learning machine (ELM) for computational efficiency. Fifteen wire arc AM experiments and nine simulations are designed for thin walls with a fixed length and unidirectional printing of each layer. The test results indicate that the proposed prediction method could construct the thermal field of a yet-to-print layer within 0.1 s on a low-cost desktop computer (Intel Core i7-3770 CPU @ 3.40GHz processor, 24.0 GB RAM). Meanwhile, the method has acceptable generalization capability in most cases from lower layers to higher layers in the same simulation, as well as from one simulation to a new simulation on different AM process parameters. More importantly, after fine-tuning the proposed method with limited experimental data, the relative errors of all predicted temperature profiles on a new experiment are sufficiently small, which demonstrates the applicability and generalization of the proposed thermal field prediction method in online applications for metal AM.

Modeling plays an important role in the additive manufacturing (AM) process and quality control. In practice, however, only limited data are available for each product due to the relatively high AM cost, which brings challenges in building either a high-quality physics-based or data-based model. Transfer learning (TL) is a new and promising group of approaches where the model of one product (source) may be reused for another product (target) with limited new target data. This paper focuses on reviewing applications of TL in AM modeling to help advance research in this area. First, notations, definitions, and categories of TL methods are introduced along with their application scenarios. Then current applications of TL in AM modeling are summarized along with their limitations. Based on reviewed applications, recommendations are given on how to apply TL for a certain AM problem, from the perspectives of source domain determination, TL method selection, target data generation, and data preprocessing. Finally, future research directions about TL in AM modeling are discussed in the hope to explore more potential of TL in improving the AM model quality with limited data.

To further reduce the computational expense of metamodel-based design optimization (MBDO), a novel sequential radial basis function (RBF)-based optimization method using virtual sample generation (SRBF-VSG) is proposed. Different from the conventional MBDO methods with pure expensive samples, SRBF-VSG employs the virtual sample generation mechanism to improve the optimization efficiency. In the proposed method, a least squares support vector machine (LS-SVM) classifier is trained based on expensive real samples considering the objective and constraint violation. The classifier is used to determine virtual points without evaluating any expensive simulations. The virtual samples are then generated by combining these virtual points and their Kriging responses. Expensive real samples and cheap virtual samples are used to refine the objective RBF metamodel for efficient space exploration. Several numerical benchmarks are tested to demonstrate the optimization capability of SRBF-VSG. The comparison results indicate that SRBF-VSG generally outperforms the competitive MBDO methods in terms of global convergence, efficiency, and robustness, which illustrates the effectiveness of virtual sample generation. Finally, SRBF-VSG is applied to an airfoil aerodynamic optimization problem and a small Earth observation satellite multidisciplinary design optimization problem to demonstrate its practicality for solving real-world optimization problems.