https://popups.uliege.be/esaform21/index.php?id=4738 Index terms fr 0 https://popups.uliege.be/esaform21/index.php?id=2812 In the last decade, machine learning is increasingly attracting researchers in several scientific areas and, in particular, in the additive manufacturing field. Meanwhile, this technique remains as a black box technique for many researchers. Indeed, it allows obtaining novel insights to overcome the limitation of classical methods, such as the finite element method, and to take into account multi-physical complex phenomena occurring during the manufacturing process. This work presents a comprehensive study for implementing a machine learning technique (artificial neural network) to predict the thermal field evolution during the direct energy deposition of 316L stainless steel and tungsten carbides. The framework consists of a finite element thermal model and a neural network. The influence of the number of hidden layers and the number of nodes in each layer was also investigated. The results showed that an architecture based on 3 or 4 hidden layers and the rectified linear unit as the activation function lead to obtaining a high fidelity prediction with an accuracy exceeding 99%. The impact of the chosen architecture on the model accuracy and CPU usage was also highlighted. The proposed framework can be used to predict the thermal field when simulating multi-layer deposition. Wed, 24 Mar 2021 19:04:24 +0100 Mon, 12 Apr 2021 11:18:39 +0200 https://popups.uliege.be/esaform21/index.php?id=2812 https://popups.uliege.be/esaform21/index.php?id=2004 The need of solving industrial problems using faster and less computationally expensive techniques is becoming a requirement to cope with the present digital transformation of most industries. Recently, data is conquering the domain of engineering with different purposes: (i) defining data-driven models of materials, processes, structures and systems, whose physics-based models, when they exists, remain too inaccurate; (ii) enriching the existing physics-based models within the so-called hybrid paradigm; and (iii) using advanced machine learning and artificial intelligence techniques for scales bridging (upscaling), that is, for creating models that operating at the coarse-grained scale (cheaper in what respect the computational resources) enables integrating the fine-scale richness. The present work addresses the last item, aiming at enhancing standard structural models (defined in 2D shell geometries) for accounting all the fine-scale details (3D with rich through-the-thickness behaviors). For this purpose, two main strategies will be combined: (i) the in-plane-out-of-plane proper generalized decomposition -PGD- serving to provide the fine-scale richness; and (ii) advance machine learning techniques able to learn and extract the regression relating the input parameters with those high-resolution detailed descriptions. Tue, 23 Mar 2021 12:23:56 +0100 Mon, 12 Apr 2021 10:22:59 +0200 https://popups.uliege.be/esaform21/index.php?id=2004 https://popups.uliege.be/esaform21/index.php?id=3882 Manufacturing process optimisation usually amounts to searching optima in high-dimensional parameter spaces. In industrial practice, this search is most often directed by human-subjective expert judgment and trial-and-error experiments. In contrast, high-fidelity simulation models in combination with general-purpose optimisation algorithms, e.g. finite element models and evolutionary algorithms, enable a methodological, virtual process exploration and optimisation. However, reliable process models generally entail significant computation times, which often renders classical, iterative optimisation impracticable. Thus, efficiency is a key factor in optimisation. One option to increase efficiency is surrogate-based optimisation (SBO): SBO seeks to reduce the overall computational load by constructing a numerically inexpensive, data-driven approximation („surrogate“) of the expensive simulation. Traditionally, classical regression techniques are applied for surrogate construction. However, they typically predict a predefined, scalar performance metric only, which limits the amount of usable information gained from simulations. The advent of machine learning (ML) techniques introduces additional options for surrogates: in this work, a deep neural network (DNN) is trained to predict the full strain field instead of a single scalar during textile forming („draping“). Results reveal an improved predictive accuracy as more process-relevant information from the supplied simulations can be extracted. Application of the DNN in an SBO- framework for blank holder optimisation shows improved convergence compared to classical evolutionary algorithms. Thus, DNNs are a promising option for future surrogates in SBO. Mon, 29 Mar 2021 14:55:15 +0200 Thu, 08 Apr 2021 21:19:38 +0200 https://popups.uliege.be/esaform21/index.php?id=3882 https://popups.uliege.be/esaform21/index.php?id=3775 Energy prediction and starvation have become an essential part of process planning for the XXI century manufacturing industry due to cost-saving policies and environmental regulations. To this aim, the research presented in this paper details how machine learning-based algorithms can be an effective way to predict and minimize the energy consumptions in the widely spread radial-axial ring rolling (RARR) process. To analyze this bulk metal forming process, 380 numerical simulations have been developed using the commercial SW Simufact Forming 15 and considering three largely utilized materials, the 42CrMo4 steel, the IN 718 superalloy, and the AA6082 aluminum alloy. To create the database for both multi-variable regression and machine learning models, ring outer diameters ranging from 650 mm to 2000 mm and various process conditions including different sets of tool speeds and initial temperatures have been considered. For the case of the multi-variable regression model, to account for all the cross-influences between all the parameters, a second-order function including 26 parameters has been developed, resulting in a reasonable average accuracy (94 %) but also in an impractical huge equation. On the other hand, the machine learning model based on the Gradient Boosting (GB) approach allows obtaining a similar accuracy (96 %) but its compact form allows a more practical utilization and its training can be expanded almost indefinitely, by adding more results from both numerical simulations and experiments. The proposed approach allows to quickly and precisely predict the energy consumption in the RARR process and can be extended to other manufacturing processes. Mon, 29 Mar 2021 14:30:24 +0200 Thu, 08 Apr 2021 20:21:19 +0200 https://popups.uliege.be/esaform21/index.php?id=3775