Electron Beam Powder Bed Fusion (EB-PBF) is a powerful additive manufacturing technique capable of producing high-performance metal parts with complex geometries. However, inherent process instabilities can lead to defects that compromise part quality and structural integrity. Traditional non-destructive testing (NDT) methods are often costly and time-consuming, and typically require specialized equipment and/or expertise. As a promising alternative, this study explores the feasibility of using active infrared thermography (IRT), coupled with deep learning, for the automated quality assessment of EB-PBF fabricated parts. We present a case study using parts with artificially induced subsurface defects, captured through active thermo-graphic imaging. A comprehensive dataset of thermal images is generated and used to train and evaluate a customized deep learning framework based on the You Only Look Once (YOLO) architecture for automated defect detection and categorization. Our results demonstrate the potential of combining IRT with data-driven analysis to offer a fast, contactless, and scalable solution for inspecting EB-PBF parts, while also highlighting the current limitations and future directions for this emerging approach.

Detecting complex anomalies on massive amounts of data is a crucial task in Industry 4.0, best addressed by deep learning. However, available solutions are computationally demanding, requiring cloud architectures prone to latency and bandwidth issues. This work presents VARADE, a novel solution implementing a light autoregressive framework based on variational inference, which is best suited for real-time execution on the edge. The proposed approach was validated on a robotic arm, part of a pilot production line, and compared with several state-of-the-art algorithms, obtaining the best trade-off between anomaly detection accuracy, power consumption and inference frequency on two different edge platforms.

With the rise of Machine Learning (ML) and Artificial Intelligence (AI), the semiconductor industry is undergoing a revolution in how it approaches manufacturing. The SMART-IC project (DATE'24 MPP category: initial stage) works in this direction, by proposing an AI-enabled framework to support the smart monitoring and optimization of the semiconductor manufacturing process. An AI-powered engine examines sensor data recording physical parameters during production (like gas flow, temperature, voltage, etc.) as well as test data, with different goals: (1) the identification of anomalies in the production chain, either offline from collected data-traces or online from a continuous stream of sensed data; (2) the forecasting of new data of the future production; and (3) the automatic generation of synthetic traces, to strengthen the data-based algorithms. All such tasks provide valuable information to an advanced Manufacturing Execution System (MES), which reacts by optimizing the production process and management of the equipment maintenance policies. SMART-IC is a 300k€ academic project funded by the Italian Ministry of University and supported by STMicroelectronics and Technoprobe with industrial expertise and real-world applications. This paper shares the view of SMART-IC on the future of semiconductor manufacturing, the preliminary efforts, and the future results that will be reached by the end of the project, in 2025.

Laser powder bed fusion (L-PBF) is the most popular Additive Manufacturing (AM) process for metals. It builds a 3D object layer-by-layer, by spreading metal powder on top of the previous layer and selectively melting it with a laser. Despite its many advantages, large-scale production may be hampered by the large number of process parameters and the challenges associated with their optimization. We propose an automated parameter selection approach based on process signatures extracted from a parameterized simulation of the process. Specifically, we outline a rapid data-driven simulation method based on Physics-Informed Neural Network (PINN). This approach involves training a neural network to solve the partial differential equation describing the process at varying values of a parameter of interest (for example, the laser power), thus eliminating the need for repeated Finite Elements Method (FEM) simulations. Our preliminary experiments demonstrate the feasibility of our approach.

Industry 4.0 involves the integration of digital technologies, such as IoT, Big Data, and AI, into manufacturing and industrial processes to increase efficiency and productivity. As these technologies become more interconnected and interdependent, Industry 4.0 systems become more complex, which brings the difficulty of identifying and stopping anomalies that may cause disturbances in the manufacturing process. This paper aims to propose a diffusion-based model for real-time anomaly prediction in Industry 4.0 processes. Using a neuro-symbolic approach, we integrate industrial ontologies in the model, thereby adding formal knowledge on smart manufacturing. Finally, we propose a simple yet effective way of distilling diffusion models through Random Fourier Features for deployment on an embedded system for direct integration into the manufacturing process. To the best of our knowledge, this approach has never been explored before.

The early detection of anomalous behaviors from a production line is a fundamental aspect of Industry 4.0, facilitated by the collection of massive amounts of data enabled by the Industrial Internet of Things. Nonetheless, the design and validation of anomaly detection algorithms, mostly based on sophisticated Machine Learning models, heavily rely on the availability of annotated datasets of realistic anomalies, which is very difficult to obtain in a real production line. To address this problem, we introduce the Robotic Arm Dataset (RoAD), specifically designed to support the development and validation of Multivariate Time Series Anomaly Detection (MTSAD) algorithms. We collect and annotate a large number of data and metadata to characterize the motion and energy consumption of a collaborative robotic arm in a full-fledged production line and annotate a comprehensive set of healthy as well as realistic anomalies scenarios. To prove the significance of RoAD and encourage future developments, we benchmark several state-of-the-art anomaly detection algorithms on our newly introduced dataset, and we freely release it to the scientific community.

Recent developments have resulted in multiple techniques trying to explain how deep neural networks achieve their predictions. The explainability maps provided by such techniques are useful to understand what the network has learned and increase user confidence in critical applications such as the medical field or autonomous driving. Nonetheless, they typically have very low resolutions, severely limiting their capability of identifying finer details or multiple subjects. In this paper we employ an encoder-decoder architecture with skip connection known as U-Net, originally developed for segmenting medical images, as an image classifier and we show that state of the art explainable techniques applied to U-Net can generate pixel level explanation maps for images of any resolution.

MOTIVATION: Computer-aided analysis of biological images typically requires extensive training on large-scale annotated datasets, which is not viable in many situations. In this paper, we present Generative Adversarial Network Discriminator Learner (GAN-DL), a novel self-supervised learning paradigm based on the StyleGAN2 architecture, which we employ for self-supervised image representation learning in the case of fluorescent biological images. RESULTS: We show that Wasserstein Generative Adversarial Networks enable high-throughput compound screening based on raw images. We demonstrate this by classifying active and inactive compounds tested for the inhibition of SARS-CoV-2 infection in two different cell models: the primary human renal cortical epithelial cells (HRCE) and the African green monkey kidney epithelial cells (VERO). In contrast to previous methods, our deep learning-based approach does not require any annotation, and can also be used to solve subtle tasks it was not specifically trained on, in a self-supervised manner. For example, it can effectively derive a dose-response curve for the tested treatments. AVAILABILITY AND IMPLEMENTATION: Our code and embeddings are available at https://gitlab.com/AlesioRFM/gan-dl StyleGAN2 is available at https://github.com/NVlabs/stylegan2 .

Incremental learning involves Machine Learning paradigms that dynamically adjust their previous knowledge whenever new training samples emerge. To address the problem of multi-task incremental learning without storing any samples of the previous tasks, the so-called Expert Gate paradigm was proposed, which consists of a Gate and a downstream network of task-specific CNNs, a.k.a. the Experts. The gate forwards the input to a certain expert, based on the decision made by a set of autoencoders. Unfortunately, as a CNN is intrinsically incapable of dealing with inputs of a class it was not specifically trained on, the activation of the wrong expert will invariably end into a classification error. To address this issue, we propose a probabilistic extension of the classic Expert Gate paradigm. Exploiting the prediction uncertainty estimations provided by Bayesian Convolutional Neural Networks (B-CNNs), the proposed paradigm is able to either reduce, or correct at a later stage, wrong decisions of the gate. The goodness of our approach is shown by experimental comparisons with state-of-the-art incremental learning methods.