In smart manufacturing, the automation of anomaly detection is essential for increasing productivity. Timeseries data from production processes are often complex sequences and their assessment involves many variables. Thus, anomaly detection with deep learning approaches is considered as an efficient and effective methodology. In this work, anomaly detection with deep autoencoders is examined. Three autoencoders are employed to analyze an industrial dataset and their performance is assessed. Autoencoders based on long short-term memory and convolutional neural networks appear to be the most promising.

This article demonstrates a methodology for the detection of foreign object impact locations on laminated composite plates using a low number of sensors. The proposed approach exploits the wave-dominated transient response of the target structure and addresses the challenges induced by the non-uniform wave propagation due to the anisotropy of composite plates on the impact localization. Captured sensor signals are processed and their frequency content is identified. Semi-analytical wave dispersion models and time-domain spectral finite element impact models encompassing the effects of laminate anisotropy on wave dispersion characteristics are employed to extract the wave velocity from their frequency content. The method for localizing the impact site is formulated as an optimization problem and involves signal correlation from different sensors, refinement of the candidate locations, and finally accurate estimation of the impact site. Numerical and experimental validations of the developed method are presented for low-velocity impacts on glass/epoxy laminated plates.

In printed circuit board (PCB) assembly, the volume of dispensed glue is closely related to the PCB’s durability, production costs, and the overall product reliability. Currently, quality inspection is performed manually by operators, inheriting the limitations of human-performed procedures. To address this, we propose an automatic optical inspection framework that utilizes convolutional neural networks (CNNs) and post-hoc explainable methods. Our methodology handles glue quality inspection as a three-fold procedure. Initially, a detection system based on CenterNet MobileNetV2 is developed to localize PCBs, thus, offering a flexible lightweight tool for targeting and cropping regions of interest. Consequently, a CNN is proposed to classify PCB images into three classes based on the placed glue volume achieving 92.2% accuracy. This classification step ensures that varying glue volumes are accurately assessed, addressing potential quality issues that appear early in the production process. Finally, the Deep SHAP and Grad-CAM methods are applied to the CNN classifier to produce explanations of the decision making and further increase the interpretability of the proposed approach, targeting human-centered artificial intelligence. These post-hoc explainable methods provide visual explanations of the model’s decision-making process, offering insights into which features and regions contribute to each classification decision. The proposed method is validated with real industrial data, demonstrating its practical applicability and robustness. The evaluation procedure indicates that the proposed framework offers increased accuracy, low latency, and high-quality visual explanations, thereby strengthening quality assurance in PCB manufacturing.