Abstract Metamaterials (MM) offer a promising solution for footwear design by enabling spatially tunable mechanical responses associated with comfort-related considerations. However, the current application of MMs in footwear manufacturing remains relatively limited. Furthermore, effective optimization of MM structures under comfort-related mechanical considerations remains a largely unexplored space. This study explores a framework that teams finite element analysis (FEA) and deep learning (DL) to optimize MM-based shoe sole designs by using the individual plantar pressure distribution data. Specifically, the FEA is deployed to analyze the mechanical response of the sole, and it establishes links among pressure, deformation and geometry, while a DL model is developed to map plantar pressure to MM lattice geometry. In our proposed method, a multimodal convolutional neural network (CNN) model is designed and trained to predict the optimized lattice-rod radii in MM based on multimodal input, including the pressure image, rod coordinates, and deformation data. The DL-optimized MM structures are subsequently validated through FEA simulations. With this proposed approach, this study conducted a case study relevant to potential footwear applications under a deformation-based surrogate objective. The results show that the proposed method achieves significant improvement in deformation-pressure alignment compared with a benchmark design. In summary, the proposed FEA-DL framework enables direct generation of optimized MM designs from user-specific plantar pressure and a surrogate design objective. This synergistic approach offers a promising pathway to achieve a customizable, efficient, high-performance, resource-conserving inverse design.

Abstract Multivariate time series (MTS) data are central to engineering systems, where streams of sensor signals enable trend analysis, anomaly detection, and the discovery of operating patterns—supporting predictive maintenance and quality control. Yet analyzing MTS remains challenging: faults are rare, severely imbalanced, and embedded in long contexts with overlapping modes. These challenges demand models that capture extended dependencies and focus on the most informative intervals. To address this, we design a hybrid deep-learning architecture integrating recurrent units for long-horizon dynamics, attention layers for critical time steps, and temporal convolutions for multi-scale local patterns. We propose GAT-Net (GRU–Attention–Temporal Convolutional Network), a robust end-to-end MTS anomaly detector. Because severe class imbalance remains the dominant barrier, we extend the backbone with a generative augmentation strategy, developing DA-GAT-Net (Data-Augmented GAT-Net). It uses a segment-aware conditional GAN to synthesize rare fault segments conditioned on local context and inserts them chronologically, preserving temporal continuity and cross-sensor correlations. We evaluate our frameworks on an industrial benchmark dataset. They outperform state-of-the-art baselines, with the generative extension yielding the largest gains in recall and F1-score. We also show that sensitivity rises sharply with a moderate amount of synthetic anomalies but levels off beyond a balanced ratio, underscoring the need for controlled, structure-preserving generation. Overall, this study highlights how attention-guided temporal modeling combined with generative augmentation can enhance fault detection in engineering systems where missed anomalies carry high operational costs.

Abstract Remaining Useful Life (RUL) prediction is a critical technology in prognostics and health management. Rotating machinery plays an indispensable role in industrial processes, and accurate RUL prediction for such equipment is essential to ensure production continuity, optimize maintenance strategies and reduce operational costs. In recent years, deep learning has emerged as a powerful tool for the analysis of multi-source monitoring data derived from rotating machinery, owing to its exceptional proficiency in feature extraction and non-linear modelling. This approach can effectively excavate the deeper-level information within the data, thereby improving prediction accuracy. Consequently, deep learning demonstrates significant application value and broad development prospects in RUL prediction for rotating machinery. This paper provides a comprehensive review of the application of deep learning techniques in RUL prediction for rotating machinery. Firstly, it introduces a unified framework for RUL prediction based on deep learning methodologies. The article then elaborates on the theoretical foundations of several key learning methods, including autoencoders, convolutional neural networks, long short-term memory, gated recurrent units, as well as emerging models such as transformers, temporal convolutional networks and graph neural networks. Subsequently, the paper summarizes the applications of these deep learning methods in RUL prediction for rotating machinery, evaluating the specific problems each method addresses based on their distinct characteristics. Finally, it discusses the significant challenges that deep learning faces within this domain and outlines potential directions for future research.

Abstract Individualized product customization aims to satisfy specific customer requirements through configuration design. However, most existing methods treat customer requirements as static inputs and struggle to accommodate their dynamic and unpredictable evolution. To address this limitation, this article proposes a customer requirement-driven dynamic evolution method for product models based on similar cases. The method introduces attribute nodes as intermediate representation units to unify requirement modeling and product modeling. Attribute node re-instantiation rules are defined to regenerate attribute values when requirements change, while design constraints ensure feasibility and consistency during product evolution. Customer requirements are divided into basic and dynamic types, enabling continuous customer participation beyond initial configuration. Basic requirements and customer preferences are used to retrieve similar product cases, and both basic and dynamic requirements subsequently drive product case evolution toward customer expectations. A case study of an air purifier product demonstrates the feasibility and effectiveness of the proposed method in supporting dynamic product evolution and customer participation.

Abstract Extreme learning machine (ELM) has gained significant attention for its remarkable generalization performance and fast processing speeds in recent years. However, its sensitivity to non-Gaussian noise and outliers has been a notable limitation. In this article, we incorporate correntropy loss and modeling error distributions in ELM training and propose a robust ELM that minimizes the mean and variance of modeling errors with maximum correntropy criterion, aimed at improving modeling performance in a noisy environment. Correntropy, as a robust generalized nonlinear similarity measure, is developed to capture the variance of errors in the modeling process, while the integration of the half-quadratic optimization technique ensures efficiency of the proposed ELM training. The robust and parameter sensitivity analysis of the proposed framework further reinforces its reliability, while experimental results on benchmark datasets confirm its superior generalization performance, even in the presence of varying ratios of outliers.

Abstract Engineering changes are inevitable in product design, especially during the development of complex products, where they often affect multiple-domain engineering entities. Traditional single-domain change models cannot capture the multidimensional nature of products; thus, they may fail to effectively predict change propagation paths or assess impacts. To address this challenge, a multilayer network change model is proposed, consisting of the assembly, feature, face, and parameter layers. This model enables the coupled propagation of changes between parameters and structures. The first three layers represent the product's geometric topology and shape, whereas the parameter layer defines functional relationships among parameters. A parameter change prediction method based on the breadth-first search (BFS) algorithm is introduced to automatically identify and optimize change propagation paths within the parameter layer. Meanwhile, through the interlayer connections between parameters and structures, the structural entities affected by parameters can be identified. A graph isomorphism algorithm is used to accurately locate and identify structural topology change entities. Structural topology change propagation among parts is predicted based on their mating entities. The identified change entities are further used to assess their impact on manufacturing processes. A software system based on the multilayer network change model is developed, and a case study is presented to demonstrate its application in product change design, which shows that the affected manufacturing entities can be pinpointed even when design parameters are changed, and change solutions can be determined in a time-consuming and cost-effective way.

Abstract Reconstructing high-fidelity fluid flow fields from sparse sensor measurements is vital for many science and engineering applications but remains challenging because of the dimensional disparities between state and observational spaces. Due to such dimensional differences, the measurement operator becomes ill conditioned and noninvertible, making the reconstruction of flow fields from sensor measurements extremely difficult. Although sparse optimization and machine learning address the above problems to some extent, questions about their generalization and efficiency remain, particularly regarding the discretization dependence of these models. In this context, deep operator learning offers a better solution as this approach models mappings between infinite-dimensional function spaces, enabling superior generalization and discretization-independent reconstruction. We introduce a deep operator-learning model that is trained to reconstruct fluid flow fields from sparse sensor measurements. Our deep-learning model employs a branch–trunk network architecture to represent the inverse measurement operator that maps sensor observations to the original flow field, a continuous function of both space and time. Our validation has demonstrated that the proposed deep-learning method consistently achieves high levels of reconstruction accuracy and robustness, even in scenarios where sensor measurements are inaccurate or missing. Furthermore, the operator-learning approach enables the capability to perform zero-shot super-resolution in both spatial and temporal domains, offering a solution for rapid reconstruction of high-fidelity flow fields.

Abstract Recent advances in wireless networking technologies, the reliable application of fault diagnostic and prognostic techniques, and rapid advances in machine learning algorithms have led to widespread use and benefits for Internet-of-Things applications. These advances have also made it possible for devices, previously used for monitoring at enterprise levels, to be applied to monitoring at the individual equipment level. Devices such as networked sensors and radio frequency identification (RFID) tags are increasingly used to detect, track, and monitor equipment in a variety of applications. While the benefits are quite promising, there are significant challenges in the adoption and deployment of such networked sensors in terms of design, operation, and economic justification. Selecting sensors with appropriate specifications is imperative for capturing high-quality information. This article presents a goal programming approach that utilizes sensor manufacturers' datasheets to specify and select a heterogeneous set of sensors. The usefulness of the proposed goal programming approach is demonstrated through its ability to concurrently handle the specification and selection of heterogeneous sensors in intricate systems. This differs from prior approaches, which treat sensor selection and specification disjointly during system design. The main contribution of this study is the development of a mathematical model to support flexible design needs and distinguishes itself from previous approaches in the field, and the use of dynamic specification bounds allowing users to explore design trade-offs and adjust system targets as needed. As a result, the proposed goal programming approach enables a more thorough and integrated design strategy for selecting sensors for intricate equipment.

Abstract With the continuous growth of personalized product demands and the upsurge in new customer orders, the arrival of new jobs and machine breakdowns due to overloading have emerged as common production disturbances, inevitably affecting the production plan. To maintain the stability of the scheduling for dynamic flexible job shops with machine breakdown and new job arrival, this article proposes a robust scheduling method that is designed with a flexible network structure and dual-action chained cooperative decision-making mechanism based on deep reinforcement learning (FD-DRL). First, a flexible neural network structure is innovatively constructed, which embeds the feature vector into operation nodes to design a dynamic production state extraction method with graph neural networks (GNN). Second, the dual-action chained cooperative decision-making mechanism is established for agents, who consider the new and remaining operations overall to maximize the utilization of machine idle time. Finally, through training and verification, the effectiveness and advancement of the proposed FD-DRL method are verified by comparing with heuristic/meta-heuristics and the static model of deep reinforcement learning (DRL).

Abstract With the integration of information technology and manufacturing industries, manufacturing is expected to enter the stage of service collaboration on platforms. The immutability and decentralization of blockchain can provide a trustworthy environment for manufacturing service collaboration (MSC). Manufacturing services with varying processes and standards lead to significant quality of service (QoS) differences, making single-chain networks unable to meet diverse ledger demands, potentially delaying subsequent manufacturing procedures on platforms. Therefore, this article proposes a blockchain-enabled MSC framework based on the Hyperledger Fabric's multichain architecture. A platform chain Org allocation mechanism based on the enterprise operation data is designed, and manufacturing services are grouped based on the QoS to ensure consistency in each service chain. Additionally, generalized stochastic Petri nets is employed to model the relevant stages of manufacturing service execution and log recording process on blockchain to analyze the system performance, and an improved gray wolf optimizer (GWO) is adopted to identify the most suitable parameters of the blockchain. The proposed optimizer achieves significant enhancements in exploration performance and stability. In addressing the problem proposed in the article, the optimizer obtains comparable solutions while reducing iterations by approximately 9%, significantly lowering simulation costs. Experiments validate the effectiveness of the proposed mechanism and algorithms.