Archives of Computational Methods in Engineering最新文献

Estimating fatigue damage is essential to ensure the safety of mechanical structures. In this paper, recent experimental and computational approaches for fatigue crack growth in metals are discussed from the industrial viewpoint. First, experimental studies for obtaining the accurate threshold stress intensity factor range ΔK_th and the effect of hydrogen on the fatigue crack growth are reviewed. In particular, we discuss the relationship between the loading frequency and the magnitude of oxide-induced crack closure, the methodology of achieving an accurate ΔK_th value and the difference between fatigue crack growth curve in gaseous hydrogen and that in air. Moreover, key factors to be considered for computation of actual fatigue crack growth behavior are reviewed. Second, computational methods on fatigue crack propagation are surveyed, where those of complex crack propagation phenomena in real mechanical structures, including crack separation and merging, are studied. Especially, we focus on the effect of the models of crack front line on the choice of computational methods.

Structural health monitoring (SHM) has witnessed a transformative evolution with the integration of Artificial Intelligence (AI) and data analytics. This review synthesizes recent developments in the realm of AI-powered SHM, elucidating key findings and emphasizing the pivotal role of these technologies in shaping the future of infrastructure monitoring. The review highlights the efficacy of AI in processing and analyzing vast structural datasets, leading to improved detection, diagnosis, and prediction of structural issues. Machine learning algorithms contribute to a proactive approach, enabling the identification of subtle patterns indicative of deterioration. The symbiosis of AI and SHM not only enhances accuracy in anomaly detection but also holds promise in revolutionizing maintenance strategies. This abstract encapsulates the significance of AI and data analytics in SHM, concluding with insights into future research directions to address challenges and unlock untapped potentials in this dynamic field.

The pushover method is a simplified yet effective seismic analysis tool that estimates the nonlinear behavior of structures under increasing lateral loads, commonly used in performance-based earthquake engineering. This review discusses the development, applications, and advancements in both monotonic and cyclic pushover methods, which are essential tools in seismic analysis. Monotonic pushover methods, which include non-adaptive approaches such as the Capacity Spectrum Method and N2 method, are widely used for their simplicity and practicality, particularly in low- to mid-rise buildings. However, these methods fail to account for higher-mode effects and complex structural behavior, especially in taller or irregular structures. To address these limitations, adaptive methods have been developed to improve accuracy by adjusting the lateral load distribution and accounting for changes in stiffness and dynamic properties as the structure deforms. These methods show better correlation with nonlinear time-history analysis, the gold standard in seismic assessment. On the other hand, the cyclic pushover method has been introduced to consider the dynamic cyclic loading and address stiffness and strength degradation of structural components–factors often overlooked by monotonic methods. Despite its advantages, the pushover method has limitations, and it should not be over-relied upon, as it may provide rapid but superficial predictions of structural behavior. However, its application in risk-based and loss assessment frameworks has expanded its potential for future use. This review highlights the versatility of pushover methods in seismic design and retrofitting, emphasizing their evolving role in improving the accuracy and reliability of structural assessments, contributing to safer and more resilient buildings in earthquake-prone regions.

Computer vision techniques have gained great traction in civil infrastructure inspection and monitoring. This paper conducted a systematic review of recent data-driven computer vision algorithms in structural damage detection published during the past 5 years. The theories of prevalent computer vision models are first reviewed with an emphasis on the progressive innovation in algorithms’ architecture. Then, recent applications of computer vision models for structural damage evaluation are discussed, which are classified into different structural categories by their material types (i.e., concrete, steel, masonry, timber) at three hierarchical levels including damage recognition, localization, and quantification. In particular, the paper also highlights the current state of using computer vision for damage assessment of timber structures, which remains under-explored compared to concrete and steel structures. Next, the paper scrutinizes existing structural damage inspection guidelines to identify key technological gaps between the capability of existing computer vision methods and manual inspection practices in the field. Finally, the paper summarizes existing challenges and recommends future research opportunities including the integration of computer vision methods with multimodal large language models, sensor-fusion, and mobile inspection approaches.

Digital fabrication technology, commonly known as 3D printing or additive manufacturing (AM), has revolutionized the conversion of digital designs into physical objects through layer-by-layer material deposition. This technology has gained extensive adoption across industries such as aerospace, healthcare, automotive, energy, and electronics, enabling mass customization, sustainable manufacturing, and the development of advanced composite materials. This study employs a scientometric approach to analyze global research trends in polymer composite AM from 2013 to 2023, assessing its growth, impact, and collaborative dynamics. The field has experienced a notable annual research growth rate of 50.92%, with a substantial increase in publications (532 papers) and active participation from 6,675 authors. The global research impact is evident, with an average citation rate of 21.94 citations per paper and international collaborations accounting for 23.53% of the total research output. India has emerged as a leading contributor in digital fabrication research, producing the highest number of published articles and research collaborations. Additionally, advancements in Design for Additive Manufacturing (DfAM), multi-material printing, functionally graded materials, and AI-driven process optimization have significantly improved mechanical, thermal, and electrical properties of polymer composites. This study serves as a comprehensive resource for researchers, engineers, and industry professionals, offering insights into emerging trends, material developments, process innovations, and future directions in AM of polymer composites.

Building Information Modelling (BIM) extends its utility to infrastructure management during the operational phase and can evolve into a Digital Twin (DT) when coupled with specific technologies or systems. In Engineering, Construction, and Operations (EC&O), BIM and DTs are strongly interconnected research topics. Especially for bridges, this relationship is represented by Bridge Information Modelling (BrIM) and Bridge Digital Twin (BDT). However, while this connection is recognised, it lacks developments regarding modelling strategies or data flow and integration. Therefore, the purpose of this study is to conduct a review of the current state of BrIM as an extension of BIM and its relationship with BDT, encompassing strategies for creating BrIM models of existing bridge assets. Additionally, it will explore integrating technologies or systems for structural performance monitoring and management (SPMM) to form BDTs. A systematic review was conducted using PRISMA protocol. Of the 3459 articles that were initially retrieved from a query of academic databases, 152 were assessed and classified manually, and 128 of these were selected for full content review. Analysis of the selected articles demonstrated the growing value of BDTs in SPMM of bridges, evolving from BrIM. Along with release of IFC4.3, BrIM development initiatives include IFC entity extension, IFC property sets usage, ontology development, and OpenBrIM implementation. Point cloud approaches are the most prevalent among different as-is BrIM modelling techniques, while parametric and data-driven approaches are gaining traction. Key challenges to BDT adoption, with respect to technological integration include interoperability, real-time performance, model updates, cost, and skill gaps.

Diffusion Models (DMs) have recently emerged as a highly effective category of deep generative models, achieving exceptional results in various domains, including image synthesis, video generation, and molecule design. This survey provides a comprehensive analysis of the expanding body of research on this topic. The primary objective of this study is to investigate the architecture and requirements of generative artificial intelligence systems. Initially, an analysis of the prerequisites and frontier ideas for the implementation of generative AI systems is performed. To clarify the operational mechanisms of the methodology, the design choices of DMs are thoroughly examined, covering aspects such as refinement, parallel generation, editing, in-painting, and cross-domain generation. This study extensively reviews fundamental DMs and their diverse applications in fields such as computer vision (CV), natural language processing (NLP), image synthesis, and interdisciplinary applications (scene generation, 3D vision, video modeling, medical image diagnosis, time-series analysis, audio generation, 3D molecule generation etc.) in other scientific domains. A comparative study for all the works that use generative AI methods for various downstream tasks in each domain is performed. A comprehensive study on datasets is also carried out. Finally, it discusses the limitations of current methods, as well as the need for additional techniques and future directions in order to make meaningful progress in this area.

This paper proposes a novel multiple time-weighted residual methodology with new insights to enable the design of generalized linear multi-step algorithms in computational dynamics. Leveraging single, double, and triple time-weighted residuals in single, two, and three-field forms, respectively, we develop a new generation of Generalized Single-Step Single-Solve algorithms for second-order time-dependent systems. This approach yields the GS4-II(_{p}), GS4-II(_{p,q}), and GS4-II(_{p,q,r}) computational frameworks, offering analysts a wide bandwidth of design options. Based on the proposed theory, we introduce the V0(^*_{text {TSS}}) schemes, which exhibit numerical properties comparable to those of the existing V0(^*) and traditional schemes, while offering the added benefit of the truly self-starting feature. The much coveted ZOO(_m) schemes (zero-order overshooting with m roots) are also synthesized to achieve second-order time accuracy in all variables, unconditional stability, zero-order overshooting, controllable numerical dissipation/dispersion, and minimal computational complexity. The relationship between the newly proposed computational frameworks and existing methods is analyzed via a comprehensive overview to date, most of which are included as subsets in the newly proposed methodology. Therefore, the multiple time-weighted residual methodology provides a new insight and in-depth understanding of the advances in the literature, showcasing the significance of the proposed theory. Finally, numerical examples from multidisciplinary applications, encompassing multi-body dynamics, structural dynamics, and heat transfer, are presented to substantiate the proposed methodology.

Sign Language Recognition (SLR) serves as a pivotal application of machine learning (ML) and deep learning (DL), enabling seamless automated communication between individuals with hearing impairments and the hearing population. Globally, there are approximately 7, 000 unique sign languages (SLs), characterized by diverse hand gestures, body movements, and facial expressions. These inherent variations add complexity to SLR systems, driving researchers to develop automated SLR (ASLR) frameworks to facilitate effective communication. To address the challenges posed by these variations, ASLR systems employ a range of advanced ML and DL methodologies to enhance accuracy. This study conducted a comprehensive review of 988 research articles retrieved from the SCOPUS database over the past two decades, employing relevant keywords to identify and analyze the prevailing trends in SLR research. The review provides a detailed evaluation of cutting-edge ML and DL techniques for hand gesture-based SLR, covering key aspects such as image acquisition, pre-processing, segmentation, feature extraction, and classification. The findings highlight that ensemble learning methods and transformer-based models outperform traditional approaches in terms of accuracy and robustness. Additionally, this study outlines critical challenges, open research questions, and potential future directions, offering valuable insights into advancing this field.

Additive manufacturing (AM) is a dynamic manufacturing process that provides new opportunities for creating products with intricate shapes and structures. AM, often known as Three Dimensional (3D) printing, has gained significant attention due to its technological developments, and the incorporation of artificial intelligence (AI) has further transformed its environment. This work aims to present the role of AI in various AM technologies and their industrial applications, highlighting the evolution of AM from a prototyping tool to standard manufacturing technology for final products. This review discusses the different AM technologies such as powder bed fusion (PBF), binder jetting (BJT), directed energy deposition (DED), and fused deposition modelling (FDM). This paper also covers artificial intelligence applications in design, process parameter optimization, quality control, material processing, reprocessing, and recycling. The outcomes reveal that the utilization of techniques like data acquisition coupled with Machine Learning (ML) algorithms is a foundational element bridging AM and AI. In addition, this review also addresses current challenges related to AI's role in advancing the evolution of AM technology while discussing potential areas for future research.