We investigate the capabilities of transfer learning in the area of structural health monitoring. In particular, we are interested in damage detection for concrete structures. Typical image datasets for such problems are relatively small, calling for the transfer of learned representation from a related large-scale dataset. Past efforts of damage detection using images have mainly considered cross-domain transfer learning approaches using pre-trained ImageNet models that are subsequently fine-tuned for the target task. However, there are rising concerns about the generalizability of ImageNet representations for specific target domains, such as for visual inspection and medical imaging. We, therefore, evaluate a combination of in-domain and cross-domain transfer learning strategies for damage detection in bridges. We perform comprehensive comparisons to study the impact of cross-domain and in-domain transfer, with various initialization strategies, using six publicly available visual inspection datasets. The pre-trained models are also evaluated for their ability to cope with the extremely low-data regime. We show that the combination of cross-domain and in-domain transfer persistently shows superior performance specially with tiny datasets. Likewise, we also provide visual explanations of predictive models to enable algorithmic transparency and provide insights to experts about the intrinsic decision logic of typically black-box deep models.

Bridge infrastructure managers are facing multiple challenges to improve the availability and serviceability of ageing infrastructure, while the maintenance planning is constrained by budget restrictions. Many research efforts are ongoing, for the last few decades, ranging from development of bridge management system, decision support tools, optimisation models, life cycle cost analysis, etc. Since transport infrastructures are deeply embedded in society, they are not only subject to technical requirements, but are required to meet the requirements of societal and economic developments. Therefore, bridge maintenance planning should accommodate multiple performance goals which need to be quantified by various performance indicators. In this paper, an application of Multi-Attribute Utility Theory (MAUT) for bridge maintenance planning is illustrated with a case study of bridges from the Netherlands road network. MAUT seeks to optimise multiple objectives by suggesting a trade-off among them and finally assigns a ranking to the considered bridges. Moreover, utility functions of MAUT appropriately account for the involved uncertainty and risk attitude of infrastructure managers. The main contribution of this study is in presenting a proof-of-concept on how MAUT provides a systematic approach to improve the decision-making of maintenance planning by making use of available data, accommodating multiple performance goals, their uncertainty, and preferences of infrastructure managers.

This paper introduces a comprehensive framework for the development of optimal multi-year maintenance plans for a large number of bridges. A maintenance plan is said to be optimal when, within the given budget, a maximum number of bridges can be maintained in the best possible year, achieving maximum performance with minimum socio-economic impact. The framework incorporates heuristic rules, multi-attribute utility theory, discrete Markov chain process, and genetic algorithms to find an optimal balance between limited budgets and performance requirements. The applicability of the proposed framework is illustrated on an extensive case study of highway bridges. The framework enables asset owners to execute various planning scenarios under different budget and performance requirements, where each resulting plan is optimal. The focus of this study has mainly been on highway bridges, however the framework is general and can be applied to any other infrastructure asset type.

The digital shift in society is making continuous growth of data. However, choosing a suitable storage architecture to efficiently store, process, and manage data from numerous sources remains a challenge. Currently, there are three storage architecture generations in practice, and the most recent one is Data Lakehouse. Given its novelty, limited research has been done into the rationale behind its introduction, strengths, and weaknesses. In order to fill this gap, this study aims to investigate the secure value (comparative strengths) of the data lakehouse architecture compared to data warehouse and data lake architectures. After conducting a comprehensive systematic literature review, we propose a data storage evolution model showing the comparative strengths and weaknesses of data warehouse, lake, and lakehouse architectures. With the use of the proposed model and expert interviews, this study demonstrates the secure value of the data lakehouse compared to the preceding architectures. In addition, the study presents a high-level view of the overlapping strengths of data Lakehouse with both data warehouse and data lake. In essence, the artifact produced by this study can be used to explain the rationale behind the evolution of data storage architectures. Further, the proposed model will help the practitioners in studying the trade-off between different architectures to offer recommendations. Finally, authors acknowledge that this study has several limitations, such as the limited sample size for the interviews and the bias due to the use of qualitative research approach. However, all the available measures were taken to minimize the effects of these limitations.

Optimal maintenance is one of the key concerns for asset-intensive industries in terms of reducing downtime and occurring costs. The advancement of data-driven technologies, affordable computing powers, and growing amounts of data introduced a paradigm with the name of predictive maintenance (PdM). PdM seeks to find out an optimal moment for the maintenance of an asset, where no early intervention leads to undue extra cost, and no late maintenance activity poses a safety risk. With the instrumentation of the cyber-physical system on assets, PdM transforms a typical structure into a smart structure that can send warnings in cases of near failure states. However, several practical challenges hamper the adoption of PdM solutions within industries. This article outlines a typical PdM modeling framework and its key components. Additionally, the adoption challenges, along with alternatives for implementation of the PdM solution are provided. This article concludes by offering several research directions that can accelerate the PdM adoption procedure.

Research into explainable artificial intelligence (XAI) methods has exploded over the past five years. It is essential to synthesize and categorize this research and, for this purpose, multiple systematic reviews on XAI mapped out the landscape of the existing methods. To understand how these methods have developed and been applied and what evidence has been accumulated through model training and analysis, we carried out a tertiary literature review that takes as input systematic literature reviews published between 1992 and 2023. We evaluated 40 systematic literature review papers and presented binary tabular overviews of researched XAI methods and their respective characteristics, such as the scope, scale, input data, explanation data, and machine learning models researched. We identified seven distinct characteristics and organized them into twelve specific categories, culminating in the creation of comprehensive research grids. Within these research grids, we systematically documented the presence or absence of research mentions for each pairing of characteristic and category. We identified 14 combinations that are open to research. Our findings reveal a significant gap, particularly in categories like the cross-section of feature graphs and numerical data, which appear to be notably absent or insufficiently addressed in the existing body of research and thus represent a future research road map.