As artificial intelligence advances toward cognitive capabilities, Large Model (LM) technologies have emerged as a novel paradigm for resolving the complexities of multi-source heterogeneous data in Maritime Emergency Management (MEM). This study presents a comprehensive survey of LMs throughout the MEM lifecycle, analyzing technical trajectories and application prospects. By characterizing the unstructured and sparse nature of maritime data, we elucidate the mechanisms of Large Language, Vision, and Multimodal Models, assessing their adaptability to maritime scenarios. The paper details specific applications across three critical stages: pre-event risk prevention and preparedness, during-event emergency response and decision support, and post-event accident investigation and analysis. Beyond application mapping, we critically interrogate bottlenecks in engineering deployment, specifically the “data silo” effect, the challenges of long-tail distribution, the risks associated with model hallucinations, and the constraints of shipboard computational resources. The review concludes by outlining future research frontiers, arguing that the development of domain-specific foundation models, deep semantic multimodal fusion, and the advancement of trustworthy and explainable AI (XAI) are pivotal for the profound integration of LMs into maritime safety systems.

With the acceleration of urbanisation, the problem of ground settlement in tunnel construction has become a serious challenge. Aiming at the existing ground settlement monitoring and resilience management problems such as limited monitoring range, neglected spatio-temporal characteristics, and poor combination of prediction results and resilience management, this paper proposes a two-stage spatio-temporal prediction and resilience optimization for tunnel-induced ground settlement. Specifically, firstly, this paper constructs a comprehensive monitoring architecture integrating SBAS-InSAR technology and Internet of Things (IoT) technology, which realises accurate and comprehensive monitoring of ground settlement. Secondly, a two-stage spatio-temporal prediction method of ground settlement is proposed: based on the wide-area spatio-temporal data acquired by SBAS-InSAR technology, the spatio-temporal transformer model is used to make the preliminary prediction. Then, with the small-area variables collected by IoT, the parameter-seeking optimisation algorithm based on the Grid Search-Particle Swarm is used in conjunction with the Time Convolutional-Bidirectional Long and Short-Term Memory Network (GR-PSO-TCN-BiLSTM) model to correct the prediction error. In terms of resilience optimisation, this paper proposes a multi-stage resilience enhancement strategy based on ground settlement prediction, which combines prevention importance, degradation importance and recovery importance, aiming to maximise the resilience of tunnel-induced ground settlement area. Finally, an empirical analysis using the traffic along the Zhengzhou Metro as an example verifies the effectiveness of the proposed method. These results indicate that coupling wide-area remote sensing with local IoT correction can substantially improve settlement prediction accuracy and provide actionable guidance for maintenance prioritization, thereby enhancing the robustness and recovery capability of metro systems.

Within the context of capability-based system-of-systems (SoS), the technology selection involves multiple stakeholders interactively participating in decision-making. In this article, a novel approach based on graph model for conflict resolution (GMCR) is proposed to handle the interactive technology selection decision across capability domains. First, a GMCR methodology based decision analysis framework for technology selection is presented, which allows decision-makers (DMs) with distinct risk attitudes, preference knowledge, and degrees of foresight to independently and interactively participate in technology selection. Then, the technology selection model is established by a four-step procedure that incorporates improved fuzzy option prioritization, followed by systematical technology selection analysis to provide strategic insights for identifying potential mutually accepted technology portfolios. Finally, an illustrative example is used to demonstrate the applicability and effectiveness of the proposed approach.

Deep Neural Networks (DNNs) are susceptible to adversarial attacks, which give rise to risks in real-world applications. Understanding the mechanism of DNNs against adversarial attacks and evaluating adversarial robustness are of great interest. However, existing interpretable methods hinder the understanding of DNNs' behavior under attack by utilizing redundant information like non-robust features. Additionally, most studies use adversarial attacks to evaluate the adversarial robustness of models. The lack of interpretability of these approaches makes it difficult for evaluators to trust and track the results obtained, which hinders the improvement of model robustness. This study proposes an explainable method to evaluate adversarial robustness based on robust paths. Specifically, during the process of model feature extraction, robust features, and non-robust features are separated through the Information Bottleneck (IB) theory. The robust features enable the identification of robust neurons in DNNs, facilitating the construction of robust pathways. By analyzing these robust paths, a deeper understanding of the decision behavior of the robust model under attack can be obtained. In addition, a quantitative metirc of adversarial robustness called Robust Neuron Coverage (RNC) is proposed. Finally, the superiority of the proposed method is validated through comprehensive experiments.

When a conflict occurs, the disputants involved and interested third parties usually expect to reach the desired equilibrium. To achieve this goal, inverse graph model for conflict resolution is an effective way to make the state of interest an equilibrium by ascertaining the required preferences. However, specifying crisp cost or effort of changing preferences over states can be challenging for decision makers (DMs) and third parties. As a result, a new inverse preference optimization model using interval optimization is introduced into the graph model by considering the uncertain cost of preference adjustment. First, the preference adjustment cost with uncertainty is formulated using interval number. Then, pessimistic preference ordering and DMs’ degrees of risk tolerance are utilized to compare cost intervals. After that, an inverse preference optimization model with uncertain adjustment cost is established. Finally, an illustrative example of the bulk water export conflict in Canada is presented to demonstrate the feasibility and effectiveness of the proposed approach.

Aiming at problems such as incomplete, inconsistent, and inaccurate requirements that often occur in the process of obtaining high-end equipment functional requirements, this paper presents a requirement quality improvement method. Referring to the agile development theory of requirements engineering, the quality improvement method constructs a functional requirement model based on user stories, defines the concept of functional requirement quality, designs functional requirement quality evaluation criteria, and constructs a functional requirement quality evaluation process. A case study and sensitivity analysis of new energy vehicle requirement development are conducted to confirm the feasibility and effectiveness of the method, and the experimental results show the superiority of the proposed method in improving the quality of high-end equipment functional requirements.

Time series is a common type of data that widely exists in various real-world scenarios, such as traffic flow data, network KPI, financial data, which can be regarded as time series. Anomaly detection in time series is an interesting research topic with a wide range of real-world applications, such as network intrusion detection, traffic situation monitoring and sensor error detection. In the real-world scenario, the frequency of anomalies is very low and little anomalous sample is available for analysis. Therefore, unsupervised methods are usually used for anomaly detection. In this paper, based on spectrum analysis and time series decomposition, an unsupervised deep framework for anomaly detection in time series data is designed. First, we decompose the original time series into trend series, seasonal series and residual series based on spectrum analysis. Then, prediction models based on long short-term memory (LSTM) networks and convolutional neural networks (CNNs) are designed to predict the trend series and seasonal series, respectively, and the residual series is reconstructed based on a Gaussian distribution. Next, the time series data are reconstructed by superimposing the predicted trend series and seasonal series and reconstructed residual series. Finally, we compare the original time series with the reconstructed time series and detect anomalies according to a certain threshold. The method proposed in this paper integrates prediction-based and reconstruction-based methods, and the experimental results on four datasets demonstrate the excellent performance of our method.

A belief-based preference structure along with its associated elicitation approach is incorporated into the graph model for conflict resolution (GMCR) to model and analyze the multistakeholder strategic conflicts involving ambiguous evidences, incomplete information, and nonlinear causal relationships. More specifically, the relative preference is first redefined using a belief structure capable of capturing uncertainties of various types such as vagueness and/or ignorance in subjective judgments regarding the complex real-world conflicts. Next, a flexible and realistic methodology based on evidential reasoning is put forward to elicit the belief preference information over feasible states within the GMCR model. Then, 16 stability definitions (solution concepts) are extended to accommodate diverse uncertainties in preferences and facilitate the informed conflict analysis. The application and interpretation of the foregoing preference structure and associated stability definitions are demonstrated with an illustrative example.

MOTIVATION: Modeling microbiome systems as complex networks are known as the problem of network inference. Microbial association network inference is of great significance in applications on clinical diagnosis, disease treatment, pathological analysis, etc. However, most current network inference methods focus on mining strong pairwise associations between microorganisms, which is defective in reflecting the comprehensive interactive patterns participated by multiple microorganisms. It is also possible that the microorganisms involved in the generated network are not dominant in the microbiome due to the mere focus on the strength of pairwise associations. Some scholars tried to mine comprehensive microbial associations by association rule mining methods, but the adopted algorithms are relatively basic and have severe limitations such as low calculation efficiency, lacking the ability of mining negative correlations and high redundancy in results, making it difficult to mine high-quality microbial association rules and accurately infer microbial association networks. RESULTS: We proposed a microbial association network inference method 'MANIEA' based on the improved Eclat algorithm for mining positive and negative microbial association rules. We also proposed a new method for transforming association rules into microbial association networks, which can effectively demonstrate the co-occurrence and causal correlations in association rules. An experiment was conducted on three authentic microbial abundance datasets to compare the 'MANIEA' with currently popular network inference methods, which demonstrated that the proposed 'MANIEA' show advantages in aspects of correlation forms, computation efficiency, adjustability and network characteristics. AVAILABILITY AND IMPLEMENTATION: The algorithms and data are available at: https://github.com/MaidiL/MANIEA.

Modern military operations in high-tech information warfare settings are dynamic processes involving various types of combat systems connected via multiple channels, which can be abstracted as a type of complex temporal combat network (TCN). The equipment contribution analysis in TCNs is of significant military value for optimizing operation process planning and improving network resilience in complex electromagnetic battlefields. This paper presents an integrated framework called capability oriented equipment contribution analysis (CECA) to analyze key equipment in TCNs. Specifically, a temporal network model is proposed to characterize TCNs by considering their dynamic nature and the heterogeneity capabilities of different types of functional entities during military operations. Based on this model, an operation capability contribution index is proposed to measure the contribution of each piece of equipment involved in executing operational tasks. Finally, the reliability and effectiveness of the CECA are demonstrated based on a TCN case study. This paper provides a detailed and precise quantitative analysis of the equipment contributions in TCNs, which yields useful insights for operation guidance and designing a more resilient combat system-of-systems.

A novel approach based on genetic algorithm (GA) is put forward to determine the possible relative preference required to reach the desired equilibria for the focal decision-makers (DMs) or the third party from the inverse perspective. More specifically, a framework of inverse graph model for conflict resolution (GMCR) modified from original GMCR incorporating GA's procedure is proposed to provide DMs with strategic insights toward inverse problems with conflict resolution. By taking full advantage of the optimization and search capability of GA, an improved preference calculation method is developed to help DMs focus limited resources on visionary strategies. Finally, an illustrative example is applied to demonstrate the applicability of the proposed approach in practice.

The identification and selection of emerging technologies has always been a hot field concerned by countries, armed forces and enterprises. Selecting emerging technologies from huge amounts of data is helpful to grasp technological frontiers and technological advantages. We use the unmanned system papers collected in Web of Science (WoS) database as datasets. Firstly, the bibliographic coupling network is constructed. And then the key technologies in the field of unmanned systems are identified by using the complex network community detection algorithm. Finally, the emerging technologies in the field of unmanned systems are screened according to the four indicators of novelty, popularity, influence and growth. We have successfully identified 113 key technologies in the field of unmanned systems and selected 10 of them as emerging technologies. The effectiveness and feasibility of the method have been verified by the evaluation of the research team, which is of great significance for the identification, assessment and prediction of technology.

The combat system-of-systems (CSoSs) in high-tech information warfare, composed of different types of combat systems connected through multiple interconnections, can be abstracted as a type of complex heterogeneous network. Research on the disintegration of the operational capability of a heterogeneous combat network (HCN) is of significant military value for optimizing the planning of operation process and improving network survivability in a complex electromagnetic battlefield. Accordingly, this paper presents an integrated framework called HCN operational capability disintegration based on link prediction (OCDLP) for modeling and solving the disintegration problem regarding the operational capability of an HCN. More specifically, a heterogeneous network model of a combat network under incomplete information is first established, considering different types of functional entities and information flows. Based on this model, a link prediction model is then applied to recover the original combat network structure while taking advantage of the observed information. Next, a disintegration evaluation metric of the operational capability of an HCN is elicited to evaluate the attack efficiency after a link prediction. Last, the reliability and effectiveness of the OCDLP are illustrated through a case study. The results provide useful insight into the operation guidance and design of a more resilient CSoSs.

Remaining useful life (RUL) estimation of aero-engine is a significant part of the aircraft reliability improvement and operational cost reduction. In order to solve the waste of performance data, in this study, a belief rule-based inference methodology using the evidential reasoning (RIMER) based RUL estimation process is proposed. Firstly, we take the principal component analysis (PCA) and the differential evolution (DE) algorithm to optimize the structure and parameter of the belief rule base (BRB), and then estimate the RUL of aero-engine through evidential reasoning (ER) algorithm. The practical case is investigated to validate the feasibility and efficiency of the proposed method. The experimental results show the RIMER based RUL estimation of aero-engine can support decision-making in the fields of aircraft prognostic and health management.

Search and Rescue at sea (SAR) is an important guarantee of maritime security. The similarity among search and rescue situations is the vital basis of generating instant search and rescue plan. On the basis of the process of maritime search and rescue, the influential factors of SAR situation are analyzed. Through establishing search and rescue sequence, a similarity measure model of maritime search and rescue situation is proposed based on concept lattice. Example analysis is involved.

The military technology planning is a key driving force for improving combat effectiveness. With regard to technological management support, the research on the architecture has been relatively weak. This paper proposed a model based technological management framework for defense acquisition. By analyzing the core decision elements in the technological management process, we define the domain-specific meta-model of the framework to capture the different needs of stakeholders. And then, six products of the technology view (TV) and a specific development process are designed to organize the concepts and relationships into models to support decision-making from different perspectives. The data consistency of models will be ensured by the meta-model. As a consequence, technical analysis activities can be performed based on the data and models in this technology architecture. This model-based approach ensures the availability, integrity, shareability, and compatibility of the architecture. Ultimately, the technology of shipborne unmanned helicopters (SUH) was explored by the ModelLink to demonstrate the applicability and effectiveness of the architecture.

A development planning approach combining game theory and network model is proposed to address the strategy selection and evolution for weapons system-of-systems (WSoS) characterized by an iterative and competitive development process between countries. More specifically, the development planning framework, including game player, strategy definition, and constraints (e.g., time and money), is first described, in which combat network is constructed to present the structure and evolution of WSoS. Next, the selection process of development planning strategy is studied based on damage accumulation and mitigation related to WSoS confrontation. Then, a competitive coevolution algorithm (CEA), reflecting that WSoS evolves along with the strategy selection change, is designed to find the optimal development strategy. Last, an illustrative example is used to demonstrate the feasibility and validity of the proposed approach.

The problem of network disintegration has broad applications and recently has received growing attention, such as network confrontation and disintegration of harmful networks. This paper presents an optimized disintegration strategy model for complex networks and introduces the GA optimization method into the network disintegration problem to identify the optimal disintegration strategy, which is a heuristic optimization algorithm and rarely applied to the study of network robustness. The efficiency of the proposed solution was verified by comparing it with other disintegration strategies used in a ER network with multi-edges. Numerical experiments suggest that our solution can improve the effect of network disintegration and that the “best” choice for edge failure disintegration can be identified through global searches. Our understanding of the optimal disintegration strategy may also shed light on a new property of the edges within network disintegration and deserves additional study.

A novel approach based on the graph model for conflict resolution (GMCR) methodology is proposed to address the problem of multistakeholder system portfolio decision analysis encountered in architecting a system of systems (SoS) with desired capabilities. More specifically, a flexible four-process framework for capability-based SoS architecting containing interactive portfolio decision analysis to promote multistakeholder design negotiations on system portfolio selections is presented. By taking full advantage of the inherent realistic and flexible design of the GMCR paradigm, an interactive portfolio decision analysis approach is designed to facilitate the systematic modeling and analysis of system portfolio decisions at the SoS level in order to achieve potential compromises among all key stakeholders having disparate preferences and interacting according to different conflict behavior patterns. This approach permits the prediction of possible mutually agreeable system portfolios for SoS architecture development. Last, the feasibility of the proposed approach is demonstrated using an illustrative example.