Search

Nonuniform thermal behavior in lithium-ion battery packs can accelerate aging, leading to inconsistent cell performance. If not adequately monitored and managed, this heating can give rise to unwanted side reactions, fires, and explosions, underscoring the criticality of temperature field reconstruction. In recent years, data-driven methods have gained popularity for addressing the temperature field reconstruction problem. However, many existing data-driven approaches require retraining when system parameters change, such as the initial temperature distribution or working conditions. This article presents a deep transfer operator learning method named physics-informed adversarial networks. The model architecture incorporates transformer blocks to capture comprehensive time and space features. Additionally, to enhance interpretability and generalization, the model introduces two effective mechanisms: 1) the integration of thermal partial differential equations to ensure compliance with physical laws; and 2) the application of domain adversarial mechanism in transfer learning to extract domain-invariant feature representations. These mechanisms enable the model to effectively reconstruct the temperature field, even in unencountered scenarios during training. The proposed method is validated under real-world energy storage working conditions, demonstrating superior performance compared to state-of-the-art deep learning methods. Notably, the approach exhibits excellent performance even when confronted with the limited availability of training data.

Individuals often make decisions in a social environment where social influence can impact on people’s decision-making domains such as online purchasing, political voting and voluntary vaccine uptake. Social influence can be recognised as the intentional or unintentional change in an individual’s belief, perception, or behaviours caused by an information diffusion process embedded in a social network. However, there is limited research on how this diffusion process is shaped by the topology or structure of the social network. This work provides an exploratory and systematic analysis of how decision-making outcomes in a population can be affected by both the structure of the social network and the starting node of where new information starts to diffuse. Simulation results considering three common network structures highlight how social networks with clear community structures lead to a larger absolute impact on decision-making outcomes and networks where the social connections follow a preferential attachment rule show the largest relative impact than the others. The results also suggest scenarios in which introducing new pieces of information to the social network can facilitate the information diffusion process and produce a more significant impact in terms of the overall population decision-making process.

In group decision-making (GDM) problems, the elicitation of preferences from decision makers (DMs) and consensus reaching process are two critical issues to be resolved. Many kinds of preference relations have been proposed in the recent two decades. In this paper, we propose a novel pairwise comparison structure called fuzzy-uncertainty-based distributed preference relation (FUDPR), which could not only indicate the intensity of preferred, non-preferred, indifferent and uncertain information of one alternative against another simultaneously, but also has the ability to represent local ignorance expressed by DMs. The FUDPR introduces the set of consecutive evaluation grades into the distributed preference relation (DPR) where only discrete single evaluation grades and global ignorance can be modeled in the preference relation. The missing values in FUDPR are estimated by the defined score function. Two kinds of consensus reaching process (CRP) are then proposed, namely soft and hard consensus mechanism, respectively, which are suitable for different decision-making situations. Both of them consider the individual dissimilarity, trust relations and self-confidence in the designed feedback mechanism. Comparative analysis is performed between the proposed method and other state-of-the-art pairwise comparison GDM models. An illustrative example of life cycle sustainability assessment (LCSA) is proposed to demonstrate the rationality and effectiveness of the proposed method in supporting real-world GDM problems.

So far, there has been only oneoptimization objective in belief rule base (BRB) learning-related studies, either modeling accuracy via parameter learning, modeling complexity via structure learning, or a unified objective representing them both. This article proposes a new cooperative and distributed multiobjective approach for heterogeneous BRB (CDMO-BRB) to simultaneously optimized its structure and parameters. The new CDMO-BRB approach is implemented in a parent–child framework. In the parent process, cooperative optimization is equipped with a multipopulation strategy and another redundant-gene strategy to simultaneously optimize the structure and parameters of BRBs in heterogeneous structures. In the child process, distributed learning is designed to improve optimization efficiency and avoid early convergence and overfitting. A benchmark case on pipeline leakage detection is studied. Case study results show that the structure and parameters of BRB can be optimized simultaneously by pushing forward the Pareto front. In comparison with previous studies, CDMO-BRB has produced superior performance. Moreover, distributed learning is also separately validated.

Ferries are usually used for transporting passengers and vehicles among docks, and any deviation of the course can lead to serious consequences. Therefore, transportation ferries must be watched closely by local maritime administrators, which involves much manpower. With the use of historical data, this article proposes an intelligent method of integrating artificial potential field with Bayesian Network to trigger deviation warnings for a ferry based on its trajectory, speed and course. More specifically, a repulsive potential field-based model is first established to capture a customary waterway of ferries. Subsequently, a method based on non-linear optimisation is introduced to train the coefficients of the proposed repulsive potential field. The deviation of a ferry from the customary route can then be quantified by the potential field. Bayesian Network is further introduced to trigger deviation warnings in accordance with the distribution of deviation values, speeds and courses. Finally, the proposed approach is validated by the historical data of a chosen ferry on a specific route. The testing results show that the approach is capable of providing deviation warnings for ferries accurately and can offer a practical solution for maritime supervision.

For medical diagnosis, fuzzy Dempster-Shafer theory is extended to model domain knowledge under probabilistic and fuzzy uncertainty. However, there are some information loss using discrete fuzzy sets and traditional matching degree method. This study aims to provide a new evidential structure to reduce information loss. This paper proposes a new intuitionistic fuzzy evidential reasoning (IFER) approach which combines intuitionistic trapezoidal fuzzy numbers and inclusion measure to improve the accuracy of representation and reasoning. The proposed approach has been validated by a stroke diagnosis. It is shown that the IFER approach leads to more accurate results.

Abstract: Belief rule based (BRB) system provides a generic inference framework for approximating complicated nonlinear causal relationships between antecedent inputs and output. It has been successfully applied to a wide range of areas, such as fault diagnosis, system identification and decision analysis. In this paper, we provide analytical and theoretical analyses on the inference and approximation properties of BRB systems. We first investigate the unified multi-model decomposition structure of BRB systems, under which the input space is partitioned into different local regions. Then we analyse the distributed approximation process of BRB systems. These analysis results unveil the underlying inference mechanisms that enable BRB systems to have superior approximation performances. Furthermore, by using the Stone–Weierstrass theorem, we constructively prove that BRB systems can approximate any continuous function on a compact set with arbitrary accuracy. This result provides a theoretical foundation for using and training BRB systems in practical applications. Finally, a numerical simulation study on the well-known benchmark nonlinear system identification problem of Box–Jenkins gas furnace is conducted to illustrate the validity of a BRB system and show its inference and approximation capability. [Copyright &y& Elsevier]

Gradient search based neural network training algorithm may suffer from local optimum, poor generalization and slow convergence. In this study, a novel Memetic Algorithm based hybrid method with the integration of “extremal optimization” and “Levenberg-Marquardt” is proposed to train multilayer perceptron (MLP) networks. Inheriting the advantages of the two approaches, the proposed “EO-LM” method can avoid local minima and improve MLP network learning performance in generalization capability and computation efficiency. The experimental tests on two benchmark problems and an application example for the end-point-prediction of basic oxygen furnace in steelmaking show the effectiveness of the proposed EO-LM algorithm.

With the QCD sum rules approach, we study the newly discovered doubly heavy baryon Ξcc++. We analytically calculate the next-to-leading-order (NLO) contribution to the perturbative part of the JP = 1/2+ baryon current with two identical heavy quarks, and then reanalyze the mass of Ξcc++ at the NLO level. We find that the NLO correction significantly improves both scheme dependence and scale dependence, whereas it is hard to control these theoretical uncertainties at leading order. With the NLO contribution, the baryon mass is estimated to be mΞcc++ = 3.66-0.10+0.08 GeV, which is consistent with the LHCb measurement. [ABSTRACT FROM AUTHOR]

Abstract The threat level assessment of a suspect object is instrumental for developing combatant responses and effecting situational awareness. In practical conditions, it suffers from four challenges: (1) inputs are gathered from heterogeneous sources, which could also result in heterogeneous formats and types, (2) historic records are insufficient owing to limited number of suspect objects and under-developed detection technologies and devices, (3) missing information resulted from inadequacy of input gathering, exchanging, and storage, and (4) the assessment process should be accessible to experts and decision makers to provide reasonability. In this study, the belief rule base (BRB) under a disjunctive assumption (disjunctive BRB) is applied because it can handle heterogeneous and insufficient information as well as provide access to experts and decision makers. Furthermore, a new disjunctive BRB spreading approach with six steps is proposed to specifically handle the missing information challenge. First, multiple sub-BRBs are initialized and then optimized using heterogeneous and insufficient information. Sub-BRBs are in different belief structures, which are part of the complete belief structure. Second, a self-organizing map (SOM)-based procedure is proposed to integrate multiple sub-BRBs into one final BRB (f-BRB) in the complete belief structure with consideration of the weights of sub-BRBs. A practical threat level assessment case is studied to validate the efficiency of the proposed disjunctive BRB spreading approach. The case study results show that (1) optimization can help improve the modeling accuracy of sub-BRBs to different degrees while initial sub-BRBs have shown varied performances, and (2) f-BRB produces the highest modeling accuracy on both historic records and newly detected suspect objects. [ABSTRACT FROM AUTHOR]