As the classical analytics in the field of data-driven fault monitoring, time neighborhood preserving embedding (TNPE), dynamic inner canonical correlation analysis (DiCCA), and slow feature analysis (SFA) provided th...
详细信息
As the classical analytics in the field of data-driven fault monitoring, time neighborhood preserving embedding (TNPE), dynamic inner canonical correlation analysis (DiCCA), and slow feature analysis (SFA) provided three different choices for characterizing time-serial variation inherent in sequential samples. Considering the unsupervised nature of these three algorithms as well as their variants, it could be more appropriate to jointly exploit time-serial variation from multiple perspectives in a comprehensive manner. This recognition then motivates us to propose a novel dynamic modeling algorithm titled as joint time-serial variation analysis (JTSVA) for fault monitoring. The proposed JTSVA aims to extract dynamic latent variables (DLVs) with respect to a joint integration of time-manifold embedding, latent auto-regressive, and slow-varying capabilities own by TNPE, DiCCA and SFA, respectively. Furthermore, an additional orthogonality constraint is further assigned to the problem definition of JTSVA, so that the extracted DLVs could have enhanced discriminant in uncovering valuable information for satisfactory fault monitoring performance. Finally, the superiority of JTSVA in fault monitoring, in terms of false alarm rate and fault detection rate, is validated through comparative experiments on two industrial-scale examples, i.e., the Tennessee Eastman benchmark process and a real-world multiphase flow facility.
The analysis of heterogeneous effects on traffic crashes is crucial for understanding their causal mechanisms and enhancing targeted safety management strategies. However, current methodologies for modeling crash hete...
详细信息
The application of physics-informed neural networks (PINNs) in fault-tolerant control (FTC) systems of electric vehicles has gathered considerable interest in using underlying physics to improve the fault diagnosis an...
详细信息
The application of physics-informed neural networks (PINNs) in fault-tolerant control (FTC) systems of electric vehicles has gathered considerable interest in using underlying physics to improve the fault diagnosis and mitigation process. PINNs, which include the governing physical equations in the neural network training process, allow for accurate modeling of the EV components, such as motors and inverters. This review aims to evaluate neural networks, especially PINNs, for fault diagnosis and FTC development in the context of EVs. It includes neural network structures, algorithms for training, methods based on physical analogies, and the application of physical principles to enhance the algorithms. The comparative analysis presents the merits of PINNs against conventional techniques, including PID, LQR, and Kalman Filters, regarding model fitness, data utilization, adaptability, computational footprint, resilience, and extensibility. Future research directions include extension works of PINNs integrating them into conventional approaches, dynamic adaptation, multidisciplinary, and EV self-powered systems.
作者:
Meng, XuranCao, YuanZou, DifanUmich
Dept Biostat Ann Arbor MI 48109 USA HKU
Dept Stat & Actuarial Sci Hong Kong 999077 Peoples R China HKU
Inst Data Sci Dept Comp Sci Hong Kong 999077 Peoples R China
Gradient regularization, as described in Barrett and Dherin (in: International conference on learning representations, 2021), is a highly effective technique for promoting flat minima during gradient descent. Empirica...
详细信息
Gradient regularization, as described in Barrett and Dherin (in: International conference on learning representations, 2021), is a highly effective technique for promoting flat minima during gradient descent. Empirical evidence suggests that this regularization technique can significantly enhance the robustness of deep learning models against noisy perturbations, while also reducing test error. In this paper, we explore the per-example gradient regularization (PEGR) and present a theoretical analysis that demonstrates its effectiveness in improving both test error and robustness against noise perturbations. Specifically, we adopt a signal-noise data model from Cao et al. (Adv Neural Inf process Syst 35:25237-25250, 2022) and show that PEGR can learn signals effectively while suppressing noise memorization. In contrast, standard gradient descent struggles to distinguish the signal from the noise, leading to suboptimal generalization performance. Our analysis reveals that PEGR penalizes the variance of pattern learning, thus effectively suppressing the memorization of noises from the training data. These findings underscore the importance of variance control in deep learning training and offer useful insights for developing more effective training approaches.
AimDynamic cancer control is a current health system priority, yet methods for achieving it are lacking. This study aims to review the application of system dynamics modeling (SDM) on cancer control and evaluate the r...
详细信息
AimDynamic cancer control is a current health system priority, yet methods for achieving it are lacking. This study aims to review the application of system dynamics modeling (SDM) on cancer control and evaluate the research *** were searched in PubMed, Web of Science, and Scopus from the inception of the study to 15 November 2023. Inclusion criteria were English original studies focusing on cancer control with SDM methodology, including prevention, early detection, diagnosis and treatment, and palliative care. Exclusion criteria were non-original research, and studies lacking SDM focus. analysis involved categorization of studies and extraction of relevant data to answer the research question, ensuring a comprehensive synthesis of the field. Quality assessment was used to evaluate the SDM for cancer *** studies were included in this systematic review predominantly from the United States (7, 43.75%), with a focus on breast cancer research (5, 31.25%). Studies were categorized by WHO cancer control modules, and some studies may contribute to multiple modules. The results showed that included studies comprised two focused on prevention (1.25%), ten on early detection (62.50%), six on diagnosis and treatment (37.50%), with none addressing palliative care. Seven studies presented a complete SDM process, among which nine developed causal loop diagrams for conceptual models, ten utilized stock-flow charts to develop computational models, and thirteen conducted *** review's macrofocus on SDM in cancer control missed detailed methodological analysis. The limited number of studies and lack of stage-specific intervention comparisons limit comprehensiveness. Detailed analysis of SDM construction was also not conducted, potentially overlooking nuances in cancer control *** in cancer control is underutilized, focusing mainly on early detection and treatment. Inconsistencies suggest a need for sta
Context: Automated analysis of code review comments (CRCs) can aid in highlighting frequently discussed issues by reviewers from large repositories. Topic modeling is a promising approach to analyzing large natural la...
详细信息
ISBN:
(纸本)9783031783852;9783031783869
Context: Automated analysis of code review comments (CRCs) can aid in highlighting frequently discussed issues by reviewers from large repositories. Topic modeling is a promising approach to analyzing large natural language repositories. However, CRCs contain natural language text and code references;thus, data pre-processing and topic modeling approaches must be carefully selected. Objective: This work aims to discuss the various decisions taken and considerations involved in the analysis of CRCs. Method: We utilized 5,560 CRCs from an open-source system to study the decisions and considerations faced during the analysis of CRCs using topic modeling, followed by an evaluation of the interpretability of identified themes by a domain expert. Results: We report several observations and challenges in improving the quality of the identified themes, including choices regarding the pre-processing, topic modeling parameters, embedding model, and objective measures of coherence used, which impact the subjective interpretability of the identified themes. Conclusions: This work offers unique considerations, and the impact of these decisions can facilitate future studies in conducting topic modeling-based analyses of CRCs. Future studies can utilize the technical demonstrator to explore the interpretability of the topics generated from CRCs.
This paper introduces Multistep-ahead and Interpretable Sequential modeling Scheme (MISMS), a pioneering approach for long-term dynamic forecasting that enhances model interpretability in process industries. MISMS is ...
详细信息
This paper introduces Multistep-ahead and Interpretable Sequential modeling Scheme (MISMS), a pioneering approach for long-term dynamic forecasting that enhances model interpretability in process industries. MISMS is distinguished by its innovative integration of multivariate analysis, multistep-ahead forecasting, and sequential model interpretability. The scheme incorporates knowledge embedding and Exploratory dataanalysis (EDA) to structure multiple variables and extract their temporal attributes. Based on this groundwork, sequence-to-sequence (Seq2Seq) submodels are developed to capture domain-wide dynamics rather than one single value. Crucially, MISMS has pioneered the application of sequence interpretation for model decomposition and transparency, improving interpretability. The proposed scheme employs different temporal models in practical scenarios and can ensure a better prediction with long-term accuracy and stability. Additionally, discussions on the model explainer's expected and unexpected outcomes are conducted, providing potential avenues for future research in process industries.
Planetary gearboxes (PGs) serve as vital transmission links in rotating machinery, and diagnosing faults within them is crucial for effective maintenance. Traditional deep learning methods often operate as "black...
详细信息
Planetary gearboxes (PGs) serve as vital transmission links in rotating machinery, and diagnosing faults within them is crucial for effective maintenance. Traditional deep learning methods often operate as "black boxes," offering limited transparency in interpreting results, especially when analyzing the complex vibration signals of PGs. To address this issue, this paper proposes a co-modulation model combined with a hybrid resolution strategy (CHRS), leveraging amplitude modulation (AM) and frequency modulation (FM) intensities, to enhance the interpretability of fault diagnosis. First, a more comprehensive and adaptable expression of the co-modulation model is developed to describe gear faults. Second, CHRS links the model's generated signal with the actual monitoring data, establishing an intrinsic connection between the mathematical model and the data. An updating mechanism based on partial differential analysis is established for model parameter estimation. A partial differential-based updating mechanism is employed for model parameter estimation, enabling the quantitative analysis of model coefficients (including AM and FM), even with a limited number of training samples. Finally, the support vector machine (SVM) is employed to train and test these model parameters, facilitating the identification of different fault types through experimental data, thus validating the effectiveness of CHRS. In summary, CHRS significantly improves the interpretability of PG fault diagnosis by enhancing both the modelingprocess and quantitative analysis of vibration signals.
In recent years, deep learning has revolutionized fields such as computer vision, speech recognition, and natural language processing, primarily through techniques applied to data in Euclidean spaces. However, many re...
详细信息
In recent years, deep learning has revolutionized fields such as computer vision, speech recognition, and natural language processing, primarily through techniques applied to data in Euclidean spaces. However, many real world applications involve data from non-Euclidean domains, where graphs naturally represent entities and their complex interdependencies. Traditional machine learning methods have often struggled to process such data in an effective manner. Graph Neural Networks represent a crucial advance in the use of deep learning to interpret and extract knowledge from graph-based data. They have opened up new possibilities for tasks such as node categorization, link inference, and comprehensive graph analysis. This paper provides a detailed analysis of Graph Neural Network (GNN) methodologies, emphasizing their architectural diversity and wide ranging applications. GNN models are systematically categorized into fundamental frameworks such as message passing paradigms, spectral and spatial methods, and advanced extensions such as hypergraph neural networks and multigraph approaches. This paper also explores domains such as social network analysis, molecular biology, traffic forecasting, and recommendation systems. In addition, it emphasizes some critical open challenges, including scalability, dynamic graph modeling, and robustness against noisy or incomplete data. The paper concludes with a proposal for future research directions to improve the scalability, interpretability, and adaptability of GNNs in this fast-evolving field.
Time delay is an inherent characteristic of real-world phenomena which may affect the system's characteristic. The systems including delay are known as time-delay systems, they are represented using delay differen...
详细信息
Time delay is an inherent characteristic of real-world phenomena which may affect the system's characteristic. The systems including delay are known as time-delay systems, they are represented using delay differential equations. modeling, discretisation, stability and control design for time-delay systems are still challenging in modern control theory. This paper systematically overviews available discretisation methods of linear and nonlinear time-delay systems. Emphasis is placed on illustrating fundamental results and recent progress on discretisation methods for delay systems. Numerous methods for the discretisation of linear and nonlinear systems considering input delays, state or output delays in the system's dynamics have been presented. A particular attention will be paid to illustrate effects of the discretisation process on the stability of discretised systems. Examples of mathematical descriptions, problems, and performance analysis for delay systems are presented. The presentation of discretisation methods is as easy as possible, focussing more on the main ideas and mathematical concepts by analogy. Finally, some possible future research directions to be tackled by researchers in this field are discussed.
暂无评论