Search

Accurate performance condition evaluation has a pivotal role in maintaining the operating reliability and preventing damage to complex electromechanical systems (CESs), which is still a challenging task. The uncertain features fusion inspired method is developed by utilizing the data-graph conversion, texture analysis, and improved evidence fusion. Unlike the conventional continuous time-series analysis-based methods, the 2D color-spectrums related to the performance conditions are constructed without information losing, and texture features of spectrums are extracted and fused to realize evaluation. The effectiveness of the proposed method is verified by actual evaluation applications. Moreover, the proposed method provides a new idea for large-scale high-dimensional data processing, decision making, uncertainty handling, and other engineering applications. • A texture based method for performance condition evaluation is proposed. • Monitoring data is converted into 2-D color spectrum without information losing. • Texture features based on self-similarity and dissimilarity are extracted. • Non-parametric features fusion was proposed to handle with uncertain information. [ABSTRACT FROM AUTHOR]

• A physics-guided, data-refined fault root cause tracing framework is proposed. • A physics-guided initial hierarchical network model is established. • Operation data-refined model updating algorithm is designed. • A novel bidirectional reasoning-based fault node ranking strategy is proposed. • A fault root cause tracing application of an offshore wind turbine is conducted. Fault root cause tracing (FRCT) is critical for the safety assurance of complex electromechanical systems. However, it is still a challenging task due to the complexity, uncertainty and time-varying characteristics of limited known fault development and propagation mechanism. Therefore, this paper proposed a physics-guided, data-refined FRCT framework. First, a physics-guided hierarchical fault root cause tracing network (HFTN) model is defined and constructed based on the statistics fault mechanism while considering fault development and propagation characteristics including network, hierarchy, and uncertainty. Second, an operation data-refined algorithm is designed to update the initial model, where Wasserstein Generative Adversarial Network and Long Short-Term Memory-based local anomalies detection, and statistical failure laws-based global dynamic fault mechanism reflection are introduced. Third, a novel bidirectional probabilistic reasoning strategy is developed to rank the real-time probabilities of fault causes in HFTN, which combines both faults reverse diagnostic and forward predictive knowledge to improve the results stability. The research is evaluated by an offshore wind turbine FRCT application, the research-assisted dynamic reliability analysis and identification of compound fault are also explored for potential application. This research combines common and individual properties of faults, has excellent accuracy and interpretability, and is expected to support integrated research of system safety. [ABSTRACT FROM AUTHOR]

Evaluation of the condition of complex electromechanical systems in the process industry is one of the most important purposes of condition monitoring, and is an indispensable step to ensure safe operation and comprehensive coverage capabilities of a system. However, there are still difficulties in obtaining a precise evaluation result from uncertain, incomplete and even conflicting system monitoring data, and this is a key step for condition evaluation. Since evidence theory has shown high efficiency in handling uncertain information, an evidence fusion-based framework for condition evaluation has been presented in this paper to improve the certainty and precision of evaluation decisions by fusing features extracted from different sources of evidence. The proposed framework contains key points for condition evaluation that are driven by data, and evidence fusion is at the core of this method. First, the frame of discernment has been automatically constructed using time-series based clustering. Second, a kernel density estimation based non-parametric method for determining the basic probability assignment of evidence has been proposed. After combination, the conditions can be evaluated using pignistic probability. An actual condition evaluation requirement of complex electromechanical systems in the process industry has been used to verify the effectiveness of the proposed framework and to compare it with existing methods. This framework can handle common problems of condition evaluation and overcome some drawbacks of other existing similar methods since no particular distribution is assumed and a prior knowledge of system conditions is not required. Furthermore, it can be flexibly used in many engineering applications. [ABSTRACT FROM AUTHOR]

• A novel F2G model is proposed for the reliability modeling of coupled faults. • The functional logics and fault logics are integrated in F2G model. • All the standardized meta models and calculation rules used in F2G are defined. • The proposed method is validated by empirical case. Reliability analysis plays a crucial role in revealing the failure causes and determining the improvement measures for reliability growth. However, reliability analysis of complex equipment with coupled faults still corresponds to a challenging task, due to unclear coupling mechanism and unsuitable analysis model. A down-top, deductive modeling method, named as fault-function graph (F2G), is proposed. First, the meta models are defined to normalize all the atomic faults, coupling relations and coupling forms in modeling. Next, an initial fault model is constructed based on typical fault-relations and coupling forms. Furthermore, the functional hierarchy of fault determined by IDEF0 is appended. Lastly, the rigorous modeling rules and computing processes are explained based on an actual case. As a graphical modeling method, it handles the coupling faults by integrating the system functional and fault information. Exploiting the advantages of conventional models, the coupling relations are quantified, and the false relations are detected based on functional constraints. Therefore, the proposed method can be used flexibly in the reliability modeling of coupled faults. Moreover, it provides a foundation for the comprehensive and dynamic reliability analysis and the failure mechanism mining of complex equipment, and it can be used in other engineering applications as well. [ABSTRACT FROM AUTHOR]

• We propose an effective combination method based on Dempster–Shafer Theory for ensemble classifiers. • The proposed method measures the outputs of member classifiers using a combinational weight. • The combinational weight consists of objective weight and subjective weight. • The objective weight is defined based on the support degree among classifiers. • The subjective weight is defined based on support degree of within classifier. Aiming at the poor performance of individual classifier in the field of fault recognition, in this paper, a new ensemble classifier is constructed to improve the classification accuracy by combining multiple classifiers based on Dempster–Shafer Theory (DST). However, in some specific cases, especially when dealing with the combination of conflicting evidences, the DST may produce counter-intuitive results and loss its advantages in combining classifiers. To solve this problem, a new improved combination method is proposed to alleviate the conflicts between evidences and a new ensemble technique is developed for the combination of individual classifiers, which can be well used in the design of accurate classifier ensembles. The main advantage of the proposed combination method is that of making the combination process more efficient and accurate by defining the objective weights and subjective weights of member classifiers' outputs. To verify the effectiveness of the proposed combination method, four individual classifiers are selected for constructing ensemble classifier and tested on Tennessee-Eastman Process (TEP) datasets and UCI machine learning datasets. The experimental results show that the ensemble classifier can significantly improve the classification accuracy and outperforms all the selected individual classifiers. By comparison with other combination methods based on DST and some state-of-the-art ensemble methods, the proposed combination method shows better abilities in dealing with the combination of individual classifiers and outperforms the others in multiple performance measurements. Finally, the proposed ensemble classifier is applied to the fault recognition in real chemical plant. [ABSTRACT FROM AUTHOR]