This paper presents a novel algorithm to assess the health status in monitoring sensor data using a kernel-based support vector machine (SVM) approach. Today, accurate fault prediction is a key issue raised by maintenance. In particular, automatically modelling the normal behaviour from condition monitoring data is probably one of the most challenging problems, specially when there is limited information of real faults. To overcome this difficulty, a data-driven learning framework based on nonparametric density estimation for outlier detection and ν-SVM for normality modelling, with optimal bandwidth selection, is proposed. A health score based on the log-normalisation of the distance to the separating hyperplane is also provided. Experimental results obtained when analysing the propagation of a critical fault over time in a marine diesel engine demonstrate the validity of the algorithm. The predictions of normality models learned were compared to those of the k-nearest neighbours (kNN) method. Low false positive rates on healthy data and improved prediction capacities are achieved.

Kernel-based support vector machines for automated health status assessment in monitoring sensor data

A circuit breaker fault diagnosis method comprises the steps of obtaining the working state historical data of a circuit breaker in a power grid, wherein the historical data comprises the electrical quantity data and the non-electrical quantity data; utilizing the historical data as the training samples; and establishing a circuit breaker fault diagnosis model by combining the to-be-measured circuit breaker data and based on a hierarchical Bayesian model.

Circuit breaker fault diagnosis method

In the power energy industry there is a growing interest in measuring and improving the reliability and security of the distribution energy grid. In this process, the power quality (PQ) disturbances have become a major concern due the uncertainty associated to possible affectations of these kinds of disturbances to the user loads and the utility operation. As consequence, probabilistic and graphical methods have to be implemented in order to locate the most probable cause and effect of certain PQ disturbance helping the analyst in the decision making process. Bayesian networks are a useful tool to include the fault knowledge model in probabilistic inference algorithms to determinate the most probable causes and effects of a PQ disturbance in a modeled situation. This paper presents the development of a tool for PQ diagnostics implemented with National Instruments LabVIEW®. The tool is applied for SAG diagnostics in a given scenario; the results reveals that the tool can diagnose the disturbance with high accuracy in each simulation, and represents a good complement for PQ measuring instruments.

Smart Diagnosis of power quality disturbances using Bayesian networks

Maintainability is an important general quality characteristic of products. Insufficient maintainability will lead to long maintenance time and high maintenance cost, thus affecting the availability of products. Maintainability verification is an important means to ensure maintainability meets design requirements. However, the cost of traditional real maintainability verification method is very high, and the virtual maintenance method has insufficient verification accuracy due to the lack of large maintenance force feedback when the human body is moving. In order to reduce the evaluation error and test sample size, the paper conducts maintainability verification based on the mixed physical and virtual maintainability test scenarios. Aiming at the problem that traditional methods are difficult to deal with the real test information and synchronous virtual simulation information in the test process, this study proposes a virtual–real fusion maintainability evaluation algorithm based on adaptive weighting and truncated SPOT (Sequential Posterior Odd Test) method. It can weigh real test information and virtual human simulation information adaptively to obtain a virtual–real fusion maintainability test sample. Then, the SPOT method is used to evaluate the maintainability of small samples. The adjustment of valve clearance, replacement of air filter element and replacement of starting motor maintenance tasks of ship engine are taken as examples for demonstration. The virtual–real fusion and virtual maintainability verification methods are respectively used for verification, and compared with the physical maintenance scenario constructed by 3D printing, indicating that the accuracy of virtual–real fusion maintainability test verification is 89%, while the virtual maintainability verification is only 33%. 

/pdf/virtual-real-fusion-maintainability-verification-based-on-1daxqorn.pdf

Virtual–real fusion maintainability verification based on adaptive weighting and truncated spot method

There is a complex coupling between the health status of the equipment, the maintenance program, and the mission performed. Scientific and rational maintenance programs need to be planned according to the specific health status in order to ensure the mission indicator requirements of the equipment. For the mission-oriented equipment maintenance decision optimization problem, a method of mission-oriented equipment maintenance program evaluation and optimization is studied in this paper. Firstly, differentiated maintenance programs are developed based on the health status of the components; then the mission success probability and mission reliability under each maintenance program are evaluated with the help of Monte Carlo numerical simulation algorithms; at the same time, the sensitivity analysis list generated by the simulation process is used to identify the current maintenance problems and to provide guidance for improving the maintenance program; finally, a maintenance assistance decision system is designed and a case study is used to verify the scientific and operational properties of the proposed method. 

Research on Mission-oriented Equipment Maintenance Program Evaluation and Optimization Method

A method for the maintainability evaluation at design stage using maintainability design attributes

Magnetically-coupled resonant wireless power transfer (MCR-WPT) technology is one of the most promising technologies to power embedded sensors. In many practical situations, embedded sensors are located in a metal environment. The nearby metal materials will affect the performance of the MCR-WPT system. In this paper, the equivalent circuit model of the MCR-WPT system in metal environment is established and the factors affecting the transmission efficiency is analyzed. Then, simulation analysis is performed using Maxwell software and Simplorer software. Two cases are considered in the simulation study: the coil is close to a metal plate and the coil is close to a metal shaft. Simulation results show that metal materials will change the coil inductance and interfere with the magnetic field distribution of the MCR-WPT system, which in turn affects the transmission efficiency of the system.

Modeling and Analysis of Wireless Power Transfer System in Metal Environment

The invention provides a method for detecting the positions of binary rocker switches based on image processing The method comprises the following steps of: A, acquiring rocker switch images; B, extracting rocker switch head images based on the characteristics of the acquired rocker switch images; C, positioning respective rocker switch working areas; D, according to the coordinates of respective rocker switch working areas, cutting the rocker switch head area images to obtain the sub-images of the respective rocker switch head areas; E, optimizing the sub-images acquired in the step D; and F, determining the states of respective rocker switches by comparing the rocker switch head pixel area coordinates and image center coordinates in respective sub-images The method can analyze the rocker switches to determine the positions of the rocker switches, and provides a reference for diagnosing switch position abnormality to prevent risks

Method for detecting positions of binary rocker switches based on image processing

Based on multi-signal model, a novel diagnostic Bayesian networks auto-construction method is proposed in this paper. Combined with information entropy theory, a diagnostic strategy with the ability of sequential fault diagnosis, multi-fault diagnosis and coupling faults diagnosis is designed. In order to deal with uncertain problems and increase diagnostic accuracy, the self adjustment method of Diagnostic Bayesian Networks (DBN) is also discussed. The effectiveness of the proposed method is verified by the application on a diesel engine test and diagnosis process.

Diagnostic Bayesian Networks auto-construction and diagnostic strategy design based on multi-signal model

The invention provides a construction method for a reconfigurable system Bayesian network. The construction method comprises the following steps that: (1) establishing a model for the correlation of the fault and the test of the reconfigurable system; (2) representing the correlation of the fault and the test of the reconfigurable system; (3) determining a reconfigurable system Bayesian network structure; and (4) determining a reconfigurable system Bayesian network parameter, and carrying out representation by a tree structure. The construction method has the beneficial effects that the programmed construction method for the reconfigurable system Bayesian network is put forward, the storage and the retrieval of the Bayesian network parameter are optimized, a new solution is provided for solving the construction problem of the reconfigurable system Bayesian network, and the application of the Bayesian network in a fault diagnosis field can be accelerated.

Shigang Zhang

Papers

A method for the maintainability evaluation at design stage using maintainability design attributes

Modeling and Analysis of Wireless Power Transfer System in Metal Environment

Method for detecting positions of binary rocker switches based on image processing

Diagnostic Bayesian Networks auto-construction and diagnostic strategy design based on multi-signal model

Construction method for reconfigurable system Bayesian network