Fault Diagnostics Research Articles

AUTOMATED IOT-CONNECTED ON-BOARD FAULT DETECTION IN FAN-COILS: PROTOTYPE CONSTRUCTION AND PRELIMINARY TESTING

The paper focuses on design and implementation of an IoT connected on-board automated fault detection and diagnostics prototype (AFDD) for a nonspecific fan-coil, in turn part of HVAC system and of a distributed digital collaboration framework used in Facility Management. A research on common IoT architecture and maintenance strategies has been carried out besides the theoretical development of a Fault detection diagram on all the typical faults in fan-coil units. A real fan-coil was then inspected to point out its construction details and the points to be monitored. Then it was equipped with the prototype AFDD system. All the components and sensors needed to build the AFDD prototype are commonly available. The design and implementation of automated fault detection and diagnostics (AFDD) for HVAC fan-coils systems fully exploits distributed computing for remote and smart system monitoring, anomaly detection and eventually fault diagnostics to improve maintenance management through the integration of a large number of data locally gathered by smart sensors. Experimental results on the prototype are given about some recurrent fan-coil anomalies. Local intelligence allows a quick and on-site anomaly detection and fault diagnostic, as proven by running the prototype AFDD equipped fan-coil: it could help managing and scheduling maintenance, reducing time-to-fix together with indirect and direct costs, if network connected. Feeding the network with relevant data about the anomalies extracted by the local intelligence allows sharing the information at every level, also in order to statistically rate HAVC components service life and reliability.

An Improved Time-Varying Morphological Filtering and Its Application to Bearing Fault Diagnosis

Morphological filtering (MF) is a typical fault feature extraction technology, which has been widely used in rolling bearing fault diagnostics. The structural element (SE) length has an important influence on feature extraction and noise removal, and the filtering result obtained by the filtered signal under a single SE length or the weighted average of different filtered signals under partial SE lengths is limited and insufficient. To address this problem, a transient feature extraction method called improved time-varying morphological filtering (ITVMF) is proposed for bearing fault diagnosis, wherein the SE length is adaptively determined according to the inherent characteristics of vibration signals. Moreover, the autocorrelation envelop spectrum (AES) is employed to further eliminate fault-unrelated components. The main innovation of this method is to automatically extract bearing fault-related transient features using a novel time-varying SE (TVSE) based on autocorrelation while effectively eliminating the interference of random impulses and strong background noise on bearing fault features. The developed ITVMF-AES is investigated and validated on the simulation signals, accelerated bearing degradation data, and artificial experimental test data, and compared with three existing feature extraction methods. The results demonstrate that the developed methodology is more efficient in excavating transient impulse features and identifying bearing faults than the comparison methods, and has potential engineering application value.

A time-series turbofan engine successive fault diagnosis under both steady-state and dynamic conditions

In recent years there has been a growing interest in gas turbine fault diagnosis, especially under dynamic conditions, due to the evolving operating profile of gas turbines and the need to deploy computationally efficient and high-precision diagnostic solutions in real-time. One of the main challenges of fault diagnosis in real-time is the power imbalance between the compressor and turbine that occurs during transient operation. In addition, the heat soakage phenomenon characterizing the transient conditions has a substantial impact on the accuracy of the diagnosis. Finally, any sudden failure that might happen during transient operating conditions creates an additional challenge to fault diagnostics. The present study proposes a gas turbine diagnostic approach based on time-series measurements encapsulating steady-state and transient operating conditions. Specifically, the introduced novel approach is capable of quantifying the surplus/deficit of the power between the compressor and the turbine by utilizing the time-series data representing the observed deviations in the shaft rotational speed in order to determine the power balance in the shaft. The maximum diagnostic errors for constant fault and sudden failure are less than 0.006% during the dynamic maneuver. The results demonstrate and illustrate that the proposed method could effectively and accurately diagnose the severity of aero-engine faults at both steady-state and transient conditions. Therefore, this study has great potential for gas turbine practitioners since the diagnosis under transient conditions in real-time can enhance the capability of engine online condition monitoring and improve the condition-based maintenance of gas turbine assets.

Integration of Nonlinear Dynamics and Machine learning for Diagnostics of a Single-Stage Gear Box


 
 
 The current study concerns diagnostics of a one-stage gear- box based on the integration of physics and machine learn- ing. A physics-based model of this system is developed, then a nonlinear dynamic analysis is performed. The accuracy of the model is validated by comparing fundamental phenomena observed in synthetic and experimental data. To address the diagnostics problem synthetic data are generated for faulty and healthy conditions. Further, physics-informed features are extracted from the phase space of the dynamic system. It is shown that these features are highly informative about the health condition of the system. Also, their advantages over purely statistical features are demonstrated by a feature ranking technique. Subsequently, they are used as inputs in a machine learning model that is developed and optimized for fault diagnostics. The performance of the proposed method is investigated from different aspects, e.g., the accuracy of fault classification, robustness to noise, and generalization to unseen scenarios.
 
 

In-Service Power Transformer Life Time Prospects: Review and Prospects

Power transformers are essential assets in power system networks that must be meticulously monitored throughout their operational life cycle. Given that a substantial percentage of in-service power transformers around the world have reached the end of their projected technical lifespan, utilities are adopting various transformer condition-based maintenance strategies to minimize potentially devastating equipment failure. Thus, further usage of aging fleet of power system assets can be improved by gaining a better understanding of life threatening factors. In-service power transformers and their auxiliary components are susceptible to a variety of operational risks, which can compromise their performance and efficiency, thus resulting in devastating failures, financial losses, and power outages. Hence, it is necessary to investigate transformer life time issues so as to fully grasp the operational and performance status in order to reduce failures and operating costs. This paper presents a unified framework of attributes pertaining to life time issues for transformer operation evaluation together with a summary of recent findings in order to explore the current status and progress of this rapidly progressing field. The failure statistics analysis is presented first, followed by an examination of the transformer’s insulation degradation and aging mechanism. Furthermore, emphasis was on detailing the commonly adopted models and strategies in transformer condition evaluation, fault diagnostics, and remnant life estimation. Despite significant advancements in ascertaining transformer health diagnosis and prognosis technologies, including test accuracy, quick, precise fault localization, and fault type classification, there are still several shortcomings that require additional research, and thus future research perspectives are also discussed.

Grey box modeling of supermarket refrigeration cabinets

Aiming to enable robust large-scale fault diagnostics and optimized control for supermarket refrigeration systems, a data-driven grey box model for an evaporator and its surrounding cooling cabinet (or room) is presented. It is a non-linear model with two states: the cabinet temperature and the refrigerant mass in the evaporator. To demonstrate its applicability, data with one-minute sampling resolution from ten evaporators in a supermarket in Otterup (Denmark) was used. The model parameters were estimated using a Kalman filter and the maximum likelihood method. Since the dynamical properties of the cabinets constantly change as goods are added and removed, the parameters were re-estimated for each night, over a period of approximately 2.5 years. The model is validated through a statistical analysis of the residuals and the importance of the ongoing re-estimation of parameters is highlighted. Furthermore, the physical meaning of the estimated parameters is discussed and potential applications for characterization and classification of cabinets are demonstrated, by showing how they can be differentiated as either open- or closed cabinets or rooms, using only the estimated heat transfer coefficients and heat capacities. For a selected case it is shown that the estimated parameter values are close to physics derived values, and that the accuracy measured by the standard errors of the estimates is approximately ±10% relative to the estimated values. The analysis demonstrates that the model is robust, accurate and reliable in terms of estimating physically meaningful parameters and it is therefore appropriate for large-scale implementation.

Sensor cost-effectiveness analysis for data-driven fault detection and diagnostics in commercial buildings

Data-driven building fault detection and diagnostics (FDD) is heavily dependent on sensors. However, common sensors from Building Automation Systems are not optimized to maximize accuracy in FDD. Installing additional sensors that provide more detailed building system information is key to maximizing the performance of FDD solutions. In this paper, we present a sensor cost analysis workflow to quantify the economic implications of installing new sensors for FDD using the concept of sensor threshold marginal cost (STMC). STMC does not represent actual sensor cost. Rather, it represents a target cost based on the economic benefit that would be realized through improved FDD performance and one or more specified economic criteria. We calculate STMCs for multiple possible fault types and use fault prevalence information to aggregate STMCs into a single dollar value to determine the cost-effectiveness of a potential sensor investment. We conducted a case study using Oak Ridge National Laboratory's Flexible Research Platform (FRP) test facility as a reference. The case study demonstrates the feasibility of the analysis and highlights the key cost considerations in sensor selection for FDD. The results also indicate that identifying and installing the few key sensor(s) is critical to cost-effectively improve FDD performance.

Controlled generation of unseen faults for Partial and Open-Partial domain adaptation

New operating conditions can result in a significant performance drop of fault diagnostics models due to the domain shift between the training and the testing data distributions. While several domain adaptation approaches have been proposed to overcome such domain shifts, their application is limited if the fault classes represented in the two domains are not the same. To enable a better transferability between two different domains, particularly in setups where only the healthy data class is shared between the two domains, we propose a new framework for Partial and Open-Partial domain adaptation based on generating distinct fault signatures with a Wasserstein GAN. The main contribution of the proposed framework is the controlled data generation with two characteristics. Firstly, previously unobserved target faults can be generated by having only access to healthy target and faulty source samples. Secondly, distinct fault types and severity levels can be generated precisely. The proposed method is especially suited for extreme domain adaption settings that are particularly relevant in the context of complex and safety-critical systems, where only one class is shared between the two domains. We evaluate the proposed framework on Partial as well as Open-Partial domain adaptation tasks on two bearing fault diagnostics case studies. In the evaluated case studies the proposed methodology demonstrated superior results compared to other methods, particularly in the presence of large domain gaps. The experiments conducted in different label space settings (Partial and Open-Partial) showcase the versatility of the proposed framework.

Artificial Intelligence and 3D Scanning Laser Combination for Supervision and Fault Diagnostics.

In this work, we combine some of the most relevant artificial intelligence (AI) techniques with a range-resolved interferometry (RRI) instrument applied to the maintenance of a wind turbine. This method of automatic and autonomous learning can identify, monitor, and detect the electrical and mechanical components of wind turbines to predict, detect, and anticipate their degeneration. A scanner laser is used to detect vibrations in two different failure states. Following each working cycle, RRI in-process measurements agree with in-process hand measurements of on-machine micrometers, as well as laser scanning in-process measurements. As a result, the proposed method should be very useful for supervising and diagnosing wind turbine faults in harsh environments. In addition, it will be able to perform in-process measurements at low costs.

Comprehensive Review on Modelling, Estimation, and Types of Faults in Solar Photovoltaic System

Solar photovoltaic (SPV) system fault diagnostics is vital in advanced supervision because it can alert users to catastrophic failure or greater risks. To provide green and clean energy using solar, it is mandatory to analyse various faults associated with photovoltaic system which can result in energy deficit and system breakdown and may lead to fire hazards which are often difficult to avoid. Hence, as an endeavour to improve the efficiency level, more study beginning with modelling of SPV system with its parameter estimation and types of SPV faults is aimed in this work.

A Review of Open-Circuit Switch Fault Diagnostic Methods for Neutral Point Clamped Inverter

Due to numerous advantages, a neutral point clamped (NPC) inverter is a preferred choice for high-power applications and renewable technology. The reliability of the NPC inverter is a major concerning factor during the assessment of system performance as power semiconductor switches are vulnerable to abnormal conditions. Open-circuit (OC) switch faults are not as dangerous as short circuit (SC) faults but eventually have enough potential to cause cascaded failure to other components in the system and thus need to be supervised carefully. The OC faults result in a distortion of voltage and current signals in the NPC converter. Based on these signals, over the past few years, many efforts have been made to identify and localize the OC switch fault to the switch level in the NPC topology. In this paper, a review of different OC switch fault diagnostic methods is provided. Starting from the NPC inverter operation under healthy and faulty conditions, the various possible and unavailable switching states along with the deviation in pole voltage under different switch fault conditions is discussed. Then, based on the approach used for system-based fault detection, the OC fault detection methods are classified. The various OC methods are further discussed on the basis of signal, i.e., current, voltage or a combination of both signals used as a signature for fault detection. Emphasis is given to the principle involved, diagnostic variables utilized, the implementation approach and the diagnostic time required. Finally, the approaches are tabulated so as to provide a quick reference for NPC fault diagnostics.

Multisource Partial Transfer Network for Machinery Fault Diagnostics

Multisource Partial Transfer Network for Machinery Fault Diagnostics

Vibration based fault diagnostics in a wind turbine planetary gearbox using machine learning

To reduce wind turbine operations and maintenance costs, we present a machine learning framework for early damage detection in gearboxes based on the cyclostationary and kurtogram analysis of sensor data. The application focus is fault diagnostics in gearboxes under varying load conditions, particularly turbulent wind. Faults in the gearbox rotating components can leave their signatures in vibrations signals measured by accelerometers. We analyze data stemming from a simulated vibration response of a 5 MW multibody wind turbine model in a healthy and damaged scenarios and under different wind conditions. With cyclostationary and kurtogram analysis applied on acquired sensor data, we generate two types of 2D maps that highlight signatures related to the fault damage. Using these maps, convolutional neural networks are trained to identify faults, including those of small magnitude, in test data with a high accuracy. Benchmark test cases inspired by an NREL study are tested and faults successfully detected.

Special Issue on the Application of Machine Learning and Artificial Intelligence in Fault Diagnostics

The rapid development of data science and associated artificial intelligence (AI) methods has seen a substantial increase in interest in their application to anomaly detection, fault diagnostic, and prognostic challenges across a wide range of industrial and civil applications. Such approaches may well be the complement that is sought for conventional physical model-based and statistical approaches which often struggle to achieve the desired performance when dealing with complex engineering systems. Researchers have started to apply a range of machine learning and AI-based methods to the large-scale, multi-dimensional data that is often associated with large-scale sensor systems, particularly those which involve IoT devices. There is clear scope for the further development of such approaches, such as deep learning, transfer learning methods, and AI models, to enhance the performance of condition monitoring and associated technologies, and this is the key motivation for this Special Issue. This Special Issue on the Application of Machine Learning and Artificial Intelligence in Fault Diagnostics contains 6 papers.

Simulation Tool to Evaluate Fault Diagnosis Techniques for DC-DC Converters

This paper presents a tool that evaluates the applicability of a fault diagnostics technique (FDT) for Switch-Mode Non-Isolated DC-DC converters (SMNI DC-DC). The proposed tool simulates the operation of a Type II controller for a buck converter under different operating conditions and, simultaneously, implements the FDT. In this way, it is possible to evaluate the applicability of the FDT in real-time. The proposed tool is implemented in Python, and requires the partitioning of the code into different modules. The first module is responsible for modeling the different states of the converter. The second module contributes to the production of the PWM signal. The concepts of symmetry and asymmetry are implicit to the signal processing methods used in the design of the FDTs that compose the third module. Finally, all of the modules are properly included in the main function, which represents the fourth module. The validation of the proposed tool is carried out by comparing the results with the LTspice simulator and with an experimental prototype built for this purpose. Different operating conditions are considered. In addition to the proposed tool, an algorithm for the design of both the power section and the converter compensator is presented.

Quantum Networks with Multiple Service Providers: Transport Layer Protocols and Research Opportunities

A quantum Internet for communicating information encoded in quantum systems over large distances would enable a host of new technologies to be deployed. While exciting progress has been made in small-scale quantum networks, a global quantum Internet will require - like the classical Internet - communicating in a multiple quantum Internet service provider (multi-qISP) setting, with protocols that are oblivious to the global network topology. Here we give a brief overview of quantum networks aimed at those new to the field, present two network-oblivious transport layer protocols for enabling high fidelity communication and performing fault diagnostics, and highlight a number of research opportunities in this nascent field of research.

Motor Bearings Fault Classification using CatBoost Classifier

Induction motors are used in all industries and are the major element of energy consumption. Faults in motor degrade the motor efficiency and result in more energy consumption. Bearing faults are reported to be the major reason for the motor breakdown and a lot of papers have been reported to focus on bearing fault diagnostics. However, low classification accuracy is the main hurdle in adopting the available fault classification algorithms. This paper has presented a novel classification algorithm using the Catboost classifier and timedomain features. The developed algorithm was tested on the laboratory test setup. The fault classification accuracy of 100 % was achieved through the proposed method.

Fault Feature Extractor Based on Bootstrap Your Own Latent and Data Augmentation Algorithm for Unlabeled Vibration Signals

Given that vibration fault signals collected from industrial circumstances are usually insufficient and have no labels, supervised learning networks cannot be directly applied to recognize fault types in this case. Hence, automatic feature extraction of unlabeled data is urgently needed. In this study, an automatic fault feature extractor (AFFE) based on the contrastive learning algorithm—Bootstrap Your Own Latent (BYOL) network, which can extract fault features automatically without needing labeled information—is proposed. A data augmentation method for vibration signals is studied because it is critical to the contrastive learning algorithm. This study determines a data augmentation combination that can help AFFE achieve excellent performance in extracting features from unlabeled bearing fault data. To verify the validity of the proposed method, we utilize some labeled data (5% of samples with labels in a dataset) to fit linear classifiers, which are combined with the proposed AFFE to extract a feature. The aim is to predict the accuracy score of feature classification for the remaining data. The case study demonstrates that the fault features extracted using AFFE can achieve a high accuracy score of 94.81%. Therefore, the proposed AFFE-BYOL is a promising diagnostic fault feature extraction scheme to process unlabeled vibrational data.

Integrated Approach to Diagnostics of Failures and Cyber-Attacks in Industrial Control Systems

This paper is concerned with the issue of the diagnostics of process faults and the detection of cyber-attacks in industrial control systems. This problem is of significant importance to energy production and distribution, which, being part of critical infrastructure, is usually equipped with process diagnostics and, at the same time, is often subject to cyber-attacks. A commonly used approach would be to separate the two types of anomalies. The detection of process faults would be handled by a control team, often with a help of dedicated diagnostic tools, whereas the detection of cyber-attacks would be handled by an information technology team. In this article, it is postulated here that the two can be usefully merged together into one, comprehensive, anomaly detection system. For this purpose, firstly, the main types of cyber-attacks and the main methods of detecting cyber-attacks are being reviewed. Subsequently, in the analogy to “process fault”—a term well established in process diagnostics—the term “cyber-fault” is introduced. Within this context a cyber-attack is considered as a vector containing a number of cyber-faults. Next, it is explained how methods used in process diagnostics for fault detection and isolation can be applied to the detection of cyber-attacks and, in some cases, also to isolation of the components of such attacks, i.e., cyber-faults. A laboratory stand and a simulator have been developed to test the proposed approach. Some test results are presented, demonstrating that, similarly to equipment/process faults, residua can be established and cyber-faults can be identified based on the mismatch between the real data from the system and the outputs of the simulation model.

Appraising machine learning classifiers for discriminating rotor condition in 50W–12V operational wind turbine for maximizing wind energy production through feature extraction and selection process

Wind energy is one of nature’s most valuable green energy assets, as well as one of the most reliable renewable energy supplies. Wind turbine blades convert wind energy into electric energy. Wind turbine blades range in size from 25 to 120 m, depending on the demands and efficiency necessary. Owing to ambient influences and wide structures, the blades are subject to various friction forces that might harm the blades. As a result, the generation of power and the shutdown of turbines are both affected. Downtimes are reduced when blades are detected on a regular basis, according to structural health management. On the 50-W, 12-V wind turbine, this research investigates the use of vibration signals to anticipate deterioration. The machine learning (ML) method establishes a nonlinear relationship between selected important damage features and the related uniqueness measures. The learning algorithm was trained and tested based on the excellent state of the edge. To forecast blade faults, classifier models, such as naive Bayes (NB), multilayer perceptron (MLP), linear support vector machine (linear_SVM), one-deep convolutional neural network (1DCNN), bagging, random forest (RF), XGBoosts, and decision tree J48 (DT) were used, and the results were compared according to their parameters to propose a better fault diagnostics model.