No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
A review on machinery diagnostics and prognostics implementing condition-based maintenance
Abstract
Condition-based maintenance (CBM) is a maintenance program that recommends maintenance decisions based on the information collected through condition monitoring. It consists of three main steps: data acquisition, data processing and maintenance decision-making. Diagnostics and prognostics are two important aspects of a CBM program. Research in the CBM area grows rapidly. Hundreds of papers in this area, including theory and practical applications, appear every year in academic journals, conference proceedings and technical reports. This paper attempts to summarise and review the recent research and developments in diagnostics and prognostics of mechanical systems implementing CBM with emphasis on models, algorithms and technologies for data processing and maintenance decision-making. Realising the increasing trend of using multiple sensors in condition monitoring, the authors also discuss different techniques for multiple sensor data fusion. The paper concludes with a brief discussion on current practices and possible future trends of CBM.
... At the same time, a variety of technological innovations in this area have reached a stage of maturity that may spur the uptake of novel strategies and policies, consequently helping to address the required drive for efficiency in aircraft maintenance. A primary example is Condition-Based Maintenance (CBM), a maintenance policy that can be defined as "preventive maintenance which includes assessment of physical conditions, analysis and the possible ensuing maintenance actions" [3] or "a maintenance program that recommends maintenance actions based on the information collected through condition monitoring" [4]. CBM and its constituent technologies help identify and prevent unscheduled maintenance, facilitate substituting maintenance tasks or extension of task intervals, and enable the optimization of maintenance schedules at the fleet level [5,6]. ...
... A third stream of research assesses the (potential) impact of CBM through cost-benefit analysis [22,[24][25][26]). Several studies have provided empirical findings pointing out that CBM reduces asset downtime and total maintenance costs compared to other maintenance strategies [4,27], especially when predicted failures can be turned into scheduled maintenance and clustered with existing activities. However, these findings are established after implementation, leading to a set of research that aims to enable an a-priori assessment of CBM costs and benefits. ...
... The study of methods and approaches to facilitate CBM implementation has historically taken somewhat of a backseat when compared to the development of detection, diagnostics, and prognostics models, as well as subsequent decision support models. This is reflected in early considerations of CBM implementation, where Jardine et al. [4] identified three key steps for every CBM program: (1) data acquisition, (2) data processing, and (3) maintenance decision-making. Data acquisition relates to "the process of collecting and storing useful information, such as process and event data, preferably in a centrally accessible system" [13]. ...
Condition-Based Maintenance (CBM) is a policy that uses information about the health condition of systems and structures to identify optimal maintenance interventions over time, increasing the efficiency of maintenance operations. Despite CBM being a well-established concept in academic research, the practical uptake in aviation needs to catch up to expectations. This research aims to identify challenges, limitations, solution directions, and policy implications related to adopting CBM in aviation. We use a generalizable and holistic assessment framework to achieve this aim, following a process-oriented view of CBM development as an aircraft lifecycle management policy. Based on various inputs from industry and academia, we identified several major sets of challenges and suggested three primary solution categories. These address data quantity and quality, CBM implementation, and the integration of CBM with future technologies, highlighting future research and practice directions.
... In contrast, predictive maintenance includes the help of the IoT when it comes to maintenance. Jardine et al. [14] define condition-based maintenance as a maintenance program that provides maintenance recommendations and decisions based on the information obtained through condition monitoring and works in three steps, namely • Data acquisition, • Data processing, and • Maintenance decision making. ...
... In contrast, predictive maintenance includes the help of the IoT when it comes to maintenance. Jardine et al. [14] define condition-based maintenance as a maintenance program that provides maintenance recommendations and decisions based on the information obtained through condition monitoring and operates in three steps. ...
... There are two main limitations to continuous monitoring: (i) the high costs associated with continuous monitoring, which arise because many specialized machines are needed; and (ii) the possibility of obtaining inaccurate information because the continuous flow of data leads to increased noise [14]. ...
Predictive maintenance is one of the most important topics within the Industry 4.0 paradigm. We present a prototype decision support system (DSS) that collects and processes data from many sensors and uses machine learning and artificial intelligence algorithms to report deviations from the optimal process in a timely manner and correct them to the correct parameters directly or indirectly through operator intervention or self-correction. We propose to develop the DSS using open-source R packages because using open-source software such as R for predictive maintenance is beneficial for small and medium enterprises (SMEs) as it provides an affordable, adaptable, flexible, and tunable solution. We validate the DSS through a case study to show its application to SMEs that need to maintain industrial equipment in real time by leveraging IoT technologies and predictive maintenance of industrial cooling systems. The dataset used was simulated based on the information on the indicators measured as well as their ranges collected by in-depth interviews. The results show that the software provides predictions and actionable insights using collaborative filtering. Feedback is collected from SMEs in the manufacturing sector as potential system users. Positive feedback emphasized the advantages of employing open-source predictive maintenance tools, such as R, for SMEs, including cost savings, increased accuracy, community assistance, and program customization. However, SMEs have overwhelmingly voiced comments and concerns regarding the use of open-source R in their infrastructure development and daily operations.
... Thus, predictive maintenance (i) reduces cost [5] because maintenance tasks are carried out when and where they are truly needed, not just by schedule, and (ii) increases safety by avoiding dangerous events during operation. This type of maintenance has received increased attention in the literature during the last decades [6][7][8], due to its application in a variety of industrial sectors, such as the aerospace [9,10], manufacturing [11], and railway [12] industries. ...
... where the matrix U n is precisely the mode matrix along the n-th dimension appearing in (9). Then, the tensor S is computed as: ...
A very fast reduced order model is developed to monitor aeroengines condition (defining their degradation from a baseline state) in real time, by using synthetic data collected in specific sensors. This reduced model is constructed by applying higher-order singular value decomposition plus interpolation to appropriate data, organized in tensor form. Such data are obtained by means of an engine model that takes the engine physics into account. Thus, the method synergically combines the advantages of data-driven (fast online operation) and model-based (the engine physics is accounted for) condition monitoring methods. Using this reduced order model as surrogate of the engine model, two gradient-like condition monitoring tools are constructed. The first tool is extremely fast and able to precisely compute the turbine inlet temperature ‘on the fly’, which is a paramount parameter for the engine performance, operation, and maintenance, and can only be roughly estimated by the engine instrumentation in civil aviation. The second tool is not as fast (but still reasonably inexpensive) and precisely computes both the engine degradation and the turbine inlet temperature at which sensors data have been acquired. These tools are robust in connection with random noise added to the sensor data and can be straightforwardly applied to other mechanical systems.
... The good performance of railway vehicle wheels is an essential requirement for ensuring that a railway network is kept highly reliable and safe. The current wheel preventive maintenance schemes are susceptible to unplanned maintenance tasks, which lead to operational bottlenecks and increased cost due to the need for non-automated and time-consuming inspections, unscheduled disturbance to the timetable, and potentially untimely wheel replacement [1]. Railway vehicle wheel efficiency is strongly connected with the wheel/rail geometric parameters [2], which must remain within standard nominal ranges. ...
... Based on the results of recent studies [24][25][26][27], it is evident that the effects of hollow worn wheels on the dynamics of a vehicle are detectable via on-board vibration measurements. This fact constitutes the motivation for the development of automated condition monitoring units for railway vehicle wheels that may be incorporated into a broader PHM system [1]. To this end, the on-board determination of wheel conicity and thus the monitoring of potentially hollow worn wheels is attempted in [28] through simulations with a simplified single wheelset model. ...
The problem of the prompt detection of early-stage hollow worn wheels in railway vehicles via on-board random vibration measurements under normal operation and varying speeds is investigated. This is achieved based on two unsupervised statistical time series (STS) methods which are founded on a multiple-model (MM) framework for the representation of healthy vehicle dynamics. The unsupervised MM power spectral density (U-MM-PSD) method employs Welch-based PSD estimates for wheel wear detection and the unsupervised MM autoregressive (U-MM-AR) method for the parameter vectors of multiple AR models. Both methods are assessed via two case studies using thousands of test cases. The first case study includes Monte Carlo simulations using a SIMPACK-based detailed railway vehicle model, while the second is based on field tests with an Athens Metro train. Wheel wear detection is pursued using lateral or vertical vibration signals from the bogie or the carbody of a trailed vehicle traveling with three different speeds (60, 70, 80 km/h) using wheels under healthy conditions or with early stage hollow wear. Both methods exhibit remarkable performance with the U-MM-AR method to achieve the best overall results, reaching correct detection rates of even 100% with false alarm rates below 5% based on a single accelerometer either on the carbody or bogie.
... Given the diversity and volume of the reaction signals, it is almost impossible to recognize fault patterns straightforwardly. Because of this, the fundamental components of a typical fault diagnostic system typically include the following: data processing (feature extraction) and fault detection [2][3][4]. The pre-processing of input patterns by feature extraction algorithms is the foundation on which the vast majority of common intelligent fault diagnosis systems are constructed [5]. ...
... The following graphic in Figures three (3) and four(4) provides a summary comparison of publications that compare AI approaches from each of the three decades, as referenced in section IV because there is no one learning algorithm that can consistently perform better than other algorithms across all datasets. Therefore, the publications chosen for the review either use a single learning method or combine it with another learning algorithm. ...
Distributed generation (DG) are gaining popularity in power systems due to its environmental and economic benefit versus other technologies. Despite the benefits of DG and their operation, discussions need to be done on the potential problems and impact of the increased penetration when they get integrated into the grid. This discussion is vital for distribution engineers and designers as studies have shown that the application of local DG into distribution grids has consequences for the protection system and protection issues concerning distributed generation. The primary purpose of power system protection is to ensure the safe operation of power systems hence the safety of the people and types of equipment. This review paper aims to explore the different fault detection techniques and their evolution in implementation over the past three decades. This review study is required to investigate the different algorithms implemented for distribution network faults detection methods in power systems networks and micro-grid networks.
... Measurement of vibration and acoustic emissions are well-proven methods of providing signals which carry complete diagnostic information on machine conditions. Vibration sensors have been used to monitor engine condition and failures in marine and aviation applications [17]. For brevity, the exact measurement techniques and data analysis methods for condition-based maintenance using vibration monitoring will not be discussed here. ...
... In the field of short-circuit fault diagnosis in ship's integrated power propulsion systems, the commonly used research methods can be classified into three categories: analytical model-based methods [6], data-driven methods [7], and qualitative model-based methods [8]. The specific classification is shown in Table 1 [9]. ...
Synchronous generators with three phases are crucial components of modern integrated power systems in ships. These generators provide power for the entire operation of the vessel. Therefore, it is of paramount importance to diagnose short-circuit faults at the generator terminal in the ship’s power system to ensure the safe and stable operation of modern ships. In this study, a generator terminal short-circuit fault diagnosis method is proposed based on a hybrid model that combines the Multi-Level Wavelet Decomposition Network, Deep-Gated Recurrent Neural Network, and Fully Convolutional Network. Firstly, the Multi-Level Wavelet Decomposition Network is used to decompose and denoise the collected electrical signals, thus dividing them into sub-signals and extracting their time-domain and frequency-domain features. Secondly, synthetic oversampling based on Gaussian random variables is employed to address the problem of imbalance between normal data and fault data, resulting in a balanced dataset. Finally, the dataset is fed into the hybrid model of the Deep-Gated Recurrent Neural Network and Fully Convolutional Network for feature extraction and classification of faults, ultimately outputting the fault diagnosis results. To validate the performance of the proposed method, simulations and comparative analysis with other algorithms are conducted on the fault diagnosis method. The proposed algorithm’s accuracy reaches 96.82%, precision reaches 97.35%, and the area under curve reaches 0.85, indicating accurate feature extraction and classification for identifying short-circuit faults at the generator terminals.
... Condition-based and predictive maintenance strategies have long been advocated for their potential reduction in un-planned, costly breakdowns and avoidance of unnecessary repairs (Nunes et al., 2023). Condition-based maintenance utilizes on-line sensing techniques from which asset degradation information may be inferred (Jardine et al., 2006). Maintenance actions are triggered by the condition of an asset, thereby saving labour, and reducing downtime, reducing costs, and increasing production. ...
Digital twins entail the entangled use of a software representation of a real asset with engineering sensors to communicate the state and behaviour of an asset. This work focuses on how digital twin technology may specifically proffer maintenance management in rail networks and rolling stock through their value adding property, servitization. These digital services are structured through a portfolio of digital twin
service patterns, which hinge on the generic configuration of digital twin building blocks to deliver a specific value. This idea springboards off the iconic concept, "design patterns" in object orientated programming which entail standardized high-level solutions to commonly recurring problems. Design templates are provided for specific services that mirror asset behaviour, provide virtual sensing, detect anomalies, and match the fingerprint of a specific response. The service patterns are presented through a graphical model that depicts the detail of cyber-physical interactions.
... The applicability of the PHM process is independent of the specific asset or component under analysis, showing validity throughout different indenture levels [11,[23][24][25], including the equipment/asset itself, like a CNC machine tool, a centrifugal pump, or an industrial fan, and the subsystems/components, such as bearings, drive belts, or rotors. Moreover, the reviews propose specific insights on each level of the ISO 13374-1, sometimes showing different but compatible aggregation of the levels. ...
Digitisation offers manufacturing companies new opportunities to improve their operations and competitiveness in the market by unleashing potentialities related to real-time monitoring and control of operating machines. Through condition-based and predictive maintenance, the knowledge about the health state and probability of failure of the machines is improved for better decision-making. Amongst them, CNC machine tools do represent a complex case from a maintenance viewpoint as their operations are ever-changing and their high reliability brings to a lack, or limited set, of run-to-failure data. To address the problem, the research work proposes an operations-aware novelty detection framework for CNC machine tools based on already-in-place controllers. The framework is based on statistical modelling of the behaviour of the machine tools, namely through gradient boosting regression and Gaussian mixture models, to identify the health state considering varying operations through time. The proposed solution is verified on sixteen multi-axis CNC machine tools in a large manufacturing company. The results show that the proposed solution can effectively support maintenance decisions by informing on the health states while discerning between varying operations and abnormal/faulty states of interest. This solution represents a brick in a cloud-edge-based industrial information system stack that can be further developed for shop floor-integrated decision-making.
... Moreover, CBM involves time-based maintenance and monitoring of the system's and components' health [12 , 13] . The CBM technique involves failure's prognostic effects for enhancing the system's life [13] . Jamshidi et al. [14] proposed a CBM technique for railway infrastructure to effectively control the equipment's degradation. ...
Condition-based maintenance involves activities that are conducted based on the equipment's performance. Continuous monitoring of equipment will ensure that it will be maintained according to a relevant activity plan. This paper proposes a maintenance framework to analyze the application of statistical control charts for condition-based maintenance of electrical generators. The proposed framework consists of four components that collaboratively determine a performance threshold for a given piece of electrical equipment. Based on the slow progression and dynamics of mechanical failures, Long Short-term Memory (LSTM) and Useful Remaining Life (URL) models were used to assist in the maintenance decision-making process. The analysis is based on detecting the dynamics of the process parameters, including vibration, noise, and temperature, based on relevant control charts. With the help of experimental methodology, failures in the performance modes and defined modes are measured. Then empirical analysis reveals how control charts respond to failure detection. The results show that X-bar consistently demonstrates failure detection capability, while R charts sometimes fail when data deviates from normality.
Moreover, heat monitoring surpassed vibration and noise in failure detection, where temperature control charts successfully identified failure. The overall results support the significant role of statistical charts in decision-making regarding condition-based maintenance for electrical equipment like generators.
•
Application of statistical control charts for condition-based maintenance of electrical generators.
•
Detecting dynamics of the process parameters, including vibration, noise, and temperature, based on relevant control charts.
•
Long Short-term Memory (LSTM) and Useful Remaining Life (URL) models were used to assist in the maintenance decision-making process
... The recommendation of maintenance actions based on machinery health information collected through condition monitoring is known as Condition-Based Maintenance (CBM) (Jardine et al. 2006). Thus, it is a decision-making strategy to enable real-time diagnosis of impending failures and prognosis of future equipment health (Peng et al. 2010). ...
Condition-Based Maintenance (CBM) is a well-known strategy that organizations implement to prevent failures of their physical assets. For that, Fault Detection and Diagnosis (FDD) processes need to be implemented successfully. Nevertheless, this stage requires the structuring of expert knowledge regarding the potential failure modes and their observability and failure data. Without a supporting tool, this setup can be difficult for many organizations. In this context, this paper proposes the Failure Mode and Observability Analysis (FMOA) to support fault detection and diagnosis implementation in asset management. The proposed method is a variation of the Failure Mode and Effects Analysis (FMEA) that analyses the potential failure modes of selected systems and correlates them with relevant properties for FDD. The proposed FMOA method was demonstrated through a case study based on a Brazilian hydroelectric power plant. The results obtained showed that the method can support organizations in the study of selected systems, equipment, and components for the implementation of fault detection and diagnosis. It contributes to the organizations to implement a CBM strategy as it organizes the knowledge base about the physical assets while correlating the potential failure modes with properties of FDD.
... MIMOSA also addresses the standardisation of condition monitoring and aspects of Condition Based Maintenance (CBM) to ensure consistency and reliability procedures (Campos, 2009). CBM comprises three fundamental tasks: gathering data, analysing data, and using the analysed data to make informed maintenance decisions (Jardine et al., 2006). An overview of the concept of data-driven maintenance based on OSA-CBM is presented in Figure 4. Sensor and information and communication technology (ICT) advancements have opened up possibilities for objective condition monitoring and data collection, leading to increased data availability and the development of data processing algorithms that are cost-effective and efficient (Kumari et al., 2022). ...
... Subsequently, frequency domain techniques are taken as alternative choices to describe fault patterns in another respect, as they have a better ability to discover and separate the frequency components. In this class, the most extensively utilized technique is FFT, i.e., fast Fourier transform [9,10]. Thus, in the frequency domain, some features, including the root variance of frequency, the frequency root mean square, and the frequency center, have been extracted by FFT and engaged in bearing fault diagnosis. ...
In bearing fault diagnosis, machine learning methods have been proven effective on the basis of the heterogeneous features extracted from multiple domains, including deep representation features. However, comparatively little research has been performed on fusing these multi-domain heterogeneous features while dealing with the interrelation and redundant problems to precisely discover the bearing faults. Thus, in the current study, a novel diagnostic method, namely the method of incorporating heterogeneous representative features into the random subspace, or IHF-RS, is proposed for accurate bearing fault diagnosis. Primarily, via signal processing methods, statistical features are extracted, and via the deep stack autoencoder (DSAE), deep representation features are acquired. Next, considering the different levels of predictive power of features, a modified lasso method incorporating the random subspace method is introduced to measure the features and produce better base classifiers. Finally, the majority voting strategy is applied to aggregate the outputs of these various base classifiers to enhance the diagnostic performance of the bearing fault. For the proposed method’s validity, two bearing datasets provided by the Case Western Reserve University Bearing Data Center and Paderborn University were utilized for the experiments. The results of the experiment revealed that in bearing fault diagnosis, the proposed method of IHF-RS can be successfully utilized.
... CBM is a maintenance action and decision-making program which recommends suitable actions according to the condition monitoring information by utilising the prognostic methods for more reliable and cost-effective maintenance [18]. ...
It is necessary to develop advanced fault prognostics techniques for improving
maintenance planning and scheduling, productivity and enhancing the system’s
reliability, and avoiding unexpected shutdowns and the associated additional
expenses. Such techniques include advanced failure prediction and remaining useful
life (RUL) estimation of mechanical components comprising mechanical systems with
complicated structures. The prediction algorithms must be accurate to support the
decision-making in maintenance and implement the required corrections; otherwise,
it would be difficult to detect any system anomalies and make the appropriate
decisions properly.
Such advanced fault prognostic techniques are required to enhance the technical
efficiency of the maintenance system in an Industry 4.0 workplace, and to develop
advanced predictive maintenance (PdM) strategy which considered one of the main
enablers of Industry 4.0 technologies.
The implementation of the PdM strategy requires other assisting tools such as
enterprise resources planning (ERP), supervisory control and data acquisition system
(SCADA), computerised maintenance management system (CMMS), sensors, and
real-time condition monitoring systems, supported by the necessary software and
graphical user interface (GUIs) units to facilitate the monitoring and assist
maintenance experts in speeding up the decision-making process.
This dissertation presents diagnostics and prognostics methods for mechanical
components based on their life degradation data. The fault prognostics are based on
the component’s failure prediction, remaining useful life (RUL) estimation, reliability
and life degradation probability estimation. The prognostics models were developed
based on different supervised machine learning (ML) techniques, mainly artificial
neural networks (ANNs), including various training algorithms, network
architectures, activation functions and computation methods. The proposed models
require prior knowledge of the mechanical system and the analysed components, such
as their general mechanism, inputs and outputs, for more effective modelling and
accurate model design. The main assumption of the analysed data is that component
degradation consists of multiple cycles of degradation along with a perfect
maintenance replacement strategy. In terms of the fault prognostics method, the
results showed its ability to predict real degradation data using NASA Ames milling
dataset and achieved good results compared to a published similar study.
The neuron-by-neuron training algorithm (NBN) and the unique network
architectures of fully connected networks (FCN) or arbitrarily connected networks
(ACN) were used as the main components of the developed prognostic models due to their high training capabilities for handling systems having complex structures. A
special network design, an accumulative step in the network, was added to increase
the models’ ability to keep the monotonously increasing trend of the degradation
process and to achieve highly accurate failure predictions and RUL estimations. The
life degradation was estimated using a proposed probabilistic ANN model in which
the maximum likelihood estimation method (MLE) was used to estimate the
distribution parameters of each successive degradation point. The results show the
high capability of the proposed model to predict successive degradation points, model
the life degradation and estimate the distribution parameters of the predicted
successive degradation points in the case of a perfect replacement strategy. Moreover,
a method was developed using the prediction results to estimate the component’s
reliability over successive time slots and replacement cycles. The proposed method
provides reasonable estimates and predictions compared to the theoretical
calculations in the case of preventive maintenance (PM) optimisation in terms of
reliability and availability estimation and PM cost minimisation.
The ANN-NBN model was applied to another industrial application: predicting the
efficiency of a degraded wind turbine gearbox. A performance monitoring method
was proposed to monitor the decreasing efficiency of a degraded gearbox based on
the predicted power output and the predicted gearbox efficiency combined with the
cumulative summation (CUSUM) change detection algorithm. The monitoring results
showed a high degree of prediction and successful monitoring of the state change
resulting from the decreased gearbox efficiency.
As the result of a comparative study of using different types of ML techniques for
failure prognostics and degradation path prediction, the FCN and ACN network
architectures combined with the NBN training algorithm confirmed its high
prediction capability and high prediction success rate compared to other models.
The proposed methods were successfully tested and verified using several case
studies from simulation datasets. The results were encouraging and showed that the
proposed methods could be used as specialised diagnostics and prognostics
methodologies and software tools that can efficiently predict the degradation process
of mechanical components in different industrial applications.
... Physical assets, no matter how well designed, are subject to deterioration over time since they are operating under certain stress or load in the real environment [1]. This deterioration eventually results in equipment failures, leading to high corrective maintenance costs, unplanned downtimes, and production losses. ...
This paper studies the joint production/inventory and condition-based maintenance control for a multi-product manufacturing system with setup and maintenance times under stochastic product demands. The problem is modelled as a semi-Markov decision process (SMDP). The objective is to find a joint production and maintenance policy that minimizes the long run expected discounted cost including setup, holding, lost sales, preventive and corrective maintenance costs. A Q-learning method with state aggregation (QLA) is proposed to find near-optimal policies for large-scale problems that cannot be solved to optimality due to the curse of dimensionality. The numerical results show that QLA provides well-performing policies in a reasonable computational time.
... There are two things worth noting regarding the presented framework in this paper: First, it is referred to as a "diagnostic" framework, while in fact it is a "Generic Anomaly Detection Framework". Diagnostics, as Jardin et al. [10] pointed out, incorporates the steps of fault detection, isolation and identification. Anomaly detection only deals with a part of it, namely fault detection. ...
Spacecraft systems collect health-related data continuously, which can give an indication of the systems’ health status. While they rarely occur, the repercussions of such system anomalies, faults, or failures can be severe, safety-critical and costly. Therefore, the data are used to anticipate any kind of anomalous behaviour. Typically this is performed by the use of simple thresholds or statistical techniques. Over the past few years, however, data-driven anomaly detection methods have been further developed and improved. They can help to automate the process of anomaly detection. However, it usually is time intensive and requires expertise to identify and implement suitable anomaly detection methods for specific systems, which is often not feasible for application at scale, for instance, when considering a satellite consisting of numerous systems and many more subsystems. To address this limitation, a generic diagnostic framework is proposed that identifies optimal anomaly detection techniques and data pre-processing and thresholding methods. The framework is applied to two publicly available spacecraft datasets and a real-life satellite dataset provided by the European Space Agency. The results show that the framework is robust and adaptive to different system data, providing a quick way to assess anomaly detection for the underlying system. It was found that including thresholding techniques significantly influences the quality of resulting anomaly detection models. With this, the framework can provide both a way forward in developing data-driven anomaly detection methods for spacecraft systems and guidance relative to the direction of anomaly detection method selection and implementation for specific use cases.
... The basic models can be constructed via standard statistical inference methods from historical data like the Maximal-Likelihood method, Bayesian inference, or, in the case of latent variables in the model, the Expectation-Maximization algorithm. The reason for constructing the models is to use them in the subsequent analyses in which the maintenance is optimized [48] and [83] to ensure that the system will deliver the required functionality. ...
This article reviews and analyzes the approaches utilized for monitoring cutting tool conditions. The Research focuses on publications from 2012 to 2022 (10 years), in which Machine Learning and other statistical processes are used to determine the quality, condition, wear, and remaining useful life (RUL) of shearing tools. The paper quantifies the typical signals utilized by researchers and scientists (vibration of the cutting tool and workpiece, the tool cutting force, and the tool’s temperature, for example). These signals are sensitive to changes in the workpiece quality condition; therefore, they are used as a proxy of the tool degradation and the quality of the product. The selection of signals to analyze the workpiece quality and the tool wear level is still in development; however, the article shows the main signals used over the years and their correlation with the cutting tool condition. These signals can be taken directly from the cutting tool or the workpiece, the choice varies, and both have shown promising results. In parallel, the Research presents, analyzes, and quantifies some of the most utilized statistical techniques that serve as filters to cleanse the collected data before the prediction and classification phase. These methods and techniques also extract relevant and wear-sensitive information from the collected signals, easing the classifiers’ work by numerically changing alongside the tool wear and the product quality.
Prognostic and health management (PHM) plays a vital role in ensuring the safety and reliability of aircraft systems. The process entails the proactive surveillance and evaluation of the state and functional effectiveness of crucial subsystems. The principal aim of PHM is to predict the remaining useful life (RUL) of subsystems and proactively mitigate future breakdowns in order to minimize consequences. The achievement of this objective is helped by employing predictive modeling techniques and doing real-time data analysis. The incorporation of prognostic methodologies is of utmost importance in the execution of condition-based maintenance (CBM), a strategic approach that emphasizes the prioritization of repairing components that have experienced quantifiable damage. Multiple methodologies are employed to support the advancement of prognostics for aviation systems, encompassing physics-based modeling, data-driven techniques, and hybrid prognosis. These methodologies enable the prediction and mitigation of failures by identifying relevant health indicators. Despite the promising outcomes in the aviation sector pertaining to the implementation of PHM, there exists a deficiency in the research concerning the efficient integration of hybrid PHM applications. The primary aim of this paper is to provide a thorough analysis of the current state of research advancements in prognostics for aircraft systems, with a specific focus on prominent algorithms and their practical applications and challenges. The paper concludes by providing a detailed analysis of prospective directions for future research within the field.
Steel has been known to humanity for 4,000 years. The level of per capita consumption of steel is treated as one of the most important indicators of socioeconomic development and living standards of the people in a country. Steel rerolling mills, also called reduction mills, are widely distributed across the world. The major concerns in the steel rerolling mill are misroll and cobble, which adversely affect most of the technoeconomic parameters, including the mill utilization, material yield, mill productivity, overall equipment effectiveness, overall line efficiency, and specific heat consumption of the reheating furnace. This paper discusses the major issues that lead to misroll and cobble in the rerolling mills, and these issues were identified via a shop floor analysis and breakdown records at an iron and steel rerolling mill. After the identification of the root causes of misroll, solutions are suggested.
Due to their reliance on several distributed nodes, big data systems are notoriously fragile. Maintaining availability, reliability, and continuous performance in the face of failures is the primary function of fault-tolerant systems. Fault prediction is an important subject for industry because it enables businesses to make major time and expense savings by offering efficient methods for predictive maintenance. To that end, there were two objectives: first to develop prediction techniques that would detect failures in doors from diagnostic data at an early stage; and second, to describe failures in terms of characteristics that differentiate them from normal behaviour. Some elements of the solution suggested merit special consideration. They provide the foundation for an efficient data pre-processing technique in which the action of a system is described in a specific timeframe by a set of appropriate statistics. This method significantly mitigates problems relating to data noise and errors, allowing an efficient outer detection. In our opinion, all of this is the basis of a general approach for advanced prognostic systems. In no default scenarios, so whether the percentage goes up or down, the outcome of the percentage is consistent with the rational regression percentage curve. The defect can be detected with a visualized data representation and with the percentage variation. We notice that ignition timer cylinders 2 and 3 in time interval 20–23 are not read in conformity with the sample regression form, and that the other sensors read on the graph intersect in calculation to get better accuracy.
The application of Artificial Intelligence (AI) approaches in industrial maintenance for fault detection and prediction has gained much attention from scholars and practitioners. This survey systematically assesses and classifies the state-of-the-art algorithms applied to data-driven maintenance in recent literature. The taxonomy provides a so far not existing overview and decision aid for research and practice regarding suitable AI approaches for each maintenance application. Moreover, we consider trends and further research demand in this area. Finally, a newly developed holistic maintenance framework contributes to a practice-oriented implementation of AI and considers crucial managerial aspects of an efficient maintenance system.
Many approaches for fault detection in industrial processes has been presented in the literature, with approaches spanning from traditional univariate statistics to complex models able to encode the multivariate nature of time-series data. Although the vast corpus of works on this topic, there is no public benchmark shared among the community that can serve as a testbench for these methods, allowing researchers to evaluate their proposed approach with other state of the art approaches. In this paper we present the Industrial Robot Anomaly Detection (IndRAD) dataset as a benchmark for evaluating fault detection algorithms on industrial robots. The dataset is composed by 13 nominal trajectories and 3 trajectories with structural anomalies. We also propose a protocol to inject sensory anomalies in clean data. The dataset, code to reproduce these experiments and a leaderboard table to be used for future research are available at https://github.com/franzsetti/IndRAD.KeywordsFault detectionanomaly detectiontime series analysisindustrial robots
Lean Management and its tools have been widely used for years. Lean Management aims at streamlining the flow of value while continually seeking to reduce the resources required to produce a given set of products. Although the adoption of Lean is not a new concept, few organizations fully understand the philosophy behind its practices and principles. The relationship between Industry 4.0 and Lean Management has been increasingly evidenced in operations management research. To create a better understanding, the main point of interest for this work is to investigate the link and integration between Industry 4.0 and Lean Management, as well as examine its implications on performance and the environmental factors influencing these relationships in some companies especially focusing on the mining industry. Based on the literature review, a questionnaire was created about Lean Management and Industry 4.0, which was applied in some companies in Brazil and Hungary, most of them from the mining industry. The aim of this paper is to evaluate the application of combining both methodologies, Lean Management and Industry 4.0. The unique contribution of the paper is to see the common areas of Lean and Industry 4.0 where there are research and knowledge, but the application level at the companies is low.
With the development of fly-by-wire controls, electro-mechanical actuators (EMAs) have become essential components for aircraft. In order to guarantee operational safety and reliability of EMA, prognostics and health management (PHM) can be utilized to acquire reliable prediction information on potential failures before they occur. Furthermore, as a significant procedure of PHM, construction of health indicators (HIs) enables assessment of performance degradation. Due to the advantages of low computational complexity and strong interpretability, the multivariate state estimation technique (MSET) has become one of the mainstream methods for HI extraction. However, given various operation conditions of EMA, it is difficult to select appropriate distance metric of MSET that accurately calculate health state of EMA, which will further lead to accuracy losses of HI. To solve above problem, an improved multivariate state estimation technique with a composite operator (CO-MSET) for EMA HI extraction is proposed in this paper. Firstly, monitored parameters under different operation conditions of EMA are used to construct observation matrix and memory matrix. Secondly, a composite nonlinear operator with different optimization weights is introduced to calculate estimates. Finally, the output of extracted HI will be further obtained by calculating the residual vector. To validate the effectiveness of the proposed method, experiments are conducted on the dataset from NASA’s flyable electro-mechanical actuator (FLEA). Experimental results illustrate that the proposed method has a better performance on HI extraction for EMA, which is suitable for EMA health state representation under various operation conditions.
Cone crushers of medium and fine crushing are used at mining enterprises for crushing materials with quite a wide range of physical and mechanical properties. The review and analysis of scientific and technical literature on crushing process regularities in cone crushers has determined the following: the dependence between material grain size distribution and crusher space profile parameters, wear rate of mobile and stationary cone liner, establishment of an optimum crusher space profile, minimisation of high manganese steel consumption, optimisation of crushing equipment repair and maintenance costs. The condition of the liner mantle is one of the determining factors in changing the qualitative and quantitative characteristics of the fine crushing process. It is established that a significant failure number of cone crushers, namely 96% lead to unscheduled repairs. Research objective. In order to implement a new approach in monitoring the liner mantle state. Methods. A complex approach was used, which includes: scientific analysis and generalization of previously published researches. The theory of fuzzy logic and fuzzy sets and methods of the system analysis made a methodological basis of researches. Novelty. The possibility of using the methods of artificial intelligence in the assessment of the liner mantle state in cone crushers has been implemented. Result: The approach in controlling the liner mantle wear in fine and medium cone crushers is defined.KeywordsMineralCrusherLiner mantleWearArtificial intelligenceModellingNeural network
This paper studies a canonical general scheduling model that captures the fundamental trade-off between processing jobs and performing diagnostics (testing). In particular, testing reveals the required processing time and urgency of need-to-schedule jobs to inform future scheduling decisions. The model captures a range of important applications. Prior work focused on special cases (e.g., jobs with independent and identically distributed processing time) to devise optimal policies. In contrast, the current paper studies the most general form of the model and describes two simple heuristics to solve it; adaptive weighted shortest processing time is an adaptive generalization of Smith’s rule that optimally solves several important extensions of previously studied models, whereas index policy optimally solves a closely related stochastic optimization bandit problem. The latter achieves an approximation guarantee that quickly approaches a constant factor that is bounded by two as the number of jobs grows and approaches optimally when the testing time decreases. Extensive numerical experiments suggest that our policies effectively solve the general setting (under 0.1% from optimal on average and under 10% from optimal in rare, worst-case instances).
This paper was accepted by Jeannette Song, operations management.
Supplemental Material: The data files and online appendix are available at https://doi.org/10.1287/mnsc.2023.4833 .
1. Introduction.- 2. Model-based Fault Diagnosis Techniques.- 3. System Identification for Fault Diagnosis.- 4. Residual Generation, Fault Diagnosis and Identification.- 5. Fault Diagnosis Application Studies.- 6. Concluding Remarks.- References.
In the last years new technologies and methodologies have been developed for increasing the reliability of fault diagnosis in mechanical equipment, mainly in rotating machinery. Global vibration indexes as RMS, Kurtosis, etc., are widespread known in industry and in addition, are recommended by international norms. Despite that, these parameters do not allow reaching reliable equipment condition diagnosis. They are attractive for their apparent simplicity of interpretation. This work presents a discussion about the diagnosis possibilities based on these traditional parameters. The database used comprises rolling bearings vibration signals taking into account different fault conditions, several shaft speeds and loading. The obtained results show that these global vibration parameters are limited regarding correct fault diagnosis, especially in initial faults condition. As an alternative method a new technique is proposed. This technique seeks to obtain a global parameter that makes better characterization of fault condition. This methodology, named Residual Energy, uses integration of the difference between the power spectrum density of the fault condition and the normal one. The results obtained with this technique are compared with the traditional RMS and Kurtosis.
Condition monitoring allows maintenance actions to be based on changes of a trend of machine or process parameters. The identification, monitoring and fusion of data from key parameters can provide greater confidence in decisions. Data fusion is a relatively new term to the condition monitoring community. In defence and other applications the field is mature and has seen extensive application. The advances in methods and frameworks for applications, in terms of condition monitoring, include: • the crystallisation of a cohesive scheme for problem definition; • structured solution selection; • comparisons with dissimilar application fields which have similar problem structures or solution methods; • the blending of quantitative and qualitative methods. This paper briefly reviews architectures or frameworks, and draws examples from manufacturing and plant applications. The chief application shown establishes the key vibration monitoring parameters for equipment within a paper mill and aids the maintenance optimisation process.
This paper explores a method to assess assets performance and predict the remaining useful life, which would lead to proactive maintenance processes to minimize downtime of machinery and production in various industries, thus increasing efficiency of operations and manufacturing. At first, a performance model is established by taking advantage of logistic regression analysis with maximum-likelihood technique. Two kinds of application situations, with or without enough historical data, are discussed in detail. Then, real-time performance is evaluated by inputting features of online data to the logistic model. Finally, the remaining life is estimated using an ARMA model based on machine performance history; degradation predictions are also upgraded dynamically. The results such as current machine running condition and the remaining useful life, are output to the maintenance decision module to determine a window of appropriate maintenance before the machine fails. An application of the method on an elevator door motion system is demonstrated.
Ferrographic analysis is required in order to detect wear particles in lubricating oil automatically, because the customary approach takes a great deal of time. We propose a new method to detect wear particles in lubricating oil in order to diagnose bearings, by means of local spatial frequency analysis using the wavelet transform. The Gabor function and cylindrical Gabor function are used as the mother functions of the wavelet transform in this paper. The Gabor function is effective in detecting particles which distribute along the lines of magnetic force on the ferrogram slide. The cylindrical Gabor function can detect circular particles. To discriminate the particles, we apply fuzzy system theory to the image transformed by two Gabor functions and show the effectiveness of this method.
The On-line Operator Aid SYStem (OASYS) has been developed to support the operator’s decision making process and to ensure the safety of a nuclear power plant by providing operators with proper guidelines in a timely manner, according to the plant operation mode. The OASYS consists of four systems such as a signal validation and management system (SVMS), a plant monitoring system (PMS), an alarm filtering and diagnostic system (AFDS), and a dynamic emergency procedure tracking system (DEPTS). The SVMS and the PMS help operators to maintain a plant in a condition to withstand the adverse events during a normal operation condition. The AFDS covers the abnormal events until it exceeds the limit range of reactor trip signals, while after a reactor trip, the DEPTS aids operators with proper guidelines so as to shut down safely.
The OASYS uses a rule-based expert system and fuzzy logic. The rule-based expert system is used to classify the predefined events and track the emergency operating procedures (EOPs) through data processing, and the fuzzy logic is used to generate the conceptual high-level alarms for the prognostic diagnosis and to evaluate the qualitative fuzzy criteria used in the EOPs. Evaluation results show that the OASYS is capable of diagnosing plant abnormal conditions and providing operators appropriate guidelines with fast response time and consistency.
The proposed system is implemented on a SUN-4/75 workstation using C language and Quintus prolog language. Currently, the OASYS is installed in the realtime full scope simulator for validation. After sufficient validation, the OASYS will be installed in the main control room for the unit one nuclear power plant at Young Gwang.
This paper is concerned with a research about fuzzy neural networks and application in fault diagnosis of rotary machine. To build robust fault diagnosis by neural networks, fuzzy neural networks are proposed. Fuzzy neural networks can memorize fault patterns and recognize fault patterns by association, and have good tolerance to unstable practical sample data. Through application in monitoring and fault diagnosis of pump sets of oil plants, it was verified that fuzzy neural networks are effective to handle practical sample data and make accurate fault diagnosis.
This paper reveals some experiences gained during the development of a knowledge-based expert system for on-line fault detection and diagnosis. The symptom tree model and fault-consequence digraph developed in our previous research showed that they can be applied effectively in real-time fault diagnosis of plants, particularly for large and complex ones.
The second edition of this highly successful text focuses on the major changes that have taken place in this field in recent times. Data Acquisition Techniques Using PCs, Second Edition, recognises that data acquisition is the core of most engineering and many life science systems in measurement and instrumentation. It will prove invaluable to scientists, engineers, students and technicians wishing to keep up with the latest technological developments. Teaches the reader how to set up a PC-based system that measures, analyzes, and controls experiments and processes through detailed design examples. Geared for beginning and advanced users, with many tutorials for less experienced readers, and detailed standards references for more experienced readers. Fully revised new edition discusses latest programming languages and includes a list of over 80 product manufacurers to save valuable time.
In this paper, we study the on-line parameter estimation problem for a partially observable system subject to deterioration and random failure. The state of the system evolves according to a continuous-time homogeneous Markov process with a finite state space. The state of the system is hidden except for the failure state. When the system is operating, only the information obtained by condition monitoring, which is related to the working state of the system, is available. The condition monitoring observations are assumed to be in continuous range, so that no discretization is required. A recursive maximum likelihood (RML) algorithm is proposed for the on-line parameter estimation of the model. The new RML algorithm proposed in the paper is superior to other RML algorithms in the literature in that no projection is needed and no calculation of the gradient on the surface of the constraint manifolds is required. A numerical example is provided to illustrate the algorithm.
The analysis of vibration spectra is an extensively used tool in machinery condition-based maintenance looking for early failure prevention and diagnosis. The cluster analysis is a statistical technique widely used in a variety of scientific fields and rarely employed in the investigation of machine vibrations. This paper tries to examine the potential of the cluster analysis in the detection of machine failure and compares its advantages and disadvantages in the discrimination of the data obtained from a series of vibration spectra. Results of the application of this technique in failure detection of two common components in rotating machinery (rolling bearings and spur gears) are presented. The vibration spectra from these elements, having in each case a known failure, are obtained from the test rig hold in the Mechanics Department of UNED. From a series of observations (vibration amplitude at different frequencies), they are classified into groups as homogeneous as possible, according to the different measured variables. The results obtained have the sufficient precision as to show that cluster analysis is an efficient criterion in the discrimination of the state of rotating machinery.
During the run-up process of rotating machinery, there is large amount information in the rotor system, which does not appear in the steady wheeling stage. Hence the run-up process faults diagnosis is an active subject of research nowadays. This paper shows how 2-dimension hidden Markov model can be used to build a fault diagnosis system based on time-frequency analysis to the process signal. The topology of this model and its parameters are introduced. Meanwhile how to select the features by time-frequency analysis is discussed here. Finally, the new model was tested with experimental data collected from Benty-Nevada rotor experimental system (Model 24755) and the new method was found to outperform previously proposed classifiers based on the average spectrum of the machinery fault signal with much more correct classifications.
This chapter introduces signals and the mathematical tools needed to work with them, and combines discussions of analog signals, discrete signals, digital signals and the methods to transition from one of these realms to another. All that it requires of the reader is a familiarity with calculus. There are a wide variety of examples. They illustrate basic signal concepts, filtering methods, and some easily understood techniques for signal interpretation. The first section introduces the terminology of signal processing, the conventional architecture of signal processing systems, and the notions of analog, discrete, and digital signals. It describes signals in terms of mathematical models – functions of a single real or integral variable. Next, we cover the two basic signal families: analog and discrete, respectively. Later, we discuss sampling and interpolation. Next we cover periodicity, and foremost among these signals is the class of sinusoids. The chapter concludes with a discussion of the mathematics that arises in the detailed study of signals. A summary, a list of references, and a problem set complete the chapter.
The focus of the paper is the optimization of condition-based maintenance decisions within the contexts of physical asset management. In particular, the analysis of a preventive replacement policy of the control-limit type for a deteriorating system subject to inspections at discrete points of time is presented. Cox's PHM with a Weibull baseline hazard function and time dependent stochastic covariates is used to describe the failure rate of the system. The methods of estimating model parameters and the calculation of the optimal policy are given. The structure of the decision-making software EXAKT is presented. Experience with collecting, processing and using real oil and vibration data is reported.
Condition-based maintenance (CBM) is an effective preventive maintenance policy since the policy enhances system reliability and decreases the unwanted breakdown of the target maintained system and the maintenance cost. We propose a simultaneous optimisation method of the inspection time vector, which is a time series of inspection, and a warning level of the target system under the constraint of preventive replacement probability. The developed CBM model considers the residual life loss, which is related to the average remaining time from the preventive replacement conduction time until the expected breakdown time of a maintained system, as well as the incurred replacement and inspection costs, whereby the long-run average incurred cost per unit time is derived as an objective function for minimisation. An algorithm to solve the optimal inspection time vector and optimal warning level is established. Numerical examples to which the constructed CBM model and the solution algorithm have been applied are shown with a variety of parameters in the CBM model, together with the comparison of the results obtained in breakdown maintenance and optimal time-based maintenance.
The following describes a spectroscopic approach for achieving condition-base maintenance (CBM) through the monitoring of lubricant degradation in machinery. The approach uses infrared and fluorescence spectroscopy to monitor lubricant condition as a function of changing spectral variance. The use of two different spectroscopies has been pursued because varying CBM requirements within the industrial and military communities can be met with the complimentary advantages of each approach. Because the spectroscopic methods are inherently multivariate, the authors have also chosen to pursue a multivariate analysis of the data. The authors results show that multivariate infrared spectroscopy offers certain advantages over multivariate fluorescence spectroscopy and vice versa. It is shown that the two approaches fulfill different CBM requirements. This is an important and useful result because the CBM requirements for monitoring a given lubricant can vary greatly from one field application to the next. The authors have monitored a synthetic aircraft lubricant, Hatco Corporation (MIL-L-23699D), that is found in the rotary gearboxes of several military helicopters. The authors show that when used in conjunction with multivariate analysis techniques, such as PCA, that both infrared and fluorescence spectroscopy can provide an ideal means of monitoring the condition of a lubricant as it degrades. The authors confirm that infrared absorption can provide a means of monitoring the breakdown of the additive package, the decomposition of an ester functionality, and the production of a degradation breakdown product as a function of degradation time. The authors also show that fluorescence emission can provide a rapid, low-cost, easy-to-implement means of monitoring the 'integrated' condition of a lubricant as a function of degradation time.
Vibratory effects are sensed quite easily by most people, they are tiring and unpleasant, at the extreme they can be frightening. It is not surprising that vibration analysis occupies a prominent place in machine diagnostics.
Prognostics, which refers to the inference of an expected time-to-failure for a mechanical system, is made difficult by the need to track and predict the trajectories of real-valued system parameters over essentially unbounded domains, and by the need to prescribe a subset of these domains in which an alarm should be raised. In this paper we propose a novel technique whereby these problems are avoided: instead of physical system or sensor parameters, sensor-level test-failure probability vectors (bounded within the unit hypercube) are tracked; and via a close relationship with the TEAMS suite of modeling tools, the terminal states for all such vectors can be enumerated. To perform the tracking, a Kalman filter with associated interacting multiple model switching between failure regimes is proposed, and simulation results indicate that performance is promising.
Foreword. Preface. 1. Introduction/Background 2. The Wavelet Transform. 3. Practical Resolution, Gain, and Processing Structures. 4. Wavelet Theory Extentions and Ambiguity Functions. 5. Linear Systems Modelling with Wavelet Theory. 6. Wideband Scattering and Environmental Imaging. Related Research. References. Subject Index.
Vibration based machine fault diagnosis is widely adopted in machine condition monitoring. Since a machine is usually composed of many mechanical components, during the machine running, each component will generate its vibration and transmit to other components thru the shaft or linkages. Hence, the vibration signal collected from a sensor is the aggregation of all generated vibrations. To enhance the accuracy in vibration based machine fault diagnosis, the vibration generated by each component must be isolated and identified. In this paper, the performance of blind-source-separation (BSS) in separating various mixed sources is discussed. The BSS based method of second order statistics (SOS) has been applied to separate the aggregated vibration signals generated from a number of mechanical components. To verify the effectiveness of the BSS based SOS, a number of experiments were conducted using both simulated data and vibration generated form the industrial machines. The results show that the BSS possesses the ability to separate both artificially and naturally mixed signals. Such ability is definitely welcome in the fields of condition monitoring and maintenance. Moreover, the paper also discusses the advantages and disadvantages of the algorithm in the applications of machine fault diagnosis and future improvements.
This paper provides an overview of the current available technologies for automated machinery condition evaluation and fault diagnosis within an overall plant asset management system. The paper presents a basic overview of an integrated plant asset management system, and focuses on the available technologies for automated diagnostics including statistical analysis of data, parametric model diagnosis, non-parametric model diagnosis (artificial neural networks), and rule-based diagnostics including expert systems and fuzzy logic. The current state-of-the-art and the expected realistic future developments are discussed.
Statistical pattern recognition is a term used to cover all stages of an investigation from problem formulation and data collection through to discrimination and classification, assessment of results and interpretation. This chapter introduces some of the basic concepts in classification and describes the key issues. It presents two complementary approaches to discrimination, namely a decision theory approach based on calculation of probability density functions and the use of Bayes theorem, and a discriminant function approach. Many different forms of discriminant function have been considered in the literature, varying in complexity from the linear discriminant function to multiparameter nonlinear functions such as the multilayer perceptron. Regression is an important part of statistical pattern recognition. Regression analysis is concerned with predicting the mean value of the response variable given measurements on the predictor variables and assumes a model of the form. Bayes' theorem; regression analysis; statistical process control
We examine a replacement problem for a system subject to stochastic deterioration. Upon failure the system must be replaced by a new one and a failure cost is incurred. If the system is replaced before failure a smaller cost is incurred. The failure of the system depends both on its age and also on values of a diagnostic stochastic process observable at discrete points of time. Cox’s proportional hazards model is used to describe the failure rate of the system. We consider the problem of specifying a replacement rule which minimizes the long-run expected average cost per unit time. The form of the optimal replacement policy is found and an algorithm based on a recursive computational procedure is presented which can be used to obtain the optimal policy and the optimal expected average cost.
A closer look has been taken into repair and maintenance actions (RMAs) that affect systems in an industrial environment. It has been discussed that the classical definition of 'repair' is complex. Virtually every RMA affects the system in some way. To model repairable systems, a concept is needed that reflects the state of the system incorporating the scale of RMAs. A few concepts and their possible incorporations are proposed. The repair and maintenance indicator type 'a' (RMIa) is introduced as the weighted accumulated operating time of all or the most important parts of the system. The indicator type 'b' (RMIb) is introduced as a state of a system depending on its age and RMA history. The indicator type 'c' is introduced to reflect the degree of repair of an RMA, These indicators can be useful for the modelling and comparison of repairable systems. Several methods to incorporate repair and maintenance indicators into intensity process models for repairable system reliability are discussed. The indicators may be used as covariates in a proportional intensity model (PIM), or as reduction factors in a virtual age process model.
A new approach to the development of a more robust diagnosis and
prognostic capability is described. This approach is based fusion of
sensor-based and model-based information. These micromechanical models
must account for initial flow size distribution and other
microstructural parameters describing initial component condition. A
specific application of this approach is addressed, the diagnosis of
mechanical failure in meshing gears. Four specific issues are considered
are these issues are presented.
Vibration-based machine condition monitoring incorporates a number of machinery fault detection and diagnostic techniques. Many machinery fault diagnostic techniques use automatic signal classification in order to increase accuracy and reduce errors caused by subjective human judgment. In this paper, fuzzy logic techniques have been applied to classify frequency spectra representing various rolling element bearing faults. The frequency spectra representing a number of different fault conditions have been processed using a variety of fuzzy set shapes. The application of basic fuzzy logic techniques has allowed fuzzy numbers to be generated which represent the similarity between frequency spectra. Correct classification of different bearing fault spectra was observed when the correct combination of fuzzy set shapes and range of membership domains were used. The problem of membership overlapping found in previous studies, where classifying individual spectrum with respect to spectra that represent true fault classes was not conclusive, has been overcome. Further work is described which will extend this technique to other classes of machinery using generic software.
A prototype condition monitoring and diagnostic system has been developed for compression refrigeration plants, which can be used under variable operational conditions. Based on a combination of causal analysis, expert knowledge and simulated failure modes, a failure mode symptom matrix has been created. Healthy system behaviour is predicted based on a regression analysis model. Using multi-valued (or ‘fuzzy’) logic, real-time recognition of failure modes, at an early stage, proved to be possible. Future developments for improvement of diagnostic systems in compression refrigeration plants are discussed.RésuméOn a mis au point un système prototype de suivi et de diagnostic pour les installations de refroidissement à compression, qui peut être utilisé dans plusieurs conditions de fonctionnement. En se fondant sur la combinaison d'une analyse causale, sur la connaissance des experts et sur les modes de simulation des pannes, on a créé une matrice dite de ‘symptôme’ en matière de panne. On prv́oit le comportement du système en bon état en se basant sur un modèle d'analyse par régression. En utilisant la logique floue, la détection précoce des pannes en temps réel s'est avérée possible. On examine les réalisations futures qui permettront d'améliorer ces systèmes de diagnostic dans les installations de refroidissement à compression.
This paper is concerned with diagnostic methods for gears and bearings, the goals of which are to detect incipient structural and metallurgical faults, characterize their nature and severity, and isolate them to particular components. Such information is invalu- able for prognostic purposes. The methodology presented herein focuses on the analysis of vibration signals induced by gears and bearings and, specifically, on a means of extracting from such vibration patterns amplitude and phase modulation signals. Such modulation terms, we conjecture, are simple indicators of the severity of a particular type of localized component defect. We show that such amplitude and phase modulation information can be extracted using a neural network computational methodology that relies on nothing more than knowledge of the bearing geometry and the frequency tones at which a given type of defect will manifest itself. We present numerical simulation examples and show that the technique requires extremely few a priori assumptions. Introduction: The topics addressed herein pertain to the general problem of machinery prognostics, i.e., estimating the remaining useful life of machinery components based on knowledge of their present condition and assumptions about future usage. The ability to diagnose particular machine components and understand the implications of diagnostic information vis-` a-vis future usage of the machine is of immense practical importance in industrial, commercial, and military applications. Component failures contribute to the majority of machine operation safety incidents and are responsible for a substantial fraction of the on-going maintenance costs accruing to machinery usage. This underscores the need for condition-based maintenance (CBM), wherein knowledge regarding the condition of machinery, rather than fixed time intervals, is used to schedule maintenance, thus averting unnecessary inspections. The quality of the diagnostic information acquired (i.e., fault detection, type and sever- ity characterization, and component isolation) plays a paramount role in determining the ecacy of prognostication. Use of inaccurate diagnostic information in, say, a prognostic decisional algorithm driving a CBM policy for a fleet of helicopters may result in ill-fated missions. Prompt and accurate detection of structural and metallurgical faults in machinery components (e.g., gears and bearings) is an important category of such diagnostic informa- tion that is our focus herein. Almost all physical techniques for detecting the onset of such defects, e.g., fatigue cracks or surface spalling, involve the measurement of abnormal vibration signals transmitted through structural media due to the presence of the defect. Vibration sensors (e.g., accelerometers) may be used to measure these signals. Sensitive spectral analysis and astute signal processing of such sensor data is critical, chiefly because in practical machinery operational scenarios, there are an enormous number of vi- brations that interfere with one another and result in an observed signal that appears bewildering and uninformative. The main problem is therefore one of discovering the nee- dle (i.e., spectral content that provides useful and pertinent information about an incipient fault) in the haystack (i.e., the raw vibration signal). A number of sophisticated techniques
We present a generic methodology for machinery fault diagnosis through pattern recog- nition techniques. The proposed method has the advantage of dealing with complicated signatures, such as those present in the vibration signals of rolling element bearings with and without defects. The signature varies with the location and severity of bearing defects, load and speed of the shaft, and different bearing housing structures. More specifically, the proposed technique contains effective feature extraction, good learning ability, reli- able feature fusion, and a simple classification algorithm. Examples with experimental testing data were used to illustrate the idea and effectiveness of the proposed method. @DOI: 10.1115/1.1687391#
A physics-based approach for diagnostics and prognostics using integrated observers and life models is presented. Observers are filters based on physical models of machine- fault combinations and use measured machine signatures to identify and characterize the state of a machine. Observers are adaptively deployed as a machine wears and can be coupled with one another to handle interacting conditions and faults. The scheme is detailed using the fault of a cracked rotor shaft that interacts with gravity and imbalance. Observers for shaft cracking and imbalance are presented. The observers provide machine condition and fault strengths to life models used to determine remaining machine life. A life model based on the Forman crack growth law of linear elastic fracture mechanics is developed to determine the number of machine cycles remaining until catastrophic failure.
This paper presents a novel methodology for the diagnosis and prognosis of crucial gear faults, such as gear tooth fatigue cracking. Currently, an effective detection of tooth cracking can be achieved by using the autoregressive (AR) modeling approach, where the gear vibration signal is modeled by an AR model and gear tooth cracking is detected by identifying the sudden changes in the model's error signal. The model parameters can be estimated under the criteria of minimum power or maximum kurtosis of model errors. However, these model parameters possess no physical meaning about the monitored gear system. It is proposed that the AR model be replaced by a gear dynamics model (GDM) that contains physically meaningful parameters, such as mass, damping and stiffness. By identifying and tracking the changes in the parameters, it is possible to make diagnosis and prognosis of gear faults. For example, a reduction in mesh stiffness may indicate cracking of a gear tooth. Towards physical model-based prognosis, an adaptive (or optimization) strategy has been developed for approximating a gear signal using a simplified gear signal model. Preliminary results show that this strategy provides a feasible adaptive process for updating model parameters based on measured gear signal.
Complex dynamical systems such as aircraft, manufacturing systems, chillers, motor vehicles, submarines, etc. exhibit continuous and event-driven dynamics. These systems undergo several discrete operating modes from startup to shutdown. For example, a certain shipboard system may be operating at half load or full load or may be at start-up or shutdown. Of particular interest are extreme or "shock" operating conditions, which tend to severely impact fault diagnosis or the progression of a fault leading to a failure. Fault conditions are strongly dependent on the operating mode. Therefore, it is essential that in any diagnostic/prognostic architecture, the operating mode be identified as accurately as possible so that such functions as feature extraction, diagnostics, prognostics, etc. can be correlated with the predominant operating conditions. This paper introduces a mode identification methodology that incorporates both time- and event-driven information about the process. A fuzzy Petri net is used to represent the possible successive mode transitions and to detect events from processed sensor signals signifying a mode change. The operating mode is initialized and verified by analysis of the time-driven dynamics through a fuzzy logic classifier. An evidence combiner module is used to combine the results from both the fuzzy Petri net and the fuzzy logic classifier to determine the mode. Unlike most event-driven mode identifiers, this architecture will provide automatic mode initialization through the fuzzy logic classifier and robustness through the combining of evidence of the two algorithms. The mode identification methodology is applied to an AC Plant typically found as a component of a shipboard system.
A simple methodology for identification and quantification of nonlinear effects such as Coulomb friction and backlash is desired for use in condition based maintenance programs for both structural and machine based applications. Typically, structural applications are passive and undergo small vibratory motion when an external excitation is presented to the system. A spring-mass system was used as the structural example. Machine applications are typically active and motion is excited by internal actuation of large motion within the system. An industrial SCARA robot was used as the machine based example. The Hilbert transform was tested for detection and quantification of Coulomb friction in both systems.