Search

In the oil and gas industry, statistical methods have been used for corrosion analysis for various asset systems such as pipelines, storage tanks, and so on. However, few industrial standards and guidelines provide comprehensive stepwise procedures for the usage of statistical approaches for corrosion analysis. For example, the UK HSE (2002) report "Guidelines for the use of statistics for analysis of sample inspection of corrosion" demonstrates how statistical methods can be used to evaluate corrosion samples, but the methods explained in the document are very basic and do not consider risk factors such as pressure, temperature, design, external factors and other factors for the analyses. Furthermore, often the industrial practice that uses linear approximation on localised corrosion such as pitting is considered inappropriate as pitting growth is not uniform. The aim of this research is to develop an approach that models the stochastic behaviour of localised corrosion and demonstrate how the influencing factors can be linked to the corrosion analyses, for predicting the remaining useful life of components in oil and gas plants. This research addresses a challenge in industry practice. Non-destructive testing (NDT) and inspection techniques have improved in recent years making more and more data available to asset operators. However, this means that these data need to be processed to extract meaningful information. Increasing computer power has enabled the use of statistics for such data processing. Statistical software such as R and OpenBUGS is available to users to explore new and pragmatic statistical methods (e.g. regression models and stochastic models) and fully use the available data in the field. In this thesis, we carry out extreme value analysis to determine maximum defect depth of an offshore conductor pipe and simulate the defect depth using geometric Brownian motion in Chapter 2. In Chapter 3, we introduce a Weibull density regression that is based on a gamma transformation proportional hazards model to analyse the corrosion data of piping deadlegs. The density regression model takes multiple influencing factors into account; this model can be used to extrapolate the corrosion density of inaccessible deadlegs with data available from other piping systems. In Chapter 4, we demonstrate how the corrosion prediction models in Chapters 2 and 3 could be used to predict the remaining useful life of these components. Chapter 1 sets the background to the techniques used, and Chapter 5 presents concluding remarks based on the application of the techniques.

Prognostics and health management (PHM) is a major tool enabling systems to evaluate their reliability in real-time operation. Despite ground-breaking advances in most engineering and scientific disciplines during the past decades, reliability engineering has not seen significant breakthroughs or noticeable advances. Therefore, self-awareness of the embedded system is also often required in the sense that the system should be able to assess its own health state and failure records, and those of its main components, and take action appropriately. This thesis presents a radically new prognostics approach to reliable system design that will revolutionise complex power electronic systems with robust prognostics capability enhanced Insulated Gate Bipolar Transistors (IGBT) in applications where reliability is significantly challenging and critical. The IGBT is considered as one of the components that is mainly damaged in converters and experiences a number of failure mechanisms, such as bond wire lift off, die attached solder crack, loose gate control voltage, etc. The resulting effects mentioned are complex. For instance, solder crack growth results in increasing the IGBT’s thermal junction which becomes a source of heat turns to wire bond lift off. As a result, the indication of this failure can be seen often in increasing on-state resistance relating to the voltage drop between on-state collector-emitter. On the other hand, hot carrier injection is increased due to electrical stress. Additionally, IGBTs are components that mainly work under high stress, temperature and power consumptions due to the higher range of load that these devices need to switch. This accelerates the degradation mechanism in the power switches in discrete fashion till reaches failure state which fail after several hundred cycles. To this end, exploiting failure mechanism knowledge of IGBTs and identifying failure parameter indication are background information of developing failure model and prognostics algorithm to calculate remaining useful life (RUL) along with ±10% confidence bounds. A number of various prognostics models have been developed for forecasting time to failure of IGBTs and the performance of the presented estimation models has been evaluated based on two different evaluation metrics. The results show significant improvement in health monitoring capability for power switches. Furthermore, the reliability of the power switch was calculated and conducted to fully describe health state of the converter and reconfigure the control parameter using adaptive algorithm under degradation and load mission limitation. As a result, the life expectancy of devices has been increased. These all allow condition-monitoring facilities to minimise stress levels and predict future failure which greatly reduces the likelihood of power switch failures in the first place.

The increasing popularity of electric vehicles (EVs) is driven by their compatibility with sustainable energy goals. However, the decline in the performance of energy storage systems, such as batteries, due to their degradation puts EVs and hybrid electric vehicles (HEVs) at a disadvantage compared to traditional internal combustion engine (ICE) vehicles. The batteries used in these vehicles have limited life. The degradation of the battery is accelerated by the operating conditions of the vehicle, which further reduces its life and increases the reliability and economic concerns for the vehicle’s operation. The aging mechanism inside a battery cannot be eliminated but can be minimized depending on the vehicle’s operating conditions and different control mechanisms that can alter the operating conditions. Different operating conditions affect the aging mechanism differently. Knowing the factors and how they impact battery capacity is crucial for minimizing degradation. This dissertation presents the detailed degradation mechanism inside the battery and the major factors responsible for the degradation, along with their effects on the battery during the operation of EVs. Then, to abate the degradation mechanism, a prognostic-based control framework (PBCF) for HEVs is proposed. Also, this framework reduces the overall cost of operating HEVs by taking into account the degradation of the energy storage systems. The strategy utilizes a degradation forecasting model of energy storage systems to predict their degradation paths. Analytical and data-driven approaches are used to find the degradation path of the energy storage systems as follows: Markov Chain Model and Neural Network Model. These two models employ distinct datasets to validate the feasibility of the proposed strategy. The predicted degradation rate is then used to control the HEV via its energy management (EM) system in order to reduce the degradation of energy storage systems. During the simulation, three distinct operating scenarios are developed to observe their effects on battery degradation and the response of the proposed control strategy PBCF.

Batteries, such as Lithium-ion battery, have the advantages of high specific energy, low pollution and high safety, and thus have become one of the main power sources for new energy automobiles. The State of Health (SOH) and Remaining Useful Life (RUL) of a power battery are the most important performance index in the battery system. The SOH of the power battery is different from other battery parameters such as voltage, current, internal resistance and temperature in that it cannot be obtained through direct measurement by any equipment or instrument. Besides, these electrical parameters will degrade with the use of batteries. For this reason, the accurate online estimation of the SOH and RUL of the battery has become one of the key challenges in the battery management system. This dissertation focuses on investigating key battery performance indicators such as internal resistance, capacity and self-discharge of Ni-H2 battery and lithium battery. The research also concentrates on the SOH estimation of the batteries through the development of Extended Kalman Filter (EKF) algorithm, the online estimation strategy of Dual Extended Kalman Filter (DEKF) and comprehensive assessment of Particle Filter (PF) and Support Vector Regression (SVR). Data-driven models, such as Autoregressive Moving Average (ARMA) model, artificial neural network, and hybrid models are implemented to provide both the degradation analysis and performance prediction for NiH2 battery cells. The prediction model can also help reduce the false alarm rate of the existing battery online monitoring system. The long-term behavior of pressure, as a degradation indicator, is modeled to help understand the battery aging behavior. Combining the electrochemical battery model, a state equation and an output equation for the SOC estimation of the lithium battery are established. Each parameter in the SOC system is observed by using the EKF algorithm. The system simulation results indicate that the SoC estimation of the lithium battery in the EKF algorithm has good precision and accuracy. Furthermore, to solve the problem that the initial parameters of the battery model in the online SOH estimation of the battery cannot be determined in advance, the DEKF algorithm is introduced. Two independent EKFs are established to estimate the state of battery system and the parameters, respectively, and mutually update their states and parameters. The film resistance and discharging capacity are estimated to represent battery’s SOH. The advantages of this proposed method are two-fold: (1) implementing physics-based models to provide physical interpretation of Lithium-ion battery cell, and (2) utilizing dual models to maintain the long-term accuracy of estimates. Finally, a novel battery SOH monitoring model is built to analyze the proposed degradation parameters and to predict the RUL by updating its probability distribution. The Support Vector Regression-Particle Filter (SVR-PF) algorithm is implemented in the research work to make improvement over the standard PF, which has the degeneracy phenomenon. The SVR-PF shows improved estimation and prediction capability compared to the standard PF.

Abstract: Maintenance management in wind energy industry has great impact on overall wind power cost. Optimizing maintenance strategies can substantially reduces the cost and makes wind energy more competitive among the energy resources. Due to the extreme conditions of remote or offshore sites where the wind turbines are installed, corrective maintenance and time-based preventive maintenance are the most adopted strategies in the wind industry in recent years. However, there is need to further reduce wind power cost via maintenance strategy improvement to increase its competitiveness. Industry and research community have been focusing on various maintenance strategies to save the maintenance cost. This thesis is devoted to developing cost-effective maintenance strategies for wind farms, focusing on conventional time-based maintenance optimization, and prognostics and condition-based maintenance (CBM) optimization within the CBM strategy framework. Studies are performed on improving corrective maintenance and time-based preventive maintenance strategies, which are currently widely adopted in wind industry. Opportunistic maintenance methods are proposed, which take advantage of economic dependencies existing among the wind turbines, and corrective maintenance chances, to implement preventive maintenance simultaneously. Imperfect preventive maintenance actions are considered as well. The methods demonstrate the immediate benefits of saving the overall maintenance cost for a wind farm. In the more advanced CBM strategy, the health conditions of components are monitored and predicted, based on which maintenance actions are scheduled to prevent unexpected failures while reducing the maintenance costs. Prognostic techniques are essential in CBM. In particular, the wind direction and speed around wind turbines are changing over time, which leads to instantaneously time-varying load applied to the wind turbines rotors. With focus on gearbox failure due to the gear tooth crack, an integrated prognostics method is developed considering instantaneously varying load condition. The numerical examples demonstrate that the gearbox remaining useful life prediction considering time-varying load is more accurate compared to existing methods under constant-load assumption. In a subsequent extended study, uncertainty in gear tooth crack initiation time is further considered for wind turbine gearbox prognostics method development. The method provides more accurate gearbox remaining useful life prediction compared to the results without considering time-varying load condition. This thesis also proposes a CBM method considering different turbine types and lead times, as well as the production loss during the shutdown time. The capability to accurately estimate the average maintenance cost for a wind farm with diverse turbines is a key contribution of the proposed method. In addition, this thesis accounts for the inaccuracy in the simulation-based algorithms that most complex problems are solved with. A numerical method for CBM optimization of wind farms is developed to avoid the variations in CBM cost evaluation, which leads to a smooth cost function surface and benefits the optimization process. The research in this thesis provides innovative methods for maintenance management in the wind power industry. The developed methods will help to significantly reduce the overall maintenance cost within either conventional maintenance or CBM strategies that the wind farm owners may apply. It will improve the competitive advantage of the wind energy, and promote a clean and sustainable energy future for the society in Canada and worldwide.

With the advancement of sensor technology and data processing capacities, various types of high volume data are available for process monitoring and prognostics in manufacturing systems. In a broaching process, a multi-toothed broaching tool removes material from the workpiece by sequential engagement and disengagement of multiple cutting edges. The quality of the final part, including the geometric integrity and surface finish, is highly dependent upon the broaching tool condition. Though there has been a considerable amount of research on tool condition monitoring and prognostics for various machining processes, the broaching process is unique in the following aspects: 1) a broaching process involves multiple cutting edges, which jointly contribute to the final part quality; 2) the resharpening and any other process adjustments to the tool can only be performed with the whole broaching tool or at least a whole segment of the tool replaced. The overarching goal of this research is to explore how engineering knowledge can be used to improve process monitoring and prognostics for a complex manufacturing process like broaching. This dissertation addresses the needs for developing new monitoring and prognostics approaches based on various types of data. Specifically, the research effort focuses on 1) the use of in-situ force profile data for real-time process monitoring and fault diagnosis, 2) degradation characterization for broaching processes on an individual component level based on image processing; and 3) system-level degradation modeling and remaining useful life prediction for broaching processes based on multiple images.

In the field of engineering, it is important to understand different engineering systems and components, not only in how they currently perform, but also how their performance degrades over time. This extends to the field of prognostics, which attempts to predict the future of a system or component based on its past and present states. A common problem in this field is the estimation of remaining useful life, or how long a system or component functionality will last. The well-known datasets for this problem are the PHM and C-MAPSS datasets. These datasets contain simulated sensor data for different turbofan engines generated over time, and have been used to study estimations of remaining useful life. In this thesis, we study estimations of remaining useful life using different methods of data analytics, preprocessing, post-processing, different target functions used for training models, and their combinations. We compared their performance primarily using scores from the 2008 Prognostics and Health Management Competition. Our basic feedforward neural network outperforms previously competitors and other modern methods. Our results also gave us a ranking between the top 10 and 15 based on a 2014 benchmark. We have improved on the results of previously, published methods, primarily focusing on the Prognostics Health Management competition score of our results for better comparisons.

Abstract: Prognostics in engineering field is dedicated to predicting how long further a component or a system will perform their intended functions before failure. Prognostics is an essential building block in condition based maintenance. Accurate prediction of the component remaining useful life provides valuable information to decision making on maintenance planning, mission planning and logistics. Preventive actions based on remaining useful life prediction can dramatically avoid unscheduled downtime, reduce operational risk and cost, and improve the safety of the working environment. This thesis is devoted to developing integrated prognostics methods for the remaining useful life prediction of a specific component by integrating physics of failure and condition monitoring data. The first contribution of the thesis is that by combining physics and data effectively, the proposed method overcomes the limitations of existing prognostics approaches, which are mainly either physics based or data-driven. To account for the uncertainty in failure times of units in population, parameters are treated as random variables in physical degradation models. By noticing the uniqueness of the failure time for a specific unit, this study utilizes Bayesian inference to reduce the uncertainty in model parameters, which leads to a more accurate prediction on remaining useful life of the specific unit. This thesis also proposes an integrated prognostics method for the component operating under time-varying operating conditions. The capability to directly relate the load to the degradation rate is a key advantage of the proposed method over the existing data-driven methods when dealing with time-varying operating conditions. This is the second main contribution of this thesis. To cater to real-time applications of condition based maintenance, an efficient spectral method named polynomial chaos expansion is investigated for uncertainty quantification in prognostics. The proposed method is able to accelerate the uncertainty quantification in the integrated prognostics method and the computational efficiency is significantly improved, which is the third main contribution of this thesis. In addition, this thesis accounts for two important factors when developing integrated prognostics method: uncertainty in damage initiation time and shock in the degradation. These two factors have not been explicitly considered for prognostics purpose in the existing research. By simultaneously adjusting both the damage initiation time and the model parameters, the prediction accuracy is improved. The failure time reduction caused by the shock is accommodated by identifying a virtual damage initiation time. This work consists of the fourth main contribution. The integrated prognostics methods developed in this thesis are applied to spur gears. Two types of failure modes are considered. One is the tooth fracture due to bending stress and the other one is the surface wear due to sliding contact. Validation is conducted using a run-to-failure experiment on a planetary gearbox.

Information from fault codes associated with a component may be used as an indicator of its health. A fault code is defined as a timestamp at which a component is not operating according to recommended guidelines. The type of fault codes which are relevant for this analysis represent mild or moderate deviations from normal behavior, rather than those requiring immediate repair. Potentially, fault codes may be used to determine the Remaining Useful Life (RUL) of a component by predicting its failure time, which will improve safety and reduce maintenance costs associated with the component. In this dissertation, methods have been developed to integrate the degradation information from fault codes into an existing prognostic parameter to improve the estimation of RUL. Optimization methods such as gradient descent were used to weight each fault code based on their relevance to degradation. Furthermore, topic models, a document analysis and clustering technique, were used as both a dimension-reduction method and fault mode isolation. Methods developed for this dissertation were applied to two real-world data sets, an actuator system and monitored signals from a motor accelerated degradation experiment. The best estimation of RUL for the actuator system was a topic model with a mean absolute error of 6.41% of the data received, and the best estimation of RUL for the motor accelerated degradation experiment was 5.7% of the average lifetime of the motors. The primary contributions of this research includes a method to construct a prognostic parameter from fault codes alone, the integration of degradation information from fault codes into an existing prognostic parameter, the use of topic models in reliability analysis of fault codes, and a software suite that performs these functions on generic data sets.

The present thesis presents an insight to the problem of battery aging diagnosis andprognosis problem for automotive applications (PHEV vehicles). The work has been carried out at Center for Automotive Research (CAR) USA. The problem itself is not a trivial one and it requiters certain systems theory tools that have been well developed in the literature to be carried out on an experimental data set that has been obtained at CAR by aging LI-ION batteries throughout various real time driving cycles profiles. Estimating the Remaining Useful Life (RUL) of a component is at the center of the problem of prognosis and health management. It is a useful tool that informs the engineer with how much time it is left before the functionality of the competent is lost. The problem of prediction is to deal with multiple sources of uncertainties such as noises in the systems, degradation in the component’s functionality.

In recent years, a significant amount of research has been directed towards the development of prognostic methodologies to forecast the future health state of an engineering system assisting condition based maintenance. These prognostic methods, having furthered the maintenance practices for mechanical systems, have yet to be applied to historic masonry structures, many of which stand in an aged and degraded state. Implementation of prognostic methodologies to historic masonry structures can advance the planning of successful conservation and restoration efforts, ultimately prolonging the life of these heritage structures. This thesis presents a review of prognostic concepts and techniques available in the literature as applied to various engineering disciplines, and evaluates the well-established prognostic techniques for their applicability to historic masonry structures. Challenges of adapting the existing prognostic techniques to historic masonry are discussed, and the future direction in research, development, and application of prognostic methods to masonry structures is highlighted. One particular prognostic technique, known as support vector regression, has had successful applications due to its ability to compromise between fitting accuracy and generalizability (i.e. flatness) in the training of prediction models. Optimal tradeoff between these two aspects depends on the amount of extraneous noise in the measurements, which in civil engineering applications, is typically caused by loading conditions unaccounted for in the development of the prediction model. Such extraneous loading, often variable with time affects the optimal tradeoff. This thesis presents an approach for optimally weighing fitting accuracy and flatness of a support vector regression model in an iterative manner as new measurements become available. The proposed approach is demonstrated in prognostic evaluation of the structural condition of a historic masonry coastal fortification, Fort Sumter located in Charleston, SC. A finite element model is used to simulate responses of a casemate within the fort considering differential settlement of supports. Within the case study, the adaptive optimal weighting approach proved to have increased prediction accuracy over the non-weighted option.

The degradation process of a complex system may be affected by many unknown factors, such as unidentified fault modes, unmeasured operational conditions, engineering variance, environmental conditions, etc. These unknown factors not only complicate the degradation behaviors of the system, but also lower the quality of the collected data for modeling. Due to lack of knowledge and incomplete measurements, certain important context information (e.g. fault modes, operational conditions) of the collected data will be missing. Therefore historical data of the system with a large variety of degradation patterns will be mixed together. With such data, learning a global model for Remaining Useful Life (RUL) prediction becomes extremely hard. This has led us to look for advanced RUL prediction techniques beyond the traditional global models. In this thesis, a novel RUL prediction method inspired by the Instance Based Learning methodology, called Trajectory Similarity Based Prediction (TSBP), is proposed. In TSBP, the historical instances of a system with life-time condition data and known failure time are used to create a library of degradation models. For a test instance of the same system whose RUL is to be estimated, similarity between it and each of the degradation models is evaluated by computing the minimal weighted Euclidean distance defined on two degradation trajectories. Based on the known failure time, each of the degradation models will produce one RUL estimate for the test instance. The final RUL estimate can then be obtained by aggregating the multiple RUL estimates using a density estimation method. A case study using the turbofan engine degradation simulation data supplied by NASA Ames is provided to study the performance of TSBP. In this study, the TSBP method has demonstrated significant improvement in performance over a Neural Network based prediction method.

The rotating machinery health assessment and prediction task can be divided into several aspects, since determining and predicting the level of degradation is a function of the signal processing and feature extraction methods, the machine learning algorithm used for health assessment, and the regression methods or other methods used for predicting the future health state. The first aspect of this research was focused on the evaluation of the logistic regression, self-organizing map, and statistical pattern recognition health assessment algorithms. The benchmarking and evaluation of these three algorithms was conducted using vibratory and electrical data collected in a test-bed setting from a new and degraded automotive alternator component. Each of the three algorithms are evaluated based on their accuracy for calculating the bearing, diode and stator winding health, with the use of the logistic regression method providing the best performance accuracy with a combined false positive and false negative rate of 5%. The second aspect of this research was focused on the evaluation of signal processing and feature extraction techniques for rotating machinery health assessment, facilitated by the use of data collected from a bearing test-rig with rolling element bearings of different levels of induced damage. The high frequency envelope technique and the empirical mode decomposition signal processing methods are compared on the basis of whether each method provides an enhanced level of indicators that can determine the health of rolling element bearings. Each of the methods is compared to the traditional Fourier transform and time domain feature extraction method to determine if the additional signal processing and computations provide more robust features for bearing health assessment. The empirical mode decomposition and envelope feature extraction methods provide a more robust set of features for detecting the lower levels of bearing damage compared to the features extracted from the Fourier transform of the time signal. The last contribution focuses on the prediction aspect of the rotating machine health assessment framework. The accuracy of the remaining useful life prediction is calculated as a function of the prediction steps ahead as well as the current health status of the system and compared for different failure thresholds. This provides a quantitative constraint on when to use this prediction method, in that the accuracy of the remaining life prediction using this method provides accurate results for predictions once the detected level of degradation is 25% or greater.