Browsing by Author "Bennett, Ian"

Now showing 1 - 10 of 10
  • No Thumbnail Available
    ItemOpen Access
    Abrupt fault remaining useful life estimation using measurements from a reciprocating compressor valve failure
    (Elsevier, 2018-11-24) Loukopoulosa, Panagiotis; Pilidisa, Pericles; Bennett, Ian; Zolkiewski, George; Li, Xiaochuan; Mba, David
    One of the major targets in industry is minimisation of downtime and cost, and maximisation of availability and safety, with maintenance considered a key aspect in achieving this objective. The concept of Condition Based Maintenance and Prognostics and Health Management (CBM/PHM) which is founded on the principles of diagnostics and prognostics, is a step towards this direction as it offers a proactive means for scheduling maintenance. Reciprocating compressors are vital components in oil and gas industry, though their maintenance cost is known to be relatively high. Compressor valves are the weakest part, being the most frequent failing component, accounting for almost half maintenance cost. To date, there has been limited information on estimating Remaining Useful Life (RUL) of reciprocating compressor in the open literature. This paper compares the prognostic performance of several methods (multiple linear regression, polynomial regression, Self-Organising Map (SOM), K-Nearest Neighbours Regression (KNNR)), in relation to their accuracy and precision, using actual valve failure data captured from an operating industrial compressor. SOM technique is proposed to be employed for the first time as a standalone tool for RUL estimation. Furthermore, two variations on estimating RUL based on SOM and KNNR respectively are proposed. Finally, an ensemble method by combining the output of all aforementioned algorithms is proposed and tested. Principal components analysis and statistical process control were implemented to create T^2 and Q metrics, which were proposed to be used as health indicators reflecting degradation processes and were employed for direct RUL estimation for the first time. It was shown that even when RUL is relatively short due to instantaneous nature of failure mode, it is feasible to perform good RUL estimates using the proposed techniques.
  • No Thumbnail Available
    ItemOpen Access
    Canonical Variable Analysis and Long Short-term Memory for Fault Diagnosis and Performance Estimation of a Centrifugal Compressor
    (Elsevier, 2018-01-03) Li, Xiaochuan; Duan, Fang; Mba, David; Bennett, Ian; Loukopoulosa, Panagiotis
    Centrifugal compressors are widely used for gas lift, re-injection and transport in the oil and gas industry. Critical compressors that compress flammable gases and operate at high speeds are prioritized on maintenance lists to minimize safety risks and operational downtime hazards. Identifying incipient faults and predicting fault evolution for centrifugal compressors could improve plant safety and efficiency and reduce maintenance and operation costs. This study proposes a dynamic process monitoring method based on canonical variable analysis (CVA) and long short-term memory (LSTM). CVA was used to perform fault detection and identification based on the abnormalities in the canonical state and the residual space. In addition, CVA combined with LSTM was used to estimate the behavior of a system after the occurrence of a fault using data captured from the early stages of deterioration. The approach was evaluated using process data obtained from an operational industrial centrifugal compressor. The results show that the proposed method can effectively detect process abnormalities and perform multi-step-ahead prediction of the system’s behavior after the appearance of a fault.
  • Thumbnail Image
    ItemUnknown
    Canonical variate analysis, probability approach and support vector regression for fault identification and failure time prediction
    (IOS Press, 2018-06-22) Li, X.; Duan, Fang; Bennett, Ian; Mba, David
    Reciprocating compressors are widely used in oil and gas industry for gas transport, lift and injection. Critical compressors that compress flammable gases and operate at high speeds are high priority equipment on maintenance improvement lists. Identifying the root causes of faults and estimating remaining usable time for reciprocating compressors could potentially reduce downtime and maintenance costs, and improve safety and availability. In this study, Canonical Variate Analysis (CVA), Cox Proportional Hazard (CPHM) and Support Vector Regression (SVR) models are employed to identify fault related variables and predict remaining usable time based on sensory data acquired from an operational industrial reciprocating compressor. 2-D contribution plots for CVA-based residual and state spaces were developed to identify variables that are closely related to compressor faults. Furthermore, a SVR model was used as a prognostic tool following training with failure rate vectors obtained from the CPHM and health indicators obtained from the CVA model. The trained SVR model was utilized to estimate the failure degradation rate and remaining useful life of the compressor. The results indicate that the proposed method can be effectively used in real industrial processes to perform fault diagnosis and prognosis.
  • No Thumbnail Available
    ItemOpen Access
    Canonical variate residuals-based contribution map for slowly evolving faults
    (Elsevier, 2019-02-23) Li, Xiaochuan; Yang, Xiaoyu; Yang, Yingjie; Bennett, Ian; Collop, Andy; Mba, David
    The superior performance of canonical variate analysis (CVA) for fault detection has been demonstrated by a number of researchers using simulated and real industrial data. However, applications of CVA to fault identification of industrial processes, especially for faults that evolve slowly, are not widely reported. In order to improve the performance of traditional CVA-based methods to slowly developing faults, a novel diagnostic approach is put forward to implement incipient fault diagnosis for dynamic process monitoring. Traditional CVA fault detection approach is extended to form a new monitoring index based on indices, Hotelling’s T2, Q and a canonical variate residuals (CVR)-based monitoring index Td. As an alternative to the traditional CVA-based contributions, a CVR-based contribution plot method is proposed based on Q and Td statistics. The proposed method is shown to facilitate fault detection by increasing the sensitivity to incipient faults, and aid fault identification by enhancing the contributions from fault- related variables and suppressing the contributions from fault-free variables. The CVR-based method has been demonstrated to outperform traditional CVA-based diagnostic methods for fault detection and identification when validated on slowly evolving faults in a continuous stirred tank reactor (CSTR) system and an industrial centrifugal pump.
  • No Thumbnail Available
    ItemOpen Access
    Condition monitoring of rotating machines under time-varying conditions based on adaptive canonical variate analysis
    (Elsevier, 2019-06-06) Li, Xiaochuan; Yang, Yingjie; Bennett, Ian; Mba, David
    Condition monitoring signals obtained from rotating machines often demonstrate a highly non-stationary and transient nature due to internal natural deterioration characteristics of their constituent components and external time-varying operational conditions. Traditional multivariate statistical monitoring approaches are based on the assumption that the underlying processes are linear and static and are apt to interpret the normal changes in operating conditions as faults, which would result in high false positive rates. On the other hand, the development of robust diagnostic techniques for the detection of incipient faults remains a challenge for researchers, given the difficulty of finding an appropriate trade-off between a low false positive ratio and early detection of emerging faults. To address these issues, this paper proposes a novel adaptive fault detection approach based on the canonical residuals (CR) induced by the combination of canonical variate analysis (CVA) and matrix perturbation theory for the monitoring of dynamic processes where variations in operating conditions are incurred. The canonical residuals are calculated based upon the distinctions between past and future measurements and are able to effectively detect emerging faults while still maintaining a low false positive rate. The effectiveness of the developed diagnostic model for the detection of abnormalities in industrial processes was demonstrated for slow involving faults in case studies of two operational industrial high-pressure pumps. In comparison with the variable-based and canonical correlation-based statistical monitoring approaches, the proposed canonical residuals-adaptive canonical variate analysis (CR-ACVA) fault detection method has demonstrated its superiorities by the detailed performance comparisons.
  • Thumbnail Image
    ItemOpen Access
    Instantaneous failure mode remaining useful life estimation using non-uniformly sampled measurements from a reciprocating compressor valve failure
    (Elsevier, 2018) Loukopoulosa, Panagiotis; Zolkiewski, George; Bennett, Ian; Sampath, Suresh; Pilidisa, Pericles; Li, X.; Mba, David
    One of the major targets in industry is minimisation of downtime and cost, and maximisation of availability and safety, with maintenance considered a key aspect in achieving this objective. The concept of Condition Based Maintenance and Prognostics and Health Management (CBM/PHM) , which is founded on the principles of diagnostics, and prognostics, is a step towards this direction as it offers a proactive means for scheduling maintenance. Reciprocating compressors are vital components in oil and gas industry, though their maintenance cost is known to be relatively high. Compressor valves are the weakest part, being the most frequent failing component, accounting for almost half maintenance cost. To date, there has been limited information on estimating Remaining Useful Life (RUL) of reciprocating compressor in the open literature. This paper compares the prognostic performance of several methods (multiple linear regression, polynomial regression, Self-Organising Map (SOM), K-Nearest Neighbours Regression (KNNR)), in relation to their accuracy and precision, using actual valve failure data captured from an operating industrial compressor. The SOM technique is employed for the first time as a standalone tool for RUL estimation. Furthermore, two variations on estimating RUL based on SOM and KNNR respectively are proposed. Finally, an ensemble method by combining the output of all aforementioned algorithms is proposed and tested. Principal components analysis and statistical process control were implemented to create T^2 and Q metrics, which were proposed to be used as health indicators reflecting degradation processes and were employed for direct RUL estimation for the first time. It was shown that even when RUL is relatively short due to instantaneous nature of failure mode, it is feasible to perform good RUL estimates using the proposed techniques.
  • Thumbnail Image
    ItemOpen Access
    An intelligent diagnostic and prognostic framework for large-scale rotating machinery in the presence of scarce failure data
    (Sage, 2019) Li, Xiaochuan; Yang, Xiaoyu; Yang, Yingjie; Bennett, Ian; Mba, David
    In this work, a novel diagnostic and prognostic framework is proposed to detect faults and predict remaining service life of large-scale rotating machinery in the presence of scarce failure data. In the proposed framework, a canonical variate residuals–based diagnostic method is developed to facilitate remaining service life prediction by continuously implementing detection of the prediction start time. A novel two-step prognostic feature exploring approach that involves fault identification, feature extraction, feature selection and multi-feature fusion is put forward. Most existing prognostic methods lack a fault-identification module to automatically identify the fault root-cause variables required in the subsequent prognostic analysis and decision-making process. The proposed prognostic feature exploring method overcomes this challenge by introducing a canonical variate residuals–based fault-identification method. With this method, the most representative degradation features are extracted from only the fault root-cause variables, thereby facilitating machinery prognostics by ensuring accurate estimates. Its effectiveness is demonstrated for slow involving faults in two case studies of an operational industrial centrifugal pump. Moreover, an enhanced grey model approach is developed for remaining useful life prediction. In particular, the empirical Bayesian algorithm is employed to improve the traditional grey forecasting model in terms of quantifying the uncertainty of remaining service life in a probabilistic form and improving its prediction accuracy. To demonstrate the superiority of empirical Bayesian–grey model, existing prognostic methods such as grey model, particle filter–grey model and empirical Bayesian–exponential regression are also utilized to realize machinery remaining service life prediction, and the results are compared with that of the proposed method. The achieved predictive accuracy shows that the proposed approach outperforms its counterparts and is highly applicable in fault prognostics of industrial rotating machinery. The use of in-service data in a practical scenario shows that the proposed prognostic approach is a promising tool for online health monitoring.
  • Thumbnail Image
    ItemOpen Access
    Multidimensional prognostics for rotating machinery: A review
    (SAGE, 2017-02-06) Li, Xiaochuan; Duan, Fang; Mba, David; Bennett, Ian
    Determining prognosis for rotating machinery could potentially reduce maintenance costs and improve safety and avail- ability. Complex rotating machines are usually equipped with multiple sensors, which enable the development of multidi- mensional prognostic models. By considering the possible synergy among different sensor signals, multivariate models may provide more accurate prognosis than those using single-source information. Consequently, numerous research papers focusing on the theoretical considerations and practical implementations of multivariate prognostic models have been published in the last decade. However, only a limited number of review papers have been written on the subject. This article focuses on multidimensional prognostic models that have been applied to predict the failures of rotating machinery with multiple sensors. The theory and basic functioning of these techniques, their relative merits and draw- backs and how these models have been used to predict the remnant life of a machine are discussed in detail. Furthermore, this article summarizes the rotating machines to which these models have been applied and discusses future research challenges. The authors also provide seven evaluation criteria that can be used to compare the reviewed techniques. By reviewing the models reported in the literature, this article provides a guide for researchers considering prognosis options for multi-sensor rotating equipment.
  • No Thumbnail Available
    ItemOpen Access
    Reciprocating compressor prognostics of an instantaneous failure mode utilising temperature only measurements
    (Elsevier, 2017-12-14) Loukopoulosa, Panagiotis; Zolkiewski, George; Bennett, Ian; Sampath, Suresh; Pilidisa, Pericles; Duan, Fang; Sattar, Tariq; Mba, David
    Reciprocating compressors are critical components in the oil and gas sector, though their maintenance cost is known to be relatively high. Compressor valves are the weakest component, being the most frequent failure mode, accounting for almost half the maintenance cost. One of the major targets in industry is minimisation of downtime and cost, while maximising availability and safety of a machine, with maintenance considered a key aspect in achieving this objective. The concept of Condition Based Maintenance and Prognostics and Health Management (CBM/PHM) which is founded on the diagnostics and prognostics principles, is a step towards this direction as it offers a proactive means for scheduling maintenance. Despite the fact that diagnostics is an established area for reciprocating compressors, to date there is limited information in the open literature regarding prognostics, especially given the nature of failures can be instantaneous. This work presents an analysis of prognostic performance of several methods (multiple linear regression, polynomial regression, K-Nearest Neighbours Regression (KNNR)), in relation to their accuracy and variability, using actual temperature only valve failure data, an instantaneous failure mode, from an operating industrial compressor. Furthermore, a variation for Remaining Useful Life (RUL) estimation based on KNNR, along with an ensemble technique merging the results of all aforementioned methods are proposed. Prior to analysis, principal components analysis and statistical process control were employed to create T^2 and Q metrics, which were proposed to be used as health indicators reflecting degradation process of the valve failure mode and are proposed to be used for direct RUL estimation for the first time. Results demonstrated that even when RUL is relatively short due to instantaneous nature of failure mode, it is feasible to perform good RUL estimates using the proposed techniques.
  • No Thumbnail Available
    ItemOpen Access
    Valve Failure Prognostics In Reciprocating Compressors Utilizing Temperature Measurements, PCA-based Data Fusion And Probabilistic Algorithms
    (IEEE, 2019-07-09) Loutas, Theodoros; Loukopoulos, Panagiotis; Georgoulas, George; Bennett, Ian; Mba, David
    In the present paper, temperature measurements are utilized to develop health indicators based on principal component analysis towards the probabilistic estimation of the Remaining Useful Life (RUL) of reciprocating compressors in service. Temperature degradation histories obtained from thirteen actual valve failure cases constitute the training data in a data-driven prognostic approach. Two data-driven prognostic methodologies are presented and proposed based on probabilistic mathematical models i.e. Gradient Boosted Trees (GBTs) and Non-Homogeneous Hidden Semi Markov Models (NHHSMM). The training and testing process of all models is described in detail. RUL prognostics in unseen data are obtained for all models. Beyond the mean estimates of the RUL, the uncertainty associated with the point prediction is quantified and upper/lower confidence bounds are also estimated. Prediction estimates for twelve real-life failure cases are presented and the pros and cons of each model’s performance are highlighted. Several metrics are utilized to assess the performance of the prognostic algorithms and conclusions are drawn regarding the prognostic capabilities of each of them.