Access the full text.
Sign up today, get DeepDyve free for 14 days.
V. Venkatasubramanian (2005)
Prognostic and diagnostic monitoring of complex systems for product lifecycle management: Challenges and opportunitiesComput. Chem. Eng., 29
(2003)
based on multifractal and MTS method
D Djurdjanovic (2003)
109International Journal of Advanced Engineering Informatics, Special Issue on Intelligent Maintenance Systems, 17
Jinqiu Hu, Laibin Zhang, W. Liang, Cunjie Guo (2012)
Intelligent Risk Assessment for Pipeline Third-Party InterferenceJournal of Pressure Vessel Technology-transactions of The Asme, 134
J. Lee, Chiao-Lin Deng, Chiachi Tsan, Mohamed AbuAli (2010)
Product life cycle knowledge management using embedded Infotronics: methodology, tools and case studiesInt. J. Knowl. Eng. Data Min., 1
Jinqiu Hu, Laibin Zhang, W. Liang, Zhaohui Wang (2009)
Quantitative HAZOP Analysis for Gas Turbine Compressor based on Fuzzy Information FusionSystems Engineering - Theory & Practice, 29
Jinqiu Hu, Laibin Zhang, W. Liang (2012)
An adaptive online safety assessment method for mechanical system with pre-warning functionSafety Science, 50
D Liu (2013)
509Acta Aeronautica et Astronautica Sinica, 34
V. Venkatasubramanian (2011)
Systemic failures: Challenges and opportunities in risk management in complex systemsAiche Journal, 57
Shicun Wang, Zhi Zu, Zenhao Li, Ticusen Ruan, Songshan Yang (1983)
Acta Aeronautica et Astronautica Sinica
Tangbin Xia, L. Xi, Jay Lee, Xiaojun Zhou (2011)
Optimal CBPM policy considering maintenance effects and environmental conditionThe International Journal of Advanced Manufacturing Technology, 56
Adam Markowski, M. Mannan, Agata Bigoszewska (2009)
Fuzzy logic for process safety analysisJournal of Loss Prevention in The Process Industries, 22
A. Markowski, M. Mannan, A. Kotynia, D. Siuta (2010)
Uncertainty aspects in process safety analysisJournal of Loss Prevention in The Process Industries, 23
Zhang Jiuxing (2013)
Quantitative Risk Evaluation Methods for Multi-factor Coupling Complex Flight Situations
Jinqiu Hu, Laibin Zhang, W. Liang (2013)
Dynamic degradation observer for bearing fault by MTS–SOM systemMechanical Systems and Signal Processing, 36
Jinqiu Hu, Laibin Zhang, Lin Ma, W. Liang (2011)
An integrated safety prognosis model for complex system based on dynamic Bayesian network and ant colony algorithmExpert Syst. Appl., 38
Geunwoong Yun, W. Rogers, M. Mannan (2009)
Risk assessment of LNG importation terminals using the Bayesian–LOPA methodologyJournal of Loss Prevention in The Process Industries, 22
D. Djurdjanović, J. Lee, J. Ni (2003)
Watchdog Agent - an infotronics-based prognostics approach for product performance degradation assessment and predictionAdv. Eng. Informatics, 17
J H Holland (1992)
17Daedalus, 121
Mary O’Connor, Susan Mitchell, M. Mannan (2006)
Designing Resilient Engineered SystemsChemical Engineering Progress, 102
Fangji Wu, Tianyi Wang, Jay Lee (2010)
An online adaptive condition-based maintenance method for mechanical systemsMechanical Systems and Signal Processing, 24
D Djurdjanovic, J Lee, J Ni (2003)
Watchdog Agent-an infotronics based prognostics approach for product performance assessment and predictionInternational Journal of Advanced Engineering Informatics, Special Issue on Intelligent Maintenance Systems, 17
N. Medellín-Castillo, R. Leyva-Ramos, R. Ocampo-Pérez, Ramon Cruz, A. Aragón-Piña, J. Martínez-Rosales, R. Guerrero-Coronado, L. Fuentes-Rubio (2007)
Adsorption of Fluoride from Water Solution on Bone CharIndustrial & Engineering Chemistry Research, 46
J. Lee, J. Ni, D. Djurdjanović, Hai Qiu, H. Liao (2006)
Intelligent prognostics tools and e-maintenanceComput. Ind., 57
J Hu (2012)
165Journal of Pressure Vessel Technology, Transactions of the ASME, 134
Jinqiu Hu, Laibin Zhang, Zhaohui Wang, W. Liang (2009)
The application of integrated diagnosis database technology in safety management of oil pipeline and transferring pump unitsJournal of Loss Prevention in The Process Industries, 22
Yanjun Wang, H. West, M. Mannan (2004)
The Impact of Data Uncertainty in Determining Safety Integrity LevelProcess Safety and Environmental Protection, 82
J Hu, L Zhang, W Liang (2009)
Mechanical incipient fault detection based on multifractal and MTS methodPetroleum Science, 6
Jinqiu Hu, Laibin Zhang, Lin Ma, W. Liang (2010)
An integrated method for safety pre-warning of complex system
Xiaole Yang, M. Mannan, Mary O’Connor, A. McFerrin (2010)
An uncertainty and sensitivity analysis of dynamic operational risk assessment model: A case studyJournal of Loss Prevention in The Process Industries, 23
S. Rahman (2011)
Reliability Engineering and System Safety
Jinqiu Hu, Laibin Zhang, W. Liang (2012)
Opportunistic predictive maintenance for complex multi-component systems based on DBN-HAZOP modelProcess Safety and Environmental Protection, 90
Xiaole Yang, M. Mannan (2010)
The development and application of dynamic operational risk assessment in oil/gas and chemical process industryReliab. Eng. Syst. Saf., 95
S. Rajaraman, J. Hahn, M. Mannan (2004)
A methodology for fault detection, isolation, and identification for nonlinear processes with parametric uncertaintiesIndustrial & Engineering Chemistry Research, 43
S M Mitchell (2006)
39Chemical Engineering Progress, 102
Jinqiu Hu, Laibin Zhang, W. Liang, Zhaohui Wang (2009)
Incipient mechanical fault detection based on multifractal and MTS methodsPetroleum Science, 6
(2012)
How to prevent accidents in process
Ricardo Vieira (1992)
Complex Adaptive SystemsDaedalus, 121(1)
486 Pet.Sci.(2013)10:486-493 DOI 10.1007/s12182-013-0299-9 Safety prognostic technology in complex petroleum engineering systems: progress, challenges and emerging trends Zhang Laibin and Hu Jinqiu College of Mechanical and Transportation Engineering, China University of Petroleum, Beijing 102249, China © China University of Petroleum (Beijing) and Springer-Verlag Berlin Heidelberg 2013 Abstract: Oil and gas facilities used in the petroleum industry can be considered as complex dynamic systems in that they require different types of equipment with various causal relationships among components and process variables under monitoring. As the systems grow increasingly large, high speed, automated and intelligent, the nonlinear relations among these process variables and their effects on accidents are to be fully understood for both system reliability and safety assurance. Failures that occur during the process can both cause tremendous loss to the petroleum industry and compromise product quality and affect the environment. Therefore, failures should be detected as soon as possible, and the root FDXVHVQHHGWREHLGHQWL¿HGVRWKDWFRUUHFWLRQVFDQEHPDGHLQWLPHWRDYRLGIXUWKHUORVVZKLFKUHODWHWR the safety prognostic technology. By investigation of the relationship of accident causing factors in complex systems, new progress into diagnosis and prognostic technology from international research institutions is reviewed, and research highlights from China University of Petroleum (Beijing) in this area are also presented. By analyzing the present domestic and overseas research situations, the current problems and future directions in the fundamental research and engineering applications are proposed. Key words: Oil and gas facility, complex system, diagnostic and prognostic, coupling faults, predictive maintenance an accident happens, it could lead to a great loss, so such 1 Introduction systems are also typical high risk systems. For instance, in the Modern petroleum technological advances are creating a FDVHVRIWKHH[SORVLRQLQWKHH[DV%3&LW\7´UH¿QHU\³ rapidly increasing number of complex engineering systems, -LOLQSHWURFKHPLFDOH[SORVLRQ³´'DOLDQRLOSLSHOLQH processes, and products, such as oil-gas gathering and H[SORVLRQDQG´³3HQJODLRLO¿HOGVSLOOSRVWPRUWHP transporting system, pumps and compressors in long-distance investigations of above disasters have shown that systemic pipelines, and various chemical equipment, which have close failures rarely occur due to a single failure of a component or relationship to the oil and gas production, transportation and personnel, and in particular the main reason causing accidents processing stages. Such complex systems widely used in and their consequences lies in the complex nonlinear petroleum industry pose considerable challenges in ensuring interactions among a large number of failure causing factors. their proper design, control, safety, and management for A complexity adaptive system (CAS) theory proposed successful and continuous operation over their life cycles. by Prof. Holland (1992) is relatively new, and important It is their scale, nonlinearities, interconnectedness, and progress is being made. Researchers in the field of oil and interactions with humans and the environment that can make gas safety science began to focus on complexity science these complex systems fragile, when the cumulative effects (CS) and are trying to use it to address the above challenges, of multiple abnormalities can propagate in numerous ways to including accident-causing theory, safety evaluation system, cause systemic failures. intelligent fault diagnosis and trend forecasting. On the With long-term operation, usually more than 30 years, other hand, as the concept of the predictive maintenance was of complex petroleum engineering systems, it is quite hard proposed, instead of the corrective maintenance, condition to avoid accidents. On the other hand, with characteristics based maintenance and preventive maintenance, the focus that hazards could spread to the outside environment, once of the maintenance of petroleum facilities is beginning to understand when complex systems could fail or be vulnerable *Corresponding author. email: hujinqiu@gmail. com to certain kinds of faults and failures, and the different kinds 5HFHLYHG-XQH Pet.Sci.(2013)10:486-493 487 of fault modes. Appropriate predictive maintenance strategies Clearly, to cope with such complexity, methodologies and contingency plans can be prepared in advance this way, and automation tools to model, analyze, explain, predict, avoiding unplanned shutdown or even disastrous accidents. and control the behavior of such multiple factors as faults Therefore, the combination of the complex science and or failure in complex systems and in various environments predictive maintenance technology makes a contribution to DUHEDGO\QHHGHG2QHFHQWUDOEHQH¿WWKDWKDVFRPHWKURXJK the development of safety prognosis of complex systems. from the study of the failure propagation behavior in complex However, there still exist many challenges in this subject. petroleum engineering systems, is the need for a prognostic What are development directions in the future? These approach, with which one can anticipate faults or failures to are urgent problems that need to be addressed. If one can help managing systemic risks. In other words, one needs real- anticipate problems, rather than relying on the current “react- time safety prognostic technology that can effectively monitor and-fix” methodology for managing systemic risks, it will various aspects of process operations, and detect, diagnose, be extremely helpful to ensure the safe operation of complex predict and advise operators and engineers about incipient petroleum engineering systems, and reduce or avoid the risk abnormal events. In this way, the root causes of accidents of calamitous accidents. can be restrained, and the hazard effect can be reduced by controlling its failure propagation behavior, so the safety 2 Failure propagation behavior in complex resilience of the complex system can be improved. petroleum engineering systems 3 International research progress and The complexity of accident causes and consequences in challenges in safety prognostic technology a complex petroleum engineering system is related to the limitation of subjective cognitive ability and also the objective According to the definition of the united nations complexity of accidents themselves. Both aspects have a international strategy for disaster reduction early warning close relationship to all kinds of sophisticated nonlinear is the emergency warning of impending disasters. In safety interactions among parts or units in the system. Based on the engineering discipline, its connotation includes emergency study of the principle of integration and interconnection, it is warning against possible impending accidents, and it can also indicated that the interaction is the true ultimate cause of the include the extension of warning against secondary accidents system failures. Complex unsafe situations are likely to arise after a period of time which would be caused by the first from undesirable and dysfunctional interactions among the accident. Therefore, one needs safety prognostic technology components, subsystems, feedback loops, humans, and the to model and predict emergent behavior in complex systems HQYLURQPHQWʊQRWMXVWIURPWKHIDLOXUHRIDVLQJOHFRPSRQHQW and evaluate the future hazard effects caused by an initial or an operator mistake. Therefore, investigation into the abnormal event to provide early warning, and offer adequate interaction of structures and hierarchy in systems becomes methods or action to control accidents with minimum losses. the premise and foundation to analyze the safety situation of In order to ensure system safety by early warning, one complex systems. needs to address the crucial challenge of being able to predict Again and again, investigations have shown that the how changes or dysfunctional interactions in a complex number of accidents caused by coupling factors is alarming. engineering system or its environment would propagate There are always several layers of failures, ranging from low- through the entire system, i.e., how one makes a prediction of level components to senior units to subsystems, which have the future event consequences from the behavior of the parts led to major disasters. Surveys showed that 92% accidents to an effective description of the whole system behavior. were caused by multiple factors, and on average each accident Safety prognostic technology includes prognostic analysis were caused by more than 4.4 basic abnormal events, and and prognostic control. Prognostic analysis is the activity by WKHKLJKHVWZDVFDXVHGE\EDVLFDEQRUPDOHYHQWV/LX monitoring, identifying, diagnosis, assessment and giving et al, 2013). The failure propagation behavior caused by early warning alarm of the complex system when there exist interactions and presented by multiple basic abnormal events fault phenomena. The prognostic control is the reaction of has some characteristics such as strong randomness and prognostic analysis, through a series of management activities elusiveness, which can lead to disastrous consequences. such as correction, prevention and control for a degrading The failure propagation behavior can be explained as a trend caused by an initial abnormal event. complicated and dynamic network of activities of multiple Recent progress has promising implications for basic abnormal events both in time and space. For instance, fundamental research and engineering application of safety ¿UVWDQDEQRUPDOHYHQWKDSSHQVDQGIHFWVDIDIHZVXEV\VWHPV prognosis in complex petroleum engineering systems for ZKLFKKDYHDFORVHUHODWLRQZLWKLWLQWKHFRPSOH[V\VWHP better product quality, inherently safer designs, abnormal then these affected subsystems fail or become degraded and events management and reliable process operations. will further influence other related subsystems by certain 3.1 Failure dynamics and fragility in complex FRXSOLQJLQWHUDFWLRQIXUWKHUVRPHNH\VXEV\VWHPVZKLFK adaptive systems play an important role in keeping the system’s main function, will fail caused by former multiple factors. Therefore, as the )RUWKHSDVWWZHQW\\HDUVWKH/,36/DERUDWRU\IRU number of failed key subsystems increases, the whole system Intelligent Process Systems) group at Purdue University will partly lose control or finally become paralyzed, which KDVIRFXVHGRQULVNLGHQWL¿FDWLRQDQDO\VLVDQGPDQDJHPHQW could lead to a disastrous accident with inestimable loss. in complex systems, developing a variety of solutions 488 Pet.Sci.(2013)10:486-493 using knowledge-based systems, neural networks, uncertainty, which causes loss of confidence in the results statistical methods, analytical models and hybrid systems 5DMDUDPDQHWDODQJ:HWDO0DUNRZVNLHWDO HQNDWDVXEUDPDQLDQ9 %\LQYHVWLJDWLRQLQWRYDULRXV 2010), and also to perform dynamic operational risk analysis accidents in different forms in different domains, the failure DQJDQG0DQQDQDE < considering effect of aging patterns and their interactions have been explored and of equipment, duration of testing intervals, and the capability investigated in a broader perspective of the potential fragility of a system to recover from disturbances to the normal state. of all complex engineering systems. We need to go beyond )RULQVWDQFH%D\HVLDQEHOLHIQHW%%1 $/23IX]]\ORJLF analyzing these events as independent one-off accidents PHWKRGXQ<HWDO0DUNRZVNLHWDO has been and to study all these accidents from a common systems used by MKOPSC for data processing to simulate causal engineering perspective so that the commonalities as well chains for the study of failure propagation behavior. as the differences in failure propagation behavior can be In order to prevent disastrous accidents, MKOPSC thoroughly understood, in order to have safer design and presented that real-time decision making for risk management better risk control in the future. needed to be developed based on progress of complexity 7KHUHVHDUFKE\/,36SURFODLPHGWKDWWKHQH[WJUDQG science, multi-perspective modeling and hybrid intelligent challenge is in the creation of next generation prognostic systems. Research into safety prognostic, high-level and diagnostic systems can sense and monitor complex prediction and control are required for: 1) safe component equipment and processes in real time, identify degradation IXQFWLRQLQJDQG FRUUHFWDQGVDIHLQWHUDFWLRQEHWZHHQ in performance, predict potential failure scenarios, components. diagnose actual failures, recommend and/or take corrective 3.4 Intelligent operation support system for chemical maintenance or control actions (Venkatasubramanian, 2011). equipment 3.2 Prognostics and decision support tools for By studying the major chemical accidents which have predictive maintenance KDSSHQHGLQUHFHQW\HDUVLQ-DSDQ6X]XNLDQG0XQHVDZD The Center for Intelligent Maintenance Systems (2012) proposed an intelligent operation support system, (IMS) at the University of Cincinnati is focused on which can effectively prevent accidents. The intelligent frontier technologies in remote monitoring, prognostics operation support system is able to calculate the effect of technologies, and intelligent decision support tools (DST) fault propagation in abnormal situations and give appropriate and has developed the trademarked Watchdog Agent® information to operators. It will help operators to make quick prognostics tools for e-maintenance systems (Djurdjanovic judgments for safety. This system predicts process variables HWDO/HHHWDO ,WVSURJQRVWLFWHFKQRORJ\ in abnormal situations using a simulator. The states of all contains four categories of analytical tools that assess and equipment are input to the simulator in order to calculate predict performance or failures of machines and processes, process variables correctly. by extracting and analyzing performance-related features Although in many aspects good progress has been from inputs such as sensor data, controller signals, expert made overseas, there are few reports or literature in China knowledge, etc. Prediction results are then used for concerning the development of safety prognostic in complex maintenance decision making and infrastructure operations. petroleum engineering systems. However, with growing Decision Support Tools (DSTs) developed by IMS demands for oil and natural gas in China, system risk facilitate maintenance operations in the most production- continuously increases, which leads to major accidents efficient manner when one or more machines are likely occurring frequently, so it is quite obvious that in every aspect to fail, according to the prediction made by prognostic of China’s petroleum industry further development of safety DOJRULWKPV/HHHWDOX:HWDO;LDHWDO prognostic technology is greatly needed. They prioritize maintenance work-orders and balance limited Considering the characteristics of petroleum equipment resources by minimizing possible losses in productivity and oil and gas production, operations and processing due to unplanned downtime. Future challenges indicated procedures, the future research focus of safety prognostic by IMS will focus on degradation analysis and prediction technology in complex systems in China should pay attention to achieve and sustain near-zero breakdown performance to: 1) the discovery of incipient faults or indistinct abnormal with self-maintenance capabilities for improved reliability, HYHQWV WKHSUHGLFWLRQRIPXOWLSOHIDXOWSURSDJDWLRQSDWKV productivity, and asset utilization. 3) the prediction of the future developing trends of single or FRPELQHGDEQRUPDOVWDWHVDFFRUGLQJWRUHODWLYHYDULDEOHVDQG 3.3 Resilience engineering in complex systems 4) the reduction of the consequences of failure interaction by dynamic approaches. The Mary Kay O’Connor Process Safety Center (MKOPSC) at Texas A & M University is world-renowned 4 Achievements in safety prognostic in process safety, develops safe processes, equipment, procedures and management strategies to minimize losses technology in China University of Petroleum within the processing industry. It has been focusing on (Beijing) risk management, consequence analysis and risk resilience engineering (Mitchell and Mannan, 2006). They are engaged In recent years, considerable progress has been made in improving existing methods to handle the relatively large in safety prognostic technology in China University of Optimal feature vector Pet.Sci.(2013)10:486-493 489 Petroleum (Beijing). In this paper, we summarize our is totally lost, but the components can still continue to work contributions to this industrially important and intellectually EHIRUH¿QDOIDLOXUH7KHQRQOLQHDULQWHUDFWLRQDPRQJPXOWLSOH exciting area. These contributions are aimed at improving the failures can be induced when there exist incipient faults, accuracy of fault diagnosis in complex systems, rationality which will cause more components even the whole system to of fault causal reasoning, and the promptness of accident fail with great loss. consequence control. The main progress lies in early incipient Therefore, in order to implement early incipient fault fault prediction and diagnosis methods, multilevel integrated diagnosis, two problems should be recognized: 1) the diagnosis and prognosis technology for coupling faults, characteristics of early fault phenomena are usually very and dynamic safety assessment and predictive maintenance weak, and often submerged in heavy noise, which cause technology and engineering applications. the incipient faults to be hard for field engineers to notice The results of these studies can have two layers of DQGLGHQWLI\DQG WKHVDPSOHVFRUUHVSRQGLQJWRDEQRUPDO meaning: 1) the application of the integrated diagnosis and or fault states are quite hard to obtain, which make the prognosis technologies in the specific complex petroleum LGHQWL¿FDWLRQPRGHOYHU\GLI¿FXOWWRHVWDEOLVKDFFXUDWHO\ engineering systems to obtain specific and useful hazard Aiming to overcome above challenges, novel effective scenario and safety pre-control measures, and promotes the prognostic and diagnostic systems for monitoring, analyzing research progress in the discipline of the complex petroleum incipient faults and predicting their degradation trend has HQJLQHHULQJV\VWHPV E\WKHVWXG\RIDVPXFKVSHFLILF been proposed and applied in complex petroleum engineering failure propagation behavior in complex systems as possible, systems (Hu et al, 2009a) (shown in Fig. 1). These approaches the prognostic mechanisms summarized from them can be can be used to identify early incipient faults effectively, and popularized in more complex engineering systems in the to help to control the development of failure propagation petroleum industry. This becomes the theoretical foundation behavior. for research into universal nonlinear interaction mechanisms The significance of the research can be presented as of multiple faults and failures. follows: )RUWKH¿UVWWLPHDTXDQWLWDWLYHGLDJQRVLVLQGH[EDVHG 4.1 Early incipient fault diagnosis and prediction on multifractal spectra and generalized dimensions for methods incipient faults in complex systems has been put forward considering the multifractal features of early incipient faults. With the exception of a few sudden failures, most failure The relationship between the local scaling and the overall processes occurring in complex systems are gradual, as the characteristics of the fractal phenomena of incipient fault has component function will deteriorate gradually until control Condition monitoring Calculation of multifractal spectrum Fault degradation observing SOM based degradation Vibration signal containing bearing identification running states 15 0 10 0 P MTS model -5 0 -1 0 0 Original multifractal -1 5 0 Calculation of Mahalanobis 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 Ti m e / ( m s ) feature vector: distance; Construction of 50 0 40 0 measurement scale 30 0 P 20 0 X={x x ..., x } O 10 0 1, 2, 9 -1 0 0 -2 0 0 ^¨ ĮI ¨ I , H } max m -3 0 0 -4 0 0 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 T ime / (ms ) 60 0 {D(q Ň q íí` 40 0 Calculation of generalized SOM based degradation 20 0 Feature Validation of dimension 0 assessment m=4 optimization Mahalanobis -2 0 0 2 m=5 -4 0 0 m=6 by two-level space 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 m=7 Orthogonal m=8 -2 m=9 array m=10 -4 m=11 m=12 -6 m=13 m=14 -8 Selection of model parameters m=15 m=16 -10 m=17 (data length, delay time, embedding m=18 -12 dimension) m=19 -14 m=20 tao=1 tao=2 tao=3 tao=4 500 500 500 23 4 5 6 7 8 ln(c(q,r)) 0 0 0 -500 -500 -500 -500 -500 0 500 -500 0 500 -500 0 500 -500 0 500 tao=5 tao=6 tao=7 tao=8 500 500 500 500 0 0 0 -500 -500 -500 -500 -500 0 500 -500 0 500 -500 0 500 -500 0 500 tao=10 tao=9 tao=11 tao=12 500 500 500 500 0 0 0 0 -500 -500 -500 -500 -500 0 500 -500 0 500 -500 0 500 -500 0 500 Feedback information for parameter Failure mode mapping; modification Predictive Degradation process maintenance plans trajectories Fig. 1 Schematic of the early incipient fault diagnosis and prediction methods 2 2 Amplitude/ m/s2 Amplitude/ m/s Amplitude/ m/s ln(r) 490 Pet.Sci.(2013)10:486-493 been established. a considerable proportion of them have a very low occurrence 2) The nonlinear relationship between the selection probability, which usually make it hard for safety engineers scheme of multifractal features and diagnosis accuracy has WR¿QGWKHDFWXDOIDXOWURRWFDXVHVDQGWRFKRRVHWKHEHVWWLPH been established, based on which a statistic optimization for repair. method that only uses a few samples with abnormal states is To meet the urgent need to solve the above challenges, DOVRLPSOHPHQWHGDQGKHQFHWKHLQÀXHQFHWKDWWKHGLDJQRVWLF we proposed multilevel integrated diagnosis and prognosis process is affected by the number of samples and personal technology for coupling faults (shown in Fig. 2), which can subjective factors can be largely reduced. effectively identify the real root causes of multiple faults 3) A diagnostic and prognostic method based on self- in complex systems, and predict the residual life under the organization mapping (SOM) for early degradation in conditions of components with dependent failures. complex mechanical systems has been further established (Hu The significance of the contribution can be presented as et al, 2013). Multifractal features also play an important role follows: in the degradation observing process for the identification 1) The interaction and dependency among entities in of incipient faults. By self-organization mapping the current VSHFL¿FFRPSOH[V\VWHPVDUHFRQVLGHUHGDQGWKHQSUHVHQWHG GHJUDGDWLRQVWDWHFDQEHLGHQWL¿HGDQGWKHGHJUDGDWLRQWUHQG by the model of multiple failure propagation paths. can be tracked dynamically which will help to predict future 2) By integrating the HAZOP model, degradation model, hazards or failures in advance. dynamic Bayesian network (DBN) model, monitoring model, 4) An integrated diagnostic platform has been established assessment model, risk evaluation model and the prediction for incipient faults in oil pipelines and transferring pump units model, a hybrid and integrated safety prognosis model (ISPM) (Hu et al, 2009b) , which can be used to simulate different is developed based on dynamic Bayesian networks (Hu et al, fault modes, positions and various levels of degradation. 2010) to reduce fault occurrence probability effectively no longer relying on the assumptions of failure independence. 4.2 Multilevel integrated diagnosis and prognosis 3) An ant colony algorithm is further used on the basis technology for coupling faults of the DBN model in ISPM to search for the most probable fault propagation path with an estimated risk value (Hu In order to control the probability of fault occurrence, et al, 2011), which helps to take safety-related actions in system reliability or performance prediction is an effective time to control fault influential ranges and reduce the fault way to track the degradation of systems in the future and consequences. make a proper proactive maintenance plan to avoid failure 4) This technology has been largely applied on many RUUHGXFHWKHIDXOWLQÀXHQWLDOUDQJH,QFRPSOH[HQJLQHHULQJ complex petroleum engineering systems, such as gas turbine systems, most single faults have multiple propagation compressor systems in long transportation pipelines (Hu et al, paths, and different propagation paths may lead to different 2012c), large scale chemical facilities, oil-gas gathering and consequences, some of which may gradually recover by its transporting systems. These cases show that this technology self-control system, while others may further cause adjacent is able to improve the accuracy and efficiency of safety components’ failure and eventually lead to catastrophic management for multi-component and multi-hazard complex accidents by failure propagation behavior. Meanwhile not all systems, providing adequate advised safety-related actions, of the compatible fault propagation pathways derived from contingency plans and proactive maintenance plans. traditional risk analysis would happen in the real world, since Condition Piping and monitoring Knowledge of systems and instrumentation diagram processes Historical database (event data, maintenance records, etc.) Condition variables DBN-HAZOP Modeling Space Current hidden states HAZOP Degradation dimension Potential hazards Safety analysis analysis Deterioration Possible reasons assessment Processes Possible consequences Safety measurements Causal Future running states relationships Time variant evolution Future performance Safety Dynamic Future reliabilities prediction Bayesian Remaining useful life Temporal network Predictive maintenance dimension plans Fig. 2 Multilevel integrated diagnosis and prognosis technology for coupling faults Pet.Sci.(2013)10:486-493 491 4.3 Dynamic safety assessment and predictive An integrated dynamic assessment model based on ‘‘3-D’’ time perspective is further presented to integrate the historical, maintenance technology of complex system current and future safety performance of the system in a unit The safety status of a complex system is usually framework, considering both assessment and pre-warning of determined by its historical, current and future states together. the system functions (Hu et al, 2012a). Other technologies The safety assessment process of such system should have are the quantitative safety evaluation model based on dynamic and time variant characteristics, which helps to track fuzzy information fusion (Hu et al, 2009c), intelligent risk the dynamic states of the system and predict future probable assessment for pipeline third-party interference (Hu et al, danger in advance. In order to overcome the disadvantages 2012b) and a new opportunistic maintenance model with a of traditional static safety assessment approaches, the results fault preventive defense function (Hu et al, 2012c). These from which are often delayed and prone to produce false technologies are helpful to track and predict the safety status alarms, an adaptive online safety assessment method is of a system dynamically and to discover potential faults in proposed (shown in Fig. 3), which consists of a dynamic time, and to help to restrain the fault symptoms successfully adaptive weighting scheme. by proactive maintenance. Real-time observables ConstruĐƟ on of safety index Minimax deviation Least MSE Quntitative refinement of approach safety indicators approach Index preprocessing Non-“maximal-type” Eliminate the impact of indicators are converted disserent dimensions and Uniformization Standardization to “maximal-type”. orders of magnitude Using various weighting mechanisms AHPBW CVBW MCBW MDBW IEBW ODBW Optimal fusion weighting scheme based on game theory Dynamic assessment from “3-D” time perspective Fig. 3 Dynamic safety assessment and predictive maintenance technology of complex systems &RHI¿FLHQWRIYDULDWLRQEDVHGZHLJKWLQJ$QDO\WLFKLHUDUFK\SURFHVVEDVHGZHLJKWLQJ&9%:$+3%: 0D[LPXPGHYLDWLRQEDVHGZHLJKWLQJ0XOWLSOHFRUUHODWLRQFRHI¿FLHQWEDVHGZHLJKWLQJ0'%:0&%: IEBW: Information entropy-based weighting, ODBW: Optimal distance-based weighting) prediction. Therefore, further progress is needed to address 5 Conclusions and future directions newer aspects of the safety prognostic in these petroleum Although in many aspects good progress has been made, engineering systems. The intellectual challenges associated incidents have not stopped occurring and it is quite obvious with these questions can be categorized into four classes. that in every aspect further development of knowledge and 1) The breakthrough from qualitative to quantitative technology of safety prognostic is needed. The problems research H[LVWLQJLQFXUUHQWUHVHDUFKZRUNFDQEHVXPPDUL]HGDV³¿YH The research of safety prognostic usually can be divided PRUH¿YHOHVV´SKHQRPHQDPRUHIRFXVRQFRPSRQHQWIDXOWV LQWR¿YHOHYHOV¿UVWWRPRQLWRUWKHIDXOWV\PSWRPVVHFRQGO\ OHVVRQV\VWHPIDXOWVPRUHIRFXVRQTXDOLWDWLYHVWXG\OHVV to identify the fault modes, on the third level root cause of RQTXDQWLWDWLYHDQDO\VLVPRUHIRFXVRQVLQJOHIDXOWVOHVVRQ the fault should be diagnosed, on the fourth level the damage PXOWLSOHIDXOWVPRUHIRFXVRQVWDWLFULVNOHVVRQG\QDPLF GHJUHHRIIDXOWVKRXOGEHHYDOXDWHGDQGRQWKH¿IWKOHYHOWKH ULVNPRUHIRFXVRQFXUUHQWULVNDVVHVVPHQWOHVVRQULVN 492 Pet.Sci.(2013)10:486-493 future developing trend of the current fault will be predicted to establish a consciousness of failure prevention defense. and the residual life will be calculated to give warning in Therefore, the dynamic tracking of the logistics, energy advance. coordination and harmful energy conversion of system input If the first to the third level of study can be called and output, will be the focus of safety prognosis in complex qualitative research, then the fourth and fifth level are systems. considered as quantitative research. On the basis of qualitative Because all of the operation processes, environmental analysis, quantitative safety prognostic research can reveal factors, operation and the degradation processes of units in the regularity of the occurrence, development, evolution and complex systems are dynamic, further research is needed to propagation of fault states in complex systems. This provides study deeply how to automatically update the forecasting a foundation for further reliability assessment and life methods and the model structure and parameters according prediction for complex systems. to changes in the external environment, operation adjustment Further research is needed to pursue this line of and its dynamic degradation processes. By understanding the quantitative safety prognosis using online monitoring and mapping relationship between the evolution and phenomena IDXOWRUDEQRUPDOHYHQWLGHQWL¿FDWLRQPRGHOGULYHQRUGDWD of failure or degradation, the safety prognostic work can driven evaluation of degradation, prediction of fault trends be able to reflect the safety characteristics of the system and their effect on other related units and the remaining useful dynamically in real time. life. In brief, in the long run considerable technological help 2) The breakthrough from research on single faults to will come from the above future progress in taming the multiple fault prognostics complexity of petroleum engineering systems, which will The prognostic of single faults is now mainly to rely result in more effective safety prognostic and diagnostic on signal processing methods, which are usually easy to systems for monitoring, analyzing, and controlling systemic implement, however in industrial uses, there are strong risks, implementing the transformation from the traditional limitations of its low accuracy, weak generalization ability “fail & fix” maintenance practices to “predict & prevent”. and bad versatility, which severely restricts its application However, getting there will require innovative thinking, especially in complex engineering systems. bolder vision, and realizing some breakthroughs in and about For such complex systems, the causes of the failure are the petroleum safety engineering industry. usually not single, but include many factors which associate Acknowledgements DQGLQWHUDFWZLWKHDFKZKLFKRWKHUEULQJPRUHGLI¿FXOWLHVIRU the prognostic work. Therefore, from the perspective of single The project is supported by the Natural Science faults, it will lead to overlooked risks or false alarms. Further )RXQGDWLRQRI&KLQD*UDQW1R WKH([FHOOHQW research is needed to pursue this line to accurately identify Doctoral Dissertation Supervisor Project of Beijing (Grant FRPSRVLWHIDXOWVDQGSUHGLFWWKHLUPXWXDOLQÀXHQFHWUHQGDQG YB20111141401), the Program for New Century Excellent impact on the overall operation of the system. Talents in University (NCET-12-0972) and PetroChina 3) The breakthrough from the research on component ,QQRYDWLRQ)RXQGDWLRQ*UDQW1R' DQG faults to system faults Beijing Natural Science Foundation (3132027) and also Another area where progress is needed is multi- Supported by Science Foundation of China University of perspective modeling for system failures, as interactions <-5& 3HWUROHXP1R among the units of the complex system are the essential reason of the occurrence of a failure or accident. Component References based prognostic analysis can usually identify induced faults, 'MX UGMDQRYLF'/HH-DQG1L-DWFKGRJ: $JHQW²DQLQIRWURQLFV while the inherent hazard still exists without elimination. based prognostics approach for product performance assessment The reason is that component based prognostic studies DQGSUHGLFWLRQ,QWHUQDWLRQDO-RXUQDORI$GYDQFHG(QJLQHHULQJ often ignore the interaction among components, the external Informatics, Special Issue on Intelligent Maintenance Systems. 2003. environment and the operation adjustment, which have a lot LQÀXHQFHRQWKHGHJUDGDWLRQSURFHVVRIFRPSRQHQWVPDNLQJ ODQG-+&RPSOH[DGDSWLYHV\VWHPV'DHGDOXV +RO the results too optimistic and unrealistic. -+X =KDQJ//LDQJ:HWDO0HFKDQLFDOLQFLSLHQWIDXOWGHWHFWLRQ Further research is needed to pursue this line of multi- based on multifractal and MTS method. Petroleum Science. 2009a. perspective modeling of complex petroleum engineering 6(2): 208-216 -=KDQJ/+X/LDQJ:HWDO7KHDSSOLFDWLRQRILQWHJUDWHGG LDJQRVLV systems, developing different views of complex systems database technology in safety management of oil pipeline and from the perspectives of structure, behavior and function, WUDQVIHUULQJSXPSXQLWV-RXUQDORI/RVV3UHYHQWLRQLQWKH3URFHVV predicting the failure propagation behavior of the system ,QGXVWULHVE under various conditions, both normal and abnormal, and its -+X =KDQJ//LDQJ:HWDO4XDQWLWDWLYH+$=23DQDO\VLVIRUJDV ¿QDOLPSDFWRQWKHLQWHQGHGIXQFWLRQ turbine compressor based on fuzzy information fusion. Systems 4) The breakthrough from static to dynamic risk (QJLQHHULQJ²7KHRU\ 3UDFWLFHF prediction -+X =KDQJ//LDQJ:HWDO$QLQWHJUDWHGPHWKRGIRUVDIHW\SUH As noted earlier, the prognostic work, which is different ZDUQLQJRIFRPSOH[V\VWHP6DIHW\6FLHQFH from traditional fault diagnosis and assessment, aims to carry -=KDQJHW/$QDO/LDQJLQWHJUDWHG:VDIHW\SURJQRVLVPRGHO+XIRU out proactive control before the incipient fault stage, and complex system based on dynamic Bayesian network and ant colony Pet.Sci.(2013)10:486-493 493 parametric uncertainties. Industrial and Engineering Chemistry algorithm. Expert Systems with Applications. 2011. 38(3): 1431-1446 Research. 2004. 43(21): 6774-6786 -+X =KDQJ//LDQJ:HWDO$QDGDSWLYHRQOLQHVDIHW\DVVHVVPHQW Suz uki K and Munesawa Y. How to prevent accidents in process method for mechanical system with pre-warning function. Safety LQGXVWULHV"5HFHQWDFFLGHQWVDQGVDIHW\DFWLYLWLHVLQ-DSDQ7KHWK 6FLHQFHD RUOG:&RQIHUHQFHRI6DIHW\RI2LODQG*DV,QGXVWU\-XQH -=KDQJ/+X /LDQJ:HWDO,QWHOOLJHQWULVNDVVHVVPHQWIRU SLSHOLQH 2012, Seoul, Korea WKLUGSDUW\LQWHUIHUHQFH-RXUQDORI3UHVVXUHHVVHO9HFKQRORJ\7 Ven katasubramanian V. Prognostic and diagnostic monitoring of $60(E UDQVDFWLRQVRIWKH7 complex systems for product life cycle management: challenges and -=KDQJHW/DO/LDQJ:2SSRUWXQLVWLFSUHGLFWLYHPDLQWHQDQFHIRU+X RSSRUWXQLWLHV&RPSXWHUV &KHPLFDO(QJLQHHULQJ complex multi-component systems based on DBN-HAZOP model. 3URFHVV6DIHW\DQG(QYLURQPHQWDO3URWHFWLRQF Ven katasubramanian V. Systemic failures: challenges and opportunities -=KDQJ/DQG/LDQJ'\QDPLF:+XGHJUDGDWLRQREVHUYHUIRUEHDULQJ LQULVNPDQDJHPHQWLQFRPSOH[V\VWHPV$,&K(-RXUQDO fault by MTS-SOM system. Mechanical Systems and Signal 3URFHVVLQJ Wan g Y, West H H and Mannan M S. The impact of data uncertainty in -/HH'HQJ&/V7DQ&+HWDO 3URGXFWOLIHF\FOHNQRZOHGJH determining safety integrity level. Process Safety and Environmental management using embedded Infotronics: methodology, tools and Protection. 2004. 82(B6): 393-397 FDVHVWXGLHV,QWHUQDWLRQDO-RXUQDORI.QRZOHGJH(QJLQHHULQJDQ G 7:) X: Data Mining. 2010. 1(1): 20-36 method for mechanical systems. Mechanical Systems and Signal -/HH1L- 'MXUGMDQRYLF'HWDO,QWHOOLJHQWSURJQRVWLFVWRROVDQG 3URFHVVLQJ HPDLQWHQDQFH&RPSXWHUVLQ,QGXVWU\ ;L//HH-7;LD HWDO2SWLPDO&%30SROLF\FRQVLGHULQJPDL QWHQDQFH ';X/LX+=KDQJ-4XDQWLWDWLYHULVNHYDOXDWLRQPHWKRGVIRUPXOWL HIIHFWVDQGHQYLURQPHQWDOFRQGLWLRQ,QWHUQDWLRQDO-RXUQDORI factor coupling complex flight situations. Acta Aeronautica et HFKQRORJ\7$GYDQFHG0DQXIDFWXULQJ $VWURQDXWLFD6LQLFD Yan g X and Mannan M S. An uncertainty and sensitivity analysis of Mar kowski A S, Mannan M S and Bigoszewska A. Fuzzy logic for G\QDPLFRSHUDWLRQDOULVNDVVHVVPHQWPRGHODFDVH-RXUQDOVWXG\RI SURFHVVVDIHW\DQDO\VLV-RXUQDORI/RVV3UHYHQWLRQLQWKH3URFHVV /RVV3UHYHQWLRQLQWKH3URFHVV,QGXVWULHVD ,QGXVWULHV Yan g X and Mannan M S. The development and application of dynamic Mar kowski A S, Mannan M S, Bigoszewska A, et al. Uncertainty aspects operational risk assessment in oil/gas and chemical process industry. LQSURFHVVVDIHW\DQDO\VLV-RXUQDORI/RVV3UHYHQWLRQLQWKH3URFHVV E 5HOLDELOLW\(QJLQHHULQJDQG6\VWHP6DIHW\ ,QGXVWULHV *:5RJHUVXQ:<-DQG0DQQDQ065LVNDVVHVVPHQWRIDQ/1* Mit chell S M and Mannan M S. Designing resilient engineered systems. -RXUQDOLPSRUWDWLRQPHWKRGRORJ\IR$WHUPLQDOXVLQJ%D\HVLDQ/23 &KHPLFDO(QJLQHHULQJ3URJUHVV /RVV3UHYHQWLRQLQWKH3URFHVV,QGXVWULHV 5DM DUDPDQ6+DKQ-DQG0DQQDQ06$PHWKRGRORJ\IRUIDXOW (Edited by Sun Yanhua) GHWHFWLRQLVRODWLRQDQGLGHQWL¿FDWLRQIRUQRQOLQHDUSURFHVVHV ZLWK DQJDQG/HH-$QRQOLQHDGDSWLYHFRQGLWLRQEDVHGPDLQWHQDQFH
Petroleum Science – Springer Journals
Published: Oct 13, 2013
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.