Reliability evaluation method based on dynamic fault diagnosis results: A case study of a seabed mud lifting system
Introduction
With the increasing depth of offshore oil and gas development and the complexity of deep-sea environment, conventional single-gradient drilling is faced with the problem of difficulty in operation [1]. Dual gradient drilling has a smaller pressure window than conventional drilling, which reduces the risk of kick and blowout, making it more suitable for deepwater drilling [2]. Multi-gradient drilling technology is proposed on the basis of double-gradient drilling technology. Compared to dual-gradient drilling, multi-gradient drilling can better match the pressure relationship at the bottom of the seabed and reduce the number of layers needed to run casing. However, the more mature technology is still the dual-gradient drilling technology. However, the dual-gradient drilling technology still faces some problems. Because dual-gradient drilling is used in narrow-density pressure window drilling operations, when the lifting system fails, it can lead to bottomhole pressure regulation failures, resulting in loss of circulation and bottomhole collapse. In 2003, well 718-5X and well 718-7X in China suffered loss of well. Sealing materials alone cost 990,000 yuan, resulting in 15 days of lost time. Seabed mud lifting system is the key technology of dual gradient drilling. Therefore, it is necessary to timely detect the faults of the mud lifting system and predict the reliability of the system.
Fault diagnosis methods can be divided into three categories, namely model-based [3], signal-based [4,5] and data-driven [6]. The model-based approach is mainly to establish mathematical models of complex industrial systems. The signal-based method is to compare the detected signals with the prior information of the normal industrial system [7]. Such signals are mainly vibration signals, and the possible anomalies and faults can be diagnosed through time-spectrum analysis. But for the complex system and process it is difficult to collect the accurate signal pattern. The data-driven method is also called knowledge-based fault diagnosis method. Information can be obtained through statistical and non-statistical methods, such as probabilistic multivariate methods [8] and Bayesian networks (BNs). BNs are one of the most useful models in the field of probability knowledge representation and reasoning [9]. In recent years, BN has been increasingly used in the field of fault detection and diagnosis. Namaki-Shoushtari and Huang [10] proposed a Bayesian control loop diagnosis method combining historical data and fault feature knowledge process. Bennacer et al. [11] proposed a hybrid method that combines BNs and case-based reasoning to overcome the usual limitations of fault diagnosis techniques and reduce manual intervention. Liu et al. [12] proposed a framework for fault diagnosis of solar-assisted heat pump system based on incomplete data and expert knowledge of BNs. Wang et al. [13] developed a fault diagnosis method for chiller based on BNs, and evaluated the result of feature selection with this method. Cai et al. [14] proposed a new fault diagnosis method of complex electronic system based on dynamic Bayesian networks (DBNs). Zhang et al. [15] proposed a threshold-based fault diagnosis and repair scheme using DBNs model, which can obtain the spatiotemporal correlation of vehicle data and realize accurate real-time or historical fault diagnosis and repair. Amin et al. [16] proposed a DBNs based fault detection, root cause diagnosis and fault propagation path identification scheme. In the above study, the fault diagnosis method based on BNs did not consider the impact of equipment performance degradation on the diagnosis results. Since the existing research assumes that the diagnosis system is fault-free, it will cause the problem of overdiagnosis, that is, the diagnosis system mistakenly thinks that there is a fault when the system equipment is fault-free. And the diagnosis method based on DBN did not consider the influence of sensor performance degradation on the diagnosis results.
At present, the main techniques for reliability evaluation of offshore oil and gas equipment include fault tree analysis [17,18], Petri net [19] and Markov chain [20,21], etc. These methods have certain limitations in analyzing complex electromechanical control, such as state-space explosion [22] and quadratic variable [23]. With the development of DBNs, reliability evaluation technology has made great progress. Liu et al. [24] developed DBNs for a multi-component parallel system, and analyzed the influence of common cause failures on the reliability of the subsea BOP stack based on the multi-fault impact model. Chang et al. [25] proposed a method of wellhead fatigue failure risk analysis based on DBNs, aiming at predicting the probability of fatigue failure of the wellhead during its service life. Cai et al. [26] proposed a method to transform the fault tree into DBN considering repair, and analyzed the influence of human factors on offshore blowout accidents. Cai et al. [27,28] also established DBNS models of SBOP system and SBOP control system respectively based on multi-state degradation model. Obviously, the traditional reliability analysis has been unable to meet the needs of actual production. Due to the dual-gradient drilling technology used in deepwater narrow density pressure window drilling operations, the reliability of the system is required to be able to timely diagnose system faults and predict system reliability changes.
Xu et al. [8] proposed a real-time reliability prediction method for dynamic systems with implicit degradation process. The method collects measurable output data from the system and uses particle filtering to identify the hidden degradation process. Finally, the reliability of the system is predicted according to the degradation process identified. Xu et al., [9] proposed a new real-time reliability prediction method for dynamic systems, which combined online fault prediction algorithm. The fault is assumed to vary linearly. After rough estimation of time-varying faults by improved particle filter algorithm, the fault estimation sequence is regarded as a time series, and the exponential smoothing method is used to predict the faults. Wang et al. [10] proposed a general real-time reliability analysis method for Wiener process-based degradation models to address the limitations of various Wiener process-based models. For example, Zhao [] studied the change of capacitor loss with time based on Wiener linear degradation process. There are several problems with the above studies. The main limitation of stochastic simulation is that the monitored values must be scalar. Implicit degradation process recognition does not apply to the simultaneous existence of multiple degradation processes. Its signal acquisition method is not suitable for large systems, as mentioned in paragraph 2, it is difficult to acquire accurate signal patterns for complex systems and processes.
To address the issues mentioned above. In this paper, a method of fault diagnosis and reliability evaluation for large complex systems and industrial processes is proposed. In order to combine reliability evaluation with fault diagnosis better, the dynamic Bayesian network is used to build fault diagnosis network and reliability evaluation network respectively. The fault parts are diagnosed by the dynamic diagnosis network and the reliability changes are inferred from the reliability evaluation network. The performance degradation of sensors and other components is considered in our fault diagnosis network. Because the performance of components and sensors in the system degrades over time, troubleshooting results may vary from time to time. In addition, the influence of common cause failure and multi-state degradation factors on reliability is also analyzed.
Section snippets
Diagnostic and reliability evaluation methods
The traditional fault diagnosis does not consider the influence of sensor and system performance degradation on the diagnosis results. In this paper, a dynamic fault diagnosis method and reliability evaluation criteria considering the performance degradation of sensors and other equipment in the system are proposed, as shown in Fig. 1. In this model, the fault diagnosis network is dynamic. The fault diagnosis network includes fault layer and evidence layer. The nodes in the fault layer
Structural analysis of the mud lifting system
With the rise of dual-gradient drilling technology in recent years, mud lifting drilling technology has been gradually recognized by people. At present, there are five systems for the research of dual-gradient drilling technology and equipment, including the Deep Vision subsea pump system, the Shell's subsea pump system (SSPS), the Riserless mud recovery (RMR) drilling technology, the Maurer hollow ball dual-density drilling system, and the Louisiana University (LSU) gas-lift dual-density
Model analysis and verification
In this paper, the proposed method is used to study four reliability evaluation cases, as shown in Table 11. Taking Case 1 as an example, sensor PS6, SP7 and SP8 are abnormal, while the values of other sensors are normal. Fig. 10 shows the results of Case 1′s diagnosis and its reliability. The results show that X4 has failed, which verifies the accuracy of the model.
Considering the reliability degradation of sensors, a dynamic diagnostic network model is established. Fig. 11 shows the
Conclusion
In this paper, a reliability evaluation method based on fault diagnosis results is proposed. An example of mud lifting system is given to verify the correctness of the method. In this method, the influence of the degradation of sensor and system equipment on the diagnostic results is considered. The problem of over-diagnosis in static diagnosis network is avoided, that is, the sensor fault may occur at a certain moment instead of the system equipment fault.
The reliability variation of mud
CRediT authorship contribution statement
Chuan Wang: Conceptualization, Methodology, Software, Writing – review & editing. Yupeng Liu: Investigation, Writing – original draft, Visualization. Dongbo Wang: Data curtion, Investigation. Guorong Wang: Supervision. Dingya Wang: Resources. Chao Yu: Resources.
Declaration of Competing Interest
We declare that we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled.
Acknowledgements
Funding: This work was supported by National Natural Science Foundation of China (51704254), Key Research and Development Project in Key Technical Field of Sichuan Province (2019ZDZX0030), International Science and Technology Innovation Cooperation Program of Sichuan Province(2021YFH0115), International Science and Technology Cooperation Project of Chengdu(2019-GH02-00039-HZ), and Technology Innovation research and development Project of Chengdu(2019-YF05-01872-SN).
References (41)
- et al.
Model-based hazard identification in multiphase chemical reactors
J Loss Prevent Proc
(2014) - et al.
Multi-scale deep intra-class transfer learning for bearing fault diagnosis
Reliab Eng Syst Safe
(2020) - et al.
A probabilistic multivariate method for fault diagnosis of industrial processes
Chem Eng Res Des
(2015) - et al.
Reliability assessment of marine floating structures using Bayesian network
Appl Ocean Res
(2018) - et al.
Fault diagnosis for a solar assisted heat pump system under incomplete data and expert knowledge
Energy
(2015) - et al.
Feature selection based on Bayesian network for chiller fault diagnosis from the perspective of field applications
Appl Therm Eng
(2018) - et al.
Fault detection and pathway analysis using a dynamic Bayesian network
Chem Eng Sci
(2019) - et al.
Assessment of the expected number and frequency of failures of periodically tested systems
Reliab Eng Syst Safe
(2013) - et al.
Probabilistic assessments in relationship with safety integrity levels by using Fault Trees
RELIAB ENG SYST SAFE
(2008) - et al.
Application of Petri nets to reliability prediction of occupant safety systems with partial detection and repair
Reliab Eng Syst Safe
(2010)
Reliability assessment of safety instrumented systems subject to different demand modes
J Loss Prevent Proc
Reliability analysis of subsea blowout preventer control systems subjected to multiple error shocks
J Loss Prevent Proc
Approximate solution for two stage open networks with Markov-modulated queues minimizing the state space explosion problem
J Comput Appl Math
Performance evaluation of multi-state degraded systems with minimal repairs and imperfect preventive maintenance
Reliab Eng Syst Safe
Dynamic Bayesian network modeling of reliability of subsea blowout preventer stack in presence of common cause failures
J Loss Prevent Proc
Dynamic Bayesian networks based approach for risk analysis of subsea wellhead fatigue failure during service life
Reliab Eng Syst Safe
A dynamic Bayesian networks modeling of human factors on offshore blowouts
J Loss Prevent Proc
Dynamic Bayesian networks based performance evaluation of subsea blowout preventers in presence of imperfect repair
Expert Syst Appl
An intelligent chiller fault detection and diagnosis methodology using Bayesian belief network
Energ Buildings
Real-time reliability evaluation methodology based on dynamic Bayesian networks: A case study of a subsea pipe ram BOP system
ISA T
Cited by (11)
Physics-guided, data-refined fault root cause tracing framework for complex electromechanical system
2023, Reliability Engineering and System SafetyA general enhancement method for test strategy generation for the sequential fault diagnosis of complex systems
2022, Reliability Engineering and System SafetyCitation Excerpt :Good testability design enables the equipment to monitor its own status in real time, facilitates the maintenance support work, and ensures the reliable operation of the system by planning an efficient fault diagnosis scheme [1–3].
Fault analysis of dragline subsystem using Bayesian network model
2022, Reliability Engineering and System SafetyCitation Excerpt :Data-driven BNs are risk-based designs used in chemical industries to reduce on-site and off-site risks of major accidents [29], or used in international shipping to detect fraud [30]. Moreover, BNs have received considerable attention from scientists and engineers across various fields, including reliability analysis, risk management and safety analysis [1,31,32], and system fault diagnosis [33,34]. It has also gained popularity among medicine practitioners for detection of false alarm during medical diagnosis [35].
Reliability analysis of dynamic fault trees with Priority-AND gates based on irrelevance coverage model
2022, Reliability Engineering and System SafetyBayesian framework for reliability prediction of subsea processing systems accounting for influencing factors uncertainty
2022, Reliability Engineering and System SafetyCitation Excerpt :Contextually to the present study objective, Wang et al. [36] have developed a dynamic BN model for subsea Xmas tree systems with emphasis on maintenance planning. In their other study, Wang et al. [35] have proposed a fault diagnosis method using dynamic BN capable of time-varying reliability evaluation of subsea mud lifting system. In general, the focus of these techniques has been on the model itself rather than on the effect of conditions developed due to the subsea environment.