System Failures of Offshore Gas Turbine Engines in Maintenance Perspective

System Failures of Offshore Gas Turbine Engines in Maintenance Perspective

October 2016.on Biarritz, France 3rd IFAC19-21, Workshop Advanced Maintenance Engineering, Service and Technology October 19-21, 2016. Biarritz, Franc...

485KB Sizes 0 Downloads 74 Views

October 2016.on Biarritz, France 3rd IFAC19-21, Workshop Advanced Maintenance Engineering, Service and Technology October 19-21, 2016. Biarritz, France 3rd IFAC Workshop on Advanced Maintenance Engineering, Service and Technology October 19-21, 2016. Biarritz, France

Available online at www.sciencedirect.com

ScienceDirect System Failures of Offshore Gas Turbine in Maintenance Perspective IFAC-PapersOnLine 49-28 (2016)Engines 280–285 System Failures of Offshore Gas Turbine Engines in Maintenance Perspective 1 2 2 , Mario M. Machado , Diego A. P. Manguinho and Anders Valland1 Lokukaluge P. Perera System Failures of Offshore Gas Turbine Engines in Maintenance Perspective 1 2 2 1

Lokukaluge P. Perera1, Mario M. Machado2, Diego A. P. Manguinho2 and Anders Valland1 Lokukaluge P. Perera1, Mario M. Machado2, Diego A. P. Manguinho2 and Anders Valland1 1 Norwegian Marine Technology Research Institute (MARINTEK), Energy Systems and Technical Operations, Trondheim, Norway. (e-mail:  [email protected]). [email protected], 11 Norwegian Marine Technology Research Institute (MARINTEK), Energy Systems and Technical Operations, Trondheim, Norway. (e-mail: [email protected], [email protected]). Norwegian Marine Technology Research Institute (MARINTEK), Energy Systems and Technical Operations, Trondheim, Norway. (e-mail:

1

[email protected], Petróleo Brasileiro SA, (Petrobras), Rio de Janeiro, Brazil. (e-mail: [email protected]). [email protected], [email protected])

2

22 Petróleo

Petróleo Brasileiro SA, (Petrobras), Rio de Janeiro, Brazil. (e-mail: [email protected], [email protected])

2

Petróleo Brasileiro SA, (Petrobras), de Janeiro, Brazil. of (e-mail: [email protected], [email protected]) Abstract: Several systemRiofailure events a selected gas turbine engine with respective to its

maintenance actions system are considered this study. system are with derived from condition Abstract: Several failure in events of a These selected gas failure turbineevents engine respective to its monitoring (CM) data of a selected gas turbine engine and modeled into a nonhomogeneous Poisson actions are considered in this study. These system failure events are derived from condition maintenance Abstract: Several system failure events of a selected gas turbine engine with respective to its process (NHPP) under likelihood estimation. erroneous intervals noted in the monitoring (CM) data ofconsidered a selected engineVarious and modeled intodata a are nonhomogeneous Poisson maintenance actions aremaximum ingas thisturbine study. These system failure events derived are from condition CM data of the gas turbine engine and removed from the respective analysis. The CM data set is divided process (NHPP) likelihood estimation. erroneous intervals are noted in the monitoring (CM)under data maximum of a selected gas turbine engineVarious and modeled intodata a nonhomogeneous Poisson into several intervals due toengine the erroneous data intervals and the modified data set is used toset estimate CM data(NHPP) of the gas turbine and removed from theVarious respective analysis. The CM data is divided process under maximum likelihood estimation. erroneous data intervals are noted in the the parameters of the respective models of system reliability. These models represent the failure intensity into intervals due toengine the erroneous data intervals the modified dataThe set CM is used estimate the CM several data of the gas turbine and removed from theand respective analysis. datatoset is divided levels of theof gas engine during various sectors and of its cycle. data These intensity levels parameters the turbine respective models of system reliability. These models represent the failure intensity into several intervals due to the erroneous data intervals thelife modified set failure is used to estimate the consist of increasing and decreasing reliability trends and those variations are compared with system levels of the gas turbine engine during various sectors of its life cycle. These failure intensity levels parameters of the respective models of system reliability. These models represent the failure intensity faults and maintenance to observe the trends respective reasons. Finally, thecompared reasons among these consist increasing andperiods decreasing reliability and those are with system levels ofofthe gas turbine engine during various sectors of its life variations cycle. These failure intensity levels system faults and maintenance periods by considering the inputs from the maintenance crew are also faults and maintenance periods to observe the respective reasons. Finally, the reasons among these consist of increasing and decreasing reliability trends and those variations are compared with system summarized in the conclusion of this study. system faults and maintenance periods by considering the inputs from the maintenance crew are also faults and maintenance periods to observe the respective reasons. Finally, the reasons among these summarized in(International the conclusion of periods this study. system and maintenance considering the Hosting inputs from the maintenance crew are also © 2016, faults IFAC Federation of by Automatic Control) by Condition Elsevier Ltd.Monitoring, All rights reserved. Keywords: Gas Turbine Engines, System Failures, Failure Intensity, Condition summarized in the conclusion of this study. based Maintenance, Offshore PowerSystem Plants.Failures, Failure Intensity, Condition Monitoring, Condition Keywords: Gas Turbine Engines, based Maintenance, Offshore PowerSystem Plants.Failures, Failure Intensity, Condition Monitoring, Condition Keywords: Gas Turbine Engines,  based Maintenance, Offshore Power Plants. component upgrades should be initiated to improve the  1. INTRODUCTION availability the power plant.beHowever, effects component of upgrades should initiated the to ageing improve the  1. INTRODUCTION of offshore gas turbine engines may require additional availability of the power plant. However, the ageing effects System failures and maintenance actions of an offshore component upgrades should be initiated to improve the 1. INTRODUCTION to cope withHowever, fatigue issues of offshoreofactions gas turbine engines may and require additional power plant with and several gas turbineactions enginesof toansatisfy the maintenance availability the power plant. thecorrosion ageing effects System failures maintenance offshore of the components. old components maintenance actions toFurthermore, copeengines with fatigue and corrosion issues power of maintenance an oil gasactions field are considered in of offshore gas turbine maysystem require additional power plant with and several gas and turbine engines the Systemrequirements failures of to ansatisfy offshore should be replaced with new ones to improve operational this study. The industrial power plant is facilitated with four of the components. Furthermore, old system components power requirements of an oil gas field are to considered in maintenance actions to cope with fatigue and corrosion issues plant with several gas and turbine engines satisfy the availability of offshore gas ones turbine enginescomponents in some gas turbine engines/generators in production, should replaced with new to improve operational of the be components. Furthermore, old system this study. The industrial is floating facilitated with four power requirements of an power oil andplant gasafield are considered in situations. That process is associated with not only respective storage and offloading (FPSO) unit located in Campos Basin, availability of offshore gas turbine engines in some gas turbine engines/generators in a floating production, should be replaced with new ones to improve operational this study. The industrial power plant is facilitated with four maintenance costs but also health, safety and environment Rio de Jeneiro (Machado et al., 2014). These gas turbine situations. That process is associated with not only respective storage and offloading (FPSO) unit located in Campos Basin, availability of offshore gas turbine engines in some gas turbine engines/generators in a floating production, (HSE) and service quality (SQ) considerations. The engines are equipped with condition monitoring (CM) maintenance costs but also health, safety and environment Rio de Jeneiro (Machado et al., 2014). These gas turbine situations. That process is associated with not only respective storage and offloading (FPSO) unit located in Campos Basin, respective maintenance costs can have a direct relationship to (HSE) and service quality (SQ) considerations. The facilities to monitor the with system health harsh ocean engines are equipped condition monitoring (CM) maintenance costs but also health, safety and environment Rio de Jeneiro (Machado et al., 2014). under These gas turbine the respective system reliability in such situations. These respective maintenance costs can have a direct relationship to environmental conditions with appropriate maintenance (HSE) and service quality (SQ) considerations. The facilities to monitor the with system health under harsh ocean engines are equipped condition monitoring (CM) maintenance actions are a part of the overall maintenance the respective system reliability in such situations. These actions and that process is categorized as condition based respective maintenance costs can have a direct relationship to environmental conditions with health appropriate facilities to monitor the system under maintenance harsh ocean strategy of the respective andofin gas operator. In general, maintenance (CBM). Therefore, theappropriate system (i.e. maintenance actions arereliability aoilpart the overall maintenance the respective system such situations. These actions and that process is with categorized as degradation condition based environmental conditions maintenance system reliability in the oil and gas industry is categorized health condition) of each gas turbine engine is monitored strategy of the respective oil and gas operator. In general, maintenance (CBM). Therefore, the system (i.e. maintenance actions are a part of the overall maintenance actions and that process is categorized as degradation condition based under three divisions: availability, and with various sensors under CM.system Catastrophic failure in the oiloiland categorized health condition) of each gas turbine engine is monitored strategyreliability of the main respective andgas gasindustry operator.is safety In general, maintenance (CBM). Therefore, the degradation (i.e. system maintainability. Since the present economic downturns, the situations in the entire power plant Catastrophic can be isavoided by under three main divisions: availability, and with various sensors CM. failure system reliability in the oil and gas industry is safety categorized health condition) of eachunder gas turbine engine monitored oil and gas industry focuses to identify the most critical monitoring engine condition executing Since divisions: the present availability, economic downturns, the situations ineach the entiredegradation power CM. plant can beand avoided by maintainability. under three main safety and with various sensors under Catastrophic failure requirements forSince these focuses oil and gasidentify platforms, where oil and gas industry to thedownturns, most various critical appropriate maintenance actions. One should note that CBM maintainability. the present economic the monitoring each engine degradation condition and executing situations in the entire power plant can be avoided by cost-effective maintenance actions are introduced under the requirements for these focuses oil and to gasidentify platforms, where various is enabled by CM activities, where appropriate maintenance oil and gas industry the most critical appropriate maintenance actions. One should note that CBM monitoring each engine degradation condition and executing required safety and maintainability cost-effective maintenance actions introduced under the decisions/actions taken bywhere the One crew to improve the power requirementsavailability, for these oil and gas are platforms, where various is enabled by CMare activities, appropriate maintenance appropriate maintenance actions. should note that CBM considerations. required availability, safety and maintainability plant availability. It is also believed that appropriate decisions/actions taken bywhere the crew to improve the power cost-effective maintenance actions are introduced under the is enabled by CMare activities, appropriate maintenance diagnostic and prognostic tools should be to considerations. required System availability, safety plant availability. is by also thatdeveloped appropriate decisions/actions are Ittaken the believed crew to improve the power maintenance is oftenand done maintainability after complete identify the present and future health conditions of gas diagnostic and prognostic tools should be developed to system considerations. plant availability. It is also believed that appropriate failures (i.e. run-to-failure maintenance) various System maintenance is often done afterincomplete turbine engines under CMB approaches. identify theand present and future of gas diagnostic prognostic tools health should conditions be developed to industries, where the respective HSE and SQ considerations system failures run-to-failure maintenance) various System (i.e. maintenance is often done afterincomplete turbine engines under CMB approaches. identify the present and future conditions gas are neglected in some situations. HSE However, thisconsiderations approach can industries, where the run-to-failure respective and SQ The environmental effectshealth can degrade the of system system failures (i.e. maintenance) in various turbine engines under CMB approaches. be improved by considering planned system maintenance (i.e. are neglected in some situations. However, this approach can performance offshore gaseffects turbine discussed The of environmental canengines degradeasthe system industries, where the respective HSE and SQ considerations preventive maintenance) in some situations. That consists of be improved by considering planned system maintenance (i.e. previously. Therefore, appropriate maintenance actions under are neglected in some situations. However, this approach can performance offshore gaseffects turbine discussed The of environmental canengines degradeasthe system implementing periodic time maintenance intervals regardless preventive maintenance) in some situations. That consists of the required system integrity and safety levels with essential previously. appropriate maintenance under be improved by considering planned system maintenance (i.e. performanceTherefore, of offshore gas turbine engines actions as discussed implementing periodic time maintenance intervals regardless the requiredTherefore, system integrity and safety levels with essential previously. appropriate maintenance actions under preventive maintenance) in some situations. That consists of implementing periodic time maintenance intervals regardless the required system integrity and safety levels with essential Copyright © 2016 IFAC 280 2405-8963 © IFAC (International Federation of Automatic Control) © 2016, 2016 IFAC 280Hosting by Elsevier Ltd. All rights reserved. Copyright Peer review under responsibility of International Federation of Automatic Control. Copyright © 2016 IFAC 280 10.1016/j.ifacol.2016.11.048

2016 IFAC AMEST October 19-21, 2016. Biarritz, France

Lokukaluge P. Perera et al. / IFAC-PapersOnLine 49-28 (2016) 280–285

of the system health condition and improves the system availability in a majority of industrial systems. However, such preventive maintenance actions can be expensive for some industries with complex machineries (i.e. systems with a large number of subsystems and components) such as gas turbine engines. Therefore, CBM as a cost effective solution is adopted by such industries, where actual health conditions of respective systems are monitored, continuously and appropriate maintenance actions can be chosen and executed appropriately.

281

seeking control (PSC). These approaches develop mathematical models for gas turbine engines consisting various parameters that relate to the health conditions of the respective components. The parameters of such mathematical models are estimated by various algorithms with sensor measurements (i.e. pressure, temperature and rotational speed values of the respective components). Several engine health management applications that relate to GPA are presented in the respective studies of (Simon and Simon (2005), Pu et al., (2013)). Similarly, additional engine health management applications that relate to PSC are presented in the respective studies of (Espana (1994), Gilyard and Orme (1993), Orme and Schkolnim (1995), and Simon and Garg, (2009)). However, GPA and PSC based engine health management approaches in gas turbine engines encounter various industrial challenges and that are summarized as: 1) sensor noise (i.e. bias and variance values in the measurements) can degrade the parameter estimation process, 2) system parameters can have various nonlinear relationships and the respective models and sensor measurements are inadequate to identify those relationships and 3) a large number of sensors are required to estimate the total number of health parameters in gas turbine engines. Even though some solutions to such challenges are proposed in the recent literature (Simon and Garg (2009) and Xuewu et al., (2009)), the complex nonlinearities among the system parameters can still degrade GPA and PSC based health management approaches in gas turbine engines. Therefore, a system failure events based health management approach for gas turbine engines is considered in this study, where the respective failures are categorized as stochastic events. The failure intensity of a selected gas turbine engine is modeled as a nonhomogeneous Poisson process (NHPP) (Perera et al. 2015a, b) under such situations. These types of models (i.e. stochastic process) are used in many reliability studies for predicting the failures of various systems and components (Rausand and Hoyland, 2004).

In general, industrial maintenance actions can be divided into three categories of corrective, preventive, and predictive. Those actions are also executed under various maintenance strategies of run to failure maintenance (RTFM), on condition maintenance (OCM) and condition based maintenance (CBM) as discussed previously. As a summary, RTFM approaches focus on corrective measures, OCM approaches focus on corrective and preventive measures, and CBM approach focuses on all corrective, preventive and predictive measures. Therefore, CBM is considered as the most suitable approach to overcome respective diagnostic and prognostic challenges in the oil and gas industry. This study focuses on understanding system reliability of a selected gas turbine engine with respect to its maintenance actions. System reliability is quantified with respect to the failure events of the selected gas turbine engine. It is also expected that these failure events also relate to age related system degradations and that should be reflected in the respective failure intensity levels of the gas turbine engine. Hence, the respective failure intensity levels of the gas turbine engine are calculated from the CM data. The health conditions of gas turbine engines are observed under two different industrial levels. The first level consists of the top-down concept: the engine degradation with respect to its current usage level is identified and compared to the average engine performance throughout its life cycle. The second level consists of the bottom-up concept: the component health conditions and maintenance information are used to determine probable effects of the engine degradation. This study overlaps both concepts, where engine health conditions with respect to the failure intensity levels are estimated for various sectors of the system life and that information is compared with its maintenance actions. Furthermore, the respective maintenance actions and recorded information are discussed with the crew to derive the conclusions at the end of this study.

A similar concept is adopted in this study to calculate the system failure intensity levels of a selected gas turbine engine. Hence, this is a simplified approach with compared to GPA and PSC based health management approaches of gas turbine engines and that can also be used to evaluate the respective maintenance actions. The failure intensity levels of a selected gas turbine engine are captured with a nonhomogeneous Poisson process (NHPP). One should note that the parameters of the NHPP model represent the respective component health conditions of the gas turbine engine. These component health conditions also relate to various failure intensity levels of the gas turbine engine in different system age intervals. Therefore, the respective future system failures and failure transitions can also be predicted by using these models and such information can also be used to overcome diagnostic and prognostic challenges in gas turbine engines.

CM and CBM have often been a part of engine health management approaches. A summary of such engine health management approaches of gas turbines is presented Perera et al. (2015a). These approaches are often based on real-time measurements (i.e. physical parameters), event data (i.e. system failures and shutdowns) and maintenance records (i.e. overhauls and repairs). Hence, engine health management approaches can predict various system failures of gas turbine engines and prevent overall offshore power plant failures. Various mathematical models are developed under these health management approaches and divided into two sections: gas path analysis (GPA) and performance

2. SYSTEM FAILURES The respective system failure events of the selected gas turbine engine are presented in Figure 1. One should note that the failure events are presented with respect to the system 281

2016 IFAC AMEST October 19-21, 2016. Biarritz, France 282

Lokukaluge P. Perera et al. / IFAC-PapersOnLine 49-28 (2016) 280–285

Fig. 1. Actual system failure events of the gas turbine engine.

relative age (i.e. without maintenance intervals) in the same figure. These events are extracted from the respective operational and maintenance data (i.e. CM data) of the gas turbine engines. However, two erroneous data regions are identified during this analysis with repeated system failure events. Those regions are noted as data interval (int.) A and B in the same figure. It is concluded that these repeated failure situations can degenerate the system failure intensity calculations, therefore such intervals are removed from the data analysis. Further details on such erroneous data intervals, detection methodologies and removal steps are presented in Perera et al. (2015a, b). The erroneous data intervals divide the system relative age into several segments (i.e. data interval (int.) 1, 2 and 3) and the results are presented in Figure 2. These erroneous data intervals are bounded by cyan lines in the same figure. The system failure events of data interval 1 and 2 are used to calculate the respective failure intensity levels of the gas turbine engine during its relative age period (i.e. without maintenance intervals). One should note that these failure intensity levels are derived from a NHPP model and that is denoted as:  t    t  1

Fig. 2. Actual and predicted system failure events.

life. That sector of the system life is categorized as its relative age (i.e. system operation period without maintenance periods), where the system relative age is less than its absolute age. The system approximate absolute age is estimated in an internal recording system and that values is used to correct the system relative age (Perera et al., 2015a, b). However, the failure events and respective maintenance periods are recorded under the system relative age, therefore the final results are also visualized under the system relative age to simplify the respective presentation. Data interval 1 (see Figure 2) is assigned with the time interval 0 10000( Hrs) in the system relative age (i.e. without maintenance periods) and the estimated parameters of the NHPP model (i.e. the system failure intensity level) are calculated as:

(1)

ˆ  0.0040, ˆ  0.9824

where   0 and   0 are the system parameters and t is the system absolute age (i.e. the total operational life without maintenance intervals) of the gas turbine engine. One should note that offshore gas turbine engines are repaired upon various system failures, therefore those maintenance actions influence the system failure intensity levels under various sections of the system life. It is also considered that the respective system failure events are recovered by a similar number of maintained periods during this model development. These maintenance actions are considered as "minimal repair" (i.e. as bad as old) under the NHPP model. One should note that NHPP models can represent system failure intensity levels under both increasing and decreasing reliability trends. Therefore, that can facilitate to develop the most suitable and simple mathematical models for system reliability applications. A decreasing failure trend in such a model represents an improving system reliability situation possibly due to better maintenance actions. An increasing failure trend in such a model represents a decreasing system reliability situation possibly due to lack of proper maintenance actions and/or age related system degradation conditions. In general, it is expected that the respective gas turbine engines should have increasing failure trends due to harsh environmental and age related system degradation conditions.

(1)

where ˆ  1 for this data interval. Hence, some reliability improvements in the gas turbine engine are noted in this data interval (Perera et. al., 2015a,b). The actual and predicted failure events for the same data interval are presented in Figure 2 as Model 1. The actual failure events are projected into this model to compare with the estimated failure events during this period. One should note that this model also predicts the number of system failures for the pre and post sectors of the system relative age. The data interval, 10000(Hrs) 10307.5(Hrs) in the system relative age (i.e. without maintenance periods), is considered as an erroneous data region and that is removed from the data analysis. Data interval 2 is assigned with the time interval 10307.5(Hrs) 18950 (Hrs)  in the system relative age (i.e. without maintenance periods) and the estimated parameters of the NHPP model (i.e. the system failure intensity level) are calculated as: ˆ  1.0336.10-15 , ˆ  3.6771

(2)

where ˆ  1 for this data interval. Hence, some reliability degradations in the gas turbine engine are noted. The actual and predicted failure events for the respective data interval are presented in Figure 2 as Model 2. One should note that the actual failure events are also projected into this model to

It is also noted that the system event data of gas turbine engines are available only for a sector of the system 282

2016 IFAC AMEST October 19-21, 2016. Biarritz, France

Lokukaluge P. Perera et al. / IFAC-PapersOnLine 49-28 (2016) 280–285

283

Fig. 3. System maintenance periods for data interval 1 & 2.

Fig. 4. System maintenance periods and failure events for data interval 1.

intensity level) for these combined data intervals are calculated as:

compare with the estimated failure events during this period. The data interval, 18950 (Hrs) 22596(Hrs)  in the system relative age (i.e. without maintenance periods) is considered as an erroneous data region and that is removed from the data analysis. Data interval 3 is ignored from this model derivation due to its short time period.

ˆ  2.0736.10-7 , ˆ  1.9008

(3)

where ˆ  1 for this data interval and the respective calculations are presented in (Perera et. al., 2015a,b). Hence, some reliability degradation conditions of the gas turbine engine are noted. The actual and predicted failure events for the same data interval are presented in Figure 2 as Model 3. Similarly, the actual failure events are also projected into this model to compare with the estimated failure events during this period. One should note that the first part of the system absolute life (i.e. unknown CM period) of the gas turbine

The number of system failure events in the first, second and third data intervals are identified as 33, 39 and 8, respectively and the first and second erroneous data intervals are identified as 19 and 11, respectively. In the next step, data interval 1 and 2 are assigned with the same relative age time frame as two discrete data intervals. The estimated parameters of the NHPP model (i.e. the system failure 283

2016 IFAC AMEST 284 October 19-21, 2016. Biarritz, France

Lokukaluge P. Perera et al. / IFAC-PapersOnLine 49-28 (2016) 280–285

engine consists of an unknown number of system failures. However, these failure events are predicted by the respective models (i.e. model 1, 2, & 3) with respect to their failure intensity levels. Therefore, each model starts with a unique number of initial failure events in various positions of its system life (i.e. the number of initial system failures) (see Figure 2). Similarly, the number of possible failure events in each erroneous data interval are also predicted by the same models.

system failure events is relatively large with compared to other maintenance periods. This represents a situation, where the gas turbine engine was running for a relatively long period without any maintenance actions just before the 18th system failure. The maintenance periods (MP) with respect to each failure event are presented in the fifth plot (i.e. from the top) of the same figure. The results show that the 19th maintenance period has a relatively long period and that may be influenced by the 18th system failure of the gas turbine engine. However, this is categorized as an unusual situation because the largest operation period without any maintenance has occurred just before the 18th failure event but the largest maintenance period has occurred just after the 19th failure event. Therefore, the respective system faults are further investigated to understand this unusual failure and maintenance situation.

It is noted that the failure intensity level in data interval 1 (i.e. Model 1) shows some improvements (i.e. ˆ  1 ) in system reliability of the gas turbine engine. However, the failure intensity level in data interval 2 (i.e. Model 2) shows a considerable reduction in system reliability (i.e. ˆ  1 ). Furthermore, the failure intensity level in the combined model (i.e. model 3) shows less reduction (i.e. due to both data intervals) in system reliability (i.e. ˆ  1 ) with compared to the model 2. Hence, it is also concluded that the overall failure intensity level of the gas turbine engine dominates by the failure intensity (i.e. failure events) of the last part of the system life and that may have a reliability decreasing trend in generally (i.e. model 3). It is also believed that these failure intensity levels relate to system maintenance actions of the gas turbine engine. In addition, the availability of the CM data can also influence the failure intensity levels of the gas turbine engine, where a longer CM data period is required to derive an accurate system failure intensity level. As the next step, the respective maintenance actions in the respective data intervals are investigated to reason those failure intensity level variations.

Various failure events are identified with respect to subsystem and component faults of these gas turbine engines and that denote by a code system (i.e. 79.xxxx). One should note that 34 fault codes with respect to subsystem and component failures of gas turbine engines are identified under CM data. The fault codes with respect to each failure event are presented in the bottom plot of Figure 4. The results show that the both failure intervals (i.e. 18th and 19th) are having the same fault code (i.e. 79.1230) and that failure is categorized as "waste heat recovery unit (WHRU) inlet/bypass damper linkage failure." This system failure was repeated during this period and resulted in a long maintenance period of the gas turbine engine. Furthermore, the gas turbine engine was operated without any maintenance actions for a considerably long period with compared to other failure events as mentioned before.

3. SYSTEM MAINTENANCE The respective system operational and maintenance periods for data interval 1 and 2 of the gas turbine engine are presented in Figure 3. The system operational periods are divided into two data intervals (i.e. data interval 1 & 2) by considering the previous failure intensity levels. One should note that the first and second data intervals consist of 33 and 39 maintenance periods, respectively. The system relative age (i.e. with the maintenance periods) is divided into 6 time segments and that are presented in this figure to improve the visibility of maintenance periods. As the next step these maintenance periods are combined to observe maintenance frequency of data interval 1 and the results are presented in the top two plots of Figure 4. The results show that relatively small time intervals are distributed along these maintenance intervals with one large maintenance interval in the middle. The system operational periods are modeled as single instances (i.e. pulses) in the same figure. This representation is used to observe the respective maintenance actions of the gas turbine engine with respect to the failure intensity levels.

However, the same long period without any maintenance has influenced on the slight improved system reliability situation in data interval 1. A considerable large maintenance period and repeated system failures are also resulted due to such actions. Hence, it is concluded that these actions are facilitated for a considerable system degradation situation and that eventually leads to a long maintenance period. The fault code (i.e. 79.042) relates to the failure that is categorized as "liquid upstream pressure fault shutdown" is also noted during the 17th failure event. The same fault (i.e. 79.1230) is repeated in between the 26th and 27th failure situations of the gas turbine engine. However, the fault (i.e. 25th failure situation) is different from the above situation (i.e. 79.042) and relates to the fault code (i.e. 79.012) that is categorized as "start system speed: crash re-engagement shutdown." These failure situations are occurred under a relatively low maintenance period. However, it is inconclusive to say these previous faulty situations may influence on the respective system failures and maintenance periods in this gas turbine engine.

Several failure events (i.e. 33 events) are also reported during data interval 1 and those events in the system relative age are presented in the third plot (i.e. from the top) of the same figure. Then, the time difference between two consecutive failure events with respect to each failure event is presented in the fourth plot (i.e. from the top) of Figure 4. One should note that the time period between 17th and 18th

A detailed discussion that had with the maintenance crew is summarized in this section. The WHRU uses water to recover additional heat that is created by the gas turbine engine. A set of dampers are used in this process, where the heat from turbine exhaust gases is transferred to this water circulating system under gas turbine engine operational 284

2016 IFAC AMEST October 19-21, 2016. Biarritz, France

Lokukaluge P. Perera et al. / IFAC-PapersOnLine 49-28 (2016) 280–285

conditions as an heat recovery approach. In some situations, these dampers are aligned slightly differently directions and that drives gas turbine exhaust gases into the atmosphere. Such situations are categorized as the WHRU is in "bypass mode", where the exhaust gases are diverted from the stove. The operational conditions of these dampers are essential and it must be properly controlled and may not obstruct the outlet of the gas turbine (i.e. both simultaneously closed is an unwanted conditions). To avoid inadequate operational conditions, some manufacturers are adopted a mechanical link that also improves the safety of gas turbine engines. The position monitoring of this mechanical link is done by a programmable logic controller (PLC) with a status table. However, various inappropriate positions of the link are resulted in WHRU failure situations in gas turbine engine. Furthermore, the maintenance delays are often associated with this type of instrumentation and that resulted in long operational periods without proper maintenance actions. In addition, few qualified maintenance crew members are available for the required maintenance of the same equipment. One should note that a similar situation is noted in the maintenance data (see Figure 4) in this study and that resulted in a relatively longer maintenance period.

285

costs calculations should consist of the detailed information relate to engine performance deterioration, part replacementrepair rates, maintenance practices, and part conditions by considering collected, documented and analyzed service usage. These cost calculations can be used to identify the most critical and expensive components and their respective maintenance actions of gas turbine engines. Therefore, the most crucial system failures (i.e. the most critical and expensive failures) can be also identified and avoided. That information (i.e. the most crucial system failures) can be used to develop cost-effective maintenance polices for offshore gas turbine engines. ACKNOWLEDGEMENT This work is supported by the Center for Integrated Operations (IO) in the Petroleum Industry, Trondheim, Norway, under the project of Remaining Useful Life (RUL) Modeling for Gas Turbine Engines, System Integrity and Dynamic Risk Assessment (T3) and funded by the Research Council of Norway (NRC) and Petróleo Brasileiro S.A. (Petrobras). REFERENCES

It is also noted that the crew is also familiar the WHRU failures that take relatively large maintenance periods (i.e. due to breakage of the mechanical link of the damper WHRU). This link is a mechanical arm between the actuator and damper itself as discussed previously. Furthermore, this failure can also be associated with a positioning failure of the same damper which is controlled by the PLC as a broken link. However, the maintenance crew also noted that the frequency of these types of failure events are increasing with respect to the system life of gas turbine engines. The lack of knowledge on the configuration details and operational conditions of WHRU is also another reason of the frequent failure events. In a majority of such failure situations, the crew has to wait for the gas turbine engine to cool down until it reaches the safe temperature to perform maintenance work. That is also another reason for the long associated maintenance period in this system fault of the gas turbine engine. It can also be concluded that these reasons have influenced on the failure intensity levels of the gas turbine engines.

Machado, M.M. Manguinho, D.A.P.M., Valland, A., Ushakov, S., Roverso, D., and Beere, W.H.A., (2014) RUL modeling for turbo generators of a FPSO: Alternatives and challenges, in Proceedings of the Rio Oil & Gas Expo and Confer-ence, Rio de Janeiro, Brazil. (IBP 1802 14). Perera, L.P., Machado, M.M., Valland, A., and Manguinho, D.A.P. (2015a) "System Reliability of Offshore Gas Turbine Engines with Erroneous Data Conditions," In Proceedings of the 25th European Safety and Reliability Conference (ESREL 2015), Switzerland, September, 2015, pp. 1679-1688. Perera, L.P., Machado, M.M., Valland, A., Manguinho, D.A.P., (2015b), "Modelling of System Failures in Gas Turbine En-gines of Offshore Platforms," In Proceedings of the 2nd IFAC Workshop on Automatic Control in Offshore Oil and Gas Production, Florianpolis, Brazil. Simon, D., and Simon, D. L., (2005), "Aircraft turbofan engine health estimation using constrained Kalman filter". J. Eng. Gas Turbines Power, 127(2), pp. 323-328. Pu, X., Liu, S., Jiang, H., and Yu, D., (2013) "Sparse Bayesian Learning for Gas Path Diagnostics, " J. Eng. Gas Turbines Power 135(7), 071601. Espana, M. D., (1994), "Sensor biases effect on the estimation algorithm for performance-seeking controllers". Journal of Propulsion and Power, 10 (4), pp. 527-532. Gilyard, G. B. and Orme, J. S., (1993), "Performance seeking control program overview and future directions". In NASA Technical Memorandum 4531. Orme, J. S. and Schkolnim, S., (1995), "Flight assessment of the onboard propulsion system model for the performance seeking control algorithm on a f-15 aircraft". NASA Technical Memorandum 4705. Alag, G. S. and Gilyard, G. B., (1990), "A proposed Kalman filter algorithm for estimation of unmeasured output variables for an f100 turbofan engine". In NASA Technical Memorandum 4234. Simon, D. L. and Garg, S., (2009), "Optimal turner selection for Kalman filter-based aircraft engine performance estimation". In Turbo Expo, ASME. Xuewu, D., Zhiwei, G., Breikin, T., and Hong, W., (2009), "Disturbance Attenuation in Fault Detection of Gas Turbine Engines: A Discrete Robust Observer Design," Systems, Man, and Cybernetics, Part C: Applications and Reviews, IEEE Transactions on, vol. 39, no. 2, pp. 234-239. Rausand, M., and Hoyland, A., (2004), System Reliability Theory, Models Statistical Methods and Applications (2nd Edt). John Wiley & Sons, Inc., Hoboken, New Jersey, USA.

4. CONCLUSION The system failure intensity levels of a selected gas turbine engine are identified in the first part and the respective maintenance activities are compared with the same system failures in the second part of this study. The system failure intensity as the average engine performance level is categorized and calculated by the respective event data (i.e. CM data) from a selected gas turbine engine. One should note that the system failure intensity can be used to evaluate present and past maintenance activities (i.e. maintenance interventions with consistence intervals) as described in this study. Hence, these results can be used for both operational and maintenance requirements of offshore gas turbine engines. The maintenance costs can also play an important role in both operational and maintenance requirements. Such 285