- Email: [email protected]
Reliability improvement of compressors based on asset fleet reliability data
Proceedings of of the the 3rd 3rd IFAC IFAC Workshop Workshop on on Proceedings Automatic Offshore Oil Gas Proceedings of thein 3rd IFAC Workshop on Production Automatic Control Control in Offshore Oil and andAvailable Gas Production online at www.sciencedirect.com Proceedings of1,the 3rd IFAC Workshop on Production May Esbjerg, Denmark Automatic Control in Offshore and Gas May 30 30 -- June June 1, 2018. 2018. Esbjerg,Oil Denmark Automatic Control in Offshore Oil and Gas Production May 30 - June 1, 2018. Esbjerg, Denmark May 30 - June 1, 2018. Esbjerg, Denmark
ScienceDirect
IFAC PapersOnLine 51-8 (2018) 217–224
Reliability improvement of compressors Reliability improvement of based on asset fleet reliability data Reliability improvement of compressors compressors based on asset fleet reliability data based on asset fleet reliability data based on asset11 fleet reliability data 22 2,3 2,3 F. F. F. F.
Schulze Schulze Schulze Schulze
Sp¨ u n C. Sp¨ untrup* ntrup*11 J. J. G. G.3 Londo˜ Londo˜ no o2,3 C. Skourup Skourup22 2,3 1 3 1 Sp¨ u ntrup* J. G. Londo˜ n o N. F. Thornhill L. Imsland 3 1 F. Thornhill L. Imsland 1 2,3 C. Skourup2 N. Sp¨ J. G.3 Londo˜ no 1C. Skourup N.untrup* F. Thornhill 3 L. Imsland1 N. F. Thornhill L. Imsland 1 1 Department of Engineering Cybernetics, Norwegian University of 1 Department of Engineering Cybernetics, Norwegian University of 1 Engineering Cybernetics, Norwegian University of Scienceofand and Technology, 7491 Trondheim, Trondheim, Norway Science Technology, 7491 Norway 1 Department Department ofand Engineering Cybernetics, Norwegian University of 2 2 Science Technology, 7491 Trondheim, Norway AS, Oil, Gas and Chemicals, 0666 Oslo, Norway 2 ABB ABB AS, Oil, Gas and Chemicals, 0666 Oslo, Norway Science and Technology, 7491 Trondheim, Norway 2 3 3 AS, Oil, Gas and Chemicals,of Oslo, Engineering, Norway London, Department Chemical 3 Imperial Imperial London, Department of0666 Chemical 2 ABBCollege ABBCollege AS, Oil, Gas and Chemicals, 0666 Oslo, Engineering, Norway 3 Imperial College London, Department of Chemical Engineering, London, SW7 2AZ, UK 3 London,Department SW7 2AZ, of UK Imperial *College London, Chemical Engineering, London, SW7 2AZ, UK Corresponding Author, Author, contact contact via via e-mail: * Corresponding London,Author, SW7 2AZ, UKvia e-mail: * Corresponding contact e-mail: [email protected] [email protected] * Corresponding Author, contact via e-mail: [email protected] [email protected] Abstract: Abstract: Physical Physical assets assets of of the the process process industries industries include include compressors, compressors, pumps, pumps, heat heat exchangexchangAbstract: Physical assets of the process industries include compressors, pumps, heat exchangers, batch reactors and many more. A large company that operates over many sites typically ers, batch reactors and many more. A large company that operates over many sites typically Abstract: Physical assets of the process industries include pumps, heat exchangers, batchsuch reactors more. Away large that compressors, operates manyof typically manages such assetsand in a amany coordinated way as company an asset asset fleet. Strategic over planning ofsites maintenance manages assets in coordinated as an fleet. Strategic planning maintenance ers, batch reactors and many more. A large company that operates over many sites typically manages such assets in a coordinated way as an asset fleet. Strategic planning of maintenance and scheduling requires information about reliability, availability and maintainability of and scheduling requires information about reliability, availability and maintainability of the the manages such assets in a coordinated way as an asset fleet. Strategic planning of maintenance and scheduling requires information about reliability, availability and maintainability of the assets in an asset fleet. assets in an asset fleet. and information about the reliability, availability and maintainability of the assetsscheduling in anpresented assetrequires fleet. The work presented in this this paper assesses assesses reliability of centrifugal centrifugal compressors based based on The work in paper the reliability of compressors on assets in an asset fleet. The work presented in this paper assesses the reliability of centrifugal compressors based on the data data collected collected in in OREDA OREDA (Offshore (Offshore and and onshore onshore REliability REliability DAta DAta project). project). The The fault fault tree tree the The workcollected presented in this paper assesses the reliability of centrifugal compressors based on the data in OREDA (Offshore and onshore REliability DAta project). The fault tree (a top-down approach to illustrate all subsystems in a system) has been modeled by focusing (a top-down approach to illustrate all subsystems in REliability a system) has been modeled byfault focusing the data collected in OREDA (Offshore and onshore DAta project). The tree (a top-down approach to illustrate subsystems in a system) has been modeled control by focusing on the subsystems of compressor (power transmission, compressor, and on the six six main main subsystems of the the all compressor (power transmission, compressor, control and (a approach to illustrate subsystems in miscellaneous). a system) has been modeled by focusing on top-down the six main subsystems of shaft the all compressor transmission, compressor, control and monitoring, lubrication system, shaft seal system,(power and All the maintainable items monitoring, lubrication system, seal system, and miscellaneous). All the maintainable items on the six main subsystems of the compressor (power transmission, compressor, control and monitoring, lubrication system, shaft seal system, andfailure miscellaneous). All the maintainable items described in ISO 14224 are considered. Based on the rates collected in OREDA, the most described in ISO 14224 are considered. Based on the failure rates collected in OREDA, the most monitoring, lubrication system, shaft seal system, and miscellaneous). All the maintainable items described failures in ISO 14224 are considered. Based on theanalysis. failure rates in OREDA, the most prevalent have identified aa Pareto The article recommendations prevalent failures have been been identified via via Pareto The collected article gives gives recommendations described in ISO 14224 arebe considered. Based on theanalysis. failureand rates collected in of OREDA, the most prevalent failures have been identified via a Pareto analysis. The article gives recommendations which subsystems should prioritized for maintenance which types faults are which subsystems should beidentified prioritized for maintenance and which types ofrecommendations faults are likely likely prevalent failures have been via a Pareto analysis. The article gives which subsystems be prioritized maintenance and which statistical types of faults are of likely to occur. The main mainshould contribution of this this for paper is an an industry-based industry-based statistical analysis of the to occur. The contribution of paper is analysis the which subsystems should be prioritized for maintenance and which types of faults are likely to occur. The main in contribution of this papersystems. is an industry-based statistical analysis of the failure mechanisms centrifugal compressor It is expected to improve the reliability failure mechanisms in centrifugal compressor systems. It is expected to improve the reliability to occur. Thecompressor main incontribution of this paper is an industry-based statistical analysis of the failure mechanisms centrifugal compressor systems. It is expected to improve the reliability of centrifugal systems and can be implemented in industrial settings with a similar of centrifugal compressor systems compressor and can be systems. implemented industrial settings with a similar failure mechanisms in centrifugal It is in expected to improve the reliability of centrifugal compressor systems and can be implemented in industrial settings with a similar documentation system like OREDA. documentation system like OREDA. of centrifugal compressor and can be implemented in industrial settings with a similar documentation system likesystems OREDA. documentation system likeFederation OREDA.of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. © 2018, IFAC (International Keywords: Keywords: Reliability Reliability analysis, analysis, Information Information processing processing and and decision decision support, support, Fault Fault tree tree Keywords: Reliability analysis, Information processing and decision support, Fault tree analysis, Maintenance Scheduling and planning analysis, Maintenance Scheduling and planning Keywords:Maintenance Reliability analysis, Information processing and decision support, Fault tree analysis, Scheduling and planning analysis, Maintenance Scheduling and planning 1. INTRODUCTION INTRODUCTION an asset asset fleet. fleet. Usually Usually they they also also have have at at least least one one common common 1. an 1. INTRODUCTION an asset fleet. Usually they also have at least one common stakeholder (Schulze Sp¨ u ntrup and Imsland, 2018). This stakeholder (Schulze Sp¨ untrup Imsland, 2018). This 1. improving INTRODUCTION an asset fleet. Usually they also and have at least one common This paper paper is is about about the reliability reliability of of comprescompres- stakeholder (Schulze and Imsland, 2018). This might beSp¨ anuntrup engineer who is responsible responsible for This improving the stakeholder might be an engineer who is for stakeholder (Schulze Sp¨ ntrup and Imsland, This This paper on is about improving the reliability of database. compres- stakeholder sors based based on asset fleet fleet data from from a reliability reliability database. mightof anuassets engineer who is responsible for the maintenance ofbeall all assets in the fleet or 2018). a manager manager sors asset data a the maintenance in the fleet or a This based paper on is about improving the reliability compresstakeholder anassets engineer whofleet is responsible for sors asset fleet data from a reliability database. Improving the reliability of equipment equipment in the the ofprocess process in- the maintenance ofbeall in the or aentire manager who aims to tomight reduce carbon emission for the entire asImproving the reliability of in inwho aims reduce carbon emission for the assors based on asset fleet data from a reliability database. the maintenance of all assets in the fleet or a manager Improving reliability in as the process in- who dustry has has a athe major impactof onequipment aspects such such as environment, aimsFleet to reduce for located the entire asset fleet. fleet. Fleet assets carbon are not notemission necessarily located in one one dustry major impact on aspects environment, assets are necessarily in Improving reliability in as the process in- set whofleet. aims toanreduce emission forthese the entire asdustry has athe major impactof onequipment aspects such environment, safety, production availability and maintenance costs. The set Fleet assets carbon are not necessarily located in one place. From operational point of view, view, assets are safety, production availability and maintenance costs. The place. From an operational point of these assets are dustry production hasintegrity a majoravailability impact on aspects such as is environment, set one fleet. Fleet assets are not necessarily located in one safety, and maintenance costs. The technical of the complex systems mandatory place. From an operational point of view, these assets are in company. For improved data collection, multiple technical integrity availability of the complex systems is mandatory oneFrom company. For improved collection, multiple safety, production and maintenance costs. The in place. an operational pointdata of view, these and assets are technical integrity the complex systems is mandatory to ensure safe and and ofreliable reliable operations (World Economic in one company. For improved data multiple companies may form a consortium consortium tocollection, collect share to ensure safe operations (World Economic companies may form a to collect and share technical integrity of the complex systems is mandatory in one company. For improved data collection, multiple to ensure safe Achieving and reliable operations (World Forum, 2017). Achieving knowledge about the Economic likelihood companies may form to a make consortium collectwith and higher share data with with the the purpose to make better to decisions with higher Forum, 2017). knowledge about the likelihood data purpose better decisions to failures ensure safe and operations Economic companies may form to aThe consortium to collect andassets. share Forum, 2017). knowledge theactivity. likelihood of andAchieving theirreliable consequences isabout a(World key activity. In data with the purpose make better decisions higher statistical significance. data is created created fromwith the statistical significance. The data is from the assets. of failures and their consequences is a key In Forum, 2017). Achieving knowledge about theactivity. likelihood data with the purpose to make better decisions with higher of failures and their consequences is a key In companies of the oil and gas industry the current economic statistical significance. The data is created from the assets. This could concern process from the operation of the companies thetheir oil and gas industry is thea current economic couldsignificance. concern process data is from the operation of the of failuresisof and consequences key activity. In This statistical Theabout data created from the assets. companies of the oil and gas industry the current economic situation putting tight constraints on the operations This could concern process data from the operation of the equipment or information failures and maintenance situation isofputting tight constraints the operations or information about failures maintenance companies thelife oilcycle and gas industry theon economic equipment This concern dataisfrom theand operation of the situation is putting tight constraints oncurrent themaintenance operations and optimizing cost while reducing reducing maintenance equipment or information and maintenance of thecould specific asset.process This about data collected and stored for and optimizing life cycle cost while of the specific asset. This data isfailures collected and stored for situation is putting tight constraints on the operations equipment or information about failures and maintenance and optimizing life cycle cost while reducing maintenance cost. At the same time, reliability of equipment is not of the specific asset. This data is collected and stored for later use. use. later cost. At the same time, reliability of equipment is not and optimizing lifeprofit, cycle whilereliability reducing maintenance of theuse. specific asset. This data is collected and stored for cost. At the same time, cost reliability of equipment is not later just correlated to but high also reduces reduces just correlated to profit, but high reliability also Maintenance in the the context context of of asset asset fleets fleets reveals reveals benefits benefits cost. At the same time, reliability of equipment is not later use. Maintenance in just correlated profit, but and highthe reliability also which reduces the risk risk for the the to environment and the work force force which is the for environment work is Maintenance in the context of asset fleets reveals benefits in terms of the economies of scale and to identify assets just correlated to profit, but high reliability also reduces in terms of the economies of scale and to identify assets the risk for the environment and the work force which is Maintenance in the context of asset fleets reveals benefits inherent with every machine failure. failure. inherent with every machine in terms of the economies of scale and to identify assets that benefit most from it (economically), but from the risk for the environment and the work force which is inherent with every machine failure. that benefit most from it (economically), but from aa in terms of the economies of(economically), scale and toalso identify assets This article refers to assets in the meaning of engineerthat benefit most from it but from maintenance perspective challenges may arise as the inherent with every machine failure. This article refers to assets in the meaning of engineer- maintenance perspective challenges may also but arisefrom as theaa that benefit most fromthe itentire (economically), This article(Beebe, refers to assetsThis in the meaning of engineering assets assets (Beebe, 2010). This includes all the the physical maintenance challenges may also arise as the objective is to toperspective maintain fleet as good good as possible. possible. ing 2010). includes all physical objective is maintain the entire fleet as as Thisassets article refers to assetsThis in the of engineermaintenance challenges may also arise as the ing 2010). includes all the physical assets in aa(Beebe, process plant, such as meaning mechanical, process objective is toperspective maintain the entire fleet as good as possible. assets in process plant, such as mechanical, process A thorough analysis of data collected from asset failing assets (Beebe, 2010). This includes all the physical objective is to maintain the entire fleet as good as possible. assets in a process plant, mechanical, process or electrical electrical equipment. A such large as number of assets assets of aa A thorough analysis of the data collected from asset failor equipment. A large number of of A thorough analysis of the data collected from asset failures and maintenance actions can give insights on where assets in a process plant, such as mechanical, process ures and maintenance actions can give insights on where or electrical equipment. large of assets of a A thorough analysis of the data collected from asset failsimilar type and and functionAcan can be number aggregated and called called similar type function be aggregated and or electrical equipment. large of assets of a ures and maintenance actions can give insights on where similar type and functionAcan be number aggregated and called ures and maintenance actions can give insights on where similar type and function can be aggregated and called 2405-8963 © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Copyright © 2018 217 Copyright © under 2018 IFAC IFAC 217 Control. Peer review responsibility of International Federation of Automatic Copyright © 2018 IFAC 217 10.1016/j.ifacol.2018.06.380 Copyright © 2018 IFAC 217
IFAC OOGP 2018 218 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
to improve maintenance systems. This study aims to analyse the failures of topside centrifugal compressors used in offshore applications by the oil and gas industry. The findings will show where improvements in the maintenance actions have the largest impact on the reliability of current systems. Furthermore, condition monitoring systems are just able to predict failures from the first signs of failure. Research on predictive maintenance tries to predict failures even before these first signs and to obtain information about the likelihood of failure beforehand. Section 2 will show that there are open questions about how information from past failures can be used to achieve insights into the reliability of a compressor systems. This paper aims to fill that gap. The data from the Offshore and onshore REliability DAta project, OREDA (SINTEF Technology and Society et al., 2015), is used an an example, but the applied methods can be applied to similar data, too. Section 2 in this paper gives necessary background about the OREDA project, the topics of reliability and maintainability and other work related to this article. Section 3 presents the methodology used in this paper, namely for the Pareto analysis and the reliability model. Section 4 presents briefly the data of OREDA that makes up the foundation for the analysis. Section 5 presents the results of the analysis including the outcome from Pareto and statistical reliability analysis. Some findings indicate critical points that reveal serious safety risks. Furthermore, mitigating strategies are developed based on the analyses. Section 6 concludes the work and gives an outlook for future work. 2. BACKGROUND 2.1 Key terms of reliability engineering A failure mode is the manner in which an equipment or a part of an equipment can fail. By listing all failure modes for all maintainable items in a system, a failure mode and effects analysis (FMEA) can be performed. In reliability engineering, faults in complex systems are often displayed by using the fault tree method. It reflects the structure of the equipment by representing the different subsystems of an asset down to individual components. For each component it is possible to identify different failure modes. Ultimately, all possible ways that a system can fail based on the individual components should be represented in such a fault tree. The fault tree breaks a complex system down into subsystems down to a level, where no further segregation is possible. A fault tree analysis is a form of deductive failure analysis. It uses a fault tree as a topdown approach to describe the system. The failure events at the lower level are then combined by Boolean logic and a undesired state of the system is explained. 2.2 The OREDA project A reliability analsis of assets such as compressors, requires a comprehensive amount of data, preferably from an entire compressor fleet. The project organization OREDA (Onshore & Offshore Reliability Data) has been set up to collect and exchange industrial reliability data in the oil and gas industry for the participating companies operating worldwide. The database resulting from the OREDA 218
Fig. 1. Example of a fault tree. It can be extended to a higher number of branches and also more levels. The numbered circles denote the events that cause failure. project comprises reliability and maintenance data for various types of equipment that are used in the exploration and production in various geographic areas, installations and operating conditions. There exists a wide range of literature based on the data from the OREDA project. 2.3 Work based on OREDA Turan et al. (2011) conducted a study of the reliability and criticality for the case study of diving support vessel. This is done with the background of the OREDA handbook and other maintenance standards and procedures, such as the ALARP and the KP3 initiative, or the basic reliabilitycentered maintenance (RCM) and risk-based inspection (RBI) principles. The reliability assessment of Turan et al. (2011) is done by utilizing two reliability importance measures. Similar to our approach, the reliabilities of the main system and the subsystems are identified and suggestions are made how to improve the overall reliability of the various systems. This work by Turan et al. (2011) is referenced since it works with the same database, looks into the subsystems and takes information to estimate the reliability. Langseth and Lindqvist (2006) chose a statistical approach to analyze a data set from the OREDA database. The model explains the different parts of the system that can fail (termed competing risks) and assumes that the preventive maintenance is imperfect and does not result in a total repair of the equipment. The authors conclude that the model has the capability to describe the data set that they used set as well. However, the authors state they are facing a non-identifiability of the competing risk problems and cannot claim to have found the correct model but rather that the data does not reject the developed model. The main outcomes of that work are therefore likelihood parameters (such as the eagerness and thoroughness of the maintenance crew) to describe the quality of the maintenance that has been performed. Rausand and Oien (1996) used OREDA to illustrate the interpretation of basic concepts in failure analysis using an example of a gate valve. During their analysis they point out limitations in OREDA and propose enhancements, e.g. to record failure causes, failure mechanisms or root causes. Until now, this information are still not collected
IFAC OOGP 2018 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
219
in OREDA. The reason for discussing this reference is that there are already improvements suggested to OREDA. This is also part of the outcome of our work. Sandtorv et al. (1996) share their practical experience of the OREDA project (e.g. that not always the personnel with the needed knowledge is available for data collection or that it is hard to agree on specifications for complex equipment). They make recommendations for future data collection projects. On the other hand, Hameed et al. (2011) named OREDA as the state of the art for data collection in the industry and they try to identify challenges regarding reliability and maintainability data collection for other industries, such as offshore wind turbines. In the failure analysis of Haugen et al. (1997) both critical failures and degraded failures, followed by critical failures, are considered. The OREDA project supplied the necessary data for this purpose. These paper demonstrate that OREDA is a useful database, and has been used successfully for research in areas such as failure analysis and reliability. 2.4 Work with a similar purpose Choi and Chang (2016) try to improve reliability of subsea production systems. They analyse the reliability of a seabed storage tank based on a fault tree. The data is taken from components of the lowest level of the seabed storage tank. Active repair times from OREDA were taken into account and other databases were used to supplement the study. This work is relevant, since their approach is based on a fault tree and shows similarities to the method described in this article. However, they merge failure rates from different systems. This work uses data which is specific to a compressor system. A synergetic approach is chosen by Ebrahimipour and Suzuki (2006). Principal component analysis is coupled with importance analysis to improve equipment performance. The leading causes and common-cause events are identified. The topic of this paper is to identify main causes for failures, similar to the work of our article, but with different methods. Another work with an relative approach for assesing the reliability of assets is given by Corvaro et al. (2017). The authors are using a RAM methodology (Reliability, Availability, and Maintainability) for the case of reciprocating compressors. The reliability is done with a RCM study from a private company. The availability calculation is performed by means of block diagrams. The input are failure rates, which is similar to the approach in the model developed by our work. The maintainability corresponds to maintenance activities suggested by the RCM analysis. These activities are assigned to criticality values that have been calculated before. Even though Corvaro et al. (2017) use RCM to build their model and the used equipment boundaries are not given by ISO 14224, the results of the work are comparable to the method presented in this paper. 3. METHODOLOGY The work presented in this paper uses statistical reliability analysis and Pareto analysis. Fig. 2 illustrates the methodology and how these two different analyses work together. The dashed boxes indicate the logical next steps, which are outside the scope of this paper. 219
Fig. 2. Methodology for the statistical reliability analysis; dashed boxes indicates work that goes beyond the scope of this paper 3.1 Selection of data from OREDA The taxonomy classification of ISO 14224:2016 for equipment used in the petroleum, natural gas, and petrochemical industry is used as well as the failure mode records for each component as described in the OREDA database. ISO 14224:2016 “Petroleum, petrochemical and natural gas industries – Collection and exchange of reliability and maintenance data for equipment” specifies the equipment to be included when studying a compressor system. Tab. 1 presents a list of compressor subsystems and the related equipment. 3.2 Pareto analysis The Pareto principle or the 80/20 rule, states that the majority (roughly 80%) of the effects come from a minor part (roughly 20%) of the causes. A tool called Pareto analysis is build on this principle. A Pareto analysis works by listing possible causes and ordering them, usually by their estimated benefit or by their impact on a problem. In the next step, a number of the most effective actions or most promising opportunities for countermeasures are selected. This is done by the Pareto principle, approaching 20% of the action items and expecting 80% of the benefit. In the case of OREDA, the information about the maintainable items and the failure modes must be sorted, based on the number of events when a specific maintainable item failed or when a specific failure mode occurred. Next to the contribution of the items the accumulated contribution in a descending order (from high contribution to low contribution) is indicated. The focus is on the action items that reach a cumulative contribution of 80%. A Pareto analysis has been conducted on the OREDA data (failure rates for each maintainable item and each failure mode). Focusing on the causes that
IFAC OOGP 2018 220 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
Fig. 3. Example of the Simulink model for the power transmission subunit; the calculation blocks are explained in Section 3.3. result in the majority of failures will help to improve the reliability of the system. 3.3 Statistical reliability model The statistical reliability model utilizes a fault tree analysis. The compressor subunits shown in Tab. 1 and all the maintainable items listed in the OREDA database are used to create a model. Based on the reliability data from the offshore compressors a fault tree is modeled in Simulink. This is done taking the six subsystems into account individually and all maintainable items for each subsystem. The common assumption that a failure is total and partial failures are not taken into account is applied (Baig et al., 2013). A failure in one of the subsystems leads to a failure of the entire compressor system. Reliability is the “ability of an item to perform a required function under given conditions for a given time interval” (ISO 14224:2016). The relevant information in the OREDA database is in a table in OREDA (SINTEF Technology and Society et al., 2015). This table gives failure rate λi due to each of the failure modes i, and fractions pi,j of the total failure rate for each combination of failure mode i and maintainable item j. The failure modes such as ELU (External Leakage - Utility Medium) are listed in Tab. 2, and the maintainable items are listed in Tab. 1. Fig. 3 illustrates the calculation procedure for the power transmission system, which is comprised of maintainable items such as Gearbox and Coupling to the driven unit. The calculation proceeds as follows: (1) The blocks labelled as 1, 2, 3 ... 14 calculate ri,j = e−λi pi,j t where t is time. For instance, in block 5, λi is the failure rate per hour for failure mode ELU, and pi,j is 0.001, being the fraction of failures for failure mode ELU reported in OREDA for the Gearbox 220
maintainable item. Since Gearbox has five possible failure modes, blocks 6 to 9 in Fig. 3 calculate ri,j values for the other four failure modes. (2) The ri,j values are inputs for calculations of reliability of the maintainable items. For instance, the ri,j values from blocks 5-9 go into Gearbox which calculates reliability of the Gearbox maintainable item over all its failure modes as rj = 1 − Πi (1 − ri,j ). (3) Similarly, the overall reliability r of the Power Transmission system is given by combining the reliability of the individual maintainable items in the Power Transmission system, i.e. r = 1 − Πj (1 − rj ).
Fig. 3 illustrates the Power Transmission system, which is one of the six subsystems of a compressor as listed in Tab. 1. Similar reliability calculations are carried out for all the subsystems. The results are combined to give an overall system reliability. 3.4 Steps beyond the scope of this work The development of mitigating strategies and the improvement of the system are not covered in this paper. However, it is intended that the results of this work are used as input for those steps. 4. CASE STUDY Currently, data from 278 installations, 17,000 equipment units with 39,000 failure and 73,000 maintenance records are stored in the database. The OREDA handbook described this database, currently in the sixth edition which has been published during the twelfth phase of the project. OREDA data being relevant to this study comprises data from from 22 offshore turbine-driven centrifugal compressors in eight different installations. Tab. 1 lists the six subsystems of compressors and all the maintainable items for each subsystem. Tab. 2 lists all the different failure modes that have been considered.
IFAC OOGP 2018 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
221
Table 1. Subdivison of a compressor into subunits and the maintainable items of each subsystem Power Transmission -Gearbox/Variable drive -Seals -Lubrication -Coupling to the driven unit -Coupling to the driver
Compressor Unit -Antisurge System -Casing -Cylinder liner -Instruments -Shaft seals -Radial Bearing -Thrust bearing -Interstage seals -Valves -Internal piping -Balance Piston -Packing -Rotor w/ impellers
Compressors Control and Lubrication SysMonitoring tem -Wiring -Reservoir w/ heat-Control unit ing system -Actuating device -Pump w/ motor -Monitoring -Filter -Internal power sup-Cooler ply -Valves and Piping -Seals -Oil -Piping -Instruments -Sensors -Seals -Valves
Shaft Seal System -Dry gas seal -Filters -Gear -Mechanical seal -Reservoir w/ heating system -Pump -Motor -Scrubber -Seal oil -Valves
Miscellaneous -Base frame -Flange joints -Control valves -Magnetic bearing control system -Isolation valves -Coolers -Piping, pipe support and bellows -Purge air -Silencers
Table 2. List of all considered failure modes for compressors and their acronyms Acronym AIR BRD ERO ELP ELU FTS STP HIO INL LOO SER NOI OTH OHE PDE UST STD UNK VIB
Failure Mode Abnormal instrument reading Breakdown Erratic output External leakage - Process medium External leakage - Utility medium Fail to start on demand Fail to stop on demand High output Internal leakage Low output Minor in-service problems Noise Other Overheating Parameter deviation Spurious stop Structural deficiency Unknown Vibration
5. RESULTS OF THE ANALYSIS AND DISCUSSION 5.1 Pareto analysis The Pareto analysis for the maintainable items (Fig. 4) reveals that the main contributing maintainable item is Unknown (The Unknown category will be further explained in section 5.4). Further, 10 other maintainable items contribute to 80% most frequently failing items. Similarly, the Pareto analysis for the failure modes (Fig. 5) gives the insight that the most common failure mode in the boundaries of the compressor system is abnormal instrument reading. This contributes for more than 30% of all failures. In addition, seven other items contribute to 80% of the total failures. The Pareto principle is in this case not exact with about 40% of the failure modes making up 80% of the failures. However, this can be explained by the small numbers of failure modes. The results show that maintenance engineers should take care if they are basing maintenance decision on the 80/20 rule, because the analysis of the OREDA shows that this rule could be misleading in some situations. The main message of the Pareto analysis is that abnormal instrument reading should be the main focus of further improvements. 221
Fig. 4. Pareto analysis of the failures for specific maintainable items; the orange line denotes the accumulated percentage if failures, the dashed region contains minor contributors to the failures that have been removed from the graph 5.2 Statistical Reliability Analysis Fig. 6 shows the results of the statistical reliability model. It gives an insight into the reliability of each sub-unit of the centrifugal gas compressor over time based on failure rates from the OREDA. It is evident over time, that the shaft seal system, the power transmission system and the lubrication system are impacting the reliability of the system very little. Their reliability value does not drop as much as other values do. This is mainly because the raw data in the database shows the failure rate for the shaft seal system and lubrication system are small.
IFAC OOGP 2018 222 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
The database reports external leakage of process medium as a failure mode that occurs in 6.25% of all failures. Leakage of process medium is unwanted in any case. This failure mode is the most frequent of the maintainable item category Unknown (2% of all failures). This means that there is a leakage, but it is unclear which maintainable part causes this leakage. Another critical failure mode is Fail to stop on demand. In 0.5% of the cases this is due to an Unknown maintainable item. From an external perspective the high number of Unknown that are documented seem to be an error in the data collection because with an investigation after the machine failure it should be possible to find out where e.g. a leak was. Since these two combinations are reported often, it should be a goal to reduce the amount of failures in the unknown category to support the industry. Fig. 5. Pareto analysis of the failures for specific failure modes; the orange line denotes the accumulated percentage if failures, the dashed region contains minor contributors to the failures that have been removed from the graph
Usage of results While the information obtained through the Pareto analysis is qualitative and suggests where to take actions and where to improve the existing systems, the statistical reliability model gives information about how the reliability of the specific sub-unit changes over time. Maintenance planning and scheduling requires such information for deciding what to do and when to take actions. Information is needed for background tasks such as ordering of spare parts, managing the inventory of spare parts, and getting the spare parts to where they are needed. The results in this paper can be useful for the following activities:
Fig. 6. Reliability of gas compressor subunits over time; the calculation blocks are explained in section 3.3. On the other hand, the Control and Monitoring sub-unit has the largest influence on reliability. The effect is by far larger than the reliability decrease of the sub-units power transmission, the compressor itself and miscellaneous other items in the compressor system. Also, the degradation process of the Control and Monitoring subunit starts faster. The overall reliability of the system reaches 80% after about 2.56 years. According to the OREDA data this would happen if there was no maintenance at all. 5.3 Knowledge about main issues for reliability During the analysis of the data from OREDA two observations were made that have a serious safety risk linked to them. 222
• Qualitative information from the Pareto analysis to identify the maintainable items that should become part of the maintenance plan • Quantitative information from the statistical reliability model such as the change of reliability over time within statistical boundaries can increase the priority for scheduling maintenance compared to an asset with a lower likelihood of degradation. • Based on the statistical reliability model, thresholds can be set. When the reliability reaches this thresholds, maintenance shall be performed for these items. This is useful for assets that are difficult to perform predictive maintenance on. The results in this paper are based on the OREDA data. However, companies operating large asset fleets typically also have their own databases. The findings in this paper suggest that it is very important to have accurate information about the different failure modes and maintainable items of an asset. The results also show that a large number of unknown faults reduces the usefulness of asset fleet data. It would be useful for further development of the methods to investigate reliability to reduce the numbers of faults categorized as unknown. A challenge in maintenance is often the absence of information. Condition monitoring data would help to predict failures several months before the asset fails. However, for long-term planning and even for scheduling with a time horizon of six months such information is not available.
IFAC OOGP 2018 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
The reason for these comments is that condition monitoring systems are becoming more prevalent, and they enable predictive maintenance. Predictive maintenance aims to predict machine failures even before the first signs are there, e.g. vibration or noise from the asset. With the statistical rating of the reliability it is possible to define thresholds and if the assumed reliability decreases below this threshold maintenance can be expected within the coming months. This is valuable information with the potential to improve existing maintenance planning systems. 5.4 Observations about the OREDA database The Unknown maintainable item is the biggest contributor to compressor failures within all maintainable items. The authors acknowledge that this category is important and often the only option to categorize something to. Nonetheless, from an operational point of view there are also a few cases with combinations listed that are not logical. The combination of the maintainable item unknown with a known failure mode (other ) means that there is knowledge about how something failed (otherwise the failure mode unknown would be the appropriate category) without knowing which item shows this failure mode. Also, the failure mode unknown appears very often. An example for another unlogical combination that is reported it the maintainable item is unknown, but the failure mode is an abnormal instrument reading. If it is known that a reading is abnormal, it must also be known where the reading comes from. In operations it might be difficult to track the history of equipment reliably. This has various reasons. The personal that is conducting this is changing over time. So the categorization of similar incidents might be deviating a lot regarding the categorization. Labels are based on the understanding of who reports a production loss or failure. This missing objectiveness arises also from the fact that there is different maintenance personnel in the different companies that contribute to OREDA. Even within one company it is very likely that failures are categorized by different personnel. The reason for highlighting these points is that the results of the conducted analysis are sensitive to every documented failure event. If an unknown error is documented or an error in an unknown maintainable item, no suggestions for future actions can be made and the statistical reliability model becomes less accurate. Therefore, a finding of the paper is that a company maintaining an asset fleet should take pains to reduce the number of unknowns in its database since this will help to improve future reliability. 6. CONCLUSION Compressor failure data from the OREDA database was used to conduct a reliability analysis and a Pareto analysis. These analyses help to understand the offshore compressors and their reliability better. The most prevalent failures have been identified. Main drivers for failure are the subunits control and monitoring and the power transmission, followed by unknown items. Abnormal instrument 223
223
reading is the most common failure mode. Main contributors can be identified and proposals can be made of how to cope with some of the challenges in daily operation of the equipment. Recommendations were made on how to improve data collection projects. The high number of unknown failures, with some of them not being logical, is critical and companies need to investigate if and why this is happening. However, the authors are aware of the challenges in the collection of the failure events and appreciate the way of improving reliability systems within a company by data collection. Aggregating data over the boundaries of a single company helps to collect data from a broader range of operating conditions and external factors such as climate. In the long run this will help the entire industry to become safer and more reliable which on the other hand will results in higher profit margins. The output of the reliability analysis is an important input for future developments of maintenance scheduling and planning tools, as the information about the decrease of reliability within a subunit is an important decision factor. Further study is recommended for the other identified parts of the compressor system that contribute to the most failures. Within the limitations of this work possible solutions for the control and monitoring systems were introduced. From the proposed methodology in Fig. 2 the improvement to the system need to be implemented. This will result in new operational data and the benefits of the changed system will become evident. Another point for further study is a more detailed assessment of the reliability. ACKNOWLEDGEMENTS Financial support is gratefully acknowledged from the Marie Sklodowska Curie Horizon 2020 EID-ITN project “PROcess NeTwork Optimization for efficient and sustainable operation of Europe’s process industries taking machinery condition and process performance into account – PRONTO”, Grant agreement No 675215. REFERENCES Baig, A.A., Ruzli, R., and Buang, A.B. (2013). Reliability analysis using fault tree analysis: A review. International Journal of Chemical Engineering and Applications, 4(3), 169–173. Beebe, R. (2010). Definitions, Concepts and Scope of Engineering Asset Management, volume 1. Springer London. Choi, I.H. and Chang, D. (2016). Reliability and availability assessment of seabed storage tanks using fault tree analysis. Ocean Engineering, 120, 1–14. Corvaro, F., Giacchetta, G., Marchetti, B., and Recanati, M. (2017). Reliability, Availability, Maintainability (RAM) study, on reciprocating compressors API 618. Petroleum, 3(2), 266–272. Ebrahimipour, V. and Suzuki, K. (2006). A synergetic approach for assessing and improving equipment performance in offshore industry based on dependability. Reliability Engineering and System Safety, 91(1), 10–19. Hameed, Z., Vatn, J., and Heggset, J. (2011). Challenges in the reliability and maintainability data collection for
IFAC OOGP 2018 224 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
offshore wind turbines. Renewable Energy, 36(8), 2154– 2165. Haugen, K., Hokstad, P., and Sandtorv, H. (1997). The analysis of failure data in the presence of critical and degraded failures. Reliability Engineering & System Safety, 58(2), 97–107. Langseth, H. and Lindqvist, B.H. (2006). Competing risks for repairable systems: A data study. Journal of Statistical Planning and Inference, 136(5), 1687–1700. Rausand, M. and Oien, K. (1996). The basic concepts of failure analysis. Reliability Engineering and System Safety, 53, 73–83. Sandtorv, H.A., Hokstad, P., and Thompson, D.W. (1996). Practical experience with a data collection project: the OREDA project. Reliability Engineering & System Safety, 51(2), 159–167. Schulze Sp¨ untrup, F. and Imsland, L. (2018). Asset fleet management in the process industry - a conceptual model. In Submitted to: 10th Advanced Control of Chemical Processes - ADCHEM ’18. Shenyang. SINTEF Technology and Society, Universitet, N.T.N., and DNV GL (Firm) (2015). OREDA - Offshore Reliability Data Handbook 6th Edition. OREDA participants, 6th edition. Turan, O., Lazakis, I., Judah, S., and Incecik, A. (2011). Investigating the reliability and criticality of the maintenance characteristics of a diving support vessel. Quality and Reliability Engineering International, 27(7), 931– 946. World Economic Forum (2017). Digital Transformation Initiative: oil and gas industry (Whitepaper). URL http://reports.weforum. org/digital-transformation/wp-content/ blogs.dir/94/mp/files/pages/files/ dti-oil-and-gas-industry-white-paper.pdf.
224
ScienceDirect
IFAC PapersOnLine 51-8 (2018) 217–224
Reliability improvement of compressors Reliability improvement of based on asset fleet reliability data Reliability improvement of compressors compressors based on asset fleet reliability data based on asset fleet reliability data based on asset11 fleet reliability data 22 2,3 2,3 F. F. F. F.
Schulze Schulze Schulze Schulze
Sp¨ u n C. Sp¨ untrup* ntrup*11 J. J. G. G.3 Londo˜ Londo˜ no o2,3 C. Skourup Skourup22 2,3 1 3 1 Sp¨ u ntrup* J. G. Londo˜ n o N. F. Thornhill L. Imsland 3 1 F. Thornhill L. Imsland 1 2,3 C. Skourup2 N. Sp¨ J. G.3 Londo˜ no 1C. Skourup N.untrup* F. Thornhill 3 L. Imsland1 N. F. Thornhill L. Imsland 1 1 Department of Engineering Cybernetics, Norwegian University of 1 Department of Engineering Cybernetics, Norwegian University of 1 Engineering Cybernetics, Norwegian University of Scienceofand and Technology, 7491 Trondheim, Trondheim, Norway Science Technology, 7491 Norway 1 Department Department ofand Engineering Cybernetics, Norwegian University of 2 2 Science Technology, 7491 Trondheim, Norway AS, Oil, Gas and Chemicals, 0666 Oslo, Norway 2 ABB ABB AS, Oil, Gas and Chemicals, 0666 Oslo, Norway Science and Technology, 7491 Trondheim, Norway 2 3 3 AS, Oil, Gas and Chemicals,of Oslo, Engineering, Norway London, Department Chemical 3 Imperial Imperial London, Department of0666 Chemical 2 ABBCollege ABBCollege AS, Oil, Gas and Chemicals, 0666 Oslo, Engineering, Norway 3 Imperial College London, Department of Chemical Engineering, London, SW7 2AZ, UK 3 London,Department SW7 2AZ, of UK Imperial *College London, Chemical Engineering, London, SW7 2AZ, UK Corresponding Author, Author, contact contact via via e-mail: * Corresponding London,Author, SW7 2AZ, UKvia e-mail: * Corresponding contact e-mail: [email protected] [email protected] * Corresponding Author, contact via e-mail: [email protected] [email protected] Abstract: Abstract: Physical Physical assets assets of of the the process process industries industries include include compressors, compressors, pumps, pumps, heat heat exchangexchangAbstract: Physical assets of the process industries include compressors, pumps, heat exchangers, batch reactors and many more. A large company that operates over many sites typically ers, batch reactors and many more. A large company that operates over many sites typically Abstract: Physical assets of the process industries include pumps, heat exchangers, batchsuch reactors more. Away large that compressors, operates manyof typically manages such assetsand in a amany coordinated way as company an asset asset fleet. Strategic over planning ofsites maintenance manages assets in coordinated as an fleet. Strategic planning maintenance ers, batch reactors and many more. A large company that operates over many sites typically manages such assets in a coordinated way as an asset fleet. Strategic planning of maintenance and scheduling requires information about reliability, availability and maintainability of and scheduling requires information about reliability, availability and maintainability of the the manages such assets in a coordinated way as an asset fleet. Strategic planning of maintenance and scheduling requires information about reliability, availability and maintainability of the assets in an asset fleet. assets in an asset fleet. and information about the reliability, availability and maintainability of the assetsscheduling in anpresented assetrequires fleet. The work presented in this this paper assesses assesses reliability of centrifugal centrifugal compressors based based on The work in paper the reliability of compressors on assets in an asset fleet. The work presented in this paper assesses the reliability of centrifugal compressors based on the data data collected collected in in OREDA OREDA (Offshore (Offshore and and onshore onshore REliability REliability DAta DAta project). project). The The fault fault tree tree the The workcollected presented in this paper assesses the reliability of centrifugal compressors based on the data in OREDA (Offshore and onshore REliability DAta project). The fault tree (a top-down approach to illustrate all subsystems in a system) has been modeled by focusing (a top-down approach to illustrate all subsystems in REliability a system) has been modeled byfault focusing the data collected in OREDA (Offshore and onshore DAta project). The tree (a top-down approach to illustrate subsystems in a system) has been modeled control by focusing on the subsystems of compressor (power transmission, compressor, and on the six six main main subsystems of the the all compressor (power transmission, compressor, control and (a approach to illustrate subsystems in miscellaneous). a system) has been modeled by focusing on top-down the six main subsystems of shaft the all compressor transmission, compressor, control and monitoring, lubrication system, shaft seal system,(power and All the maintainable items monitoring, lubrication system, seal system, and miscellaneous). All the maintainable items on the six main subsystems of the compressor (power transmission, compressor, control and monitoring, lubrication system, shaft seal system, andfailure miscellaneous). All the maintainable items described in ISO 14224 are considered. Based on the rates collected in OREDA, the most described in ISO 14224 are considered. Based on the failure rates collected in OREDA, the most monitoring, lubrication system, shaft seal system, and miscellaneous). All the maintainable items described failures in ISO 14224 are considered. Based on theanalysis. failure rates in OREDA, the most prevalent have identified aa Pareto The article recommendations prevalent failures have been been identified via via Pareto The collected article gives gives recommendations described in ISO 14224 arebe considered. Based on theanalysis. failureand rates collected in of OREDA, the most prevalent failures have been identified via a Pareto analysis. The article gives recommendations which subsystems should prioritized for maintenance which types faults are which subsystems should beidentified prioritized for maintenance and which types ofrecommendations faults are likely likely prevalent failures have been via a Pareto analysis. The article gives which subsystems be prioritized maintenance and which statistical types of faults are of likely to occur. The main mainshould contribution of this this for paper is an an industry-based industry-based statistical analysis of the to occur. The contribution of paper is analysis the which subsystems should be prioritized for maintenance and which types of faults are likely to occur. The main in contribution of this papersystems. is an industry-based statistical analysis of the failure mechanisms centrifugal compressor It is expected to improve the reliability failure mechanisms in centrifugal compressor systems. It is expected to improve the reliability to occur. Thecompressor main incontribution of this paper is an industry-based statistical analysis of the failure mechanisms centrifugal compressor systems. It is expected to improve the reliability of centrifugal systems and can be implemented in industrial settings with a similar of centrifugal compressor systems compressor and can be systems. implemented industrial settings with a similar failure mechanisms in centrifugal It is in expected to improve the reliability of centrifugal compressor systems and can be implemented in industrial settings with a similar documentation system like OREDA. documentation system like OREDA. of centrifugal compressor and can be implemented in industrial settings with a similar documentation system likesystems OREDA. documentation system likeFederation OREDA.of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. © 2018, IFAC (International Keywords: Keywords: Reliability Reliability analysis, analysis, Information Information processing processing and and decision decision support, support, Fault Fault tree tree Keywords: Reliability analysis, Information processing and decision support, Fault tree analysis, Maintenance Scheduling and planning analysis, Maintenance Scheduling and planning Keywords:Maintenance Reliability analysis, Information processing and decision support, Fault tree analysis, Scheduling and planning analysis, Maintenance Scheduling and planning 1. INTRODUCTION INTRODUCTION an asset asset fleet. fleet. Usually Usually they they also also have have at at least least one one common common 1. an 1. INTRODUCTION an asset fleet. Usually they also have at least one common stakeholder (Schulze Sp¨ u ntrup and Imsland, 2018). This stakeholder (Schulze Sp¨ untrup Imsland, 2018). This 1. improving INTRODUCTION an asset fleet. Usually they also and have at least one common This paper paper is is about about the reliability reliability of of comprescompres- stakeholder (Schulze and Imsland, 2018). This might beSp¨ anuntrup engineer who is responsible responsible for This improving the stakeholder might be an engineer who is for stakeholder (Schulze Sp¨ ntrup and Imsland, This This paper on is about improving the reliability of database. compres- stakeholder sors based based on asset fleet fleet data from from a reliability reliability database. mightof anuassets engineer who is responsible for the maintenance ofbeall all assets in the fleet or 2018). a manager manager sors asset data a the maintenance in the fleet or a This based paper on is about improving the reliability compresstakeholder anassets engineer whofleet is responsible for sors asset fleet data from a reliability database. Improving the reliability of equipment equipment in the the ofprocess process in- the maintenance ofbeall in the or aentire manager who aims to tomight reduce carbon emission for the entire asImproving the reliability of in inwho aims reduce carbon emission for the assors based on asset fleet data from a reliability database. the maintenance of all assets in the fleet or a manager Improving reliability in as the process in- who dustry has has a athe major impactof onequipment aspects such such as environment, aimsFleet to reduce for located the entire asset fleet. fleet. Fleet assets carbon are not notemission necessarily located in one one dustry major impact on aspects environment, assets are necessarily in Improving reliability in as the process in- set whofleet. aims toanreduce emission forthese the entire asdustry has athe major impactof onequipment aspects such environment, safety, production availability and maintenance costs. The set Fleet assets carbon are not necessarily located in one place. From operational point of view, view, assets are safety, production availability and maintenance costs. The place. From an operational point of these assets are dustry production hasintegrity a majoravailability impact on aspects such as is environment, set one fleet. Fleet assets are not necessarily located in one safety, and maintenance costs. The technical of the complex systems mandatory place. From an operational point of view, these assets are in company. For improved data collection, multiple technical integrity availability of the complex systems is mandatory oneFrom company. For improved collection, multiple safety, production and maintenance costs. The in place. an operational pointdata of view, these and assets are technical integrity the complex systems is mandatory to ensure safe and and ofreliable reliable operations (World Economic in one company. For improved data multiple companies may form a consortium consortium tocollection, collect share to ensure safe operations (World Economic companies may form a to collect and share technical integrity of the complex systems is mandatory in one company. For improved data collection, multiple to ensure safe Achieving and reliable operations (World Forum, 2017). Achieving knowledge about the Economic likelihood companies may form to a make consortium collectwith and higher share data with with the the purpose to make better to decisions with higher Forum, 2017). knowledge about the likelihood data purpose better decisions to failures ensure safe and operations Economic companies may form to aThe consortium to collect andassets. share Forum, 2017). knowledge theactivity. likelihood of andAchieving theirreliable consequences isabout a(World key activity. In data with the purpose make better decisions higher statistical significance. data is created created fromwith the statistical significance. The data is from the assets. of failures and their consequences is a key In Forum, 2017). Achieving knowledge about theactivity. likelihood data with the purpose to make better decisions with higher of failures and their consequences is a key In companies of the oil and gas industry the current economic statistical significance. The data is created from the assets. This could concern process from the operation of the companies thetheir oil and gas industry is thea current economic couldsignificance. concern process data is from the operation of the of failuresisof and consequences key activity. In This statistical Theabout data created from the assets. companies of the oil and gas industry the current economic situation putting tight constraints on the operations This could concern process data from the operation of the equipment or information failures and maintenance situation isofputting tight constraints the operations or information about failures maintenance companies thelife oilcycle and gas industry theon economic equipment This concern dataisfrom theand operation of the situation is putting tight constraints oncurrent themaintenance operations and optimizing cost while reducing reducing maintenance equipment or information and maintenance of thecould specific asset.process This about data collected and stored for and optimizing life cycle cost while of the specific asset. This data isfailures collected and stored for situation is putting tight constraints on the operations equipment or information about failures and maintenance and optimizing life cycle cost while reducing maintenance cost. At the same time, reliability of equipment is not of the specific asset. This data is collected and stored for later use. use. later cost. At the same time, reliability of equipment is not and optimizing lifeprofit, cycle whilereliability reducing maintenance of theuse. specific asset. This data is collected and stored for cost. At the same time, cost reliability of equipment is not later just correlated to but high also reduces reduces just correlated to profit, but high reliability also Maintenance in the the context context of of asset asset fleets fleets reveals reveals benefits benefits cost. At the same time, reliability of equipment is not later use. Maintenance in just correlated profit, but and highthe reliability also which reduces the risk risk for the the to environment and the work force force which is the for environment work is Maintenance in the context of asset fleets reveals benefits in terms of the economies of scale and to identify assets just correlated to profit, but high reliability also reduces in terms of the economies of scale and to identify assets the risk for the environment and the work force which is Maintenance in the context of asset fleets reveals benefits inherent with every machine failure. failure. inherent with every machine in terms of the economies of scale and to identify assets that benefit most from it (economically), but from the risk for the environment and the work force which is inherent with every machine failure. that benefit most from it (economically), but from aa in terms of the economies of(economically), scale and toalso identify assets This article refers to assets in the meaning of engineerthat benefit most from it but from maintenance perspective challenges may arise as the inherent with every machine failure. This article refers to assets in the meaning of engineer- maintenance perspective challenges may also but arisefrom as theaa that benefit most fromthe itentire (economically), This article(Beebe, refers to assetsThis in the meaning of engineering assets assets (Beebe, 2010). This includes all the the physical maintenance challenges may also arise as the objective is to toperspective maintain fleet as good good as possible. possible. ing 2010). includes all physical objective is maintain the entire fleet as as Thisassets article refers to assetsThis in the of engineermaintenance challenges may also arise as the ing 2010). includes all the physical assets in aa(Beebe, process plant, such as meaning mechanical, process objective is toperspective maintain the entire fleet as good as possible. assets in process plant, such as mechanical, process A thorough analysis of data collected from asset failing assets (Beebe, 2010). This includes all the physical objective is to maintain the entire fleet as good as possible. assets in a process plant, mechanical, process or electrical electrical equipment. A such large as number of assets assets of aa A thorough analysis of the data collected from asset failor equipment. A large number of of A thorough analysis of the data collected from asset failures and maintenance actions can give insights on where assets in a process plant, such as mechanical, process ures and maintenance actions can give insights on where or electrical equipment. large of assets of a A thorough analysis of the data collected from asset failsimilar type and and functionAcan can be number aggregated and called called similar type function be aggregated and or electrical equipment. large of assets of a ures and maintenance actions can give insights on where similar type and functionAcan be number aggregated and called ures and maintenance actions can give insights on where similar type and function can be aggregated and called 2405-8963 © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Copyright © 2018 217 Copyright © under 2018 IFAC IFAC 217 Control. Peer review responsibility of International Federation of Automatic Copyright © 2018 IFAC 217 10.1016/j.ifacol.2018.06.380 Copyright © 2018 IFAC 217
IFAC OOGP 2018 218 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
to improve maintenance systems. This study aims to analyse the failures of topside centrifugal compressors used in offshore applications by the oil and gas industry. The findings will show where improvements in the maintenance actions have the largest impact on the reliability of current systems. Furthermore, condition monitoring systems are just able to predict failures from the first signs of failure. Research on predictive maintenance tries to predict failures even before these first signs and to obtain information about the likelihood of failure beforehand. Section 2 will show that there are open questions about how information from past failures can be used to achieve insights into the reliability of a compressor systems. This paper aims to fill that gap. The data from the Offshore and onshore REliability DAta project, OREDA (SINTEF Technology and Society et al., 2015), is used an an example, but the applied methods can be applied to similar data, too. Section 2 in this paper gives necessary background about the OREDA project, the topics of reliability and maintainability and other work related to this article. Section 3 presents the methodology used in this paper, namely for the Pareto analysis and the reliability model. Section 4 presents briefly the data of OREDA that makes up the foundation for the analysis. Section 5 presents the results of the analysis including the outcome from Pareto and statistical reliability analysis. Some findings indicate critical points that reveal serious safety risks. Furthermore, mitigating strategies are developed based on the analyses. Section 6 concludes the work and gives an outlook for future work. 2. BACKGROUND 2.1 Key terms of reliability engineering A failure mode is the manner in which an equipment or a part of an equipment can fail. By listing all failure modes for all maintainable items in a system, a failure mode and effects analysis (FMEA) can be performed. In reliability engineering, faults in complex systems are often displayed by using the fault tree method. It reflects the structure of the equipment by representing the different subsystems of an asset down to individual components. For each component it is possible to identify different failure modes. Ultimately, all possible ways that a system can fail based on the individual components should be represented in such a fault tree. The fault tree breaks a complex system down into subsystems down to a level, where no further segregation is possible. A fault tree analysis is a form of deductive failure analysis. It uses a fault tree as a topdown approach to describe the system. The failure events at the lower level are then combined by Boolean logic and a undesired state of the system is explained. 2.2 The OREDA project A reliability analsis of assets such as compressors, requires a comprehensive amount of data, preferably from an entire compressor fleet. The project organization OREDA (Onshore & Offshore Reliability Data) has been set up to collect and exchange industrial reliability data in the oil and gas industry for the participating companies operating worldwide. The database resulting from the OREDA 218
Fig. 1. Example of a fault tree. It can be extended to a higher number of branches and also more levels. The numbered circles denote the events that cause failure. project comprises reliability and maintenance data for various types of equipment that are used in the exploration and production in various geographic areas, installations and operating conditions. There exists a wide range of literature based on the data from the OREDA project. 2.3 Work based on OREDA Turan et al. (2011) conducted a study of the reliability and criticality for the case study of diving support vessel. This is done with the background of the OREDA handbook and other maintenance standards and procedures, such as the ALARP and the KP3 initiative, or the basic reliabilitycentered maintenance (RCM) and risk-based inspection (RBI) principles. The reliability assessment of Turan et al. (2011) is done by utilizing two reliability importance measures. Similar to our approach, the reliabilities of the main system and the subsystems are identified and suggestions are made how to improve the overall reliability of the various systems. This work by Turan et al. (2011) is referenced since it works with the same database, looks into the subsystems and takes information to estimate the reliability. Langseth and Lindqvist (2006) chose a statistical approach to analyze a data set from the OREDA database. The model explains the different parts of the system that can fail (termed competing risks) and assumes that the preventive maintenance is imperfect and does not result in a total repair of the equipment. The authors conclude that the model has the capability to describe the data set that they used set as well. However, the authors state they are facing a non-identifiability of the competing risk problems and cannot claim to have found the correct model but rather that the data does not reject the developed model. The main outcomes of that work are therefore likelihood parameters (such as the eagerness and thoroughness of the maintenance crew) to describe the quality of the maintenance that has been performed. Rausand and Oien (1996) used OREDA to illustrate the interpretation of basic concepts in failure analysis using an example of a gate valve. During their analysis they point out limitations in OREDA and propose enhancements, e.g. to record failure causes, failure mechanisms or root causes. Until now, this information are still not collected
IFAC OOGP 2018 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
219
in OREDA. The reason for discussing this reference is that there are already improvements suggested to OREDA. This is also part of the outcome of our work. Sandtorv et al. (1996) share their practical experience of the OREDA project (e.g. that not always the personnel with the needed knowledge is available for data collection or that it is hard to agree on specifications for complex equipment). They make recommendations for future data collection projects. On the other hand, Hameed et al. (2011) named OREDA as the state of the art for data collection in the industry and they try to identify challenges regarding reliability and maintainability data collection for other industries, such as offshore wind turbines. In the failure analysis of Haugen et al. (1997) both critical failures and degraded failures, followed by critical failures, are considered. The OREDA project supplied the necessary data for this purpose. These paper demonstrate that OREDA is a useful database, and has been used successfully for research in areas such as failure analysis and reliability. 2.4 Work with a similar purpose Choi and Chang (2016) try to improve reliability of subsea production systems. They analyse the reliability of a seabed storage tank based on a fault tree. The data is taken from components of the lowest level of the seabed storage tank. Active repair times from OREDA were taken into account and other databases were used to supplement the study. This work is relevant, since their approach is based on a fault tree and shows similarities to the method described in this article. However, they merge failure rates from different systems. This work uses data which is specific to a compressor system. A synergetic approach is chosen by Ebrahimipour and Suzuki (2006). Principal component analysis is coupled with importance analysis to improve equipment performance. The leading causes and common-cause events are identified. The topic of this paper is to identify main causes for failures, similar to the work of our article, but with different methods. Another work with an relative approach for assesing the reliability of assets is given by Corvaro et al. (2017). The authors are using a RAM methodology (Reliability, Availability, and Maintainability) for the case of reciprocating compressors. The reliability is done with a RCM study from a private company. The availability calculation is performed by means of block diagrams. The input are failure rates, which is similar to the approach in the model developed by our work. The maintainability corresponds to maintenance activities suggested by the RCM analysis. These activities are assigned to criticality values that have been calculated before. Even though Corvaro et al. (2017) use RCM to build their model and the used equipment boundaries are not given by ISO 14224, the results of the work are comparable to the method presented in this paper. 3. METHODOLOGY The work presented in this paper uses statistical reliability analysis and Pareto analysis. Fig. 2 illustrates the methodology and how these two different analyses work together. The dashed boxes indicate the logical next steps, which are outside the scope of this paper. 219
Fig. 2. Methodology for the statistical reliability analysis; dashed boxes indicates work that goes beyond the scope of this paper 3.1 Selection of data from OREDA The taxonomy classification of ISO 14224:2016 for equipment used in the petroleum, natural gas, and petrochemical industry is used as well as the failure mode records for each component as described in the OREDA database. ISO 14224:2016 “Petroleum, petrochemical and natural gas industries – Collection and exchange of reliability and maintenance data for equipment” specifies the equipment to be included when studying a compressor system. Tab. 1 presents a list of compressor subsystems and the related equipment. 3.2 Pareto analysis The Pareto principle or the 80/20 rule, states that the majority (roughly 80%) of the effects come from a minor part (roughly 20%) of the causes. A tool called Pareto analysis is build on this principle. A Pareto analysis works by listing possible causes and ordering them, usually by their estimated benefit or by their impact on a problem. In the next step, a number of the most effective actions or most promising opportunities for countermeasures are selected. This is done by the Pareto principle, approaching 20% of the action items and expecting 80% of the benefit. In the case of OREDA, the information about the maintainable items and the failure modes must be sorted, based on the number of events when a specific maintainable item failed or when a specific failure mode occurred. Next to the contribution of the items the accumulated contribution in a descending order (from high contribution to low contribution) is indicated. The focus is on the action items that reach a cumulative contribution of 80%. A Pareto analysis has been conducted on the OREDA data (failure rates for each maintainable item and each failure mode). Focusing on the causes that
IFAC OOGP 2018 220 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
Fig. 3. Example of the Simulink model for the power transmission subunit; the calculation blocks are explained in Section 3.3. result in the majority of failures will help to improve the reliability of the system. 3.3 Statistical reliability model The statistical reliability model utilizes a fault tree analysis. The compressor subunits shown in Tab. 1 and all the maintainable items listed in the OREDA database are used to create a model. Based on the reliability data from the offshore compressors a fault tree is modeled in Simulink. This is done taking the six subsystems into account individually and all maintainable items for each subsystem. The common assumption that a failure is total and partial failures are not taken into account is applied (Baig et al., 2013). A failure in one of the subsystems leads to a failure of the entire compressor system. Reliability is the “ability of an item to perform a required function under given conditions for a given time interval” (ISO 14224:2016). The relevant information in the OREDA database is in a table in OREDA (SINTEF Technology and Society et al., 2015). This table gives failure rate λi due to each of the failure modes i, and fractions pi,j of the total failure rate for each combination of failure mode i and maintainable item j. The failure modes such as ELU (External Leakage - Utility Medium) are listed in Tab. 2, and the maintainable items are listed in Tab. 1. Fig. 3 illustrates the calculation procedure for the power transmission system, which is comprised of maintainable items such as Gearbox and Coupling to the driven unit. The calculation proceeds as follows: (1) The blocks labelled as 1, 2, 3 ... 14 calculate ri,j = e−λi pi,j t where t is time. For instance, in block 5, λi is the failure rate per hour for failure mode ELU, and pi,j is 0.001, being the fraction of failures for failure mode ELU reported in OREDA for the Gearbox 220
maintainable item. Since Gearbox has five possible failure modes, blocks 6 to 9 in Fig. 3 calculate ri,j values for the other four failure modes. (2) The ri,j values are inputs for calculations of reliability of the maintainable items. For instance, the ri,j values from blocks 5-9 go into Gearbox which calculates reliability of the Gearbox maintainable item over all its failure modes as rj = 1 − Πi (1 − ri,j ). (3) Similarly, the overall reliability r of the Power Transmission system is given by combining the reliability of the individual maintainable items in the Power Transmission system, i.e. r = 1 − Πj (1 − rj ).
Fig. 3 illustrates the Power Transmission system, which is one of the six subsystems of a compressor as listed in Tab. 1. Similar reliability calculations are carried out for all the subsystems. The results are combined to give an overall system reliability. 3.4 Steps beyond the scope of this work The development of mitigating strategies and the improvement of the system are not covered in this paper. However, it is intended that the results of this work are used as input for those steps. 4. CASE STUDY Currently, data from 278 installations, 17,000 equipment units with 39,000 failure and 73,000 maintenance records are stored in the database. The OREDA handbook described this database, currently in the sixth edition which has been published during the twelfth phase of the project. OREDA data being relevant to this study comprises data from from 22 offshore turbine-driven centrifugal compressors in eight different installations. Tab. 1 lists the six subsystems of compressors and all the maintainable items for each subsystem. Tab. 2 lists all the different failure modes that have been considered.
IFAC OOGP 2018 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
221
Table 1. Subdivison of a compressor into subunits and the maintainable items of each subsystem Power Transmission -Gearbox/Variable drive -Seals -Lubrication -Coupling to the driven unit -Coupling to the driver
Compressor Unit -Antisurge System -Casing -Cylinder liner -Instruments -Shaft seals -Radial Bearing -Thrust bearing -Interstage seals -Valves -Internal piping -Balance Piston -Packing -Rotor w/ impellers
Compressors Control and Lubrication SysMonitoring tem -Wiring -Reservoir w/ heat-Control unit ing system -Actuating device -Pump w/ motor -Monitoring -Filter -Internal power sup-Cooler ply -Valves and Piping -Seals -Oil -Piping -Instruments -Sensors -Seals -Valves
Shaft Seal System -Dry gas seal -Filters -Gear -Mechanical seal -Reservoir w/ heating system -Pump -Motor -Scrubber -Seal oil -Valves
Miscellaneous -Base frame -Flange joints -Control valves -Magnetic bearing control system -Isolation valves -Coolers -Piping, pipe support and bellows -Purge air -Silencers
Table 2. List of all considered failure modes for compressors and their acronyms Acronym AIR BRD ERO ELP ELU FTS STP HIO INL LOO SER NOI OTH OHE PDE UST STD UNK VIB
Failure Mode Abnormal instrument reading Breakdown Erratic output External leakage - Process medium External leakage - Utility medium Fail to start on demand Fail to stop on demand High output Internal leakage Low output Minor in-service problems Noise Other Overheating Parameter deviation Spurious stop Structural deficiency Unknown Vibration
5. RESULTS OF THE ANALYSIS AND DISCUSSION 5.1 Pareto analysis The Pareto analysis for the maintainable items (Fig. 4) reveals that the main contributing maintainable item is Unknown (The Unknown category will be further explained in section 5.4). Further, 10 other maintainable items contribute to 80% most frequently failing items. Similarly, the Pareto analysis for the failure modes (Fig. 5) gives the insight that the most common failure mode in the boundaries of the compressor system is abnormal instrument reading. This contributes for more than 30% of all failures. In addition, seven other items contribute to 80% of the total failures. The Pareto principle is in this case not exact with about 40% of the failure modes making up 80% of the failures. However, this can be explained by the small numbers of failure modes. The results show that maintenance engineers should take care if they are basing maintenance decision on the 80/20 rule, because the analysis of the OREDA shows that this rule could be misleading in some situations. The main message of the Pareto analysis is that abnormal instrument reading should be the main focus of further improvements. 221
Fig. 4. Pareto analysis of the failures for specific maintainable items; the orange line denotes the accumulated percentage if failures, the dashed region contains minor contributors to the failures that have been removed from the graph 5.2 Statistical Reliability Analysis Fig. 6 shows the results of the statistical reliability model. It gives an insight into the reliability of each sub-unit of the centrifugal gas compressor over time based on failure rates from the OREDA. It is evident over time, that the shaft seal system, the power transmission system and the lubrication system are impacting the reliability of the system very little. Their reliability value does not drop as much as other values do. This is mainly because the raw data in the database shows the failure rate for the shaft seal system and lubrication system are small.
IFAC OOGP 2018 222 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
The database reports external leakage of process medium as a failure mode that occurs in 6.25% of all failures. Leakage of process medium is unwanted in any case. This failure mode is the most frequent of the maintainable item category Unknown (2% of all failures). This means that there is a leakage, but it is unclear which maintainable part causes this leakage. Another critical failure mode is Fail to stop on demand. In 0.5% of the cases this is due to an Unknown maintainable item. From an external perspective the high number of Unknown that are documented seem to be an error in the data collection because with an investigation after the machine failure it should be possible to find out where e.g. a leak was. Since these two combinations are reported often, it should be a goal to reduce the amount of failures in the unknown category to support the industry. Fig. 5. Pareto analysis of the failures for specific failure modes; the orange line denotes the accumulated percentage if failures, the dashed region contains minor contributors to the failures that have been removed from the graph
Usage of results While the information obtained through the Pareto analysis is qualitative and suggests where to take actions and where to improve the existing systems, the statistical reliability model gives information about how the reliability of the specific sub-unit changes over time. Maintenance planning and scheduling requires such information for deciding what to do and when to take actions. Information is needed for background tasks such as ordering of spare parts, managing the inventory of spare parts, and getting the spare parts to where they are needed. The results in this paper can be useful for the following activities:
Fig. 6. Reliability of gas compressor subunits over time; the calculation blocks are explained in section 3.3. On the other hand, the Control and Monitoring sub-unit has the largest influence on reliability. The effect is by far larger than the reliability decrease of the sub-units power transmission, the compressor itself and miscellaneous other items in the compressor system. Also, the degradation process of the Control and Monitoring subunit starts faster. The overall reliability of the system reaches 80% after about 2.56 years. According to the OREDA data this would happen if there was no maintenance at all. 5.3 Knowledge about main issues for reliability During the analysis of the data from OREDA two observations were made that have a serious safety risk linked to them. 222
• Qualitative information from the Pareto analysis to identify the maintainable items that should become part of the maintenance plan • Quantitative information from the statistical reliability model such as the change of reliability over time within statistical boundaries can increase the priority for scheduling maintenance compared to an asset with a lower likelihood of degradation. • Based on the statistical reliability model, thresholds can be set. When the reliability reaches this thresholds, maintenance shall be performed for these items. This is useful for assets that are difficult to perform predictive maintenance on. The results in this paper are based on the OREDA data. However, companies operating large asset fleets typically also have their own databases. The findings in this paper suggest that it is very important to have accurate information about the different failure modes and maintainable items of an asset. The results also show that a large number of unknown faults reduces the usefulness of asset fleet data. It would be useful for further development of the methods to investigate reliability to reduce the numbers of faults categorized as unknown. A challenge in maintenance is often the absence of information. Condition monitoring data would help to predict failures several months before the asset fails. However, for long-term planning and even for scheduling with a time horizon of six months such information is not available.
IFAC OOGP 2018 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
The reason for these comments is that condition monitoring systems are becoming more prevalent, and they enable predictive maintenance. Predictive maintenance aims to predict machine failures even before the first signs are there, e.g. vibration or noise from the asset. With the statistical rating of the reliability it is possible to define thresholds and if the assumed reliability decreases below this threshold maintenance can be expected within the coming months. This is valuable information with the potential to improve existing maintenance planning systems. 5.4 Observations about the OREDA database The Unknown maintainable item is the biggest contributor to compressor failures within all maintainable items. The authors acknowledge that this category is important and often the only option to categorize something to. Nonetheless, from an operational point of view there are also a few cases with combinations listed that are not logical. The combination of the maintainable item unknown with a known failure mode (other ) means that there is knowledge about how something failed (otherwise the failure mode unknown would be the appropriate category) without knowing which item shows this failure mode. Also, the failure mode unknown appears very often. An example for another unlogical combination that is reported it the maintainable item is unknown, but the failure mode is an abnormal instrument reading. If it is known that a reading is abnormal, it must also be known where the reading comes from. In operations it might be difficult to track the history of equipment reliably. This has various reasons. The personal that is conducting this is changing over time. So the categorization of similar incidents might be deviating a lot regarding the categorization. Labels are based on the understanding of who reports a production loss or failure. This missing objectiveness arises also from the fact that there is different maintenance personnel in the different companies that contribute to OREDA. Even within one company it is very likely that failures are categorized by different personnel. The reason for highlighting these points is that the results of the conducted analysis are sensitive to every documented failure event. If an unknown error is documented or an error in an unknown maintainable item, no suggestions for future actions can be made and the statistical reliability model becomes less accurate. Therefore, a finding of the paper is that a company maintaining an asset fleet should take pains to reduce the number of unknowns in its database since this will help to improve future reliability. 6. CONCLUSION Compressor failure data from the OREDA database was used to conduct a reliability analysis and a Pareto analysis. These analyses help to understand the offshore compressors and their reliability better. The most prevalent failures have been identified. Main drivers for failure are the subunits control and monitoring and the power transmission, followed by unknown items. Abnormal instrument 223
223
reading is the most common failure mode. Main contributors can be identified and proposals can be made of how to cope with some of the challenges in daily operation of the equipment. Recommendations were made on how to improve data collection projects. The high number of unknown failures, with some of them not being logical, is critical and companies need to investigate if and why this is happening. However, the authors are aware of the challenges in the collection of the failure events and appreciate the way of improving reliability systems within a company by data collection. Aggregating data over the boundaries of a single company helps to collect data from a broader range of operating conditions and external factors such as climate. In the long run this will help the entire industry to become safer and more reliable which on the other hand will results in higher profit margins. The output of the reliability analysis is an important input for future developments of maintenance scheduling and planning tools, as the information about the decrease of reliability within a subunit is an important decision factor. Further study is recommended for the other identified parts of the compressor system that contribute to the most failures. Within the limitations of this work possible solutions for the control and monitoring systems were introduced. From the proposed methodology in Fig. 2 the improvement to the system need to be implemented. This will result in new operational data and the benefits of the changed system will become evident. Another point for further study is a more detailed assessment of the reliability. ACKNOWLEDGEMENTS Financial support is gratefully acknowledged from the Marie Sklodowska Curie Horizon 2020 EID-ITN project “PROcess NeTwork Optimization for efficient and sustainable operation of Europe’s process industries taking machinery condition and process performance into account – PRONTO”, Grant agreement No 675215. REFERENCES Baig, A.A., Ruzli, R., and Buang, A.B. (2013). Reliability analysis using fault tree analysis: A review. International Journal of Chemical Engineering and Applications, 4(3), 169–173. Beebe, R. (2010). Definitions, Concepts and Scope of Engineering Asset Management, volume 1. Springer London. Choi, I.H. and Chang, D. (2016). Reliability and availability assessment of seabed storage tanks using fault tree analysis. Ocean Engineering, 120, 1–14. Corvaro, F., Giacchetta, G., Marchetti, B., and Recanati, M. (2017). Reliability, Availability, Maintainability (RAM) study, on reciprocating compressors API 618. Petroleum, 3(2), 266–272. Ebrahimipour, V. and Suzuki, K. (2006). A synergetic approach for assessing and improving equipment performance in offshore industry based on dependability. Reliability Engineering and System Safety, 91(1), 10–19. Hameed, Z., Vatn, J., and Heggset, J. (2011). Challenges in the reliability and maintainability data collection for
IFAC OOGP 2018 224 Esbjerg, Denmark. May 30 - June 1, 2018F. Schulze Spüntrup et al. / IFAC PapersOnLine 51-8 (2018) 217–224
offshore wind turbines. Renewable Energy, 36(8), 2154– 2165. Haugen, K., Hokstad, P., and Sandtorv, H. (1997). The analysis of failure data in the presence of critical and degraded failures. Reliability Engineering & System Safety, 58(2), 97–107. Langseth, H. and Lindqvist, B.H. (2006). Competing risks for repairable systems: A data study. Journal of Statistical Planning and Inference, 136(5), 1687–1700. Rausand, M. and Oien, K. (1996). The basic concepts of failure analysis. Reliability Engineering and System Safety, 53, 73–83. Sandtorv, H.A., Hokstad, P., and Thompson, D.W. (1996). Practical experience with a data collection project: the OREDA project. Reliability Engineering & System Safety, 51(2), 159–167. Schulze Sp¨ untrup, F. and Imsland, L. (2018). Asset fleet management in the process industry - a conceptual model. In Submitted to: 10th Advanced Control of Chemical Processes - ADCHEM ’18. Shenyang. SINTEF Technology and Society, Universitet, N.T.N., and DNV GL (Firm) (2015). OREDA - Offshore Reliability Data Handbook 6th Edition. OREDA participants, 6th edition. Turan, O., Lazakis, I., Judah, S., and Incecik, A. (2011). Investigating the reliability and criticality of the maintenance characteristics of a diving support vessel. Quality and Reliability Engineering International, 27(7), 931– 946. World Economic Forum (2017). Digital Transformation Initiative: oil and gas industry (Whitepaper). URL http://reports.weforum. org/digital-transformation/wp-content/ blogs.dir/94/mp/files/pages/files/ dti-oil-and-gas-industry-white-paper.pdf.
224