Knowledge based engineering approach for subsea pipeline systems' FFR assessment: A fuzzy expert system

Abstract

Purpose

- The purpose of this paper is to focus on developing a knowledge-based engineering (KBE) approach to recycle the knowledge accrued in an industrial organization for the mitigation of unwanted events due to human error. The recycling of the accrued knowledge is vital in mitigating the variance present at different levels of engineering applications, evaluations and assessments in assuring systems' safety. The approach is illustrated in relation to subsea systems' functional failure risk (FFR) analysis.

Design/methodology/approach

- A fuzzy expert system (FES)-based approach has been proposed to facilitate FFR assessment and to make knowledge recycling possible via a rule base and membership functions (MFs). The MFs have been developed based on the experts' knowledge, data, information, and on their insights into the selected subsea system. The rule base has been developed to fulfill requirements and guidelines specified in DNV standard DNV-RP-F116 and NORSOK standard Z-008.

Findings

- It is possible to use the FES-based KBE approach to make FFR assessments of the equipment installed in a subsea system, focussing on potential functional failures and related consequences. It is possible to integrate the aforementioned approach in an engineering service provider's existing structured information management system or in the computerized maintenance management system (CMMS) available in an asset owner's industrial organization.

Research limitations/implications

- The FES-based KBE approach provides a consistent way to incorporate actual circumstances at the boundary of the input ranges or at the levels of linguistic data and risk categories. It minimizes the variations present in the assessments.

Originality/value

- The FES-based KBE approach has been demonstrated in relation to the requirements and guidelines specified in DNV standard DNV-RP-F116 and NORSOK standard Z-008. The suggested KBE-based FES that has been utilized for FFR assessment allows the relevant quantitative and qualitative data (or information) related to equipment installed in subsea systems to be employed in a coherent manner with less variability, while improving the quality of inspection and maintenance recommendations.

Introduction

Knowledge-based engineering (KBE) developments focus recycling the knowledge, which has previously been stored in the minds, charts, diagrams, etc., of the personnel (i.e. in terms of experience, insights, etc.) who have been extensively involved in engineering projects (Leake et al. , 2014; da Silva et al. , 2014; Stokes, 2001). For instance, risk assessment that has been used to identify and eliminate known or potential failures in order to enhance the reliability and safety of complex systems requires the recycling of an extensive amount of engineering knowledge (Gu et al. , 2012; Ratnayake, 2014a). However, KBE has not been adopted in most engineering applications, as currently there is no generic approach for structuring engineering knowledge and representing it in such a way that it is possible to use it in different KBE platforms (Zha and Sriram, 2006). In this context, KBE is defined as "the use of advanced software techniques to capture and re-use product and process knowledge in an integrated way" (Stokes, 2001; Leake et al. , 2014). The following actors are mainly involved with KBE approaches: experts (i.e. those who "are responsible for defining the domain knowledge to be applied in the KBE system by end-users" - e.g. engineers involved in assessment); knowledge engineers (i.e. those who "structure and formalize the expert's knowledge in a consistent and unambiguous format" and who are "familiar with the field of application" in order to describe the knowledge); developers (i.e. those who "transform the formalized knowledge into operational applications" and "write the code for a particular KBE platform" - e.g. programmers or trained analysts); end-users (i.e. those who will use the final KBE application to perform a specific assessment - e.g. newly recruited engineers who are already involved in assessment and evaluations) (Stokes, 2001).

The functional failure risk (FFR) assessments of offshore oil and gas (O & G) production and process systems are currently performed using established recommended practices (e.g. DNV-RP-F116 (2009); NORSOK standard Z-008, Z008, 2011), experts' knowledge (i.e. by means of experience, insights, data, and information) (Seneviratne and Ratnayake, 2012; Ratnayake, 2014a). However, the risk assessments show significant variation from one to the other due to "uncertainty" that has been caused by: fuzziness (i.e. "lack of definite or sharp distinctions"); ambiguity (i.e. "one-to-many relationships"); discord (i.e. "disagreement in choosing among several alternatives"); and nonspecificity (i.e. "two or more alternatives are left unspecified") (Klir and Yuan, 1995). Hence, "expert" or "knowledge based" systems' development is vital to recycle established knowledge to perform risk assessments that would normally require a well-experienced human expert (Pillay and Wang, 2003; Eilouti, 2009). However, the aforementioned approach does not seek to replace the human involvement; instead, the aim is to replace the routine assessment processes for which the knowledge is well established and understood (Li et al. , 2011). For instance, these assessment processes take significant time (or effort) and yet require very little creative thought to make further improvements. Hence, the aim of KBE developments is to let the computers take on the repetitive (or dull) routines to reduce the variability that may be present in the final assessment due to diverse expertise (Gu et al. , 2012).

In addition to the aforementioned, currently there is a serious concern in the O & G industry that knowledge existing or accrued for assessment processes diminish over the time as a result of knowledge migration from one organization to another (Ratnayake, 2014b). This is mainly due to the fact that there is no formal mechanism for experienced engineers in the workforce to pass on much hard-earned knowledge (i.e. in terms of experience and insights) to the new recruits who are less experienced (Maruta, 2014). However, KBE is not expected to replace suitably qualified and experienced engineers; instead, it does provide a mechanism to hold and recycle any available knowledge (Maruta, 2014). Moreover, it demands a suitable means of representing the accrued knowledge gained from different projects, supporting new recruits and maintaining the quality of deliverables at an anticipated level with less variability.

The KBE concept routes back to the 1950s (Sandberg, 2003), with the developments during this time mainly focussing on enabling a system to have its own intelligence, known as artificial intelligence (AI). The AI techniques are mainly used to devise adaptive solving strategies that are employed to solve a broad spectrum of tasks (Pota et al. , 2014; Ciarapica and Giacchetta, 2009). However, these approaches have been put aside for a while due to the challenge of the computational complexity vs the time taken for a human to solve it. However, the concept of KBE focusses on the use of advanced software techniques to capture and recycle the knowledge in an integrated way and resurrects the possibility of using AI approaches with a human touch (Stokes, 2001; Lovett et al. , 2000). Aside from that, it focusses on minimizing routine assessment work to about 20 percent, while allowing creative work to expand to about 80 percent, reducing the overall cycle time (Mason, 2014; Stokes, 2001). For instance, membership functions (MFs), together with a rule base in the fuzzy set theory approach, enable the integration and recycling of existing knowledge that has been accrued over the quantitative and qualitative data, as well as personnel experience and skills (Pota et al. , 2014; Ratnayake, 2013; Ciarapica and Giacchetta, 2009). Hence, this manuscript illustrates the KBE development via the use of a fuzzy logic-based consistent approach, making the criticality assessment of subsea systems align with the guidelines provided in NORSOK standard Z-008 (Z008, 2011). The approach is developed to perform an initial screening of subsea equipment criticality assessment. Subsequently, it is possible to use the results of such assessments for carrying out detailed assessment.

Background for the KBE development

The safety regulations applying to petroleum activities (i.e. O & G) on the Norwegian Continental Shelf suggest classifying systems and equipment on production and/or process systems in relation to the health, safety and environmental (HSE) consequences of potential functional failures (Ptil, 2011). In this context, the responsibility for carrying out the consequence classification (i.e. based on the potential functional failures in O & G production/process equipment that can lead to serious consequences) assessment lies with the particular petroleum production and/or process facility owner's industrial organization (Ratnayake, 2014a; Wang, 2001). In essence, the asset owner's industrial organization outsources the assessment work to an engineering contractor organization which has a track record of providing expertise for performing such assessments (Ratnayake, 2012a). Usually, the assessment is performed with the help of existing data (e.g. OREDA (2009) - Offshore REliability Data) as well as experts' experience, knowledge, and insights. The assessment focusses on identifying the various fault modes with associated failure causes and failure mechanisms, estimating the probability of failure (PoF) for the individual fault mode, and assessing the potential consequence of failure (CoF).

Based on the PoF and CoF of potential failures, the initial screening (or classification) of the facilities' systems and equipment is performed (Ratnayake, 2014b; Harms-Ringdahl, 2003; Hale et al. , 1997; Ciarapica and Giacchetta, 2009). In this context, "a failure is the termination of the ability of an item to perform its required function" (DNV-RP-F116, 2009). The "failure" is "an event affecting a component or system and causing one or both of the following events: loss of component or system function; or deterioration of functional capability to such an extent that the safety of the installation, personnel or environment is significantly reduced" (DNV-OS-F101, 2012). The main focus of the initial screening (or classification) is to use it as a basis for selecting maintenance activities and maintenance frequencies, for prioritizing between different maintenance activities, for evaluating the need for resources (e.g. remotely operated vehicles (ROVs), nondestructive evaluation (NDE) tools, etc.) and finally, to carry out a detailed assessment (i.e. to recognize the need for inspection, modification, repair, etc.) for the equipment with high-risk ranking. Figure 1 illustrates the overall process involved in subsea systems' FFR assessment.

However, current practices reveal that the inconsistency (or variance) among evaluations (e.g. in assessment, planning, etc.) is significantly high (Ratnayake, 2014b). Hence, it is vital to develop a consistent approach to improve systems' safety and minimize waste (in terms of quality loss, time, etc.) (Diana, 1995). Standards such as DNV-RP-F116 (2009), DNV-OS-F101 (2012) and NORSOK standard Z-008 provide requirements and guidelines for risk-based inspection and risk-based maintenance (RBM). The standards provide directions for performing a classification of consequences due to potential failures and alternatively, revising (or recommending) necessary maintenance activities for plant systems and equipment in the Norwegian petroleum industry (Z-008, 2011). In particular, NORSOK Z-008 covers the aforementioned in the design phase, in the preparation for operations, and in the operational phase of offshore topside, subsea production and O & G terminals. The standard uses risk assessment as the guiding principle for maintenance decisions. It is suggested that the RBM decisions are carried out against defined criteria and that the criteria are in accordance with the selected asset owner's overall policy for the minimization of HSE, production, and cost-related challenges (Ptil, 2011). The current practice is to develop a risk matrix (i.e. aligned with NORSOK standard Z-008 guidelines) along with possible ranges or linguistic terms and then to perform risk assessments. As there is no means to study the boundary of each range or level of a linguistic term, engineering practice reveals final risk assessments to be suboptimal (Ratnayake, 2014a). Hence, it is vital to introduce KBE approaches to the existing evaluation processes (Chapman and Pinfold, 1999).

The role of human errors vs equipment failures

The Piper Alpha disaster almost 26 years ago and the Deepwater Horizon catastrophe four years ago reveal that there is no doubt that safety assessment and systems remain top priority (Ratnayake, 2012b; Ratnayake and Markeset, 2012). As the continuous technology evolution makes systems and operations more sophisticated, dangerous and complex, it is vital to cut back on the weakest link (i.e. people) in the safety chain (Hale, 2014). The more autonomation (i.e. letting machines work harmoniously with their operators by giving them the "human touch") there is in assessments (e.g. the use of KBE approaches) and systems operations, the more help there is to cut back the human involvement, eliminating hazards, while letting people remain involved in safety systems and in running the process (Aziz and Hafez, 2013; Hale, 2014). The recent comprehensive investigations into unwanted incidents revealed that: minimal awareness, equipment errors, human errors, risk management, organizational weaknesses, working culture at the site, inspection and maintenance, general health and safety assessments, etc. led the particular operational asset(s) to a catastrophic incident (Ratnayake, 2011). Moreover, Lardner and Fleming (1999) revealed that as a rule of thumb, 80 percent of large-scale accidents and disasters are due to a combination of both human and organizational causes, while only 20 percent of accidents are due to technical causes. DOE Standard (2009) reveals that only about 30 percent of accidents are contributed to by individual mistakes and about 70 percent of them have been caused by organizational weaknesses. Hence, considering the combination of the two aforementioned findings, it is possible to estimate that more than half (56 percent) of unwanted events are caused by organizational weaknesses leading to human errors, and only a quarter (24 percent) of unwanted events are caused by individual mistakes leading to human errors (see Figure 2).

Hence, organizations need to take measures to prevent accidents rather than merely focus on technical challenges related to failures. For instance, the investigation report about the Hercules military flight which crashed onto a mountainside in northern Sweden, killing all five officers on board, revealed that "poor routines in planning," staff being "relatively new on the job and inexperienced," "letting employees with limited experience have responsibility for considerable traffic" are some of the facts which led to the accident (Newsinenglish, 2013). The current O & G industry also suffers from frequent knowledge migration, leaving new inexperienced personnel to carry out assessment and planning and make recommendations, etc. Hence, it is vital to retain the experts' knowledge as much as possible within an organization to assure their processes are at the expected level of system function. In this context, it is possible to mitigate the effect of knowledge migration to a certain extent by incorporating KBE. Figure 3 illustrates how KBE supports continuous improvement (i.e. over the improvements made in an isolated fashion).

If an expert leaves an organization, the concept of KBE development enables his/her knowledge to be incorporated systematically into a mathematical model, which can effectively and efficiently be implemented with the help of software development. The developed software together with the suggested model can later be integrated with an existing structured data management system. The aforementioned approach supports performing inspection and maintenance assessment with less variability (i.e. with less variation from person to person), whether the personnel involved in the assessment are newly recruited or experienced.

Industrial challenge

Standards such as DNV-RP-F116 (2009), DNV-OS-F101 (2012) and NORSOK standard Z-008 provide requirements and guidelines for constructing a tailor-made risk matrix (i.e. an O & G assets owner has freedom to adapt the guidelines to fit into the organizational risk philosophy while satisfying the minimum requirements specified in the standard(s)) to carry out FFR assessment. For instance, Table I illustrates the DNV-RP-F116 (2009) approach.

However, when the assessment of FFR due to potential failures is carried out, there is no formal mechanism to incorporate data and information at the boundaries of the risk categories (i.e. alternatively at the boundaries of the ranges and levels of linguistic data). Figure 4 illustrates the general work process, which outlines systematically the breakdown of plant systems into suitable items for FFR assessment.

Major difficulties lie especially in classifying consequences along the boundary from high (H) to medium (M) and from medium to low (L), as there are no means to incorporate real data (qualitative or quantitative). For instance, along a boundary, the sudden jumps of risk classification hinder the actual level of consequence, which leads to suboptimal RBM decisions. Figure 5 illustrates an example of how equipment in a main function is assigned to standard sub-functions.

Essentially, different items of equipment or machinery (i.e. called tags) are evaluated to assess possible FFR (IEC 61508, 2000). The FFR assessment of each item of equipment in systems and sub-systems supports prioritizing the maintenance requirements (e.g. interval, spare-parts, etc.). In this context, equipment is designated with a "tag." In essence, the tag coding (or tagging) has been used to "equip an item function with a label that gives it a unique identification" (Z-DP-002, 1996). Each "tag" is evaluated based on the consequence categories and possible functional failure frequency using a tailor-made matrix (as illustrated in Table I).

The data and information for performing FFR assessments are usually gathered via expert knowledge, documentation (e.g. piping and instrumentation diagram, process flow diagram, historical data, vendor's recommendations, etc.), guidelines (e.g. DNV-RP-F116, DNV-OS-F101, NORSOK Z-008, ISO 14224, API RP 14C, etc.) and regulatory requirements (e.g. Petroleum Safety Authority (PSA) activity regulations sections 45, 46, 47 and 48), when establishing maintenance programs for new plants or updating the existing maintenance programs (Ptil, 2011). All the aforementioned are equally important to an O & G asset owner's organization as well as to the engineering service provider organization in which the FFR assessment has been performed.

Basically, input ranges to perform FFR assessments are established in a form of a matrix (i.e. also referred to as a risk matrix). In an FFR assessment, the common practice is to use qualitative data or discrete scales of input ranges. The aforementioned cause uncertainty, especially, when an assessment has to be made based on an input value (qualitative or quantitative) at the boundary of a range (Ratnayake, 2014b). This causes significant variation among the assessments made by personnel who have diverse levels of experience. This has further been exacerbated when the assessments are compelled to be made on an ad hoc basis due to the lack of a consistent mechanism. Hence, it is vital to develop KBE approaches to recycle the anticipated knowledge in a consistent manner.

Methodology

The use of FFR assessment in RBM planning in a subsea system is employed to illustrate the fuzzy expert system (FES)-based KBE approach. The FFR assessment guidelines in DNV-RP-F116 and NORSOK Z-008 have been selected to illustrate possible knowledge recycling in assuring subsea systems' functional performance (DNV-RP-F116, 2009; Z-008, 2011). The guidelines regarding the activities' regulations, section 46 of the PSA Norway, request the use of NORSOK standard Z-008 for FFR assessments in mitigating the challenges in health, working environment and safety of the O & G operational (or newly built) assets (Ptil, 2012). However, as NORSOK standard Z-008 has mostly focussed on topside rotating equipment, DNV-RP-F116 standard has been utilized in conjunction with it. In this context, NORSOK standard Z-008 provides requirements and guidelines for performing RBM and FFR assessments for plant systems and equipment in the Norwegian petroleum industry (Z-008, 2011). Moreover, DNV-RP-F116 provides guidelines for operating subsea pipeline systems safely and without "loss of component or system function" in such a way that "deterioration of functional capability to such extent that the safety of the installation, personnel or environment is significantly reduced" is avoided (DNV-RP-F116, 2009). Hence, DNV-RP-F116 guidelines have been utilized to develop the risk matrix which has later been used as the rule base. The knowledge base has been developed with the help of MFs and assessment rules (or simply a rule base).

FES

Recently, Pillay and Wang (2003) proposed modified knowledge (i.e. using experts' knowledge) based failure mode and effects analysis for estimating the risk. In this context, FESs, which are based on fuzzy logic, play a significant role. The fuzzy logic provides a form of a logic in which the variables can have degrees of truthfulness or falsehood represented by a range of values between 1 (true) and 0 (false), enabling the outcome of an operation to be expressed as a probability rather than as a certainty (Bai et al. , 2014). Consequently, fuzzy logic allows the development of knowledge-based systems by descriptive or qualitative representation of expressions such as "very low" or "very high," while incorporating symbolic statements that are more natural and intuitive than mathematical equations (Castro-Schez et al. , 2013). Hence, it is possible to use the direct opinions of multiple experts (i.e. based on a probabilistic interpretation of MFs) for aggregating the opinions of individual experts.

An FES consists of a rule base and MFs. The rule base comprises a collection of fuzzy IF-THEN rules, which are utilized by the fuzzy inference engine to determine a mapping from fuzzy sets in the input universe of discourse U [subset or is implied by]R n to fuzzy sets in the output universe of discourse V [subset or is implied by]R , based on fuzzy logic principles. The fuzzy IF-THEN rules have the form as follows:

(Equation 1)

where (Equation 2) and G j are fuzzy sets, x =(x ₁ , x ₂ , ... , x _n )T [epsilon]U and y7[epsilon]V are input and output linguistic variables which belong to the input and output universes, respectively, and j =1, 2, ... , m . The practical experience reveals that these fuzzy IF-THEN rules provide a convenient framework to incorporate human experts' knowledge. In Equation (1) each fuzzy IF-THEN rule defines fuzzy set (Equation 3), (Equation 4) ... , (Equation 5) [> or =, slanted] G j for i= 1, 2, ... , n , in the product space U ×V . Experts' opinions and data/information retrieved from different sources are taken into the mathematical model using the aforementioned rules. The main focus is to enhance the discriminating power in the RBM decision making process while associating the uncertainties related to the linguistic variables to the degree of criticality at the boundaries such as high to medium, medium to low, and so on. The rules also allow quantitative (e.g. personal safety (PS), PoF, environmental degradation (ED), etc.), qualitative and judgmental data (e.g. personnel safety) to be integrated in a uniform manner (Bowles and Pelaez, 1995; Guimaraes and Lapa, 2004).

In order to use an FES in engineering systems it is necessary to add a fuzzifier to the input and a defuzzifier to the output of the FES. The fuzzifier maps crisp points in U to fuzzy sets in U , and the defuzzifier maps fuzzy sets in V to crisp points in V . The fuzzy rule base and fuzzy inference engine are the same as those in the pure fuzzy logic system. In 1975, Mamdani built one of the first fuzzy systems which used a set of fuzzy rules supplied by experienced human operators to control a steam engine and boiler combination (Mamdani and Assilian, 1975). To date, the Mamdani approach has been successfully applied to a variety of industrial processes and consumer products (Wang, 1993).

PS, ED, production loss (PL), cost of subsea intervention (i.e. subsea invention with ROVs, existing drilling units or dedicated subsea intervention vessels), maintenance and repair activities (CSIM & R), and PoF have been selected as the input variables and FFR as the output variable. Figure 6 illustrates the overall view of the proposed fuzzy criticality assessment system (Ratnayake, 2014a).

The inherent challenges related to risk assessments are as follows: "uncertainty" that has been caused by: fuzziness (i.e. "lack of definite or sharp distinctions" due to "vagueness, cloudiness, haziness, unclearness, indistinctness, and sharpness"); ambiguity (i.e. "one to many relationships"); discord (i.e. "disagreement in choosing among several alternatives" due to "dissonance, incongruity, discrepancy, and conflict") and nonspecificity (i.e. "two or more alternatives are left unspecified" due to "variety, generality, diversity, equivocation, and imprecision") (Klir and Yuan, 1995). However, fuzzification as well as the development of MFs and a rule base with the help of experts' knowledge enables the aforementioned to be mitigated to a significant level.

The input and output variables shall consist of quantitative, qualitative and judgmental (i.e. linguistic) data. Using an appropriate MF, the user has "more confidence" that the input parameter lies in the center of the interval than at the edges. In this study, the author has incorporated Gaussian MFs (Tay and Lim, 2008), which are defined by Equation (2):

(Equation 6)

where c represents the center and [sigma] determines the width of the MFs. To model the MFs, the Gaussian combination MF (GCMF) (i.e. "gauss2mf"), which is available in MATLAB (R2012b), has been utilized (Mathworks, 2014). The function "gauss2mf" is a combination of two parameters (i.e. (c , [sigma] )) indicated in Equation (2). It follows the following syntax (Mathworks, 2014):

(Equation 7)

The first part of the function of the GCMF is specified by [sigma] ₁ and c ₁ which determines the shape of the left-most curve. The second part of the GCMF, specified by [sigma] ₂ and c ₂ , determines the shape of the right-most curve. Whenever c ₁ [< or =, slant]c ₂ , the "gauss2mf" function reaches a maximum value of 1. Otherwise, the maximum value is less than 1. The order of the parameters is as follows: ([sigma] ₁ c ₁ [sigma] ₂ c ₂ ) (Mathworks, 2014). Moreover, the other parameters of the FES that have been selected for the current analysis are as follows: "AND" operator with "minimum," "OR" operator with "maximum," "Implication" with "minimum," "Aggregation" with "maximum" and "Defuzzification" with "centroid" algorithm. A fuzzy rule base has been developed using the table-look-up approach (see Table III) to align with guidelines provided in DNV-RP-F116 (2009) and NORSOK standard Z-008 (2011). The toolbox simulator tool of MATLAB (R2012b) has been utilized to execute the suggested FES (MATLAB, 2012).

Case study, data collection, modeling, assessment and results

An illustrative case study was carried out in collaboration with an engineering contractor company which provides maintenance support services to an operator company. Hence, the existing RBM assessment process has been selected to illustrate the risk assessment approach.

MF selection

The consequence (i.e. CoF) and probability (i.e. PoF) of a functional failure have been selected as input to the FES. The FFR has been assigned as the output. There are three factors under each functional failure consequence rank. The highest value (among the factors) of the consequence due to a particular functional failure mode is selected for assessing the gravity of the consequence of functional failure. However, in the current case study, if two factors have been assessed to be equal in level of consequence, the industrial organization uses the following hierarchy: (1). PS; (4). ED; (3). PL; (2). CSIM & R. The intervals, corresponding membership values and finally MFs (i.e. for Gaussian MFs: "c and [sigma] ") were established based on experts' knowledge (i.e. based on experience and insights), company documentation, historical data, literature and the author's own experience.

Data collection

The CSIM & R activities are inherently higher for subsea systems than is the case for topside systems. The cost is often driven by the duration of the time taken to fulfill the challenge and the need of a dedicated subsea intervention vessel or rig. In this context, the sophistication and the size of the vessel or rig also play a significant role. Hence, based on the level of intervention, CoF has been categorized in relation to the cost of the type of intervention needed. Also, it is possible the impact of subsea systems' failure to be high on production, depending on the geographical location (e.g. North Sea or Barents Sea) as a result of the time taken to mobilize intervention vessels and the sophistication required for the ROV and related inspections in order to plan and execute intervention activities. As a rule of thumb, most interventions related to subsea systems can take approximately 12-19 hours or more (Zijderveld et al. , 2012).

Firstly, the ranges, ranks (i.e. for PS) and corresponding consequence severity levels have been estimated with the help of the asset owner's documentation, standards, asset owner's previous findings, data from similar kinds of applications (e.g. OREDA), and experienced personnel who have had extensive experience in subsea systems' FFR assessment and establishing or updating a maintenance program. The subsea intervention activities depend on the various depth ranges (e.g. shallow: (0-500 m); deep: (500-1,500 m); ultra-deep: > 1,500 m) and level of repair required (Zijderveld et al. , 2012). In essence, subsea intervention is required to execute the inspection and maintenance activities, when production is interrupted, and to increase the extraction rate (Zijderveld et al. , 2012). Moreover, Infield's (2009) report revealed that the demand on vessel days in relation to subsea intervention has a steady increase with medium to heavy interventions which are roughly accountable for half of the total subsea intervention requirements. Taking all the aforementioned factors into account, a tailor-made risk matrix has been developed considering the criteria indicated in DNV-RP-F116 (2009), while aligning with NORSOK Z-008 (2011) requirements. Table III illustrates risk ranking and inspection frequency in relation to different risk categories. Table IV illustrates ranks, linguistic terms, levels and ranges assigned for possible CoF (i.e. due to functional failures) and PoF (Tables II and III).

Modeling

The input variables PoF together with one of the CoFs (i.e. CSIM & R) have been utilized to illustrate the methodology. Essentially, CSIM & R plays significant role in subsea maintenance operations and consequently in making FFR assessments. Table IV illustrates the parameters of each Gaussian MF (i.e. for ED, MTBF and FFR).

Fuzzification is vital in analyzing the inputs (i.e. ranges estimated for CoFs and PoF) and outputs (i.e. FFR) close to and beyond the boundaries of different levels in making optimal subsea systems' FFR assessments. It enables suboptimal inspection and maintenance recommendations to be minimized. However, in this context, the asset owner's organizational risk philosophy influences the defining of ranges and possible membership values along the different ranges (i.e. CoF and PoF). To establish an MF plot, the author's own experience, experts' views (i.e. how they perceive and experience the influence of different systems and connected equipment on a functional failure), as well as data and information from the case study asset owner's organization and other existing sources (e.g. OREDA) have been utilized. In this case, it means that the way in which equipment and different instruments are physically connected or affect each other's operation, PoF and CoF, have been taken into consideration. Figure 7 illustrates the MF plots of CSIM & R.

Figure 8 illustrates MF plots of PoF.

Figure 9 illustrates MFs for possible FFR scenarios and corresponding ranks.

Figure 10 illustrates a rule view and example calculation of the FFR rank for selected subsea equipment (tag). The calculation has been carried out for PoF=0.00316 and CSIM & R=3.16 million euro. The suggested FES-based KBE approach calculated the FFR rank to be 17.4.

The corresponding risk level of the potential failure(s) is H (using the MFs in Figure 9). The aforementioned linguistic value along with inspection and maintenance recommendation (i.e. yearly inspection is necessary (see Table IV)) would be recorded in the final assessment report and, finally, the computer maintenance management system (CMMS) available in the asset owner's organization is updated to include the necessary RBM activities. Also, these analyses support the performance of reliability centered maintenance assessment at a later stage of a subsea system's life cycle. Table V illustrates the summary of overall FFR assessment and corresponding inspection and maintenance recommendations.

It is possible to perform similar analyses for different equipment (or tags) in a subsea system based on relevant CoF and PoF combinations. The final FFR for a certain piece of equipment (or tag) is the highest FFR (i.e. FFR=f (PoF, CoF)) that has been calculated among all possible combinations of PoF and relevant CoFs.

Discussion

Once the MFs and rules for FFR assessments have been established, then it is possible to perform the assessments in the same way based on the relevant PoFs and CoFs. The variation due to personnel experience is minimized, as the same MFs and rules are used in a selected system. This alternatively improves the quality of the FFR assessments and supports the consistency of the maintenance recommendations. As a result, the variation among different maintenance programs will be minimized. The possibility of mapping FFR three dimensionally (3D) with respect to two input variables (i.e. FFR vs PoF and one of the CoFs) enables a sensitivity analysis to be carried out. For instance, Figure 11 illustrates a 3D plot of FFR vs PoF and CSIM & R.

The 3D plot also enables an assessment of the consistency of the rules used for the assessments by examining a plot of the FFR surface over the possible combinations of the input variables. For instance, Figure 10 reveals that there are no significant inconsistencies, as there are no evident abrupt changes in the FFR for a small change in the PoF or CSIM & R. Similarly, it is also possible to model other combinations of recycling the accrued knowledge. Although Gaussian MFs have been employed in the current study, the use of triangular MFs has also been reported in numerous studies (see Pedrycz, 1994; Klim, 2004).

Conclusion

This manuscript employs a software tools-oriented KBE approach that has been developed to recycle knowledge via an FES. The importance of knowledge recycling has been explained in relation to FFR assessment. After that, the possibility of implementing such recycling is illustrated with the help of software tool Matlab (2012) (i.e. fuzzification via MFs, development of rules to perform FFR assessment and defuzzification to get the results). An illustrative case has been performed in relation to the FFR assessment of subsea pipeline systems. The suggested KBE approach minimizes the variation (i.e. by minimizing uncertainties that may cause assessments to be performed) and waste (i.e. in terms of time and other resources) in subsea systems' FFR assessment and making recommendations. It also enables the weakest link (i.e. human involvement) in safety assessments to be reduced while allowing "human touch" for further improvements with relevant experts. In particular, the aforementioned variation, waste and suboptimal assessments occur as a result of the inexperienced and unsuspecting personnel who may be involved in FFR assessments. Moreover, it is also possible to integrate the suggested approach in an existing structured information management system in an engineering services providing organization or in a CMMS in an asset owner's industrial organization.

Further research should be carried out to investigate the possibility of a dynamic approach to MFs' development to incorporate the conditions (or parameters) that should change over time in operating systems.

Figure 1 Overall process involved in subsea systems' FFR assessment

[Image omitted: See PDF]

Figure 2 Percentage contribution of organizational weaknesses, equipment failures and individual mistakes to unwanted events

[Image omitted: See PDF]

Figure 3 System function vs time: role of KBE

[Image omitted: See PDF]

Figure 4 FFR work process

[Image omitted: See PDF]

Figure 5 Illustration of equipment main function(s) and sub-functions

[Image omitted: See PDF]

Figure 6 KBE development: fuzzy FFR assessment system

[Image omitted: See PDF]

Figure 7 MF plots of CSIM&R

[Image omitted: See PDF]

Figure 8 MF plots of PoF

[Image omitted: See PDF]

Figure 9 MF plots of FFR

[Image omitted: See PDF]

Figure 10 A rule view and an assessment result

[Image omitted: See PDF]

Figure 11 3D plot of FFR vs PoF and CSIM&R

[Image omitted: See PDF]

Table I Risk matrix

[Image omitted: See PDF]

Table II Risk ranking and inspection frequencies in relation to different risk categories

[Image omitted: See PDF]

Table III Tailor-made rule base for FFR assessment

[Image omitted: See PDF]

Table IV Gaussian MF parameters for input and output variables

[Image omitted: See PDF]

Table V Use of FFR assessments for making inspection and maintenance recommendations

[Image omitted: See PDF]

Equation 1

[Image omitted: See PDF]

Equation 2

[Image omitted: See PDF]

Equation 3

[Image omitted: See PDF]

Equation 4

[Image omitted: See PDF]

Equation 5

[Image omitted: See PDF]

Equation 6

[Image omitted: See PDF]

Equation 7

[Image omitted: See PDF]

Corresponding author

Dr R.M. Chandima Ratnayake can be contacted at: [email protected]

References

References

Aziz, R.F. and Hafez, S.M. ( 2013 ), " Applying lean thinking in construction and performance improvement ",Alexandria Engineering Journal , Vol. 52 No. 4, pp. 679 - 695 .

Bai, H., Ge, Y., Wang, J., Li, D., Liao, Y. and Zheng, X. ( 2014 ), " A method for extracting rules from spatial data based on rough fuzzy sets ",Knowledge-Based Systems , Vol. 57, February, pp. 28 - 40 .

Bowles, J.B. and Pelaez, C.E. ( 1995 ), " Fuzzy logic prioritization of failures in a system failure mode, effects and criticality analysis ",Reliability Engineering & System Safety , Vol. 50 No. 2, pp. 203 - 213 .

Castro-Schez, J.J., Murillo, J.M., Miguel, R. and Luo, X. ( 2013 ), " Knowledge acquisition based on learning of maximal structure fuzzy rules ",Knowledge-Based Systems , Vol. 44, May, pp. 112 - 120 .

Chapman, C.B. and Pinfold, M. ( 1999 ), " Design engineering - a need to rethink the solution using knowledge based engineering ",Knowledge-Based Systems , Vol. 12 Nos 5-6, pp. 257 - 267 .

Ciarapica, F.E. and Giacchetta, G. ( 2009 ), " Classification and prediction of occupational injury risk using soft computing techniques: an Italian study ",Safety Science , Vol. 47 No. 1, pp. 36 - 49 .

da Silva, J.C., Matelli, J.A. and Edson Bazzo, E. ( 2014 ), " Development of a knowledge-based system for cogeneration plant design: verification, validation and lessons learned ",Knowledge-Based Systems , Vol. 67, September, pp. 230 - 243, available at: http://dx.doi.org/10.1016/j.knosys.2014.05.002

Diana, W. ( 1995 ), " Application of systems thinking to risk management: a review of the literature ",Management Decision , Vol. 33 No. 10, pp. 35 - 45 .

DNV-OS-F101 ( 2012 ),Submarine Pipeline Systems , Offshore Standard, Det Norske Veritas AS, available at: https://exchange.dnv.com/publishing/codes/download.asp?url=2012-08/os-f101.pdf (accessed June 20, 2014).

DNV-RP-F116 ( 2009 ),Integrity Management of Submarine Pipeline Systems , Det Norske Veritas Offshore Code, available at: https://exchange.dnv.com/publishing/codes/download.asp?url=2009-10/rp-f116.pdf (accessed June 20, 2014).

DOE Standard ( 2009 ),Human Performance Improvement Handbook Volume 1: Concepts and Principles , US Department of Energy AREA HFAC, Washington, DC, available at: www.hss.doe.gov/nuclearsafety/ns/techstds/standard/hdbk1028/doe-hdbk-1028-2009_volume1.pdf (accessed August 23, 2009).

Eilouti, B.H. ( 2009 ), " Design knowledge recycling using precedent-based analysis and synthesis models ",Design Studies , Vol. 30 No. 4, pp. 340 - 368 .

Gu, D-X., Liang, C-Y., Bichindaritz, I., Zuo, C-R. and Wang, J. ( 2012 ), " A case-based knowledge system for safety evaluation decision making of thermal power plants ",Knowledge-Based Systems , Vol. 26, February, pp. 185 - 195 .

Guimaraes, A.C.F. and Lapa, C.M.F. ( 2004 ), " Effects analysis fuzzy inference system in nuclear problems using approximate reasoning ",Annals of Nuclear Energy , Vol. 31 No. 1, pp. 107 - 115 .

Hale, A.R., Heming, B.H.J., Carthey, J. and Kirwan, B. ( 1997 ), " Modelling of safety management systems ",Safety Science , Vol. 26 Nos 1-2, pp. 121 - 140 .

Hale, G. ( 2014 ), "Safety automation reduces hazard levels", OE Digital, available at: www.oedigital.com (accessed July 10, 2014).

Harms-Ringdahl, L. ( 2003 ), " Assessing safety functions - results from a case study at an industrial workplace ",Safety Science , Vol. 41 No. 8, pp. 701 - 720 .

IEC 61508 ( 2000 ),International Standard IEC 61508: Functional Safety of Electrical/Electronic/Programmable Electronic Safety Related Systems , International Electrotechnical Commission (IEC), Geneva, available at: www.iec.ch (accessed June 20, 2014).

Infield ( 2009 ), "Subsea well intervention market update report to 2014", available at: www.infield.com/demodisk/pdf/well_intervention.pdf (accessed June 20, 2014).

Klim, Z.H. ( 2004 ), " Preliminary hazard analysis for the design alternatives based on fuzzy methodology ",Fuzzy Information, Processing NAFIPS '04 , ISBN: 0-7803-8376-1, Vol. 1, pp. 46 - 50 .

Klir, G.J. and Yuan, B. ( 1995 ),Fuzzy Sets and Fuzzy Logic: Theory and Applications , ISBN 0-13-101171-5, Prentice Hall PTR, Upper Saddle River, NJ .

Lardner, R. and Fleming, M. ( 1999 ), " To err is human ",The Chemical Engineer , Vol. 689, pp. 18 - 20 .

Leake, D., Maguitman, A. and Reichherzer, T. ( 2014 ), " Experience-based support for human-centered knowledge modeling ",Knowledge-Based Systems , Vol. 68, September, pp. 77 - 87, available at: http://dx.doi.org/10.1016/j.knosys.2014.01.013

Li, B.M., Xie, S.Q. and Xu, X. ( 2011 ), " Recent development of knowledge-based systems, methods and tools for One-of-a-Kind Production ",Knowledge-Based Systems , Vol. 24 No. 7, pp. 1108 - 1119 .

Lovett, P.J., Ingram, A. and Bancroft, C.N. ( 2000 ), " Knowledge-based engineering for SMEs - a methodology ",Journal of Materials Processing Technology , Vol. 107 Nos 1-3, pp. 384 - 389 .

Mamdani, E.H. and Assilian, S. ( 1975 ), " An experiment in linguistic synthesis with a fuzzy logic controller ",International Journal of Man-Machine Studies , Vol. 7 No. 1, pp. 1 - 13 .

Maruta, R. ( 2014 ), " The creation and management of organizational knowledge ",Knowledge-Based Systems , Vol. 67, September, pp. 26 - 34, available at: http://dx.doi.org/10.1016/j.knosys.2014.06.012

Mason, M.K. ( 2014 ), "Knowledge management: the essence of the competitive edge", available at: www.moyak.com/papers/knowledge-management.html (accessed July 6, 2014).

Mathworks ( 2014 ), "Fuzzy inference system modeling: Gaussian combination membership function", available at: www.mathworks.se/help/fuzzy/gauss2mf.html (accessed June 20, 2014).

MATLAB ( 2012 ),MATLAB (R2012b): Fuzzy Logic Toolbox, 1984-2012 , The MathWorks Inc .

Newsinenglish ( 2013 ), "Poor routines led to Hercules crash", available at: www.newsinenglish.no/2013/10/22/poor-routines-led-to-hercules-crash / (accessed October 30, 2013).

OREDA ( 2009 ),OREDA 5th Edition Volumes I & II, Volume 1 - Topside Equipment Volume 2 - Subsea Equipment 2009 , OREDA, ISBN 978-82-14-04830-8.

Pedrycz, W. ( 1994 ), " Why triangular membership functions? ",Fuzzy Sets and Systems , Vol. 64 No. 1, pp. 21 - 30 .

Pillay, A. and Wang, J. ( 2003 ), " Modified failure mode and effects analysis using approximate reasoning ",Reliability Engineering and System Safety , Vol. 79 No. 1, pp. 69 - 85 .

Pota, M., Esposito, M. and De Pietro, G. ( 2014 ), " Fuzzy partitioning for clinical DSSs using statistical information transformed into possibility-based knowledge ",Knowledge-Based Systems , Vol. 67, September, pp. 1 - 15, available at: http://dx.doi.org/10.1016/j.knosys.2014.06.021

Ptil ( 2011 ), "Regulations relating to conducting petroleum activities (the activities regulations)", available at: www.ptil.no/activities/category399.html #_Toc345662825 (accessed March 18, 2013).

Ptil ( 2012 ), "Guidelines regarding the activities regulations", available at: www.ptil.no/activities /category404.html#_Toc345663053 (accessed March 28, 2013).

Ratnayake, R.M.C. ( 2009 ), "Industrial asset integrity management: sustainability, balanced performance and organizational alignment", PhD thesis UiS No. 86, University of Stavanger, Stavanger, ISBN: 978-82-7644-395-0, ISSN: 1890-1387.

Ratnayake, R.M.C. ( 2011 ), " Modelling of asset integrity management process: a case study for computing operational integrity preference weights ",International Journal of Computational Systems Engineering , Vol. 1 No. 1, pp. 3 - 12 .

Ratnayake, R.M.C. ( 2012 a), " A decision model for executing plant strategy: maintaining the technical integrity of petroleum flowlines ",International Journal of Decision Sciences, Risk and Management , Vol. 4 Nos 1-2, pp. 1 - 24 .

Ratnayake, R.M.C. ( 2012 b), " Sustainable asset performance: the role of PAS 55 1 & 2 and human factors ",International Journal of Sustainable Engineering , Vol. 6 No. 3, pp. 198 - 211 .

Ratnayake, R.M.C. ( 2013 ), " Plant systems and equipment maintenance: use of fuzzy logic for criticality assessment in NORSOK standard Z-008 ",Proceedings of the IEEE International Conference on Industrial Engineering and Engineering Management, Bangkok, pp. 1468 - 1472 .

Ratnayake, R.M.C. ( 2014 a), " KBE development for criticality classification of mechanical equipment: a fuzzy expert system ",International Journal Disaster and Risk Reduction , Vol. 9, September, pp. 84 - 98 .

Ratnayake, R.M.C. ( 2014 b), " Application of a fuzzy inference system for functional failure risk rank estimation: RBM of rotating equipment and instrumentation ",International Journal of Journal of Loss Prevention in the Process Industries , Vol. 29, May, pp. 216 - 224 .

Ratnayake, R.M.C. and Markeset, T. ( 2012 ), " Asset integrity management for sustainable industrial operations: measuring the performance ",International Journal of Sustainable Engineering , Vol. 5 No. 2, pp. 145 - 158 .

Sandberg, M. ( 2003 ), "Knowledge based engineering in product development", technical report, Luleå Univeristy of Technology, Lueå, ISSN 1402-1536.

Seneviratne, A.M.N.D.B. and Ratnayake, R.M.C. ( 2012 ), " In-service inspection of static mechanical equipment on offshore oil and gas production plants: a decision support framework ",2012 IEEE International Conference on Industrial Engineering and Engineering Management , pp. 85 - 90 . doi: 10.1109/IEEM.2012.6837707.

Stokes, M. ( 2001 ),Managing Engineering Knowledge-MOKA: Methodology for Knowledge Based Engineering Applications , ISBN: 978-1-86058-295-0, Wiley-Blackwell, London .

Tay, K.M. and Lim, C.P. ( 2008 ), " On the use of fuzzy inference techniques in assessment models: part II: industrial applications ",Fuzzy Optimization and Decision Making , Vol. 7 No. 3, pp. 283 - 302 .

Wang, J. ( 2001 ), " The current status and future aspects in formal ship safety assessment ",Safety Science , Vol. 38 No. 1, pp. 19 - 30 .

Wang, L.X. ( 1993 ),Adaptive Fuzzy Systems and Control - Design and Stability Analysis , University of California Berkeley, CA and PTR Prentice Hall, Upper Saddle River, NJ .

Z-008 ( 2011 ), "Z-008 risk based maintenance and consequence classification" (Rev. 3, June 2011), NORSOK standard, available at: www.standard.no/no/Fagomrader/Petroleum/NORSOK-Standard-Categories/Z-Regularity--Criticality/Z-0082 / (accessed March 8, 2013).

Z-DP-002 ( 1996 ), "Design principles - coding system", NORSOK standard, available at: www.standard.no/PageFiles/941/Z-DP-002r2.pdf (accessed October 30, 2013).

Zha, X.F. and Sriram, R.D. ( 2006 ), " Platform-based product design and development: a knowledge-intensive support approach ",Knowledge-Based Systems , Vol. 19 No. 7, pp. 524 - 543 .

Zijderveld, G.H.T., Tiebout, J.J., Hendriks, S.M., Poldervaart, L. and Gusto, M.S.C. ( 2012 ), " Subsea well intervention vessel and systems ",Offshore Technology Conference, Houston, TX, 30 April-3 May, OTC 23161.