You are on page 1of 5

Plant Systems and Equipment Maintenance: Use of Fuzzy Logic for Criticality

Assessment in NORSOK Standard Z-008


R.M. Chandima Ratnayake
Department of Mechanical and Structural Engineering and Materials Science, University of Stavanger, Norway
(chandima.ratnayake@uis.no)

Abstract - The NORSOK standard Z-008 suggests a manuscript illustrates the use of a fuzzy logic based
criticality matrix for use in consequence classification, consistent approach for making criticality assessment
maintenance planning, inspection planning and for aligned with NORSOK standard Z-008 guidelines. Such
prioritizing work orders. When the criticality assessments analyses greatly support maintenance engineers in
are carried out using a criticality matrix, suboptimal establishing or updating preventive maintenance
classification tends to occur as there are no means to
programs.
incorporate actual circumstances at the boundary of the
input ranges or at the levels of linguistic data and criticality
categories. This manuscript suggests a fuzzy inference II. METHODOLOGY
system (FIS) to overcome the aforementioned. Membership
functions and the rule base development have been carried
out in alignment with the Z-008 standard recommended Fuzzy logic system
guidelines. A rule view and a calculation result have been
demonstrated to illustrate the methodology. A ‘pure fuzzy logic system’ consists of a fuzzy rule
base. The rule base consists of a collection of fuzzy IF–
Keywords – criticality matrix, maintenance, THEN rules. These fuzzy IF–THEN rules are utilized by
consequence classification, fuzzy inference system the fuzzy inference engine to determine a mapping from
fuzzy sets in the input universe of discourse U ⊂ Rn to
fuzzy sets in the output universe of discourse V ⊂ R based
I. INTRODUCTION on fuzzy logic principles. The fuzzy IF–THEN rules
follow the form as below:
The safety regulations pertaining to petroleum
activities [i.e. oil and gas (O&G)] on the Norwegian R(1) : IF x1 is and … xn is THEN y is G1 (1)
Continental Shelf suggest that facilities' systems and
equipment are classified in relation the health, safety and where and Gj are fuzzy sets, x = (x1,x2,…, xn)T ϵ U
environmental (HSE) consequences due to potential and y ϵ V are input and output linguistic variables which
functional failures [1]. Basically, the operator companies belong to the input and output universes, respectively, and
perform criticality assessments with the help of external j=1, 2,…, m. The practical experience reveals that these
engineering service providing (i.e. engineering contractor fuzzy IF-THEN rules provide a convenient framework to
companies) experts [2]. They identify the various fault incorporate human expert knowledge. In Eq. (1), each
modes with associated failure causes and failure fuzzy IF–THEN rule defines fuzzy set , … => Gj
mechanisms, and predict the probability of failure for the for i = 1, 2,…,n; in the product space U × V. Expert
individual fault mode. Based on the failure frequency (FF) opinions and data/information retrieved from different
and consequence of failure, the criticality classification of sources are taken into the mathematical model using the
facilities' systems and equipment is performed. In the aforementioned rules. The main focus is to enhance the
aforementioned context, the NORSOK standard Z-008 discerning power in the criticality analysis process, whilst
provides requirements and guidelines for criticality-based associating the uncertainties related to the linguistic
maintenance and consequence classification of variables to the degree of criticality at the boundaries such
maintenance activities for plant systems and equipment in as high to medium, medium to low, and so on. The rules
the Norwegian petroleum industry [3]. also allow the integration of quantitative (e.g. mean time
It is suggested that the criticality-based maintenance between failure (MTBF), amount of oil release, etc.),
activities (or tasks) are carried out against defined criteria qualitative and judgmental data (e.g. personnel safety) in
and that the criteria are in accordance with the selected a uniform manner [4].
overall company policy for the minimization of health, In order to use the ‘pure fuzzy logic system’ in
safety & environmental (HSE), production and cost engineering systems, add a fuzzifier to the input and a
related challenges [1]. The current practice in carrying out defuzzifier to the output of the pure fuzzy logic system.
criticality assessments is to use a criticality matrix (i.e. The fuzzifier maps crisp points in U to fuzzy sets in U,
aligned with NORSOK standard Z-008 guidelines) along and the defuzzifier maps fuzzy sets in V to crisp points in
with possible ranges or linguistic terms [3]. As there is no V. The fuzzy rule base and fuzzy inference engine are the
means to study the boundary of each range or level of same as those in the pure fuzzy logic system. The ‘fuzzy
linguistic term (LT), engineering practice reveals that logic system’ has also been referred to as the ‘fuzzy logic
final criticality assessments are suboptimal. Hence, this controller’ due to its usage as a controller [5]. In 1975,

978-1-4799-0986-5/13/$31.00 ©2013 IEEE 1468


Proceedings of the 2013 IEEE IEEM

Mamdani built one of the first fuzzy systems which used a to align with guidelines provided in the NORSOK
set of fuzzy rules supplied by experienced human standard Z-008. The toolbox simulator of the MATLAB
operators to control a steam engine and boiler (R2011a) tool has been utilized for implementing a fuzzy

Fuzzy criticality assessment system


Experts’ knowlege from the experienced maintenance personnel

Generation of Generation of
Crisp
Crisp inputs input Generation of output
output
membership rule base membership
functions functions

Fuzzy inference system


Rank of MTBF Fuzzy MTBF
Rank of ED Fuzzification
Defuzzification
Rank of of input Evalution Fuzzy Criticality
to crisp output
downtime cost crisp of rules conclusions rank
Fuzzy overall variables
Rank of variables
level of
personnel consequence
safety
Run time calculations
Fuzzy Fuzzy
inputs outputs

Fig. 1. Fuzzy criticality ranking system.

combination [6]. To date, Mamdani’s approach has been criticality ranking system [8].
successfully applied to a variety of industrial processes
and consumer products [7]. Fig. 1 illustrates the overall
view of the proposed fuzzy criticality assessment system. III. DATA COLLECTION, MODELING, ANALYSIS
Personal safety, environmental degradation (ED), AND RESULTS
downtime costs and MTBF have been selected as the
input variables and criticality as the output variable. They A. Membership function selection
consist of quantitative, qualitative and judgmental (i.e. The consequence and MTBF (i.e. FF) of a functional
linguistic) data. For each of the aforementioned variables, failure have been selected as input to the FIS. The
there is an associated membership function, which is criticality of a functional failure has been assigned as the
established with the help of data, information and expert output. There are three factors under each functional
opinion [6]. The fuzzification process aids fuzzifiying the failure consequence rank. However, the highest value
inputs by determining the value of the membership (among the factors) of the consequence due to the failure
functions corresponding to the different inputs. of a particular failure mode is selected for assessing the
Furthermore, instead of restricting the user to a single, gravity of the consequence of a functional failure. When
crisp, input value, this process allows an interval of values two factors have an equal level of consequence, then the
to be given, where, values near the center of the interval priorities are given as follows: 1). Personnel safety (PS);
are assumed to be 'more certain' than those near the edges, 2). Production loss (i.e. downtime cost (DTC)); 3).
and the width of the interval indicates the amount of Environmental degradation (ED) (release of containment).
uncertainty present in the different input variables. The Basically, the following questions are addressed during
aforementioned has been achieved by associating the consequence classification process: Has the
appropriate membership functions (MF) for the input consequence a direct effect on safety? Has the
variables. Using an appropriate MF, the user has ‘more consequence a direct effect on environmental
confidence’ that the input parameter lies in the center of degradation? Has the consequence a direct effect on
the interval than at the edges. In this study the author has production? Has the consequence a direct effect on
incorporated Gaussian membership functions [7] which production costs? The last two questions were merged and
are defined by Eq. (2): addressed in terms of the downtime costs.
The intervals, corresponding membership values and
(2) finally membership functions (i.e. for Gaussian
membership functions ‘c and σ’) were established based
Where, c and σ are the center and width of the on expert opinion, literature and the author’s own
membership function respectively. experience.
The fuzzy inference system (FIS) parameters were
selected as follows: ‘And’ method with ‘minimum’, ‘Or’ B. Case study
method with ‘maximum’, ‘Implication’ with ‘minimum’,
‘Aggregation’ with ‘maximum’ and ‘Defuzzification’ To illustrate the approach, a case study was carried
with ‘centroid’ algorithm. Fuzzy rule bases were out in collaboration with an engineering contractor
developed using the table-look-up approach (see Table II) company which is providing criticality assessments to an

1469
Proceedings of the 2013 IEEE IEEM

operator company. Firstly, Table I was established based sub-systems, items, etc. The MTBF is the expected
on the views of personnel who have extensive experience failure rate if no preventive maintenance activities were
in criticality assessments for establishing or updating performed.
preventive maintenance programs. In the modeling process, the established ranges and
Table I illustrates ranks, linguistic terms and/or ranks (see Table I) of input and output variables were
ranges assigned for the possible consequence of a utilized as the basis for crisp input and output variables. A
functional failure and MTBFs. These were tailor-made rule base was developed based on expert opinion as well
TABLE I
RANGES, RANKS AND LINGUISTIC TERMS FOR CONSEQUENCES AND MTBF

Consequences

Factors Levels of consequence due to a functional failure


Input variable_2

Rank 1 2 3 4 5

LT Very high High Moderate Low Very low

PS Fatality Permanent injury Serious personnel injury Medical treatment First aid

ED > 200 m³ (20-200) m³ (2 - 20) m³ (0.2 - 2) m³ < 200 litres

DTC > 20 million (4 – 20) million (0.4 – 4) million (0.1 – 0.4) million < 0.1 million

Failure frequency
Input variable_1

Rank 1 2 3 4 5

LT Very high High Moderate Low Very low

MTBF Less than 1 1 month to 1 year (12 1 year (12 months) to 5 5 years (60 months) to 30 years More than 30 years
month months) years (60 years) (360 months) (360 months)

values particular to an operator company aligned with as historical data and information retrieved from possible
NORSOK standard Z-008. They were estimated by functional failures of O&G production/process
considering the possible functional failure of systems, equipment. Table II illustrates a tailor-made rule base (i.e.

TABLE II
TAILOR MADE RULE BASE FOR CRITICALITY ANALYSIS AND RBM DECISIONS

Consequences

PS Fatality Permanent Serious personnel Medical treatment First aid


Input membership injury injury
functions
ED > 200 m³ (20-200) m³ (2-20) m³ (0.2-2) m³ < 200 litres
Failure frequency

DTC > 20 million (4-20) million (0.4-4) million (0.125-0.4) million < 0.125 million

MTBF Rank 1 2 3 4 5

Less than 1 month 1 VH VH VH VH VH

1 month to 1 year 2 VH VH H M-H M-H

1 year to 5 years 3 VH H M-H M-L L

5 years to 30 years 4 H M-H M-L L VL

More than 30 years 5 M-H M-L L VL VL

TABLE III
3 GAUSSIAN MF PARAMETERS FOR INPUT AND OUTPUT VARIABLES

Gaussian MF parameters (σ, c)


VH H M L VL
Input MTBF (0.5, 1) (0.3, 2) (0.3, 3) (0.4, 4) (0.2, 4.85)
variables ED (0.5, 1) (0.4, 2) (0.4, 3) (0.4, 3.75) (0.3, 4.75)
VH H M-H M-L L VL
Output Criticality (0.3, 0.15) (0.3, 1) (0.3, 2) (0.3, 3) (0.3, 3.75) (0.3, 4.75)
variable

1470
Proceedings of the 2013 IEEE IEEM

aligned with NORSOK standard Z-008), which has been


developed for an operator company for performing
consequence classification for rotating mechanical
Membership
equipment (e.g. pumps, turbines, etc.) and instruments
[e.g. Pressure shut down valve (PSV), emergency shut
down valve (ESV), etc.].
To illustrate the methodology, only the input
variables MTBF and possible ED due to the release of
containment were utilized. Table III illustrates the Criticality (output variable)
parameters for each Gaussian membership function for Fig. 4. MF plots of criticality.
the input variables and output variable.
The introduction of fuzziness is vital in facilitating
the analysis of the inputs (i.e. ranges of estimated MTBF, IV. RESULTS
environmental degradation measures, etc.) at the
boundaries in the process of criticality assessment and Fig. 5 illustrates a rule view and an example
risk based maintenance (RBM)-related decision making. calculation of criticality rank for a tag. The calculation
Also, at the boundary of different criticality ranges VH-H, has been carried out for the MTBF rank = 2.6 and
H-M, etc. (see Table II), it is important to meet optimal environmental degradation rank = 1.7. The criticality rank
criticality analysis results. Hence, to introduce the estimated by the fuzzy criticality ranking system is 0.818.
fuzziness, the possible ranges of each input/output and The corresponding linguistic value is H (using the MFs in
corresponding membership values (i.e. called MF plots) Fig. 4). This linguistic value would be recorded in the
have been introduced. For establishing MF plots experts’ final criticality assessment report.
opinions [i.e. how do they perceive and experience the
influence of different machinery (Note: rotating
equipment and different instruments connected to a
certain function and how they contribute to a potential
failure of that function have been taken into
consideration) on a functional failure] as well as data and
information from different sources have been utilized.
Fig. 2 illustrates MF plots of MTBF.

Membership

MTBF (input variable 1)


Fig. 2. MF plots for rank of MTBF. MTBF = 2.6 ED = 1.7 Criticality = 0.818

Fig. 3 illustrates MF plots of environmental Fig. 5. A rule view and calculation [criticality rank is 0.818 for MTBF =
2.6 and ED = 1.7].
degradation.
Similar assessments would be carried out for each tag
(which has been connected to a functional hierarchy),
focusing on possible functional failures and related
Membership consequences. The possible functional failures are
recognized by means of a functional hierarchy which is
developed by taking functions performed by a system,
sub-system or item into consideration [3]. Basically, a
technical hierarchy (also called a technical tag hierarchy)
ED (input variable 2) is available in the production and/or process facility
Fig. 3. MF plots for rank of ED.
owner’s computer maintenance management system
(CMMS). Following the guidelines of NORSOK standard
Fig. 4 illustrates MFs for possible criticality scenarios Z-008, functional hierarchy is developed. After carrying
and corresponding ranks. out all the criticality assessments of the machinery, the
results would be recorded in a particular operator
company’s CMMS and used for updating maintenance

1471
Proceedings of the 2013 IEEE IEEM

activities such as frequency of inspection, man-hours, functional failure. Basically, it provides an opportunity for
work description, etc. The assessment results also provide incorporating expert knowledge in a consistent manner.
the basis for making decisions about whether to go for a Apart from that, the suggested approach can be linked to
corrective or preventive maintenance program. currently existing structured information management
system (SIMS) (e.g. Applysorco AS in Norway, which is
V. DISCUSSION an engineering contractor company, has its own SIMS) or
CMMS.
The consistency of the rule base used for the Further research should be carried out to establish
criticality assessment is verified (see Fig. 6) by examining more refined membership functions and corresponding
a plot of the criticalities’ surface over the possible intervals to represent the real behavior of the input and
combinations of the input variables. Fig. 6 reveals that output variables.
there are no inconsistencies, as there are no evident abrupt
changes in criticality for a small change in the ACKNOWLEDGMENT
consequence or MTBF ranks. The 3D graphical
representation of MTBF, consequence, and criticality (see The authors convey their special thanks to Jawad
Fig. 6) reveals that there is a possibility of giving decimal Raza and personnel in the department of asset integrity
ranks (i.e. corresponding to real input values) instead of management, Applysorco and Sukhvir Panesar Singh,
being strictly confined to the defined consequence ranges ConocoPhillips, for the support extended in
and corresponding integer ranks. accomplishing this research work.

REFERENCES

[1] Ptil, “Regulations relating to conducting petroleum


activities (the activities regulations)”, 2011.
http://www.ptil.no/ activities/ category399.html#_
Toc345662825 (accessed on 18.03.2013).
[2] R. M. C. Ratnayake, “A decision model for executing
plant strategy: Maintaining the technical integrity of
petroleum flowlines”, International Journal of Decision
Sciences, Risk and Management, vol.4, no.1/2, pp.1-24,
2012.
[3] Z008, “Z-008 Risk based maintenance and consequence
Fig. 6. Graphical representation of consistency. classification (Rev. 3, June 2011), 2011.
http://www.standard.no/no/Fagomrader/Petroleum/NORSO
The suggested criticality analysis approach has been K-Standard-Categories/Z-Regularity--Criticality/Z-0082/
demonstrated using one of the input consequence (accessed on 08.03.2013).
[4] J. B. Bowles, and C. E. Pelaez, “Fuzzy logic prioritization
(possible ED due to a functional failure) factors and FF
of failures in a system failure mode, effects and criticality
(i.e. MTBF). It is also possible to model the other analysis”, Reliab. Eng. Syst. Safe., vol. 50, no. 2, pp. 203–
consequence factors (i.e. PS and DTC) in a similar 213, 1995.
manner. The analysis can be carried out by the use of [5] C. M. F. Lapa, and A. C. F. Guimaraes, “Effect analysis
triangular MFs [9]. However, if all of the input variables fuzzy inference system in nuclear problems using
have been taken simultaneously then the number of rules approximate reasoning”, Ann. Nucl. Energy, vol. 31, no.
required for the assessment shall be increased 107, 2004.
significantly. [6] E. H. Mamdani, and S. Assilian, “An experiment in
linguistic synthesis with a fuzzy logic controller” Int. J.
Man Mach. Stud., vol. 7, no. 1, pp. 1–13, 1975.
VI. CONCLUSION [7] L. X. Wang, Adaptive Fuzzy Systems and Control—Design
and Stability Analysis, University of California at Berkeley,
In the manuscript, a rule base development and PTR Prentice Hall, 1993.
[8] MATLAB, MATLAB 7.12.0 (R2011a): Fuzzy logic toolbox,
analysis has been carried out to align with NOSOK 1984-2011 The MathWorks, Inc., 2011.
standard Z-008 for performing criticality assessment of [9] Z. H. Klim, Preliminary hazard analysis for the design
the possible functional failure that can result from the alternatives based on fuzzy methodology, Fuzzy
failures of O&G production/process equipment. Expert Information, Processing NAFIPS '04, ISBN: 0-7803-8376-
knowledge has been utilized to develop the MFs and rule 1, vol. 1, pp. 46-50, 2004.
base.
The suggested fuzzy criticality ranking system provides
the means to incorporate the actual tag (i.e. equipment)
related input variables to assess the criticality of a

1472

You might also like