Troubleshooting of crude oil desalination plant using fuzzy expert system

Desalination 266 (2011) 162–170

Contents lists available at ScienceDirect

Desalination j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / d e s a l

Troubleshooting of crude oil desalination plant using fuzzy expert system Gholamreza Zahedi a,⁎, S. Saba a, Musleh al-Otaibi b, Khairiyah Mohd-Yusof a a b

Process Systems Engineering Centre (PROSPECT), Faculty of Chemical and Natural Resources Engineering, Universiti Teknologi Malaysia, UTM Skudai, 81310 Johor Bahru, Johor, Malaysia Department of Chemical Engineering, University of Kuwait, Safat 13060, Kuwait

a r t i c l e

i n f o

Article history: Received 25 February 2010 Received in revised form 19 July 2010 Accepted 12 August 2010 Available online 15 September 2010 Keywords: Desalination Refinery Trouble shooting Fuzzy logic

a b s t r a c t This paper investigates fuzzy troubleshooting of a complex crude oil desalination plant. First, probable plant faults and all the related information were investigated. Then, based on the collected data, proper membership functions and consistent rules were prepared. In the next step of study, the Mamdani fuzzy inference system was utilized. Finally, the proposed expert system was employed to recognize faulty performance of the plant. Comparing the proposed model with plant data, it was found that the proposed system is capable of fast and accurate trouble shooting of the plant. In addition to the trouble shooting, the system can also be used for operator training. © 2010 Elsevier B.V. All rights reserved.

1. Introduction The difficulty in troubleshooting chemical plants increases with increasing plant complexity. Major damages and negative impacts on both plant's performance and human factors can occur when there is delay in plant decision making. Henceforth, creating an automated supervision that can identify faults precisely in an efficient manner is necessary for safe and economic operation of complex plants. Systematic and consistent troubleshooting reduces operator errors and consequently, decreases maintenance and repair time. A crude oil desalination process is a complex plant which includes large, multipart and complex stages. Therefore, performing a correct troubleshooting method is an essential matter to prevent any unpredicted shutdown. Fuzzy troubleshooting (FT) provides a systematic method for diagnosing by incorporating expert knowledge acquired from experts or plant manual. The use of fuzzy expert system provides a means for dealing with an uncertain situation when existing knowledge is ill-defined [1]. A fuzzy expert system is also protected from all human emotions, like stress, and exhaustion, that can cause deficiency while troubleshooting [2]. Fuzzy troubleshooting was recently utilized in trouble shooting of chemical plants. Tarifa and Scenna, [3] presented an efficient method for identifying faults in large chemical processes. The whole chemical plant was divided into segments by using structural, functional or causal decompositions. For each segment, a signed directed graph model was used. The model outputs were used to develop a troubleshooting system of the multi-stage flash (MSF) desalination plant.

⁎ Corresponding author. Fax: +60 7 5581463. E-mail address: [email protected] (G. Zahedi). 0011-9164/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.desal.2010.08.020

Tarifa and Scenna, [4] represented a fault diagnostic system for an MSF desalination plant which prepared drinkable water in Argentina. For this purpose, they applied a real time expert system. The diagnostic system was developed in order to verify the process state (normal or abnormal). The system output was 0(abnormality) and 1 (normality) for each potential fault. Venkatasubramanian et al. [5], investigated fault diagnosis methods that were based on historic process knowledge. They compared and evaluated the various methodologies in terms of the set of desirable characteristics. Morgan et al. [6], in their study represented an expert system which was able to troubleshoot the source of milling problems. The expert system utilized a fuzzy logic-based process to take the signals and information, and recommended possible alternatives to the process to achieve high-performance milling operations. They concluded that fuzzy logic is a satisfactorily robust method to deal with the variety of dynamic analysis data encountered within the milling process. Abdul-Wahab et al. [7], developed a fuzzy logic-based expert system in order to provide real time troubleshooting advice. They particularly worked on the brine heater of the MSF plant. They found that the system was able to perform troubleshooting tasks and could be used for either on-site trouble shooting or off-line operator training. Fuzzy expert systems also have many other applications like manufacturing, decision problem, controlling and modeling [8–11]. Following a review on the oil desalter process, this paper provides a brief introduction on fuzzy logic. In the next part, membership functions for troubleshooting of the plant are presented. Finally, the validity of proposed system is proven using off-line data and plant operator experiences. Based on our literature survey, the proposed

G. Zahedi et al. / Desalination 266 (2011) 162–170

troubleshooting method on the crude oil desalter plant has not been proposed in previous studies.

2. Process description Oil desalination is a process of removing dissolved salt from oil. Salt in oil may crystallize in equipment causing blockages, or may result in catalyst deactivation in refinery reactors. In addition, salt can cause corrosion in instruments and transportation pipes. In order to prevent these major problems, the salt content of crude oil must be reduced to a certain level. The crude oil treatment method can vary from simple techniques like gravity settling to complicated method such as high voltage desalination. Selection criteria for the desalination method depend on the amount of water and salt to be removed. In this study, an industrial refinery complex crude oil desalination plant was selected for troubleshooting purpose. This plant includes several instruments and sections, and utilizes different desalination techniques [12]. Fig. 1 shows the schematic of the plant. As demonstrated in Fig. 1, at point No. 1, the crude oil emulsion, which contained water and dissolved salt, flowed to a wet tank. The emulsion may contain up to 25% water. Typically, a desalting plant should treat crude oil until the water and salt contents are reduced to 0.10% by volume and 5.0 pounds per thousand barrels (PTB), respectively. To eliminate such a large amount of water from the oil stream, a two-stage desalting system was used. In the wet tank, some of the water was separated gravitationally. The separated water flows to a waste water treatment plant or is disposed off to a disposal pit. Then, at point No. 2, an emulsifier is injected into the stream. The mixture then entered a heat exchanger (Point No. 3), where heat is recovered from the treated crude product stream (stream No. 10). The stream then flows to a water bath indirect heater (point No. 4). At point number 5, recycled water from the second stage vessel is injected into the emulsion flow. The aim of recycling water application was to minimize freshwater consumption at the mixing

163

valve (No. 6), (an induced shearing force disturbs recycled water and emulsion). A simple globe valve performs the function of a mixing valve, where an operator would set the differential pressure across the valve to be as high as possible to assure better mixing of the two fluids. Stream No. 7 leaves the mixing valve to enter the first stage desalter vessel. In the first stage desalter, the emulsion was exposed to a high voltage electrostatic field. The application of the electrostatic field caused coalescence of the dispersed water phase. As a result, the enlarged coalesced water dropped gravitationally and accumulated at the bottom of the vessel. Effluent water from the first stage (stream No. 1) left the system to a wastewater treatment plant or the disposal pit. This stream contained various impurities and salts. Treatment of the emulsion was improved in the second stage desalting vessel. Crude oil passed through a mixing valve at the entrance of the second stage vessel (Stream No. 8). The emulsion that contained residual salt was once more blended with fresh water (stream No. 9). The differential pressure across the mixing valve is usually maintained about 15 psia. Then, incompletely treated emulsion flowed into the second stage near the bottom of the vessel and, moved upward through the electrical voltage field. The recovered water was collected at the bottom of the vessel and recycled to the first stage desalter (stream 5), while the treated crude flowed from the top of the vessel (stream No. 10) and into an analyzer (stream No. 12). If the treated crude is within the desired specification, a signal was sent to the diverting valve to open the dry tank. Otherwise, the flow is directed back to the wet tank. A more detailed description of the process can be seen in [12].

2.1. Fuzzy logic This section presents the basic concepts of fuzzy set theory introduced by Zadeh [13,14]. Fuzzy sets are development of classical sets. In contrast to the classical set, or crisp set, the boundary of a fuzzy set is not precise. That is, the gradual change from nonmembership to membership. This gradual change is expressed by membership

Fig. 1. Schematic of desalination process.

G. Zahedi et al. / Desalination 266 (2011) 162–170

functions, which completely and uniquely characterizes a particular fuzzy set [15]. Fuzzy sets are basic concepts of fuzzy logic. Classical sets and subsets are well known and have been used for several years. For example, consider a set of all numbers between 0 and 10 as a universal set (set A). A lot of subsets can be extracted from this universal set. For instance, a set of numbers between 2 and 8 (set B). The borders of set B are exactly known. If all data from set A, which exist in set B are valued by 1 and the remaining numbers by 0, it can be shown which numbers is a member of set B and which one is not. Fig. 2 shows this separation. However, to determine some vague subsets, for example, a subset of “large numbers” from set A. It is clear that the largest one is 10. But numbers like 9.9 or 9 may also be classified as large numbers. What about 8 and 7? Can they be considered large, or not? Such sets, which are not exactly known, are the basic concepts of fuzzy sets. Determining the meaning of large for a number depends on the experience of the person dealing with the problem. For example, one may think that numbers less than 7 are not large. So a graph like Fig. 3 can show which numbers are members of this subset. Number 10 gets a value of 1, which means it is exactly a large number (its membership in this subset is crisp) and numbers that are less than 7 get a value of 0. That means, they are not members of a large number set. Numbers between 7 and 10 get intermediate values based on the proximity to 7 or 10. Graphs like that in Fig. 3 are known as “Membership Function (MF)” in fuzzy logic. In a classical set, each member only belongs to a set or does not belong to a set. In contrast, in a fuzzy set, each member belongs to the set to some extent. The degree of membership is identified by the membership function, which varies from 0 to 1. The membership function of a fuzzy set A is denoted by μA(x) and a fuzzy set A is indicated as: A = fðx; μA ðxÞÞjXg

ð1Þ

where X is the universe of objects [16]. Important and commonly used membership functions are trapezoidal, triangle, Gaussian and bell function. To fuzzy interface systems (FIS), the first step is identifying the fuzzy sets and subsets precisely. Then, the fuzzy rule should be created. These fuzzy rules can be extracted from expert knowledge or existing information, and can be improved via model validation. A general form of fuzzy if-then rule is: If x is A and y is B then z is C

Index of membership in set B

where, A, B and C are linguistic terms. For example, “if the weather is cloudy” and “the temperature is low” then “it will be snowy”. A, B and C can be any conversational sentence or logical sentences.

1

0 2

8

10

Set A Fig. 2. Separation of set A into two classes (members and not members).

Index of membership in set B

164

1

0 7

10

Set A Fig. 3. Separation of big numbers from set A.

By implementing fuzzy if-then rule, expert or vague knowledge is considered in a modeling scheme and helps to achieve accurate conclusion from a very simple and comprehensive model. After these steps, a FIS should be built. FIS is a framework based on the concepts of fuzzy set, fuzzy rules, fuzzy reasoning mechanism and de-fuzzification method [16]. There are various FISs where Mamdani, Sugeno-Takagi-Kang [17] and Tsukamoto fuzzy inference system are famous and have been employed in many researches [18–20]. In this paper a Mamdani FIS was employed. 3. Model development As previously mentioned, troubleshooting of a crude oil desalination plant is complicated because of complexity of the plant. The first step in creating a fuzzy troubleshooting system is to define major probable faults in the plant. This information could be obtained from the experience of professional operators, plant manual, books and literatures. Consequently, in this study, the problematic instruments which should be periodically checked for the fault were identified. These instruments and parameters are listed in Table 1. For troubleshooting purposes, based on model evaluation and simplicity of application, triangular and trapezoidal membership functions were used for the input and output variables. Triangular MF was employed for normal state and trapezoidal MF was employed for the abnormal state. Output variables were expressed as action functions, which were based on urgency factor. Urgency factor varies between −100 to 100. The triangular and trapezoidal MFs are defined as:  x−a c−x  ; ;0 triangleðx; a; b; cÞ = max min −b c−b

ð2Þ

where a, b and c are three parameters (ie the corners) of a triangular MF,     x−a d−x ; 1; ;0 trapezoidðx; a; b; c; dÞ = max min b−a d−c

ð3Þ

a, b, c and d are four parameters of a trapezoidal MF. These parameters were selected based on the information listed in Table 1. Conditions of faults and instruments distinguish the number of MFs. Therefore, these numbers should be recognized through experience and model evaluation. For each related action as output variable, the appropriate MFs should be created. The negative or positive sign of the word ‘urgent’ was related to the problems which take place in low or high range of the input variables respectively. These signs help to distinguish the right action that should be taken to remove the fault. These parameters are listed in Table 2. This step is crucial for creating an accurate troubleshooting system so that information in the instruction manuals and knowledge of experts were carefully considered. The process inputs and their

G. Zahedi et al. / Desalination 266 (2011) 162–170

165

Table 1 Input variables and their operating level.

Wet tank interface level Wet tank oil level Feed pump discharge pressure Fuel flow to the system for the desalter heater Stack temperature for the desalter heater Water bath/crude oil temperature for the desalter heater Water level of desalter Oil carried out in water for desalters (effluent water ) Tank level for wash water Flow of wash water pumps Level of degassing tank Oil in water for oil water separation Salt content in desalted crude for desalted crude

Higher level

Lower level

Set point

Accepted tolerance

2500 mm 11,000 mm 30 bar 1261 NM3/H 450 °C 95 °C 1040 mm 200 PPM 11,000 mm 47 M3/H 10,250 mm – 10 PTB

1800 mm 7000 mm. 14 bar 200 NM3/H 200 °C 74 °C 485 mm b20 PPM 2000 mm Up to 10 M3/H 3800 mm – b5 PTB

2100 mm 9000 mm 16.85 bar Based on temp. required Av. 350 °C 85 °C 762 mm b 10 7100 mm 28 M3/H 9250 mm b 20 PPM b 5PTB

2000–2300 mm 7600–1050 mm 14–20 bar 200–1261 NM3/H 200–400 °C 90 °C 635–890 mm 10–200 PPM 4000–11,000 mm – 4000–9750 mm – Normally b 20 PTB

related output membership functions for each part are depicted in Figs. 4 to 16. The next step for making a fuzzy troubleshooting program is creating fuzzy rules (Table 3). These rules express the relationship between inputs and outputs and the action which should be taken in different situations. Initially, a rule system was prepared. After creating a complete model, the output of the model was investigated. By considering the accuracy of the results, the rule system was modified to get more accurate answers. These rules are listed in Table 6. The number placed at the end of each rule sentence was the weight of each rule. This weight can be varied from zero to one, and describes the importance of the specific rule compared to the other rules. In this study, all rules have the same weight factor. Taking the wet tank interface level as an example, it can be seen that it consists of three input fuzzy MF, which are, ‘low’, ‘normal’, ‘high’ and this fuzzy set has three output actions such as: ‘−urgent’, ‘normal’, ‘+urgent’. According to Table 1 this fuzzy set is explained by three fuzzy rules which are: 1. If (interface level is low) then (action1 is −urgent) (1) 2. If (interface level is normal) then (action1 is normal) (1) 3. If (interface level is high) then (action1 is +urgent) (1) After preparing consistent rules for the desalination plant, a proper inference system should be applied. This step completes a troubleshooting program which is able to decide the degree of faulty performance of the system to assist an operator to react accurately in decisive situations. The Mamdani inference system, which is suitable for the troubleshooting was selected.

Table 2 MF parameters for output actions.

4. Validation of the trouble shooting expert system The developed fuzzy troubleshooting system was tested by several off-line tests. In these tests, some input data were fed to FIS. The expert system then verified the data and returned conclusions as normal or abnormal. For example, it is required to verify the right action when the interface level is low (i.e. the level is 1900 mm, which is below 2100 mm). The values of the input is ‘1900’, and the output given by FIS would be “−76”. Regarding the output MF for the action 1, when the output reading is in the range of −40 to 40, it means that its corresponding input value is normal and there is no need to do any troubleshooting. Following this procedure, the outputs of the proposed expert system was validated by a group of expert operators in the plant. Based on the judgment of the experts, the system performances were satisfactory. 5. Conclusion and remarks

Output MF's parameters variables -urgent

Action 1 Action 2 Action 3 Action 4 Action 5 Action 6 Action 7 Action 8 Action 9 Action 10 Action 11 Action 12 Action 12

The final step in developing a fuzzy expert system is to implement and evaluate its troubleshooting performance. To accomplish this, some input data were fed to the system. The inputs of this system consisted of the wet tank interface level, wet tank oil level, feed pump discharge pressure, fuel flow to the system for the desalter heater, stack temperature for the desalter heater, water bath/crude oil temperature for the desalter heater, water level of desalter, oil carried out in water for desalters (effluent water ), tank level for wash water, wash water pumps flow rate, level of degassing tank, oil content in water for oil water separation, salt content in desalted crude for desalted crude. The corresponding output associated with the input varied between −100 and 100, depending on the urgency weight factor. The output shows the degree of urgency which should be considered in order to perform the required troubleshooting action.

Normal

+ urgent

a

b

C

d

a

b

c

d

a

b

c

D

− 100 − 100 − 100 − 100 – − 100 − 100 – − 100 − 100 − 100 – –

− 100 − 100 − 100 − 100 – − 100 − 100 – − 100 − 100 − 100 – –

− 65 − 65 − 65 − 65 – − 65 − 65 – − 65 − 65 − 65 – –

− 40 − 40 − 40 − 40 – − 40 − 40 – − 40 − 40 − 40 – –

− 65 − 65 − 65 − 65 0 − 65 − 65 0 − 65 − 65 − 65 0 0

− 40 − 40 − 40 − 40 0 − 40 − 40 0 − 40 − 40 − 40 0 0

40 40 40 0 40 40 40 40 0 0 40 40 40

65 65 65 0 65 65 65 65 0 0 65 65 65

40 40 40 – 40 40 40 40 – – 40 40 40

65 65 65 – 65 65 65 65 – – 65 65 65

100 100 100 – 100 100 100 100 – – 100 100 100

100 100 100 – 100 100 100 100 – – 100 100 100

This paper represents a fuzzy troubleshooting program for assisting operators to decide quickly and accurately in order to prevent major economic loss or serious health injuries in a complex desalination plant. This plant is a multipart system and its troubleshooting is too convoluted. In this case, all the necessary information was available to prepare the appropriate MF, which is one of the essential steps to achieve a precise troubleshooting system. All expert knowledge was incorporated to set up rules that could explain the state of the system in a more precise model. Off-line data were used to check the performance of the proposed system. During the validation step, the results and action is carried out by the system were deemed to be satisfactory by expert operators.

166

G. Zahedi et al. / Desalination 266 (2011) 162–170

action1 normal

-urgent

0.8

0.6

0.4

0.2

0 -100

-50

0

50

normal

low

+urgent

1

degree of membershipness

degree of membershipness

interface level of wet tank high

1

0.8

0.6

0.4

0.2

0 1800

100

2000

action1

2200

2400

interface level(mm)

Fig. 4. Memberships function for interface level of wet tank and its related action.

action2 normal

-urgent

low

+urgent

1

high

normal

1

degree of membershipness

degree of membershipness

oil level of wet tank

0.8

0.6

0.4

0.8

0.6

0.4

0.2

0.2

0 -100

-50

0

50

0 7000

100

8000

action2

9000

10000

11000

oil level(mm)

Fig. 5. Memberships function for oil level of wet tank and its related action.

action3 -urgent

normal

0.8

0.6

0.4

normal

high

1

0.8

0.6

0.4

0.2

0.2

0 -100

low

+urgent

degree of membershipness

1

degree of membershipness

Discharge Pressure of feed pump

0 -50

0

action3

50

100

15

20

25

Discharge Pressure

Fig. 6. Memberships function for discharge pressure of feed pump and its related action.

30

G. Zahedi et al. / Desalination 266 (2011) 162–170

-urgent

1

fuel flow to the heater normal

degree of membershipness

degree of membershipness

action4

0.8

0.6

0.4

0.2

0 -100

167

normal

1 no fuel

0.8

0.6

0.4

0.2

0 -80

-60

-40

-20

0

0

200

400

action4

600

800

1000

1200

fuel flow (nm3/h)

Fig. 7. Memberships function for fuel flow to the heater and its related action.

action 5

normal

0.8

0.6

0.4

0.8

0.6

0.4

0.2

0.2

0

high

1

degree of membershipness

degree of membershipness

normal

normal

1

stack temperature of heater

0

20

40

60

80

0 200

100

250

action5

300

350

400

450

stack temperature (c)

Fig. 8. Memberships function for stack temperature of the heater and its related action.

action6 normal

-urgent

high

normal

1

0.8

0.6

0.4

0.8

0.6

0.4

0.2

0.2

0 -100

low

+urgent

degree of membershipness

degree of membershipness

1

water bath /crude oil temperature

-50

0

action6

50

100

0 75

80

85

90

water bath /crude oil temperature(C)

Fig. 9. Memberships function for water bath/crude oil temperature and its related action.

95

168

G. Zahedi et al. / Desalination 266 (2011) 162–170

action 7 -urgent

water level in desalter

normal

low

+urgent

degree of membershipness

degree of membershipness

0.8

0.6

0.4

high

normal

1

1

0.8

0.6

0.4

0.2

0.2

0 -100

-50

0

50

0 500

100

600

action7

700

800

900

1000

water level(mm)

Fig. 10. Memberships function for water level of desalter and its related action.

action 8 normal

+urgent

1

degree of membershipness

degree of membershipness

normal

oil carrid out in water

0.8

0.6

0.4

high

1

0.8

0.6

0.4

0.2

0.2

0

0 0

20

40

60

80

100

50

0

action8

100

150

200

oil carrid out in water(ppm)

Fig. 11. Memberships function for oil carried out in water and its related action.

action 9

1

0.8

0.6

0.4

0.8

0.6

0.4

0.2

0.2

0 -100

normal

low

normal

-urgent

degree of membershipness

degree of membershipness

1

tank level for wash water tank

-80

-60

-40

action9

-20

0

0 2000

4000

6000

8000

tank level(mm)

Fig. 12. Memberships function for tank level in wash water tank and its related action.

10000

G. Zahedi et al. / Desalination 266 (2011) 162–170

action10 -urgent

0.8

0.6

0.4

0.2

0 -100

normal

low

1

degree of membershipness

degree of membershipness

wash water pump flow normal

1

169

0.8

0.6

0.4

0.2

-80

-60

-40

-20

0 10

0

20

action10

30

40

50

pump flow(M3/h)

Fig. 13. Memberships function for wash water pump flow and its related action.

1

degasing tank level +urgent

normal

-urgent

1

degree of membershipness

degree of membershipness

action 11

0.8

0.6

0.4

low

normal

high

0.8

0.6

0.4

0.2

0.2

0 -100

-50

0

50

0 4000

100

6000

action11

8000

10000

degasing tank level(mm)

Fig. 14. Memberships function for desalting tank level and its related action.

action12

oil in water in water spration unit +urgent

normal

0.8

0.6

0.4

0.8

0.6

0.4

0.2

0.2

0

too much

normal

1

degree of membershipness

degree of membershipness

1

0 0

20

40

60

action12

80

100

0

5

10

oil in water(ppm)

Fig. 15. Memberships function for oil in water in separation unit and its related action.

15

20

170

G. Zahedi et al. / Desalination 266 (2011) 162–170

action13 +urgent

1

degree of membershipness

degree of membershipness

normal

salt in dry crude

0.8

0.6

0.4

0.2

0

normal

1

high

0.8

0.6

0.4

0.2

0 0

20

40

60

80

100

0

action13

2

4

6

8

10

salt content in crude(PTB)

Fig. 16. Memberships function for salt in dry crude and its related action.

Table 3 Fuzzy rule for the troubleshooting system. A. Fuzzy rule for interface level of wet tank. 1. If (interface level is low) then (action1 is − urgent) (1) 2. If (interface level is normal) then (action1 is normal) (1) 3. If (interface level is high) then (action1 is + urgent) (1) B. Fuzzy rule for oil level of wet tank. 1. If (oil level is low) then (action2 is − urgent) (1) 2. If (oil level is normal) then (action2 is normal) (1) 3. If (oil level is high) then (action2 is + urgent) (1) C. Fuzzy rule for discharge pressure of feed pump. 1. If (discharge pressure is low) then (action3 is − urgent) (1) 2. If (discharge pressure is normal) then (action3 is normal) (1) 3. If (discharge pressure is high) then (action3 is + urgent) (1) D. Fuzzy rule for fuel flow of the heater. 1. If (fuel flow is no fuel) then (action4 is − urgent) (1) 2. If (fuel flow is normal) then (action4 is normal) (1) F. Fuzzy rule for stack temperature of the heater. 1. If (stack temperature is normal) then (action5 is normal) (1) 2. If (stack temperature is high) then (action5 is + urgent) (1) G. Fuzzy rule for water bath/crude oil temperature of heater. 1. If (water bath/crude oil temperature is low) then (action6 is − urgent) (1) 2. If (water bath/crude oil temperature is normal) then (action6 is normal) (1) 3. If (water bath/crude oil temperature is high) then (action6 is + urgent) (1) H. Fuzzy rule for water level of desalter. 1. If (water level is low) then (action7 is − urgent) (1) 2. If (water level is normal) then (action7 is normal) (1) 3. If (water level is high) then (action7 is + urgent) (1) I. Fuzzy rule for oil carried out in water of desalter. 1. If (oil carried out in water is normal) then (action8 is normal) (1) 2. If (oil carried out in water is high) then (action8 is + urgent) (1) J. Fuzzy rule for tank level of wash water tank. 1. If (tank level is low) then (action9 is − urgent) (1) 2. If (tank level is normal) then (action9 is normal) (1) K. Fuzzy rule for pump flow of wash water tank. 1. If (flow is low) then (action10 is − urgent) (1) 2. If (flow is normal) then (action10 is normal) (1) L. Fuzzy rule for tank level of degassing tank. 1. If (degassing tank level is low) then (action11 is − urgent) (1) 2. If (degassing tank level is normal) then (action11 is normal) (1) 3. If (degassing tank level is high) then (action11is + urgent) (1) M. Fuzzy rule for oil in water of water separation unit. 1. If (oil in water(oil water separation) is normal) then (action12is normal) (1) 2. If (oil in water(oil water separation) is too much) then (action12 is +urgent) (1) N. Fuzzy rule for salt content of desalted crude. 1. If (salt content in desalted crude is normal) then (action16 is normal) (1) 2. If (salt content in desalted crude is high) then (action16 is + urgent) (1)

The way forward is to use this program on-line to add or delete some rules to obtain a more accurate system. During the application of

the fuzzy troubleshooting system, the program will produce an alarm that is attached with the troubleshooting actions. This alarm is very important to the operation team as it represents abnormality in the process of the plant.

References [1] A. Labbi, E. Gauthier, Combining fuzzy knowledge and data for neuro fuzzy modeling, Journal of Intelligent Systems 7 (1–2) (1997) 145–163. [2] E. Carrasco, F.J. Rodriguez, A. Punal, E. Roca, J.M. Lema, Rule-based diagnosis and supervision of a pilot-scale wastewater using fuzzy logic techniques, Expert Systems Application 22 (2002) 11–20. [3] E. Tarifa, N. Scenna, A methodology for fault diagnosis in large chemical processes and an application to a multistage flash desalination process: part II, Reliability Eng. System 60 (1998) 41–51. [4] E. Tarifa, J. Scenna, Fault diagnosis for a MSF using a SDG and fuzzy logic, Desalination 152 (2002) 207–214. [5] V. Venkatasubramanian, R. Rengaswamy, S.N. Kavuri, K. Yin, A review of process fault detection and diagnosis Part III: process history based methods, Computers & Chemical Engineering 27 (2003) 327–346. [6] G. Morgana, R.Q. Chengb, Y. Altintasb, K. Ridgwaya, An expert troubleshooting system for the milling process, International Journal of Machine Tools and Manufacture 47 (2007) 1417–1425. [7] A.S. Abdul-Wahab, A. Elkamel, M.A. Al-Weshahi, A.S. Al Yahmadia, Troubleshooting the brine heater of the MSF plant using a fuzzy logic-based expert system, Desalination 217 (2007) 100–117. [8] A. Kandel, Fuzzy Expert Systems, CRC Press, Orlando, 1992. [9] T.P. Hong, C.Y. Lee, Introduction of fuzzy rules and membership functions from training examples, Fuzzy Sets and Systems 84 (1996) 33–47. [10] Abdüsselam Altunkaynak, Shankararaman Chellam, Prediction of specific permeate flux during cross flow microfiltration of polydispersed colloidal suspensions by fuzzy logic models, Desalination 253 (1-3) (2010) 188–194. [11] G. Zahedi, S. Saba, H. Hashim, Prediction of ozone concentration around an industrial area using fuzzy neural network method, Submitted to Journal of Environment Management, 2009. [12] K. Mahdi, R. Gheshlaghi, G. Zahedi, A. Lohi, Characterization and modeling of a crude oil desalting plant by a statistically designed approach, Petroleum Science and Engineering 61 (2008) 116–123. [13] L.A. Zadeh, Fuzzy sets, Information and Control vol. 8 (1965) 338–353. [14] V. Novák, I. Perfilieva, Discovering the World with Fuzzy Logic, Springer- Verlag, Heidelberg, 2000. [15] G.J. Klir, B. Yuan, Fuzzy Sets and Fuzzy Logic: Theory and Applications, PrenticeHall, New York, 1995. [16] J.-S. Roger, C. Jang, T. Sun, Neuro-fuzzy modeling and control, Proceeding of the IEEE vol. 83 (No. 3) (1995) 378–400. [17] J. Casillas, O. Cordon, F. Herrera, L. Magdalena (Eds.), Interpretability Issues in Fuzzy Modeling, Springer-Verlag, Berlin, 2003. [18] E.H. Mamdani, Advanced in the linguistic synthesis of fuzzy controller, International Journal of Man-Machine Studies 8 (1976) 669–678. [19] Y. Tsukamoto, An approach to fuzzy reasoning method, Advances in fuzzy set and theory and applications, 1979, pp. 137–149. [20] M. Sugeno, Fuzzy Control, Nikkan Kogio, Tokyo, 1988.