chapter 5 real time applications -...
TRANSCRIPT
92
CHAPTER 5
REAL TIME APPLICATIONS
5.1 OVERVIEW
In Chapter 4, the proposed methodology to achieve FDI and FTC along with
numerical and simulation results were presented. For this purpose, benchmark
systems under various case studies were analysed. In this Chapter, the applications
of the proposed methodology on real time systems are discussed and presented. The
real time systems considered in this research work are the Sulphur Recovery Unit
(SRU) in petroleum refinery, Sewage Treatment Plant (STP) in waste water
treatment system and Bottle Filling Plant (BFP). In SRU unit, two main systems
namely, Main Clause Recovery Converter (MCRC) and Vaporized Liquefied
Petroleum Gas (LPG) header systems are modeled and analysed. Based on the
analysis done, the applications of the proposed methodology to achieve FDI are
discussed. Likewise, operations of STP and BFP are modeled, and analyzed in Petri
net environment. A MATLAB based GUI toolbox is then developed to estimate the
typical faults occurring in a STP. Finally, the conditions for achieving FDI along
with FTC in BFP are derived analytically and presented.
5.2 PROPOSED FDI IN SULPHUR RECOVERY UNIT (SRU)
In this section, methods adopted for modeling, analysis and FDI of the systems
in a typical SRU unit of a petroleum refinery are discussed. For this purpose, real
time data collection for a SRU in Chennai Petroleum Corporation Ltd (CPCL) is
93
done and based on the data, available techniques as discussed previously are
applied.
5.2.1 Introduction to SRU
The Chemical Process considered here is the SRU of Refinery process of Chennai
Petroleum Corporation Ltd., Chennai. The SRU forms an important and essential
unit in Refining process where extraction and removal of Sulphur from the products
of refining are carried out to improve the purity and efficiency of the products. SRU
comprises several sub-units working together to achieve the purest form of the
products, thereby reducing the content of elemental sulphur. Since the SRU
developed in CPCL works on the principle of Main Clause Recovery concept
(MCRC), the MCRC sequence working is considered here for study. The detailed
description of MCRC unit will be covered in a later section separately.
Refining Process Industry comprises various components of boilers, heat
exchangers, cooling towers, etc. and the flow of the fuel/gas is done through a series
of pumps and valves, controlled through control units placed at various locations in
the Industry. The control units take appropriate control actions through the help of
numerous sensors for controlling flow, level, pressure, temperature, etc. System
normally monitored through automatic/Manual means is subjected to frequent
malfunction and has to be rectified so as to improve the efficiency avoiding frequent
breakdowns. Moreover, the faults or malfunctions that frequently occur in the
system are either due to Valve failures or sensor failures. Since valves and sensors
are an integrated part of every sub-unit of the process, it becomes quite difficult to
identify and detect the faults arising due to these failures. Hence, it is utmost
necessary to develop algorithms to achieve effective fault diagnosis to detect valve
and senor failures.
94
In Literature, as discussed in section 1.5 of Chapter 1, there have been numerous
schemes and designs developed, and made available to achieve fault diagnosis.
Here, the concept of Petri nets is used to model the systems as DEDSs, rather than
continuous systems for the sake of simplicity. The faults occurring in valves are
classified as Transition faults and the ones occurring in sensors are classified as
Place faults in a Petri net environment. The normal condition of an event is
considered to be observable, and is thus modeled. Once the fault has occurred, the
corresponding event is termed as unobservable, and thus, the system is again
modeled based on the unobservabilty nature. By comparing the new system with the
original system, the faults occurred in the system are identified analytically.
5.2.2 Process description
The process design of SRU is based on the Main Clause Recovery Concept
(MCRC) technology, which is licensed by M/s. DELTA HUDSON
ENGINEERING LIMITED. It is a combination of the Clause process and the
extension of the Clause reaction up to the temperature at which the product sulphur
starts condensing on the catalyst itself. Extension of Clause reaction upto sulphur
dew point enhances sulphur recoveries beyond 99%. As sulphur condensation on
the catalyst leads to its activity reduction, regeneration of such portion of the
catalyst bed is required. Overview of the total SRU plant is shown in Figure 5.1.
95
Figure 5.1 Overview of SRU plant
96
5.2.2.1 Clause section
Each Train of SRU consists of a thermal reactor and four catalytic converters. The
first catalytic converter is a conventional clause converter, while the last three are
MCRC converters that alternate between a sub-dew point mode, and a regeneration
mode. With such a four-converter MCRC configuration, sulphur recovery higher
than 99 % is achievable. The salient feature of the MCRC process is that
regeneration takes place on line.
The acid gas from Amine Regeneration Unit (ARU), Hydrogen Sulphide (H2S)-
rich gas from the 1st stage Sour Water Stripping Unit (SWS) and Ammonia (NH3)-
rich gas from the 2nd stage SWS Unit is fed to the SRU to convert H2S contained in
the feed gas to elemental sulphur. The acid gas is preheated to 89°C using Low
Pressure (LP) Steam. The preheated acid gas is fed into the Main Combustion
Chamber (MCC) furnace in each train, which is already heated to 1100°C, with fuel
gas or vaporized Liquefied Petroleum Gas (LPG). Air flow to MCRC is adjusted to
get H2S to SO2 ratio as 2:1. The major part of heat generated in the furnace is
recovered by producing Main Pressure (MP) steam in the Waste Heat Boiler
(WHB). The vapours leaving WHB is further cooled to 191°C in sulphur
condenser-1 to remove elemental sulphur from the gas mixture producing LP steam.
The gas from condenser is reheated in a line burner using a slip stream of ARU gas
from Knock Out (KO) Drum as fuel. The preheated vapours from the line burner at
273°C further flows to the 1
st clause converter, where additional conversion to
sulphur takes place. The reaction gases from the converter are cooled in a reheat gas
exchanger, and sulphur condenser 2. The produced sulphur condenses and flow to
sulphur pit via sulphur lock. Overview of the MCRC converter is shown in Figure
5.2.
97
5.2.2.2 Main clause recovery section
In the MCRC section as observed in Figure 5.2, there are three catalytic stages
along with respective sulphur condensers. Operating conditions in two of the three
MCRC reactors are aimed in such a way that the adsorption of produced sulphur
takes place on the part of the catalyst present in the reactor. Such adsorption of
sulphur of the catalyst is achieved by maintaining temperature in the corresponding
reactors lower than the sulphur dew point. Removal of sulphur from product gases
of Clause reaction by way of adsorption on catalyst is to increase sulphur
conversion by moving Clause reaction in the forward direction. Additionally lower
reaction temperature also helps thermodynamic equilibrium to shift towards higher
sulphur equilibrium concentrations. Thus, overall higher sulphur conversions are
easily achieved by reducing the operating temperature in the catalytic reactors.
However, deposition of sulphur on active site of catalyst prohibits its further
activity. To regenerate the activity of such catalyst sites, sulphur present on such
site needs to be desorbed. Desorption of sulphur is carried out by operating the
reactors at a temperature higher than the sulphur vaporization temperature. The
need of operating sulphur loaded catalyst at higher temperature is met by operating
one of the three MCRC reactors at any given time under regeneration mode. Each
MCRC reactor operates in a sequence under the following operating modes: (1)
Regeneration mode, (2) Sub-dew point mode-1, and (3) Sub-dew point mode-2.
98
Figure 5.2 Overview of MCRC Section
99
As mentioned before, MCRC section includes the three sub-dew point sulphur
converters. These converters alternate between a sub-dew point mode of operation
and a regeneration mode. For the purpose of discussion, the converter operating in
the regeneration mode is said to be in the sulphur converter II position, and the
converters operating in the sub-dew point mode are said to be in No. III and No. IV
positions.
The reheated gas from the Clause section enters the No. II position sulphur
converter and regenerates the catalyst bed by vaporizing sulphur which was
previously adsorbed onto the catalyst when this converter occupied the No. III and
IV positions. Clause conversion continues in this converter even while it is
regenerating. The exit gas is cooled to remove elemental sulphur and flows directly
(without reheat) to the No. III position where sulphur condenser is operating in a
sub-dew point mode. The gas mixture from converter flows to the condenser and
the balance gas from condenser directly flow to No. IV position sulphur converter,
where additional sulphur is produced and adsorbed onto the catalyst.
The adsorbed sulphur onto the catalyst is removed from the same converter in the
next cycle. Approximately every 24 hours, one of the three MCRC converters is
changed from sub-dew point mode to regeneration mode, and another one is
changed from regeneration mode to sub-dew point mode. The cycle time may be
changed depending on the loading of sulphur onto the catalyst bed during the actual
operation.
MCRC section is equipped with three converters and three sulphur condensers.
Process gas from Clause section at a temperature of 277°C enters into the MCRC
converter-II and regenerates the catalyst at high temperature by vaporizing sulphur
100
from the catalyst pores, where sulphur was already adsorbed when the converter
was in a sub-dew point mode. Catalyst is regenerated and simultaneously Clause
reaction further proceeds to produce elemental sulphur in the regeneration mode
reactor. The process gas from MCRC Converter II flows directly to sulphur
condenser-III at a temperature of 296°C. The condensed elemental sulphur is
separated in the outlet channel of the sulphur condenser-II, and flows to sulphur pit
via sulphur lock. The balance gas exiting from sulphur condenser III at a
temperature of 143°C flows to sub-dew point converter-III. The Clause reaction
conversion to sulphur occurs at sub-dew point condition where produced sulphur is
adsorbed onto a catalyst pore structure. The leaving process gas from sub-dew
converter III flows to sulphur condenser-IV where Sulphur is condensed and flows
to sulphur pit via sulphur lock. The balance process gas at a temperature of 124°C
leaving sulphur condenser IV flows to converter-IV without re-heating. The clause
reaction continues at a sub-dew point mode to produce sulphur. The process gas
from converter-IV flows directly to condenser-V, from where condensed sulphur is
taken out to sulphur pits via sulphur seal pit. Tail gas from condenser-V flows
directly to a tail gas incinerator, and to the stack. A tail gas coalesce is not necessary
since demister in the outlet channels of the condensers are adequately designed to
separate any sulphur mist. Furthermore, adsorption of sulphur at a sub-dew point
converter catalyst depresses mist formation.
When the catalyst is on a sub-dew point, sulphur converter-III reaches a certain
weight percentage loaded with sulphur, and the converters interchange positions
with converter-II automatically by means of switching valves. The valves operate
on a time cycle control system. Interchange of position of converters-II and III
means the process gas from clause section enters first into converter-III, which is
put in regeneration mode, and the balance gas from converter-III after cooling in
condenser-IV enters into converter-II at a sub-dew point mode. The gas from
converter-II after cooling flows to converter-IV which is still at a sub-dew point
101
mode. Similarly, in the next cycle, converter-IV will be in the regeneration mode
and the other two converters will be put at a sub-dew point mode by switching
valves. Hence, at any time one converter is in regeneration mode and two converters
are in sub-dew point mode.
The switch over cycle of MCRC converters is explained as follows. Due to low
temperature reaction in the sub-dew point converters, and in order to condense as
much sulphur vapour as possible, the last three condensers operate at a low
temperature. This low temperature is obtained by generating steam at a lower
pressure (1.1 kg/cm2) than in the preceding condensers. The steam is condensed in a
steam condenser from which the condensed steam is drained back into the shell side
of the sulphur condensers. The vaporised steam from the condensers is sent to
fin-fan coolers, which consists of two fans, one is fixed drive, and the other is
variable one. The variable drive controlled by a temperature controller placed at the
outlet adjusts the speed of the fan motor. The following are the reactions taking
place in Clause and MCRC Catalytic converters:
The equilibrium clause reaction, Equation (5.1), which is exothermic at the
converters temperatures - is led to an almost total completion over an activated
alumina catalytic bed:
2H2S + SO2 3/x Sx + 2 H2O, (5.1)
where x is mainly 6 or 8 at the operating temperature of the converter.
The above reaction, given by (5.1), takes place on all four catalytic beds. The
hydrolysis of most of COS and CS2 produced in the reaction furnace is also
accomplished in the lower part of the 15th
catalytic bed according to the reactions:
COS + H2O H2S + CO2, (5.2)
102
CS2 + 2 H2O 2 H2S + CO2. (5.3)
The above mentioned endothermic reactions, Equation (5.2) and Equation (5.3),
take place due to high temperature achieved in the catalytic bed despite no
dedicated catalyst being foreseen.
The following sulphur equilibrium reactions take place continuously in vapour
phase during the combustion, the catalytic conversion and the cooling phases,
depending upon the operating temperatures:
3S2 S6, (5.4)
4S2 S8. (5.5)
Condensation of the vapour sulphur produced in thermal and catalytic conversion
steps is achieved in the waste heat boiler and sulphur condensers according to the
following reaction:
S8 (vap) 8 S1 (liq). (5.6)
Sulphur degassing is achieved in the sulphur pit; the following chemical reaction
of poly-sulphides dissociation also takes place:
H2Sx H2S + S (x-1). (5.7)
H2S absorbed in liquid sulphur is stripped and sent to the incinerator via a steam
ejector.
5.2.2.3 Thermal incinerator
In the thermal incinerator, the combustible components in the process gas from the
last condenser of MCRC section and vent gas from sulphur pit are thermally
103
oxidized at high temperature. Thermal Incinerator is common for two-train sulphur
recovery. There are two tail gas streams from two trains and one vent gas stream
from sulphur pit; those are oxidized in incinerator.
The thermal incinerator converts almost all H2S and NH3 in the tail gas so that H2S
in stack gases has concentration lower than 10 Parts Per Million (PPM) in volume.
The exit gas temperature must be kept at 750°C. The proper ratio of air to fuel is
controlled by the ratio-controller. The oxygen content of flue gases is controlled by
the oxygen analyzer. Excess air operation is essential for complete incineration of
process gases. Oxygen content of stack gases is set at 2 % vol. The flue gases are
vented to atmosphere through the vent stack. Shutdown of thermal incinerator due
to any reason will activate the shutdown of SRU.
5.2.2.4 Sulphur storage
Sulphur produced is routed to Run/Down (R/D) line through specially designed
sulphur lock, and collected in the sulphur pit. Sulphur pit is provided with degassing
facility and LP steam heating coil to keep sulphur in molten condition. Sulphur
pump pumps the product sulphur to pelletization unit, where molten sulphur is
pelletized, and stored in silos through bucket conveyor. Sulphur yard is also
provided to produce lump sulphur by pumping molten sulphur to yard by cooling
with cooling water.
5.2.2.5 Initial firing
For start-up of Main and Line Burners, and as a supplementary fuel for line
burners, in case of low H2S content of acid gas, vaporized LPG is used. The
overview of the LPG header unit is shown in Figure 5.3.
104
Liquid LPG from Battery Limits is received in a LPG vaporiser drum and is
vaporized by means of LP steam flowing into the internal coil of the drum. The
level in vaporizer drum is controlled by a level controller acting on a LPG inlet line
valve and its pressure is controlled by a pressure controller, acting on a control
valve on steam inlet to the internal coil. The vaporized LPG is sent to the vaporized
LPG Knock Out (KO) drum, where entrained liquids are separated and from where
it is distributed to main burners and to line burners of both the trains.
Figure 5.3 Overview of LPG Header Unit
105
5.2.3 Process modeling and analysis
As discussed in Chapter 2, here, process modeling is done using Petri nets
considering the whole process as DEDSs. Thus, the model comprises discrete
places and discrete transitions connected by normal arrows.
The presence of a token in a place denotes the condition for starting an operation
or event, firing of a transition denotes the start of the event, and movement of the
token from one place to another represents condition after the operation or event is
completed. Based on these conditions, the modeling of the systems is done.
Figure 5.4 and Figure 5.5 represent the equivalent systems for valve sequence
operations of Main Clause Recovery Concept (MCRC) and vaporized LPG train
modeled using Petri nets.
Dark circles in Figure 5.4 and Figure 5.5 represent the actual valves that are
present in the unit. White circles represent the conditions for the valve, i.e., whether
it is open or closed. Thus, the presence of a token in the respective white circles
shows that the corresponding valves are to be kept open for sequential operation.
The explanation for each place and transitions used for modeling in Figure 5.4 are
mentioned in Table 5.1, and for Figure 5.5, in Table 5.2.
106
Figure 5.4 Petri net model of MCRC valve sequence 1
Figure 5.5 Petri net model of vaporized LPG header train
107
The inlet valves are represented as IN1, IN2 and IN3, the outlet valves
represented as OUT1, OUT2 and OUT3, and intermediate valves are represented as
IM1, IM2, and IM3, for the simplicity of understanding.
Once successful modeling of the system is completed, the algorithms for FDI are
developed, and the corresponding faults (valve failure, sensor failure etc.) can be
detected and identified.
Table 5.1 Description of places/transitions and conditions for model shown in
Figure 5.4 P/T Description P/T Condition
Yes-Token present
No- Token absent
ON-Firing enabled
OFF-Firing disabled
Cycle
1
Cycle
2
Cycle
3
P1 Process Flow gas Yes Yes Yes
T1 Valve 1 ON ON ON ON
P2, P4, P6 Flow gas ready No No No
P3, P5, P7 Mode I, II, III No No No
T2, T3, T4 Inlet valve
1, 2, 3
ON OFF OFF
T8, T9, T10 Intermediate valve
1, 2, 3
ON ON OFF
T5, T6, T7 Outlet valve
1, 2,3
OFF OFF ON
P9, P10, P11 Inlet Valve 1, 2, 3
condition
Yes No No
P12, P13, P14 Outlet valve 1, 2, 3
condition
No No Yes
P15, P16, P17 Intermediate valve
1, 2, 3 condition
Yes Yes No
108
Table 5.2 Description of places/transitions and conditions for model shown in
Figure 5.5 S. No P/T Description
1. P1 KO drum output
2. P2, P3 Valve 1 condition and operation
3. P4, P10 Valve 2 condition and operation
4. P5, P11 Valve 4 condition and operation
5. P6, P7 Pressure switch and Reaction furnace condition
6. P8, P9 Alternate Valve 1 line condition and operation
7. P12 Furnace operation
8. P13, P14 Flare output for alternate line
9. P15, P16 Flare output for main line
10. P17, P18 Flow transmitter output and condition
11. P19, P20 Condition for flare operation and main and alternate
lines
12. P21, P22 Conditions for valve 1 in alternate line
13. P23, P24 Condition for valve 3 in flare output of main and
alternate lines
14. T1-T6 Main line valves
15. T7-T10 Alternate line valves
16. T11, T15 Flare output valves of main line
17 T14, T12 Flare output valves of alternate line
18. T13 Condition for Flow Transmitter
19. T16 Condition for valve 1 in alternate line
5.2.4 Proposed FDI
As mentioned in the earlier section, Petri nets are a powerful tool for system
description. Nevertheless, they can be an useful tool for process monitoring also.
The problem of process monitoring in any Process plant can be stated as follows:
Measurement signals come from the system at a constant scanning rate. When
processing these data, a computer based system should decide on-line in real-time,
109
if an error has occurred or not. To perform this, the computer program needs to
know the “total” process description.
Suppose it is possible to map the structure of the total process using a Petri net, the
transport of physical conservation quantity can be represented by firing of tokens. If
the conservation quantity takes only few discrete values, and if the measured signals
are not noisy, then by comparison, and using the property of reachability, errors can
be identified. This is carried out by finding the initial marking, and comparing it
with the new marking vectors developed after firing of tokens. Any change in the
sequence will result in an error in the corresponding element of the marking vector,
which gives indication of the faulty nature of the process. From the erroneous data
obtained, the corresponding faults can be easily diagnosed.
Even when the signals received from the system are noisy, the property of
observability can be considered as the criterion for diagnosis where suitable
threshold limit has to be satisfied. If the error measured as mentioned above is
higher than the limit, it means that a fault has occurred. Then suitable diagnosis is
done to classify the faults.
Normally in Petri nets, presence of place faults are similar to sensor faults and the
presence of transition faults are equivalent to valve failure. Based on these
conditions, the fault monitoring algorithm for identifying faults in a MCRC model
shown in Figure 5.4 is given as follows:
110
Step (i): With respect to the modeled control structure, the initial token content per
place is found by determining the initial marking vector denoted by
Mi(0).
Step (ii): After determining Mi(0), the actual number of tokens arising in the
running of the process is found out, which is denoted by Mi(k).
Step (iii): The difference between the marking vectors are calculated i.e. Mi(k)-
Mi(0).
Step (iv): If the difference is zero, then the system is considered as fault free.
Step (v): If not, the corresponding place fault has occurred, and an algorithm is
developed to identify the faulty place in the structure modeled earlier.
Figure 5.6 and Figure5.7 show the development of the system for fault diagnosis
modeled in a MATLAB environment.
Figure 5.6 View of original model of MCRC valve sequence 1
111
Figure 5.7 View of fault model of MCRC valve sequence 1
The presence of a place fault is diagnosed by considering and checking the
reachability of markings. For diagnosing the transition faults, evaluation of
performance measures are considered, which will be discussed in Chapter 6. The
algorithm developed in MATLAB environment is shown in Figure 5.8. Likewise,
Figure 5.9 and Figure 5.10 show the views of the simulation results obtained for a
faulty and a non-faulty process.
As observed in Figure 5.8, using the commands available in NETLAB toolbox
[81] in a MATLAB environment, the values of pre-incidence, post-incidence, and
incidence matrices are determined similar to the method discussed in section 3.2.2
in Chapter 3.
Faulty
place P10
112
Figure 5.8 View of algorithm developed for FDI in MATLAB
Once the matrices are known, the values of place markings can be found out from
(2.4) as described in section 2.2.1 in Chapter 2. Now, for any fault occurring in the
model (especially place faults), the value of place marking obtained when fault
occurs will not be the same as obtained for a non-faulty condition.
Thus, the non-presence of a null value in the difference between the place marking
vectors shows that the corresponding place is under fault. As discussed in this case,
the presence of a fault in place P10 is determined and the simulated results are
shown in Figure 5.9.
113
Figure 5.9 View of Simulation result developed in MATLAB for a faulty
process
Figure 5.10 View of Simulation result developed in MATLAB for a non-faulty
process
The modeled
system is faulty
and faulty place
is P10
The modeled
system is fault free
114
For the purpose of simulating the processes modeled using Petri nets as shown in
Figure 5.11, SIRPHYCO software [82] is utilized. The simulation window shows
time taken for firing of each valve to operate in the MCRC sequence. Moreover, the
graphs also depict the sequence of operation, i.e., initially P3 is marked followed by
P5, P7 and finally outputted through P8. Similar graphs can be obtained by
considering the other valve sequences of the MCRC.
Figure 5.11 Results for MCRC valve sequence 1
Figure 5.12 Results for MCRC valve sequence 2
115
The marking evolution of the places considering valve sequences 2 and 3 are
simulated and shown in Figure 5.12 and Figure 5.13, respectively.
Figure 5.13 Results for MCRC valve sequence 3
5.3 PROPOSED ESTIMATION BASED FDI IN SEWAGE TREATMENT
PLANT
The second real-time application considered here is a sewage water treatment plant
which is analogous to a three tank benchmark system [83]. Sewage treatment, or
domestic waste-water treatment, is the process of removing contaminants from
waste-water and household sewage, both runoff (effluents) and domestic. It includes
physical, chemical, and biological processes to remove physical, chemical and
biological contaminants. Its objective is to produce a waste stream (or treated
effluent) and a solid waste or sludge suitable amount for discharge or reuse back
into the environment. This material is often inadvertently contaminated with many
toxic organic and inorganic compounds.
116
Here, the objective is to develop model-based analysis for a typical STP plant and
to achieve estimation based FDI.
5.3.1 Process description
STP plant [84] as shown in Figure 5.14 comprises three major processes which
follow:
Primary Treatment,
Secondary Treatment, and
Tertiary Treatment.
Primary Treatment:- In the primary sedimentation stage, sewage flows through
large tanks, commonly called "primary clarifiers", or "primary sedimentation tanks"
[84]. The tanks are large enough that sludge can settle and floating material, such as
grease and oils can rise to the surface and be skimmed off. The main purpose of
primary sedimentation stage is to produce both a generally homogeneous liquid,
capable of being treated biologically and a sludge that can be separately treated or
processed.
Secondary Treatment: - The rotating disks support the growth of bacteria and
micro-organisms present in the sewage, which breakdown and stabilize organic
pollutants. To be successful, micro-organisms need both oxygen to live and food to
grow. Oxygen is obtained from the atmosphere as the disks rotate. As micro-
organisms grow, they build up on the media until they are sloughed off due to shear
forces provided by the rotating discs in the sewage. The spinning mesh wheel
develops a bio-film coating of microorganisms that feed on the suspended wastes in
the aquarium water and are also exposed to the atmosphere as the wheel rotates.
117
This is especially good at removing waste urea and ammonia urinated into the
aquarium water by fish and other species [84].
Tertiary Treatment: - The purpose of tertiary treatment is to provide a final
treatment stage to raise the effluent quality before it is discharged to the receiving
environment (sea, river, lake, ground, etc.). More than one tertiary treatment process
may be used at any treatment plant. If disinfection is practiced, it is always the final
process. It is also called "effluent polishing" [84].
Figure 5.14 Overview of a STP Plant [84]
The STP plant shown in Figure 5.14 is analogous to a three tank system as shown
in Figure 5.15. Hence, based on this consideration, modeling and analysis for the
STP plant is done by considering the plant as a three tank system. The details of the
three tank system are as follows:
118
The water to be treated is given to tank 1 through pump 1.
The fluid from tank 1 after sedimentation is passed over to tank 2 through
valve 1 with constant out flow.
After secondary treatment in tank 2, the treated water flows to tank 3.
In tank 3, “effluent polishing” takes place by the addition of disinfectant
through pump 2.
Figure 5.15 Equivalent three tank system model for the STP plant shown in
Figure 5.14
The other details with respect to the process parameters are as follows:
Tank 1 - Maximum level is 15 cm.
Tank 2 - Maximum level is 15 cm.
Tank 3 - Maximum level is 25 cm.
The considered tank is cylindrical in shape.
Valves used are on-off valve.
Level and flow sensors are used to sense their corresponding parameters.
Pump 1 has a flow rate of 3 cm3/sec.
Pump 2 has a flow rate of 2 cm3/sec.
119
Valve 1 and Valve 2 have an outflow rate of 4 cm3/sec and 5 cm
3/sec,
respectively.
Note: A raise of 1 unit in level is seen if the volume filled is 1 cm3.
In tank 2 there is a delay given for the stirrer action.
5.3.2 Process modeling and analysis
The Petri net model consists of a hybrid net structure which has five discrete
places and three continuous places as shown in Figure 5.16. The three continuous
places signify the three tank system, and the four continuous transitions with
maximal speed signify valves and pumps. The discrete place signifies the logic with
which the system should work. The movement of token in discrete places resembles
the flow of logic in the system. The first place is analogous to pump 1, and the next
is valve 1 open and valve 2 closed. The third place is valve 2 open and valve 1
closed. The fourth one is a delay, and the last one signifies the second pump. A
token in each of the places means that the corresponding condition comes true.
Figure 5.16 Petri net model of the system shown in Figure 5.15
120
The details of the corresponding places and their logic are shown in Table 5.3.
Table 5.3 Details of corresponding places for the model as shown in Figure 5.16
PLACE LOGIC
P1 Pump 1 ON
P3 Valve 1 open ,Valve 2 close
P4 Valve 1 close , Valve 2 open
P7 Pump 2 ON
P8 Delay
P2 Tank 1
P5 Tank 2
P6 Tank 3
As observed in Figure 5.16, the equivalent of pump 1 initially ON is shown by the
presence of a token in place, P1. The level in tank 1 increases in accordance with the
maximal speed of transition, T1. When the level in tank 1 reaches 15 cm, transition,
T3, fires, and token from place, P4, moves to place, P3, which means valve 1 is open
thus fluid flows from tank 1 to tank 2 where there is a delay given for the purpose
of secondary treatment. The same procedure is repeated in valve 2, thus fluid moves
to tank 3. Now pump 2 is activated and the disinfectant is allowed to flow through
the tank to reach a height of 25 cm.
Figure 5.17 shows the response of the tanks with respect to time, i.e., the response
of the places. The response of tank 1 and tank 2 are shown with respect to the
response obtained for place, P2, and place, P5. Similarly, the response of tank 3 is
obtained by considering the response of place, P6.
121
(a)
(b)
(c)
Figure 5.17 Responses of (a) tank 1, (b) tank 2 and (c) tank 3 obtained with respect
to places P2, P5 and P6
122
5.3.3 Proposed estimation based FDI
The proposed fault diagnosis algorithm uses the status signal of the devices from
the Programmable Logic Control (PLC) unit at start-up to estimate the output and
predict the fault. This is otherwise called as observer technique as discussed in
section 2.4 in Chapter 2. The observer does estimation and prediction on how the
output might turn out in case of an error.
The proposed methodology is adopted in developing a GUI toolbox in MATLAB
for performing FDI in a three tank system. Figure 5.18 shows the display of the
front panel for the three tank model developed and analysis done for the purpose of
fault diagnosis.
Figure 5.18 Overview of front panel for three tank system
123
There are mainly three panels at the front end.
Algorithm Panel
Graphical Panel
Others Panel
Algorithm Panel:- The first one deals with the algorithm part. This panel contains
two push buttons that are linked to two key algorithms. The first push button in the
panel is the index button. This button has the link to all coding, where all the
schemes, i.e., estimation, evolution, error marking, and fault diagnosis algorithm are
combined into a single program that has directional links to all the other programs.
The second push button in the panel is the one that contains link to the continuous
evolution part. This module can be executed by the push of this button.
Graphical Panel: - This panel mainly deals with the analysis graphs and evolution of
marking of each place. There are six push buttons in this panel. The first one shows
the evolution of a particular place in accordance with time, when the system is
normal. Second button is linked to the original marking of a faulty system (if it
contains any fault). The third one is the evolution of a fault obtained from the error
matrix (the difference between marking evolution of the estimate and the faulty
system). The fifth and sixth are 3 dimensional plots that are used to identify faults
easily. They provide us better perspective of the faults obtained.
Others Panel: - This panel can be easily comprehended. The index button in the
algorithm panel guides us through all the required steps to find out the fault, but this
panel provides us access to all specific algorithms, i.e., all specific modules like
discrete evolution, estimation, etc., and it gives us option to look at all the variables
at once so that we can view the results again for the purpose of documentation, etc.
124
The details of the algorithms applied in the developed GUI toolbox for achieving
FDI are as follows:
Estimation: -The main purpose of estimation is to estimate the evolution of the
marking using the observer technique. The main purpose of the estimator algorithm
is to develop an observer, observer reachabilty graph, and obtain a change in the
system with respect to the observer. The algorithm for the same is as follows:
Algorithm 1-Discrete evolution
Discrete evolution- Discrete evolution is a module developed to obtain the working
of a proper system without a fault. As quoted earlier, Petri net has a strong
mathematical background. Through the addition of old marking with the incidence
matrix column of the particular transition fired (known by the word), the new
marking of the model can be obtained. The initial conditions assumed are as
follows:
Net structure is known,
Transition firing is observable,
Initial marking is not known, and
Word is known.
The algorithm is as follows:
Step 1: Initial estimate Me = 0 , let word W = W0.
Step 2: Wait until transition, t, fires.
Step 3: Update previous estimate: M 'e = max{Me, pre(:, t)}.
Step 4: New estimate: Me = M 'e + c(:, t).
Step 5: W = transition fired Wt ; goto Step 2 until word is complete.
Algorithm 2- Continuous evolution
Continuous evolution- Continuous evolution is the module developed in order to
obtain the evolution of the continuous places. The initial conditions assumed are as
follows:
125
Initial marking of the fault-free system is known,
Word is known, and
Pre, Post, Incidence matrix is known.
The algorithm is as follows:
Step 1: Input the initial matrix M0.
Step 2: Input the transition, t, to be fired.
Step 3: The new marking, M = M0 + C (:, t).
Step 4: Repeat the procedure till the word is complete.
Algorithm 3 –Proposed FDI
Proposed FDI- The concept of FDI as explained before utilizes the estimated value
from the observer to develop the observability graph. This graph contains a variable
called observability error, which is the difference between the original marking and
the estimated marking. This error changes in the case of an occurrence of a fault. In
this way, the place where the fault has occurred is determined. The initial conditions
assumed are as follows:
Net structure is known,
Initial marking is known, and
Transition speed is known.
The algorithm is as follows:
Step 1: Input initial marking, transition speed, Mc, v.
Step 2: Add the speed to the marking M 'c = Mc + v.
Step 3: Repeat the procedure till desired height (marking) is obtained.
Step 4: Repeat the procedure for all continuous places.
Algorithm 4-Estimation
The initial conditions assumed are as follows:
Net structure is known,
Initial marking of the faulty system is known, and
The word is known.
126
The algorithm is as follows:
Step 1: Input the initial marking Mf, word W.
Step 2: The column of incidence matrix corresponding to particular transition
fired is added to the initial marking M 'f = Mf + C (:, t).
Step 3: The observability graph is developed by generating the estimation
error.
Step 4: The fault is narrowed down by generating the residue of estimate
error.
Figure 5.22 and Figure 5.23 show various graphs obtained for the purpose of
analysis. Here, three 3-dimentional plots have been plotted which are useful for the
purpose of visualizing a fault. For good understanding of various conditions of a
fault, and to analyze the performance of the algorithm, the input to the sequence is
given as [ 456321 ]. This is the sequence in which transitions fire. Once
this sequence of firing is given as the input, the system generates the estimated
marking matrix as shown in Figure 5.19.
Figure 5.19 Window showing the estimate matrix for sequence 1
Estimation is done as explained in Chapter 4. Once the estimated matrix is
found, the next procedure is to give initial marking of the proper model, i.e., the
fault-free system and the details of the marking given in Table 5.4.
127
Table 5.4 Details of marking for fault free system
Place Marking Meaning
P1 1 Pump 1 ON
P3 0 Valve 1 open ,Valve 2 close ( no app)
P4 1 Valve 1 close , Valve 2 open
P7 0 Pump 2 Off
P8 0 Delay off
Thus, the marking to be given is [ 00101 ]. When this input is given, the
module in itself develops the discrete evolution, i.e., the marking matrix of the
fault-free system, and then the error matrix is also developed as shown in Figure
5.20.
Figure 5.20 Window showing the discrete evolution of error matrix and marking
for sequence 1
Now, after these two matrices are developed, the input of the fault system is given
(The faults in this case are assumed for the purpose of simulation). Let a fault in
pump 1 be assumed, and the supply and status signal from the pump becomes 0
instead of 1, i.e., marking is [ 00100 ]. Then, the fault occurred is shown in
Figure 5.22. The details of the original marking and error marking in real time are
shown in Figure 5.21.
128
Figure 5.21 Window showing original marking and error marking in real time
The first row of the residue matrix as shown in Figure 5.21 shows a constant value
which means an error has occurred in this system as that place is said to be
deadlocked. When this input is given to the system, a fault place is located, and the
corresponding analysis graph, and place of fault is found with minimum
information about the system. The results obtained from simulation of the fault are
shown in Figure 5.22 and Figure 5.23.
Figure 5.22 Details of faults occurring with respect to time
129
Figure 5.23 3-D plot of the fault
The constant value of marking in the 3-D plot shown in Figure 5.23 shows a
typical fault. Here, it is to be noted that multiple faults can also be detected at the
same time. The details of marking to be given for detecting typical place faults are
shown in Table 5.5.
Table 5.5 Details of input sequence marking for detecting typical place faults
MARKING FAULT
[ 00100 ] 1st place (P1)
[ 00111 ] 3rd
place (P3)
[ 00001 ] 4th
place (P4)
[ 01101 ] 7th
place (P7)
[ 10101 ] 8th
place(P8)
130
5.4 PROPOSED ESTIMATION BASED FDI AND FTC IN BOTTLE
FILLING PLANT (BFP)
The third real-time application considered for study in this research work is the
Bottle Filling Plant (BFP). The main purpose to consider BFP for study in this
research is because of the fact that any BFP resembles Hybrid Dynamic Systems
(HDSs) [85]. As discussed in Chapter 2, a hybrid system is one that exhibits both
continuous and discrete dynamic behavior, i.e., a system that can
both flow (described by a differential equation) and jump (described by a difference
equation or control graph). Often, the term "hybrid dynamic system" is used, to
distinguish over hybrid systems, such as those that combine neural nets and fuzzy
logic, or electrical and mechanical drivelines. A hybrid system has the benefit of
encompassing a larger class of systems within its structure, allowing more
flexibility in modeling dynamic phenomena.
In general, the state of a hybrid system is defined by the values of the continuous
variables and a discrete control mode. The state changes either continuously
according to a flow condition, or discretely according to a control graph.
Continuous flow is permitted as long as the so-called invariants hold, while discrete
transitions can occur as soon as the given jump conditions are satisfied. Discrete
transition may be associated with events. HDSs thus can be used to emulate the
functioning of a plant more effectively. Hence, for the above reasons, modeling and
analysis of a BFP using Hybrid Petri nets is obtained to achieve FDI and FTC.
5.4.1 Process description
BFP [86] considered here as shown in Figure 5.24, includes both continuous
processes and discrete processes.
131
Discrete Processes:- The discrete processes are those which involve a jump
of control from one part to another of a system. A bottle-filling system consists of
various discrete processes, the systematic movement of bottles one by one, the
detection of sensor, Opening and closing of valves for filling, Counting of bottles,
collection and packaging of bottles.
Continuous Processes:- The continuous process is the filling of the bottle
from the reservoir. Although the flow of liquid from the reservoir to the bottle is not
continuous, it is considered to be so through “Fluidification”. Through
fluidification, the modeling of the bottle-filling system as a hybrid Petri net model
becomes easier.
Due to these reasons, the bottle-filling systems are considered to be HDSs. The
various parts of BFP which have been taken into consideration are: (1) Switch,
(2) Conveyor system, (3) IR sensor, and (4) Reservoir tanks.
Conveyor system:- A conveyor system is a common piece of mechanical
handling equipment that moves materials from one location to another. Conveyors
are especially useful in applications involving the transportation of heavy or bulky
materials. Conveyor systems allow quick and efficient transportation for a wide
variety of materials, which make them very popular in material
handling and packaging industries. Many kinds of conveyer systems are available,
and are used according to the various needs of different industries. Here, a normal
conveyor system is considered to transport the bottles in the packaging industry.
The conveyor system is controlled through a conveyor motor, and the control of the
motor helps moving the conveyor belt. Bottles are placed over the conveyor at
regular intervals and according to bottle-filling, the conveyor automatically moves
to bring the next bottle into position once a bottle gets filled.
132
IR Sensor: - Infra Red (IR) sensors are generally used for object detection
and not for distance measurement. The basic idea is to send infra red light through
IR-LEDs, which is then reflected by any object in front of the sensor. In the system
shown in Figure 5.24, IR sensor is considered to sense the presence of a bottle in the
conveyor system before bottle-filling takes place.
Reservoir Tanks: - The reservoir tanks are used to give a constant supply of
liquid to the filling system. There are reservoir systems with level monitoring and
automatic refilling. The reservoir system considered here fills itself automatically
when the amount of liquid drained through it becomes large.
Figure 5.24 Overview of a Bottle Filling Process plant [86]
133
5.4.2 Modeling and analysis of process
Here, in this research work, two different prototypes of BFP have been modeled
and analysed. The first one is a multiple BFP, and the second one is ratio filling
BFP. The details of modeling and analysis of both are explained separately in the
following subsections.
5.4.2.1 Modeling and analysis of multiple BFP
The Petri net model for multiple BFP consists of a hybrid net structure which has
five discrete places and three continuous places as shown in Figure 5.25. The three
continuous places signify the reservoir and filling nozzles, and the two continuous
transitions with maximal speed signify the valves. The discrete places signify the
logic with which the system should work. The movement of token in the discrete
places resembles the flow of logic in the system. The First place is analogous to the
main switch of the bottle-filling system. The conveyor motor is denoted by a
discrete place, P2. The motor has bottles present on the top of it, and rotates with the
capacity load. The third place is analogous to the Infra Red (IR) sensor; the IR
sensor is used to find whether the bottle is in position or not. The sixth and seventh
places form a loop for continuous bottle-filling. Places, P16, P17 act as bottle count
and the place, P14, act as a packing unit, where every two bottles are packed
together. A token in each of the places means that the corresponding condition
becomes true.
The equivalent Petri net model is shown in Figure 5.25. The model considered here
is a multiple bottle filling process, where, 30 litre capacity bottles and 40 litre
capacity bottles are filled simultaneously in two different parallel set-ups. For a
transition to fire, there should be a token in every connecting place.
134
Figure 5.25 Equivalent Hybrid Petri net model developed for the process shown in Figure 5.24
135
The algorithm is explained as follows:
Main switch is ON, i.e., there is presence of a token, and the conveyor motor
starts to run and the bottles present on top of them appear at the filling position
in both the parallel modules.
Bottle is detected when IR sensor senses the presence of a bottle and then
subsequently, the conveyor motor stops.
Filling system is initiated and the valves of the 30 litre as well as 40 litre
capacities are opened, which means liquid flows from the reservoir to the
bottles.
Delay is provided for the operation of bottle-filling completely.
Bottle count is updated for both the systems as soon as the 30 litre and 40 litre
bottles are filled.
When bottle gets filled, it arrives at the collection unit, where two bottles are
filled, and finally appear at the packaging unit, i.e., when two tokens are
collected in the place, then bottles are sent to a carton for packaging.
By the time 40 litre bottle is filled twice, the 30 litre bottle will be filled
approximately 3 times, and hence the reservoir is refilled as soon as the two
tokens are present in the bottle collection place.
136
Table 5.6 shows corresponding places and their logic with respect to the developed
model.
Table 5.6 Places and transitions with their corresponding logic
Figure 5.26 shows the response of bottle-filling with respect to time, i.e., the
response of the places, P4 and P5. The graphs as shown in Figure 5.26 indicate the
bottle-filling as well as the bottle counts that are tracked every instant of time.
Place/Transition Logic details
P1, P11 The system is switched “ON”
P2, P10 The conveyor motor starts to run
P3, P9 The IR sensor senses the bottle
P6, P7, P12, P13 Bottle filling start; valve open; delay
P16, P17 Bottle counter
P14 Bottle collection and packing
P15 Reservoir Tank refill
P5 Reservoir Tank
P4 30 litre capacity bottle filling
P5 40 litre capacity bottle filling
T5 Valve for 30 litre bottle filling
T7 Valve for 40 litre bottle filling
T15 Packing unit activation
T16 Refill for reservoir
137
Figure 5.26 Response of bottle filling Places
5.4.2.2 Modeling and analysis of ratio based BFP
Apart from the previous model as discussed in section 5.4.2.1, a ratio based BFP is
designed, modeled and analyzed through Hybrid Petri net structure and explained in
detail in this section. Here, the Petri net model consists of four continuous places
and nine discrete places as shown in Figure 5.27.
Place, P1, indicates the “start/stop” switch. Place, P2, indicates the conveyor motor,
Place, P3, is analogous to an IR sensor; Places, P5 and P6 form a loop for continuous
bottle-filling. This controls the opening and closing of valves, and thus controls the
ratio of the liquid that gets filled in the bottle. The Continuous places, P4 and P9 are
analogous to the reservoir of different liquids that need to be mixed in a ratio.
Places P8 and P10 are the filling places that gets filled according to the ratio
designed. A token in each of the places means that the corresponding condition is
138
true. The continuous transitions fired at a maximal speed signify the valves of the
filling system.
Both systems are modeled through Constant Continuous Petri Net (CCPN) which
fire at a constant maximal speed throughout the process. In the ratio BFP designing
of the speed for firing of continuous transitions is of utmost importance as it decides
the ratio of liquid to be mixed. In the considered model, the maximal speed is set in
such a way that a ratio of 1:2 is obtained, i.e., for one unit of one liquid, two units of
another liquid should be mixed together. The Petri net model of the ratio based BFP
is shown in Figure 5.27. Table 5.7 shows corresponding places and transitions and
logic they signify.
Table 5.7 Places and transitions with their corresponding logic for model
shown in Figure 5.27
Place/ Transition Logic details
P1 The system is switched “ON”
P2 The conveyor motor starts to run
P3 The IR sensor senses the bottle
P5, P6 Bottle filling start; valve open; delay
P7, P11 Mixing of liquids
P12 Bottle filling
P13 Bottle is collected and sent to
packaging
P4,P9 Reservoir Tanks of two different
liquids
P8 Liquid A
P10 Liquid B
T6,T8 Ratio adjustment
T10 Mixing of two liquids inside bottle
T11 Packing unit activated
139
Figure 5.27 Petri net model of the ratio-based BFP
140
The algorithm is explained as follows:
Main switch is ON, i.e., there is the presence of a token and the conveyor motor
starts to run, and the bottles present on top of them appear at the filling position.
Bottle detected when IR sensor senses the presence of a bottle, and then the
conveyor motor stops.
Filling system is initiated and the valves of Liquid A and Liquid B are opened,
which means the liquids A and B flow from the reservoir to the filling places
according to the firing rates of the continuous transitions, which determine the
ratio of the liquid mixture.
Delay is given for the purpose of maintaining proper ratio during the mixing of
two liquids.
Once the required ratio is obtained, i.e., for one part of liquid A, two parts of
liquid B are to be added. The bottle is sent to a collecting unit.
In that unit, when two tokens are collected in a place, the bottles are sent to a
carton for packaging.
Figure 5.28 shows the response of liquid A filling with respect to time, i.e., the
response of the place, P8.
Figure 5.28 Response of bottle filling at place, P8
141
The response of the filling at place, P10, is shown in Figure 5.29. Here, place, P10,
refers to the filling of liquid B.
Figure 5.29 Response of bottle filling at place, P10
Figure 5.30 shows another graph that displays the number of bottles filled with
respect to time. The distinct peak at a particular time instant refers to the fact that a
bottle is being filled at that time instant.
Figure 5.30 Number of bottles filled with respect to time at place P12
142
5.4.2.3 Proposed FDI and FTC
The proposed FDI and FTC block diagram as shown in Figure 5.31 comprise
fault diagnosis, fault tolerance and fault isolation. As detailed earlier once faults are
identified, they are checked if the faults are tolerant. If they are tolerant, then the
system runs normally. When the fault is intolerable, then the redundant path is
chosen and the process carries on without the system being halted. This is called
fault Isolation. The various components of the block diagram are as follows:
Optimized system model:- The optimized system model refers to the model that
is designed mathematically as well as through Petri nets according to the
requirements fulfilling the ideal conditions.
Diagnoser model:- The diagnoser model is one that is designed keeping in mind
the various probable faults that might occur in the system due to many factors
such as prolonged running, friction, level violation, etc.
Fault diagnosis: - Once the ideal model and the faulty models are designed, the
next process is comparison of the ideal model with the faulty model and finding
out reason for faults to have occurred. The various faults are taken care by
applying the fault tolerance algorithm procedures.
Fault tolerance:- Fault tolerance is one of the important steps as it checks the
extent of the fault, and prevents the system from stalling due to minor faults and
through the algorithm, a threshold is set for the faults and minor faults that do
not cross the threshold do not affect the smooth functioning of the system. Only
when the fault crosses the threshold, further steps are taken.
143
Figure 5.31 Block diagram of Fault Diagnosis and Identification (FDI) and Fault Tolerant Control (FTC)
144
The developed fault tolerance algorithm to achieve this condition is shown in
Figure 5.32.
Figure 5.32 Algorithm to achieve Fault Tolerant Control (FTC)
The condition to be satisfied for fault tolerance of BFP is given as L. q ≤ b,
where L represents the nc × n integer matrix, b represents the nc × 1 integer vector,
nc is the number of constraints and q is the marking vector of the Petri net. Based on
Check if
ideal marking –
obtained marking =0
System is in
IDEAL
working
condition
System is Faulty
Check whether the fault in the system is
within the tolerant limit.
If
L. q ≤ b
L. q ≤ b
L
represen
ts nc × n
integer
matrix,
b
represen
ts nc × 1
integer
vector,
nc is the
Fault is
within the
tolerable
limit
Faults have exceeded the tolerant limit,
and hence isolate the faulty part of the
system.
Obtain the ideal condition matrix and
current condition incidence matrix of the
system and check for error
Yes
No
Yes
No
END
145
the condition above, the ideal condition for constraint matrix L and the marking
vector q are estimated and given as
100
001
140
60
30
≤
40
30. It can be observed
here that if this condition is satisfied, the system is within the tolerable limit and the
system runs normally. (The tolerable limit is that the product (L.q) should be within
30 litre and 40 litre, respectively, and if it exceeds the value, then the bottle
overflows, and the filling process is isolated).
Once fault tolerant algorithm is developed, and the tolerance level of the fault is
checked according to the degree of tolerance, the system decides whether to tolerate
the fault or isolate it from the main process for repair.
If the fault is not within tolerance limits, fault isolation is carried out. This
process involves designing a redundant path for the process to take in case of a
conflict in the system. The design of the redundant path has been done using a
software called HPSIM [87], and this allows the faulty part to be isolated from the
main system, and the rest of the process carries out as normal. The main advantage
is that even though there is fault, the system does not come to a halt, but it continues
working. The design of the redundant path is shown by considering an example as
shown in Figure 5.33.
146
Figure 5.33 Example of fault isolation using HPSIM
As observed in Figure 5.33, there is a redundant path that is present for each
flow of control so that if any of which fails, then an alternative path is taken and the
system does not halt.
In this part of the research work, C programming has been used as a front end
interface to identify the current status of the system. A program has been developed
which compares the ideal condition markings and the obtained markings of the
bottle-filling system. This program compares these two markings of the system and
it identifies the working condition of the system, and indicates the user whether the
bottle is filled correctly, or any fault is detected in the system. The erroneous system
response is the vital analyzing tool from which the occurrence of a fault is detected.
The faults that could possibly occur are either due to overflow of the bottle due to
the conveyor not moving (Place fault), errors due to the improper functioning of the
147
IR sensor, and fault occurrence due to the improper filling valve operation
(transition fault). The graphs in Figure 5.34 show the response of faulty system.
Figure 5.34 Response of a faulty system
From Figure 5.34, it can be observed that:
Bottle overflow occurs as the marking cross 30, and it should be maintained
within the threshold, i.e., 30 markings (litres).
It is clear that when there is a fault in the system, the bottle fills up haphazardly.
148
Apart from graphical analysis, in this part of the research work, mathematical
analysis is also done and results are presented. As discussed earlier, the ideal
marking for the BFP process considered is given as 3016030
PPP 854
, where capacity
of bottle - A is 30 units, capacity of bottle-B is 40 units, P4 is the place analogous to
bottle-filling in bottle – A and P5 is analogous to bottle-filling in bottle – B.
Now, the condition for bottle overflow is given as
Faulty marking
3016032
PPP 854
,
For partial bottle-fiiling, the normal and faulty markings are given by
Normal marking
4013410
PPP 854
,
Faulty marking
3613410
PPP 854
.
Similarly, for bottle missing, the values of normal marking and faulty markings are
given as follows:
Normal marking
000000100100
PPPPPPPPPPPP 151413121110976321
Faulty marking
Marking P4 indicates bottle
overflow
Marking P8 indicates partial filling
of bottle B
149
000000100000
PPPPPPPPPPPP 151413121110976321
Marking P3 represents missing of bottle
Hence, in this Chapter, a detailed description of the major contributions made in
this research work with respect to achieve FDI and FTC in real-time systems is
presented. In Chapter 6, the details of evaluation of performance measures for
model checking and analysis are discussed.