MP- I01

AJI-I I

O- 100%
7
Distributed Centre|
System Testing
Dr. John P. Andersen
Simons-Eastern Consultants

An efficient method for doing system is to verify the system con- puts together. As implied, this
100% testing of distributed con- figuration, and applications soft- method requires modification of
trol systems (DC$) and their con- ware and hardware, including all the configuration to be tested
figuration before installation at I / O channels. In the past, several (which is obg,cted to by some cli-
the client's plant is needed. In the different methods have been used ents for obvious reasons) and that
past this was done with hond- by vendors of control equipment the controller has enough remain-
held signal generators, meters, to verify equipment operation be- ing space to implement the simu-
and pateh boards. This method is fore installation at the plant site. lation. This method can do a good
nen3 time consuming and does not
test the response of the controls to
real-time conditions. Due to Si-
mons-Eastem's concern for the
The objective of staging a system is to verify the
quality of delivered systems, a system configuration, and applications software and
project to develop an ~ficient hardware, including all I/O channels.
testing program was started. The
system as developed allows simu-
The most common method is job of checking proper operation
lation of the process in real time.
the use of hand-held meters to of interlocked loops and some
This system is implemented using
a real-time expert system, which
monitor outputs and to supply process dynamics. It is also a
signals to inputs or dedicated good technique to be used for ini-
provides the process simulation,
connected to a computer-based patch boards with patentiometers tial operator training. This
I10 subsystem, which is hard- and meters wired to the I/O. This method, however, does not verify
wired to the control system being performs the job of checking all proper assignment or operation of
tested. The expert system allows I/O channels and the proper op- I / O channels and is not easily or
the process simulations to be de- eration of simple loops. This quickly chansed in many systems.
veloped and changed quickly method becomes very tedious and Another drawback is that it as-
while testing of the DCS is in time consuming when many in- sumes the controller software
progress. This paper will describe terlocks are involved, and it does modifications mcluired for simu-
in more detail the design of the not check how the loops will re- lation will be accurately "undone"
system and its results. spond to process dynamics or or removed afterwards without
loop interaction. Because this the inadvertent introduction of
method is so cumbersome, some new software errors that go un-
vendors provide only a partial t~ted.
iNTRODUC~ON A tb~d Lecb_niqueis to develop
random check of i / 0 on the sys-
Part of Simons-Eastern's ser- tern floor as standard practice. the process simulation in an exter-
vices is the configuration and Another method configures nal computer and tie it to the con-
sta~ng of control systems for its into the controllers simple simula- trol systems data highway. The
clients. The ob~ctive of staging a tions that s o ~ w i ~ inputs and out- simulation computer then

ISSN 0019-0578/91/02/004115/$2.50© ISA 1991 VOLUME 30 ® NUMBER 2 * 1991 4~

DCS:SELECTION,IMPLEMENTATION,AND MAXIMIZATION

changes the inputs and reads the (3.) The ability to do signal control system. The strips are for
outputs from across the highway. filtering, engineering unit both analog and discreto signals.
This is not possible with all con- conversion, and signal charac- The signals are brought back to
trol systems and requires modifi- terization with minimal load- the I / O c o n v e r s i o n / i s o l a t i o n
cation of the c o n t r o l l e r ing of the simulation modules by a 19 twisted pair cable
configurations. Simulations are computer. with DB37 connectors. This al-
not easily changed by untrained (4.) Prewiring of the control sys- lows the control system to be pre-
engineers, and physical I/O oper- tem beyond the I/O capacity wired to the termination strips,
ation or assignment is not veri- of the I/O subsystem and the and, w h e n testing begins, the
fied. It is very expensive in most ability to move from one I/O twisted pair cables are moved
cases because it usually requires a group to the next quickly. from one termination strip to the
vendor-specific intelligent high- other as required.
way interface. This interface is re- (5.) The ability to automate the as- Internally, the intelligent I/O
quired to allow reconfiguration of signment of control system subsystem operates within -10 to
the controllers to look to the high- I/O channels to the I/O sub- +10 volts at a 14 bit A/D, D / A
way for receipt of I/O instead of system from the engineering resolution. However, I / O conver-
its normal I/O cards and termina- database and generate wiring sion modules can be inserted to
tions. Again, the result is that the lists for prewiring the system. a c c o m m o d a t e almost a n y re-
actual unmodified configuration (6.) The ability for any engineer quired DCS or field sensor signal
to be used in the plant is not the with minimal training to use range. For example, the I / 0 con-
same configuration tested. the system to develop simula- version modules on the analog
Simons-Eastern configures tions for checkout purposes. side can convert the 4-20 mA sig-
many different control systems nals to 1 to 5 volts and, likewise,
and is very concerned about the (7.) The ability to change simula-
tions on the fly without re- the 0 to 10 volt outputs to 4-20
quality and proper operation of mA. The system can be set up to
the systems delivered to the compiling.
look like either four-wire or two-
client's site. It is very important (8.) The ability to handle a wide wire transmitters by simply plug-
that all configuration errors and range of simulation/testing ging in a personality plug in the
hardware problems are found be- objective~ from the insertion tern~,~aLion stl~p. Millivolt sig-
fore start-up. The finding and of basic process models to full nals can also be provided, again
correcting of configuration errors dynamic unit process models by plugging in a personality plug.
in the field is time consuming and when needed. On the discrete side the isolation
expensive because of the lost pro-
modules convert 24-volt or 110-
duction they may cause. A pro-
gram was started to develop an volt signals to 5-volt level signals
HARDWARECONFIGURATION for the ASCS system. Relay mod-
efficient system that could be used
to test any control system. The The Advanced Systems Check- ules are used to simulate field
following sections will describe out Simulator (ASCS) shown in switches.
the design of the system and its Figure 1 is designed to meet the The system is currently con-
Use. above criteria. The design inte- figured for 96 analog and 128 dis-
grates standard components and crete channels. IQ could
software from multiple vendors. conceivably be expanded to a total
The base I/O subsystem is built of 20,000 points. The .,ystem is
DESIGN CONSIDERATIONS upon intelligent I / O data acquisi- configured by an AT&T 6286
tion components similar to those ATTM compatible. Whet, the I/O
The following requirements utilized in flight simulators. All control software is running, the
were placed on the design of the custom software to link the multi- I / O information is transfe._~ed_be-
system: ple systems together was done tween I / O subsystem and the AT
(1.) The ability to be directly con- in-house. via a DMA card. This information
nected to any control system's Each t e r m i n a t i v n strip (16 is then transferred to an AT&T
I/O and require no DCS con- channels) is wired to the control 6386/25 AT compatible via
f~guration modifications prior system's I/O channels in exactly Ethernet. The 6386 is runnin~a~he
to testing. the same manner as field sensors process simulation using G2 , a
and final control elements will be real-time expert ~]~,-~-~,~.~...from
(2.) The ability to concurrently test connected in the field. This veri- GENSYM®. G2 is suitable for
the interactions between multi- fies any jumper settings that have most routine simulation checkout
ple analog and di_~r_ete !oo_ps-~ to be set on the I/O cards of the requirements and can serve as a

42 ISA TRANSACTIONS
 DISTRIBUTEDCONTROL SYSTEMTESTING

jects. The interface objects trans-

AT&T 6386/25 fez" I/O information between G2
UNIX and the ASCS system using GSI
(G2 Standard Interface). The oon-
tml objects manipulate the signals
Etkernet TCP/IP from the control system and send
them back (through an interface
object) as inputs to simulate Izans-
mitter responses. Example objects
IDAq I D ~ J ! [ Data are delay, lead/lag, multiplier, in-
Collection
Subroutines ]A] ~A Prooessor tegrator, signal source generators,
DOS 3.3 etc. These simple objects can then
• T & T 6286 be used to build process simula-
tions by placing them on a work-
RPC space and connecting them
together on the screen (Rgure 2).
G2 performs the operations de-
fined by each object in real time.
Next is process objects such as

-o j I -ob:
Coow~td~mma
#l 16 Clmmme~
Cea~stea/l.miatlen
81 16
tanks, valves, pumps, etc. The ob-
jects perform the same functions
as control objects but are specific

F #2

#4
. ---'-'J
#2

#4
,,
r~
n
to the type of process equlpm~t
each defines. For example (Figure
3), a valve object could be con-
nected to an input interface object.
The flow attribute of the valve ob-
--.- P r s o m ~ / ~ Flu z ject would be calculated based on
the value of the input object. This
valve could be connected to a tank
I -,..l:l:l:tllll'lllJlllllJlJltlllllll,h,:l
I I object, which would calculate the
level in the tank based on the dif-
ference of the flow into and out of
the tank. The tank object would
then be connected to a output in-
l m e l h ~ Dam Aoqm.~uauSystem terrace object, which would send
FimzlmeffnceAdal~r
Seas] D~m Link the level signal to the control sys-
Re,M~ ~ C~
32 CbmmelAmlogC=nl tem. Process simulations are put
6t Cmnnel Dt.~-~e Cml together for checking out the
staged control systems by using
thc-se three object types.

Fioure1-SimulatorHardwareConfiguration
sophisticated expert system shell idly prototype systems using a CONTROLSYSTEMTESTING
to even more powerful differen- structured, natural language in an
tial equation solving simulation intuitive, graphic-oriented envi- An e n g i n e e r i n g database
packages when large, full unit ronment. Built into G2 is also a (DbaseVMon PCs or PowerhouserM
process models are needed. simulator that can solve algebraic, on the VAXm), which contains all
of the control system fags, termi-
difference, and first-order differ-
nations, wire tags, engineering
ential equations in real time. This units, instrument specifications,
SOFTWARE
simulation capability is used etc., is maintained on most jobs.
IMPLEMENTATION
heavily in this project. This database maintains all of the
G2 is an advanced tool used for Three primary types of objects in/ormatioa u,a~
. . . . is required by
developing real~fime expert sys- have been defined for this project: 'both the electrical and process
*.~.~n.~.=! t allows a deve!or--~r~,to rap- interface, control, and_ proce~ o ~ control depar _~m.ents,

VOLUME 30 • NUMBER 2 s 1991 43

 DCS:SELECTION.IMPLEMENTATION,AND MAXIMIZATION

UP - I01 Fr-101 I/P-102 LT-102

AII-11 AOl- 1 AII-12 A01-2

object

flow (gpm)
E3 0- ~00%
oetpet
object
t
O- 100% T

Lead/Lag Unit
Conversion

D flow (gpm)

flow (gpm)
'xD INT

Figure 2-Tenk Level ~mulation

D
Te~k
Size

The first step in beginning to

test a control system is running a I/P- 102 LT-102 I/P-101 Fr-lOl
database program to extract all AII-I1 A01-1 AII-12 AOI-I
input and output signals from the
engineering, database and assign-
ing them to the simulator's I / O
c~'~annels in a predefined pattern
based on the type of control sys-
tem being tested. The program
breaks the total number of control
system points down into groups
that the simulator I/O subsystem
can handle. It then generates a
wiring list and point definition
files. The point definition files are
loaded into G2, defining the inter-
face object point attributes (point
tag, zero, and span).
The control system is then pre- ll
wired to the termination strips.
When the first group is wired,
t ~ n g can _begjn. A p_wdefined
generic simulation knowledge
base is loaded, and then the point k'im|lFA q Teu~l* | A U ~ I ~ l m , * l ~ t l ~
i
I IWUlW ~ l g | l R L ~ V ~ I ' ~ I I I I U I G | I U I I

~-~ USATRANSACTIONS
 m ,

DISTRIBUTEDCONTROLSYSTEMTESTING

defirdtlon file i ; leaded. The pre-

defined simulation simply ties
outputs to inputs through a The first step in beginning to test a control system is
l e a d / l a g or d e a d time plus
lead/lag blocks for initial testing.
running a database program to extract all input and
For many simple flow, pressure, output signals from the engineering database and as-
and temperature loops these are signing them to the simulator's IlO channels in a
sufficient process models for DCS
checkout. For more complicated predefined pattern based on the type of control sys-
loops and loops that interact, the tem being tested.
simulations are built up from pre-
defined control blocks and pro- cess is there. Because of this, we The system has been used to ver-
cess e q u i p m e n t definitions. encourage our clients to be pres- ify the controls for a paper ma-
Likewise, motor control and other ent during the final testing pro- chine, three CTMP lines, three
discrete loops can be tested via o e ~ They help in identifying pressurized grinders, and a recov-
simple time delay models or more operating problems that may not ery boiler. It is currently sched-
elaborate simulations where com- have been accounted forin the de- uled to be used to check out the
plex batch sequences need testing. sign. They may also identify controls for a chlorine dioxide
Testing continues until all points those problems that may have plant and an entire kraft bleach
in the group are verified. The come about because of misunder- pulp mill.
simulations can be modified as s t a n d i n g s d u r i n g the design
they are running by simply add- In the future, we plan to use
phase. It is significantly more cost
ing or deleting objects and recon- more of the AI capability of the
effective to fix the p o b l e m s at
necting the signal paths. expert system to further automate
checkout ~ther than on-site. The
When testing on a group is the testing process, for example,
basic simulations are not exact
completed, I / O cab!es are moved models of the process but do a letting the system generate pro-
to the next group of termination cess disturbances and watching
good job of identifTh~g many con-
strips. This process continues figuration problems that may not for the corrective control actions
until all groups are checked. The have been found using the other to occur. If the control actions are
system can simulate approxi- described techniques. More exact incorrect, then the expert system
mately 15 to 20 loops in real time unit operations models can also be could log it with any suggestions
depending on their complexity. inserted into file ASCS system for corrective action to be taken.
All I / O is updated once per sec- when their development is war- Also, as the industry begins to
ond. The system would not be ranted. convert to smart transmitters, the
able to simulate processes that re- We have founo! the system to system will be modified to acquire
spond more quickly than this. be very benefici~ for conlxol sys- and transmit the I / O using those
tem checkout, operations, and communications protocols also.
maintenance training prior to This system will continue to grow
CONCLUSION
start-up. It successfully verifies as more process models are devel-
During the testing process the the control system I/O, and con- oped and as control systems con-
simulations look as if the real pro- troller and display configurations. tinue to change.

VOLUME 30 = NUMBER 2 = 1991 45