What Is A Distributed Control System?

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 62

What is a

Distributed Control
System?
1. Introduction
As the name suggests

The control functions are distributed


throughout the system, rather than being
processed in a single, centralized
computer.
A typical DCS will have:

Control Units (or Stations)

- that function independently of


each other and control specific
sections of plant.

Operator Stations
- that monitor the data in the
control units, providing graphical
displays and allowing the operator
to make changes.
This description is also valid for a
system comprising PLCs and PCs with
operator station software in them.

Is there a difference between the two


systems?

Is a PLC based system a DCS?


This is where we come to the second
important definition of a DCS.

A DCS is a fully integrated system with


a global database.

PLC based systems, you cannot buy different


control units and operator stations from
different vendors and put them together.
A DCS comes as a complete system, in which
the communication between each unit is
invisible to the user.

In addition,

If a point (or function block) is created in a


control unit, then the whole system knows
about it.

It is not necessary to generate a separate


database in the operator station to match
those in the control units.

The information is already there.


Historically:

Distributed Control System developed in a


chemical process environment

PLC based systems developed in the


electrical field.

PLCs have developed from relay logic

DCSs have developed from analog


controllers.
- Their ability to process analog data, and
run complex sequences is DCS main
strength,

- While their processing of relay type logic, is


similar to a PLC, it is much slower than that of a
PLC.

A PLC can process relay logic in milliseconds

The scan rate of a typical DCS control unit is one


second.

(This is fast enough of most analog control, and in fact the


processor is handling a considerable amount of analog data within
its scan.)
Another identifying feature of a DCS
- is the use of TAGNAMES.

While PLCs
- tend to use system addresses to reference
data.
What is tagname?

- A tagnames is a freely user definable name that


is applied to all function blocks and input/output
points in the control units. From then on, this
point can be accessed anywhere in the system
through its tagname.

- A tagname is more than just a variable. If it is


the name of an analog input, it will contain not
just the value, but also other information such as
signal quality, engineering range and so on. A
tagname assigned to a function block includes all
the variables, statuses and modes of that block.
2. Available Distributed
Control Systems

The number of DCSs on the market has


decreased over the last five years or so, as
larger companies have acquired smaller
companies.

This has been partly so that the larger


companies could acquire systems with
strengths that they didn't have, or simply to
reduce competition.
The table below show the current major
vendors, and which systems they have
acquired over the years.
VENDOR SYSTEM
Yokogawa Yokogawa CS1000/3000
Siemens Siemens SPPA-T2000/SPPA-T3000
Honeywell Honeywell TDC3000
Emerson Fisher Provox
Rosemount RS7
ABB Bailey
Fischer and Porter System 6
Kent
Taylor
Westinghouse Westinghouse
3. Architecture of the
Distributed Control System
A typical architecture of a DCS has 4 layers:
- I/O - Field I/O and PLCs
- Control - Control Units (e.g. FCS - Field Control
Station)
- Operation - Operator Stations (e.g. HIS - Human
Interface Station)
- Management Information - Database and
Management Stations (e.g. Exaquantum and PRM -
Plant Resource Manager)
3.1 The I/O Layer

Inputs and outputs to the field are processed at the


I/O layer. This layer comprises various interfaces as
follows:

1. Hardwired I/O - Analog (such as pressure


transmitters, modulating valves, etc) and digital
(relays and switches) inputs and outputs can be
hardwired directly from the field. The I/O interface
panel will have a range of interface cards to plug in
to process different types of inputs and outputs. This
will be discussed in more detail later.
2. Fieldbus I/O - There are various fieldbuses. The
main ones are Foundation Fieldbus, Profibus and Hart.
These allow transmitters and actuators to connect to
the I/O interface through a single digital network.
These networks can carry more than just the process
value. They can include configuration data which can
be set from the system.

3. PLC Interfaces - PLCs can be connected to the


DCS by some sort of communications interface card,
usually in the I/O interface panel, but sometimes
directly onto the control network.
3.2 The Control Layer

This is where all the control functions reside.


Control units communicate to the I/O layer and
read the inputs, process their control functions, and
write to the outputs.

Each control unit is independent of each other, so if


one fails, this does not affect the functioning of the
others. However, a control unit can read and write
data from another control unit using "peer-to-peer"
communication over the control network.
3.3 The Operation Layer

The interface to the operator is carried out in the


operation layer. Operator stations provide a graphical
interface to the functions and data in the control
units, and therefore the process. This is in the form
of graphics, trends, reports and maintenance
displays.
3.4 The Management Information Layer

All high level information not required for real-time


plant control, but necessary for long term process
management is processed in the management layer.
This layer is made up of three sub-layers:

Gateway - to read data from the control units

Database - to hold and archive data for


historical analysis

Management - to process information held in


the database
The Gateway Sub-Layer

In the past, each vendor had their own proprietary


communications protocol to enable a host computer
to access data from the control units. Now, Microsoft
in conjunction with a number of the major vendors,
has developed a communications standard called
OPC. OPC means OLE for Process Control and allows
any host computer to connect to any DCS.

OPC can reside in an operator station or in a


standalone computer.
The Database Sub-Layer

The data that is read from the DCS must be stored


somewhere, and there are a number of database
packages designed for this purpose. These include
Exaquantum (a Yokogawa product) and PI (an
independent product). These packages read data
through the OPC gateway and store the data in a
standard database format. Exaquantum uses Microsoft
SQL Server which is compatible with most
management packages. These packages also provide
other functions such as archiving data to disk, data
compression and basic reporting and display functions.

This software usually resides in a standalone PC.


The Management Sub-Layer

There are a range of different packages available that


provide different information for the user. This includes
sophisticated reporting, batch and recipe handling,
plant resource management, alarm management, plant
optimization etc. These access the data from the
historical database, but can write directly to the
control units via OPC.
4. Distributed Control
System Networks
Associated with these layers are networks that link the
layers together. The networks correspond broadly to
the layers as follows:

I/O Networks - Remote I/O bus, Fieldbus, PLC


communications
Control Network - link control units and operator
stations

Plant-wide Network - where most management


information applications reside
4.1 I/O Networks

There are several possible I/O networks depending


on the type of I/O being interfaced to:

Remote I/O bus - some control units have their


hardwired I/O interface cards integral to the control
unit (e.g. the Fischer and Porter System 6).
However, others (such as the Yokogawa
CS1000/3000) have remote I/O, that is, the I/O
interface cards that are separate from the control
station and communicate to it via a high speed
remote I/O bus.
Fieldbus networks - A fieldbus card usually
resides in an I/O slot on the hardwired I/O panel.
Different cards are available for different types of
fieldbus networks, as mentioned above.
PLC networks - As with Fieldbus, the PLC
communications card usually resides in an I/O slot on
the hardwired I/O panel. The cards have a serial
interface port, and some now have Ethernet
capability, and driver software is loaded into the card
to enable it to talk to a particular type of PLC. So for
example, Modbus, AB or Siemens protocol may be
loaded into the card to talk to any of these or
compatible PLCs. The Modbus protocol is an industry
standard, and a large range of PLCs and other devices
(analyzers, weigh scales, etc) can communicate on
this protocol.
4.2 The Control Network

The control units and operator stations


communicate to each other over a high speed data
highway known as the control network. This is
always a proprietary network and is usually token
passing to guarantee its performance. See below
for more information on token passing.
Such a network is prioritized so that certain
information can always get through:

Alarm messages - highest priority

Process information and operator commands -


medium priority

File transfer (e.g. when loading a program into


a control unit) - lowest priority
4.3 The Plant-Wide Network

All management information is available on the


plant-wide or management information network.
This is usually a standard Ethernet network which is
available on all PCs and, as the name suggests, can
be integrated into the main plant network.
5. Network Types
Ethernet

The most common network communications used is


Ethernet. The term Ethernet refers to what is called
the 'link' and 'physical' layers of the network, that is,
the cable and the method of carrying the signal.
When information, such as a file, is to be sent over
the network, a protocol such as TCP/IP will split the
file into a series of small packets with source and
destination addresses attached to each packet. This is
then passed on to Ethernet to carry this signal over
the cable to the destination PC.
The problem is, what happens if more than one PC
sends out information onto the network? Ethernet uses
a method called collision detection. If packets are sent
out at the same time by different PCs, then they will
collide. When a collision happens, all PCs are locked
from sending anything onto the network while the two
PCs that tried to send packets start random timers.
The first one to time out has control of the network
until its packet is sent.

While this has become the international standard for


office networks, there are several problems with
using it in a control environment:
Firstly, the collision method takes a lot of time to
work itself out. It takes, on average, ten times as
long to negotiate the sending of a packet as it does to
actually send it. Therefore, for a 10 Mb/s Ethernet
network, only 1 Mb/s is actually used for sending
data. Similarly, a 100 Mb/s network is really 10 Mb/s.
Secondly, there is no guarantee that a packet of data
will get through in a reasonable amount of time. If the
network is heavily loaded, and there are many
collisions, the packet will get through at some random
time according to its success with collision timings.
This could be almost immediately, but it also could be
many seconds later. This lack of determinacy is one
of the main problems in using Ethernet in a control
environment where data and alarms must be reported
immediately.
Token Passing

The majority of control systems use some proprietary


form of token passing networks. These networks
operate on the principle of a token that is passed
from device to device on the network. While a device
(or station) has the token, it has full control over the
network, and can broadcast messages, request and
send data, etc until it has finished. It then passes the
token onto the next device on the network. There is
also a time limit that a device can hold onto the
token for so that it does not hold it for too long,
stopping other devices from accessing the network.
In this way, these networks utilize 100% of the
network bandwidth, so that a 10 Mb/s network delivers
10 Mb/s of data. It is also determinate in that under
maximum load conditions, the time for all data to be
sent and received is known.
These token passing networks are considered to be
more robust for control applications. However, with the
introduction of switching hubs for Ethernet, and the
increased reliability and availability of Ethernet
components, Ethernet is becoming a more acceptable
option. Switching hubs allow point-to-point
connections between devices, reducing the amount of
collisions and thereby increasing throughput.
6. Availability and
Redundancy
An important aspect of DCS that sets them apart
from PLCs is that of availability. Because DCSs were
developed to control large processes such as oil
refineries, it is important that they are always
available, controlling the plant.
Thus, the following must not cause the DCS to stop
controlling the plant:

Hardware failure

Removal and installation of hardware

Downloading of programming changes

Other software changes


6.1 Hardware Failure

Problems due to hardware failure can be minimized in


three ways:

1. Redundancy of key hardware components

2. Reliability of hardware

3. Electrical Isolation
6.1.1 Redundancy

Redundancy means the duplication of hardware, such


that if a hardware component fails, a backup
component takes over its function.

Redundancy is an important feature of DCS and


ensures that in the event of the failure of a hardware
component (such as a CPU card), a duplicate
component can take over the functions of the failed
one. The changeover to the backup component must
be immediate and not cause any loss in control.
Not all components need to be duplicated for
redundancy; it depends on the application. For
example, in many cases, the field I/O modules do not
need to be duplicated, because the failure of one does
not usually affect the control as a whole. In some
applications, such as oil and gas, it is necessary for
everything to be duplicated, including I/O cards. But
in other cases, such as coal-fired power stations, only
a few key measurements (such as boiler pressure)
require duplicated I/O cards.
The control components that are always
duplicated are:

CPU modules
Communication modules and cables
Power Supplies

Control components that are sometimes duplicated


are:

Input/Output modules
Isolation devices

Power Sources
Operator Stations are usually not dual redundant
because they have no affect on control. In any case,
there is usually more than one operator station, so
that if one fails, the other can still provide monitoring.
Data Historians and Data Gateways are often
duplicated so that no data is lost in the event of the
failure of the Data Historian computer.

Power sources should be duplicated and most DCS


controllers accept two power inputs. Usually, one input
will be from the normal mains supply and the other
will be from a UPS, so that there is always one source
of power available.
6.1.2 Hardware Reliability

Hardware reliability is measured in MTBF Mean Time


Between Failure. Figures for each component in the
system are usually published by each DCS vendor, and
these provide information on the reliability of the DCS.

Figures are usually very high (over 1,000,000 hours


per component) as DCS hardware is designed to be
extremely reliable, often at the cost of functionality. For
example, a DCS controller will usually have a CPU chip
operating at a low frequency, usually no more than
100MHz. This is because a slower CPU is more reliable
and can operate within a greater range of temperature
conditions.
The ability of the hardware to operate in extreme
conditions is important, and environmental ratings are
always published by the vendor. A DCS should be able
to operate in temperatures up to 50 Degrees C.

6.1.3 Electrical Isolation

The best way to ensure system availability is to ensure


that no external sources can affect the controller. The
two components that are most vulnerable to external
sources are the power supplies and the I/O modules.
Isolation of I/O is achieved through interposing
isolation modules, and onboard isolation between
channels.

Contact Inputs/Outputs Contact outputs on I/O


cards are usually only transistors and can handle only
a small amount of current. Therefore, they are usually
connected to mechanical relays that interface to the
field. Mechanical relays have a much larger rating,
with standard ones for DCS application usually rated at
10A.
6.2 Removal and Installation of Hardware

If hardware components need to be replaced, either


due to component failure, or hardware upgrades, it is
important that this be done without loss of control of
the plant.

DCSs have a facility known as hot replacement,


which means that a hardware component can be
removed from and installed in the controller while the
system is live and power is being supplied to the
hardware.
For example, a CPU card that is controlling the
process can be removed online. Control immediately
switches over to the backup CPU without any loss of
control to the process. A new CPU card can then be
installed, and the existing CPU will detect the
presence of the new CPU and copy the software over
to it automatically. The system is now back to a fully
functional state.
6.3 Downloading of Programming Changes

In most PLC systems, a program must be stopped to


enable downloads to the PLC. During this period, the
PLC is not controlling the process. This is unacceptable
in many process environments, and a DCS must be
able to accept changes to its software without any
interruption to the control of the process.
Different DCSs deal with this in different ways. The
Yokogawa CS3000 DCS has two software loading
modes: online and offline load.

Online load loading software changes to the required


software modules without control interruption of the
process. Available for most software loads.

Offline load a complete software reload that shuts


down the controller during the load and resets all of
the tuning parameters. Required if changes to the
memory structure of the CPU are required (very rare).
7. DCS Hardware

7.1 Field Controller Hardware

The Field Controller Hardware usually comprises the


following components:

Back-plane and rack Power Supply


Battery (for memory backup) CPU

Communication cards and couplers

I/O Interface
The above figure is a Yokogawa CS3000 field controller (KFCS). The
components are duplicated to ensure full dual redundancy.

Terminology:
ESB I/O interface V net Data highway communications
Processor cards CPU
7.2 I/O Interface Hardware

Process I/O is interfaced to the field controller with


signal conditioning cards.

Signal conditioning cards are placed in racks which


may be part of the field controller or remote to it.
This figure shows an example of a field controller with I/O racks
connected to it. In this example, the Yokogawa CS3000 KFCS has a high
speed local I/O bus (ESB 128 Mb/s) and a standard Ethernet type
remote I/O bus (ER 10 Mb/s).
The I/O racks provide local power and an
interface to the I/O bus:
The I/O can be broadly defined as:

Analog

Digital

Communications

Analog Inputs and Outputs are I/O that vary


according to a process signal. Typically, they are 4
20 mA.
Example 1: if a pressure transmitter is connected to
an input, it will provide a signal of 4mA at 0% and
20mA at 100% of its range.

Example 2: a control valve will receive a 4-20mA


signal from the Analog Output module such that when
the output is 4mA, the valve will be closed, when it is
12mA, the valve will be at 50% and when it is 20mA
the valve will be fully open. The valve position will
modulate in between.
Digital Inputs and Outputs are switches, that is,
on/off type inputs and outputs and are often known as
Contact I/O. Typical digital inputs are limit switches,
level and pressure switches, pushbuttons. Typical
digital outputs are shut-off valve open/close
commands, motor start/stop commands.

Communications I/O are I/O from a third party


device (usually a PLC) that are acquired over a serial
or Ethernet link. These I/O are then mapped to I/O
points in the controller database.
I/O processing

There are many different types of Analog inputs and


these must be processed according to the input
type. The main types are as follows:

Current 4-20mA or 0-20mA

Voltage 1-5VDC or 0-10VDC or mV

Thermocouple a mV type temperature input

RTD a resistance type voltage input

Pulse frequency type 0-10kHz


Analog Outputs are generally 4-20mA only,
although some are 1-5VDC.

Isolation input/output channels can be isolated or


non-isolated. Isolation means that all channels are
electrically isolated, that is, they do not share a
common earth. If an I/O channel receives a large
voltage spike, then it does not affect the other
channels.
Digital inputs are generally low-power, and the
channels are non-isolated. Isolation from the field is
achieved using interposing relays.

Digital outputs are also usually low-power, non-


isolated, although on-board relays are available. This
does reduce the amount of points available on one
card, however. General practice is to use low-
power/non-isolated cards with an interposing relay
terminal board.

You might also like