The High Performance HMI Overview v2.01 (1)

High Performance Graphics to Maximize Operator Effectiveness

Version 2.0: Including a Major Case Study

Written byBill Hollifield, Principal Alarm Management and HMI Consultant, PAS

Hector Perez, HMI Product Manager, PAS

High Performance HMI | Cover© PAS 2012

IntroductionThe human-machine interface (HMI) is the collection of screens, graphic displays, keyboards,

switches, and other technologies used by the operator to monitor and interact with the

control system (typically DCS or SCADA.) The design of the HMI plays a critical role in

determining the operator’s ability to effectively manage the operation, particularly in response

to abnormal situations.

For several reasons, the current design and capability of most HMIs are far from optimal for

running complicated operations. Most of these consist simply of schematic-style graphics

accompanied by numbers. Such displays provide large amounts of raw data and almost no

real information. They provide inadequate situation awareness to the operator.

This paper concentrates on proper and effective design of the graphics used in modern

control systems.

HMIs Past and PresentBefore the advent of sophisticated digital control systems, the operator’s HMI usually

consisted of a control wall concept.

The control wall (see Figure 1) had the advantages of providing an overview of the entire

operation, many trends, and a limited number of well-defined alarms. A trained operator could

see the entire operation almost at-a-glance. Spatial and pattern recognition played a key role

in the operator’s ability to detect burgeoning abnormal situations.

The disadvantages of these systems were that they were very difficult to modify. The addition

of incremental capability was problematic, and the ability to extract and analyze data from

them was almost non-existent. The modern

electronic control systems (DCS/SCADA)

replaced them for such reasons.

When the modern systems were introduced,

they included the capability to create and

display graphics for aiding in the control

of the operation. However, there were no

guidelines available as to how to actually

create effective graphics. Early adopters

created graphics that mimicked schematic

drawings, primarily because they were

readily available.

High Performance HMI | Page 1© PAS 2012

Figure 1: Example of a Control Wall

The limited color palette was used inconsistently and screens began to be little more than

crowded displays of numbers.

Graphics such as Figures 2 and 3 were developed over 20 years ago and remain common

throughout the industry. Indeed, inertia, not cost, is the primary obstacle to the improvement

of HMIs. Engineers and Operators become accustomed to this style of graphic and are

resistant to change.

As a result, industries that use modern control systems are now running multi-million dollar

operations from primitive HMIs created decades ago, at a time that little knowledge of proper

practices and principles was available.

Figure 2: An Early Graphic Exhibiting Many Problematic Practices


As control system hardware progressed, graphics from the manufacturers began to develop

very flashy example graphics which were used for marketing purposes. While good for that,

they are quite ineffective for actually controlling a process. Many companies and projects,

however, began to use those flashy examples. The results were displays that are actually sub-

optimal for operators.

Figure 4 is an example of flashy design taken from a power generation facility to illustrate

the point. The graphic dedicates 90% of the screen space to the depiction of 3-D equipment,

vibrantly colored operation lines, cutaway views, and similar elements. However, the

information actually used by the operator consists of poorly depicted numerical data which is

scattered around the graphic, and only makes up 10% of the available screen area.

Figure 3: A Typical Crowded, P&ID-Style Graphic


There are no trends, condition indicators, or key performance elements. You cannot easily tell

from this graphic whether the operation is running well or poorly. That situation is true for

more than 90% of the graphics used throughout the industry because they were not designed

to incorporate such information. Instead, they simply display dozens to hundreds of raw

numbers lacking any informative context.

Justification for HMI ImprovementPoorly performing HMIs have been cited time and again as significant contributing factors

to major accidents. Yet our industry has made no significant change in HMI design. There is

another industry that learns from its accidents and has made phenomenal advancement in

HMI design based on new technology. That industry is avionics. Lack of situation awareness is

a common factor cited in aviation accident reports.

Figure 4: A Flashy Graphic Inappropriate for Actual Operational Control


Modern avionics feature fully-integrated electronic displays (See Figure 5). These depict all

of the important information, not just raw data, needed by the operator (i.e., pilot). Position,

course, route, engine diagnostics, communication frequencies, and automated checklists are

displayed on moving maps with built-in terrain proximity awareness. Real-time weather from

satellite is overlaid on the map. Detailed database information on airports is available with

just a click. Situation awareness and abnormal situation detection is far improved by these

advances. This capability – impossible even a dozen years ago in multi-million dollar airliners –

is now standard on even the smallest single engine aircraft.

Since safety is significantly improved with modern HMIs, it is only logical that we would want all

operators to have access to them. Yet most companies have done little to upgrade.

There have been tests involving actual operators running realistic simulations using traditional

graphics vs. High Performance ones. PAS participated in such a test at a large power plant,

sponsored by the Electric Power Research Institute (EPRI). The results were consistent with a

similar test run by the ASM® (Abnormal Situation Management) Consortium on an ethylene

plant. The test showed the high performance graphics provided significant improvement in the

detection of abnormal situations (even before alarms occurred), and significant improvement in

the success rate for handling them. In the real world, this translates into a savings of hundreds of

thousands of dollars per year.

Figure 5: Garmin G1000® Avionics Package in a Small Plane


Proper Graphic PrinciplesIneffectively designed graphics are very easy to find. Simply search the internet for images

under the category “HMI”. Problems with these graphics include:

• Primarily a schematic representation

• Lots of displayed numbers

• Few trends

• Spinning pumps/compressors, moving conveyors, animated flames, and similar

distracting elements

• Brightly colored 3-D vessels

• Highly detailed equipment depictions

• Attempts to color code piping with contents

• Large measurement unit callouts

• Bright color liquid levels displaying the full width of the vessel

• Lots of crossing lines and inconsistent flow direction

• Inconsistent color coding

• Misuse of alarm-related colors

• Limited, haphazard navigation

• A lack of display hierarchy

Ineffective graphics encourage poor operating practices, such as operating by alarm.

By contrast, High Performance graphics have:

• A generally non-schematic depiction except when functionally essential

• Limited use of color, where color is used very specifically and consistently

• Gray backgrounds to minimize glare

• No animation except for specific alarm-related graphic behavior

• Embedded, properly-formatted trends of important parameters

• Analog representation of important measurements, indicating their value relative to

normal, abnormal, and alarm conditions

• A proper hierarchy of display content providing for the progressive exposure of

detailed information as needed

• Low-contrast depictions in 2-D, not 3D

• Logical and consistent navigation methods

• Consistent flow depiction and layout to minimize crossing lines

• Techniques to minimize operator data entry mistakes

• Validation and security measures

• Embedded information in context (via right-click menus or similar methods) such as

alarm documentation and rationalization, standard operating procedures, etc.


Data or Information?A primary difference of high performance graphics is the underlying principle that, wherever

possible, operational values are shown in an informational context and not simply as raw

numbers scattered around the screen.

As an example, consider this depiction of a compressor (see Figure 6). Much money has

been spent on the purchase of instrumentation. Yet, unless you are specifically trained and

experienced with this compressor, you cannot tell if it is running at peak efficiency or is about

to fail.

The mental process of comparing each number to a memorized mental map of what is good is

a difficult cognitive process. Operators have hundreds (or even thousands) of measurements

to monitor. Thus the results vary by the experience and memory of the operator, and how

many abnormal situations they have experienced with this particular compressor. Training new

operators is difficult because the building of these mental maps is a slow process.

Adding more numbers to a screen like this one does not aid in situation awareness; it actually

detracts from it.

“Information is data in context made useful.”

West East

Drive: 232.2 amps

Cooler

W. Vibration: 2.77 E. Vibration: 3.07

2.77MSCFH

155.2 °F 108.2 °F 166.1 °F55.7 psig

135.1psig

190.5 psig

Oil 155.2 °FOil 85.1 psi

65.1 °F

Figure 6: All Data, No Information


By contrast, these numbers can be represented in a bank of analog indicators, as in Figure 7.

Analog is a very powerful tool because humans intuitively understand analog depictions. We

are hard-wired for pattern recognition.

With a single glance at this bank of properly designed analog indicators, the operators can tell

if any values are outside of the normal range, by how much, and the proximity of the reading

to both alarm ranges and the values at which interlock actions occur.

In just a second or two of examination, the operator knows which readings, if any, need further

attention. If none do, the operator can continue to survey the other portions of the operation.

In a series of short scans, the operator can be fully aware of the current performance of their

entire span of control.

The knowledge of what is normal is embedded into the HMI itself, making training easier and

facilitating abnormal situation detection – even before alarms occur, which is highly desirable.

Similarly, depiction of PID controllers is accomplished with the addition of easily scanned

setpoint, mode, and output information as in Figure 8.

Figure 6: All Data, No Information

Coolgpm

RECYCLE COMPRESSOR K43

Alarm Indicator

Desirable Operating Range

AlarmRange

AlarmRange

2

Suctpsig

Interpsig

Dschpsig

SuctdegF

InterdegF

DschdegF

E. Vibmil

N. Vibmil

W. Vibmil

MotorAmps

Oilpsig

OildegF

42.7 38.7 93.1 185 95 120 170 12 8 9 170 80 290

Interlock Threshold


ColorColor must be used consistently. There are several types of common color-detection

deficiency in people (red-green, white-cyan, green-yellow). For this reason:

Color, by itself, is never used as the sole differentiator of an important condition or status.

Most graphics throughout the world violate this principle. A color palette must be developed,

with a limited number of distinguishable colors used consistently.

Bright colors are primarily used to bring or draw attention to abnormal situations, not normal

ones. Screens depicting the operation running normally should not be covered in brightly

saturated colors, such as red or green pumps, equipment, valves, etc.

When alarm colors are chosen, such as bright red and yellow, they are used solely for the

depiction of an alarm-related condition and functionality and for no other purpose. If color is

used inconsistently, then it ceases to have meaning.

So what about the paradigm of using bright green to depict “on” and bright red for “off”, or

vice versa if you are in the power industry? This is an improper use of color. The answer is a

depiction such as Figure 9.

Figure 8: Analog Depiction of PID Controllers

AUTO

32.1

44%

SmallMismatch

Setpoint Diamond

2

562.1

Analog Controller: Add Setpoint, Output, and Mode to standard analog indicator

Output barScaled 0-100%

AUTO485.0

2%

2

562.1AUTO485.02%

If Final Control Element hasPosition feedback:

22%

Shows any mismatch between output and actual position

Actual Position

Output

Setpoint diamond and PV pointer line up when controller is working properly


The relative brightness of the object shows its status, plus a Process Value WORD next to

it. Equipment items brighter than the background are on (think of a light bulb inside them).

Items darker than the background are off. If equipment has no status that is sensed by the

control system, but is desired to be shown on the graphic anyway, it is shown with a fill the

same as the background color.

Alarm DepictionProper alarm depiction should also be redundantly coded based upon alarm priority (color /

shape / text). Alarm colors should not be used for non-alarm related functionality.

When a value comes into alarm, the separate alarm indicator appears next to it (See Figure

10). The indicator flashes while the alarm is unacknowledged (one of the very few proper

uses of animation) and ceases flashing after acknowledgement, but remains as long as the

alarm condition is in effect. People do not detect color change well in peripheral vision, but

movement, such as flashing, is readily detected. Alarms thus readily stand out on a graphic

and are detectable at a glance.

Figure 9: Depicting Status with Redundant Coding and Proper Color Usage

(Shape is Filled Darker)

Wrong Better

Pump Not Running

PumpRunning

Wrong Better

STOPPED

RUNNING(Shape is Filled Brighter)


It is highly beneficial to include access within the HMI to the alarm rationalization information

contained in the Master Alarm Database (See Figure 11). If these terms are unfamiliar, you are

advised to read the ISA 18.2 standard for Alarm Management in the Process Industry, or read

the API RP-1167 Alarm Management Recommended Practice if you are in the pipeline industry.

Figure 10: Depiction of Alarms

122.1 degC3

Lowered efficiencyOff spec ProductionAlarm Consequences:

Priority: 3

Alarm: PVHI

Check feed composition

Feed composition variance

Adjust reflux per computation; check controller for cascade mode

Insufficient reflux

Check pressure and feed parameters vs. SOP 468-1

Pressure excursion

Adjust base steam rate

Excess steam

Corrective Actions:

Alarm Causes:

Alarm Consequences:

Response Time: <15 min

Class: Minor Financial

Setting: 320 deg F

TI-468-02 Column Overhead Temperature

Lowered efficiencyOff spec ProductionAlarm Consequences:

Priority: 3

Alarm: PVHI

Check feed composition

Feed composition variance

Adjust reflux per computation; check controller for cascade mode

Insufficient reflux

Check pressure and feed parameters vs. SOP 468-1

Pressure excursion

Adjust base steam rate

Excess steam

Corrective Actions:

Alarm Causes:

Alarm Consequences:

Response Time: <15 min

Class: Minor Financial

Setting: 320 deg F

TI-468-02 Column Overhead Temperature

Reveal Alarm Rationalization documentation on demand!

Information recalled from your Master Alarm Database

Right-click call-upShape and Size of a standard faceplate if possible

Link to Procedures

Figure 11: Linked Alarm Information


TrendsThe most glaring deficiency in HMI today is the general lack of properly implemented trends.

Every graphic generally has one or two values on it that would be far better understood if

presented as trends. However, the graphics rarely incorporate them.

Instead, engineers and managers believe claims that their operators can easily trend any

value in the control system on demand with just a click. This is incorrect in practice; a properly

scaled and ranged trend may take 10 to 20 clicks/selections to create, and usually disappears

into the void if the screen is used for another purpose!

This deficiency is easily provable; simply walk into the control room and count how many

trends are displayed. Our experience in hundreds of control rooms is that trends are vastly

underutilized and situation awareness suffers due to that.

Figure 12: Trend Depiction of Desirable Ranges

West Comp Discharge Temp °C

40.0

50.0

-60 -30-90 Hours

44.1West Comp Discharge Temp

40.0

50.0

-60 -30-90 2 Hours

West Comp Flow MSCFH

45.0

55.0

-60 -30-90 2 Hours

49.1West Comp Flow MSCFH

45.0

55.0

-60 -30-90 Hours

49.1

500015

1 Hr

Feed Water3280DrumLevel-0.5

MainSteam4150

0-10


Trends should be embedded in the graphics and appear, showing proper history, whenever

the graphic is called up. This is generally possible, but is a capability often not utilized.

Trends should incorporate elements that depict both the normal and abnormal ranges for the

trended value. There are a variety of ways to accomplish this (See Figure 12).

Level IndicationVessel levels should not be shown as large blobs of saturated color. A simple strip depiction

showing the proximity to alarm limits is better. A combination of trend and analog indicator

depictions is even better (See Figure 13).

Bar Charts Attention to detail is important. It is typical to use bar charts to show relative positions and

values. While this may be better than simply showing numbers, it is inferior to the use of

moving pointer elements since as the bar’s value gets low, the bar disappears. The human eye

is better at detecting the presence of something than its absence. The example in Figure 14 is

superior in showing relative values, besides the color improvement.

Figure 13: Vessel Levels

Better Vessel Level

Indication

Very Poor Vessel Level

Indication

Crude Feed TK-21

Trend andAnalog Level

Indication

2 Hrs 86.5%

2

0

100


Tables and ChecklistsEven tables and checklists can incorporate proper principles (See Figure 15). Consistent colors

and even status indication can be integrated.

Figure 14: Bars vs. Pointers

25% 50% 75%

ValveV-1V-2V-3X-1X-2X-3X-4S-1S-2K-1K-2

100%100%

95%88%

100%100%

75%0%

55%100%100%

100%0%

Analog Position - Better

25% 50% 75%

100%100%

95%88%

100%100%

75%0%

55%100%100%

100%0%

Analog Position – Poor Bar Graphs

ValveV-1V-2V-3X-1X-2X-3X-4S-1S-2K-1K-2


There are dozens of additional principles like these. See the References section.

Display HierarchyDisplays should be designed in a hierarchy that provides progressive exposure of detail.

Displays designed from a stack of schematic designs will not have this; they will be “flat” –

like a computer hard disk with one folder for all the files. This does not provide for optimum

situation awareness and control. A four-level hierarchy is desired.

Figure 15: Tables and Checklists

Breaker 15 Power Oil Temp 16-33Oil Pres Status Level in TK-8776 Gen System Status Comp 88 in Auto Lineup Ready Sys Status Checks Bearing Readouts Comm check Outlet Temp < 250 Cooling Flow Internal Circuit Check Bypass Closed AFS Function

OK

NOT OK OK

OK

OK

NOT OK OK

OK OK OK

OK OK OK OK

Startup Permissives

NOT OK

Air Comp

Status Mode Diag

C #1 RUNNING AUTO OK

C #2 STOPPED MAN OK

C #3 RUNNING AUTO OK

C #4 STOPPED AUTO FAULT 3


Level 1 –Operation OverviewThis is a single display showing the operator’s entire span of control, the big picture. It is an

overall indicator as to how the operation is running. It provides clear indication of the current

performance of the operation by tracking the Key Performance Indicators (See Figure 16).

Control interactions are not made from this screen.

The Figure 16 example is from a large power plant. We often hear “But it doesn’t look like a

power plant?!” Correct! Does your automobile instrument panel look like a diagram of your

engine? The display is designed so that important abnormal conditions and alarms stand out

clearly.

Figure 16: Example Level 1 Display

Stator GPM

Turb Oil °F

Hydrogenpsig

LPT-B in.hg

LPT-A in.hgMVARNet MW

Turbine – GeneratorGross MW HZ

702.1 640.1 - 5.2 60.00 0.2 49.1 351 3.1 20.11150.2

HW Lvl in.H2OB2 BPFT

Condenser–Feed Wtr DA Lvlin.H2O

3.1 0.0

9.4 775

Econ GasOut °F

Boiler BBDpH

Fans

- 0.5 6.0 7.0

Econ % O2

Sec Airin.H2OFurn in.H2O

200

Stack COppm

SO2#/MMBTU

CEMS/MISC% Opac

NOX#/MMBTU

0.921.0

05-31-1013:22:07

Total Alarms

Unit 2Overview

0

1200

600

1200

1 Hr

3000

600

Steam°F

990

Reheat°F

1005

Steampsig2400

-5

7500

0

1250

1 Hr

5

0

AirKLBH

7400

Coal KLBH

1000

FurnPres-0.5

0

5000

0

5000

1 Hr

15

-15

SteamKLBH

4750

Fd WtrKLBH

4580

DrumLvl in.

-0.5

Drum Lvlin.H2O

-0.5

2

B2 ID Stall

A2 ID Stall

A2 FD Stall

B2 FD Stall

2

AA-ONPULVStatus

A2 CWP

B2 CWPB-ONC-OND-ON

E-ONF-ONG-OFFH-ON

Alarms

E

B

F

C

G

D

H

PUMPS ANDFANS

Pump Status / Alarms

ON

ON

A2 HWP

B2 HWPON

OFF

C2 HWP

SUBFPON

ON

A2BFPT

B2BFPTON

ON

A2 ECW

B2 ECWON

ON

Fan Status / AlarmsA2 FD

ON

ON

B2 FD

B2 IDON

ON

A2 PA

B2 PAON

ONA2 ID

C-SBAC

D-SBACON

ON

Hydrogen°F

104

Econ pH

9.4

Aux Stmpsig

300

A2 BPFT

400

Cond Hdrpsig

Inst AirpsigA/F Ratio

0.45 907.1

DA WideFT.H2O

9.0

2525 2525

0

2

1

8

3 5

AUTO AUTO AUTO AUTO


Level 2 –Unit ControlEvery operation consists of smaller, sectional unit operations. Examples include a single

reactor, a pipeline segment, a distillation train, or a compressor station. A level 2 graphic

exists for each separate major unit operation. It is designed to contain all the information and

controls required to perform most operator tasks associated with that section, from a single

graphic (See Figure 17).

Level 3 –Unit Detail Level 3 graphics provide all of the detail about a single piece of equipment. These are used for

a detailed diagnosis of problems. They show all of the instruments, interlock status, and other

details. A schematic type of depiction is often desirable for a Level 3 display (See Figure 18).

Figure 17: Example Level 2 Display of a Reactor

VENT SYS

ThioniteProduct: Mid-Run

5.0 %

PRODUCT

OK

PumpsNeeded 1

SHUTDOWN

M5

VENTM5

IN OUT

Calc Diff:

-10%

+10%

Hours: 238.1

State:

1970719301

Material Balance

2.1 %

FREEZEM5

STOPPED FAULT

Main Menu

80.5

%A

15.5

%B

4.0

%CFeed Composition

76.0

77.5AUTO

33.1%

Feed MPH

12.0

11.9AUTO

22.3%

ADTV-1 MPH

4.0

4.0AUTO

44.3%

ADTV-2 MPH Reserved

Faceplate Zone

When any item on the screen is selected, the faceplate for that item appears in this reserved area.

All control manipulation is accomplished through the standardized faceplates.

Reactor M5

Analysis: Purity %

32.0 -60 -30-90 2 Hrs

40.0

Analysis: Inhibitor Concentration %

4.0 -60 -30-90 2 Hrs

6.0

80.5

Coolant:GPM °C

15.5

PurgeMCFH

4.0

Cat.Act%

Conv.Eff. %

80.5 15.5

75.0

75.9AUTO

54.3%

Lvl %

92.0

Prod MPH

45.0

45.0AUTO

54.1%

Temp °C

95.0

97.2AUTO

44.3%

Pres psig

To Coils

74.3 %

L2 M5 Startup

L2 M5 Scram

M5 Vent Sys

Daystrom Pumps

M5 Agitator

L2 Feed System

L2 Prod Recovery

M5 Interlocks

M5 Cooling Sys

---- Level 3 ----

AgitatorON 4

M5 Circ

Pump A

Pump B

RUNNING

3

ISOLATEM5

ON-OKRTAM:

Run Plan:Actual:

Feed MPH

72.0 -60 -30-90 2 Hrs

80.0

OP

Chem 1

MPH

10.0 -60 -30-90 2 Hrs

14.0

OP

Chem 2

MPH

2.0 -60 -30-90 2 Hrs

6.0

OP

40.0 -60 -30-90 2 Hrs

48.0

Temp °C

OP

L2 Compression

L2 RX Summary

I I-5A I-5B I-5C

I-5D

I I

I I-5EI I-5FI

Feed ADTV-1 ADTV-2

Temp Pres Level

Interlock Actions:Stop Feed Stop ADTV-1Stop ADTV-2Max CoolingMax Vent

OFFOFFOFFOFFOFF

Reset

0%

Tot. In:Tot Out:

%DIFF


Level 4 –Support and Diagnostic DisplaysLevel 4 displays provide the most detail of subsystems, individual sensors, or components.

They show the most detailed possible diagnostic or miscellaneous information. A “Point

Detail” display is a typical example. The dividing line between Level 3 and Level 4 displays can

be somewhat gray.

API-1165: Recommended Practice forPipeline SCADA DisplaysIn 2006, API issued a document on SCADA HMI displays. There are some inconsistencies

within that document.

Overall, the concepts incorporated in the text portion of the document are valid. It mentions

several good practices. The examples section, however, provides several depictions that are in

direct violation of the text principles.

Figure 18: Example Level 3 Display

© PAS - Confidential and Proprietary 2012 | 57

PSOAUTO

76.8 MSCFH76.088.5 %

Flow Demand

RUNNING

PSOCAS

90.090.0 %

WC Speed

65.0 °C

West Comp Discharge Temp °C

40.0

50.0

-60 -30-90 2 Hours

1st

Stage

2nd

Stage

CW

32.0 °C

20.0 °C

28.0 °C

WESTCOMP

OH

44.0 °C

90.8 %

48.0 psi

90.0 psi

1 Stgpsi

EAST COMP

Speed %

20.1 psi EAST COMP2 Stgpsi

48.0 90.0

CLR IN°C

SPEED CASCADEIN EFFECT

CLR OUT°C

65.0 32.0

Winding°C

111.0 °C

111.0

West Compressor Interlock Initiators

Comp in Overspeed

Winding Temp is High

Vibration is High

1 Stg Pres is High

2 Stg Pres is High

Suction Pres is Low

Oil Pres is Low

NO

NO

NO

West Comp Flow MSCFH

45.0

55.0

-60 -30-90 2 Hours

48.4 MSCFH

West Comp Speed %

85.0

95.0

-60 -30-90 2 Hours

SHUTDOWNWESTCOMP

RECOVERY

IDLEWESTCOMP

ISOLATEWESTCOMP

West Compressor

90.8

90.0 %90.0CAS

Tot Flow MSCFH

76.8

88.5 %76.0

AUTO

95.1 Oil psi

Main Menu

Reserved Faceplate Zone

When any item on the screen is selected, the faceplate for that item appears in this reserved area.

All control manipulation is accomplished through the standardized faceplates.

L2 Compression

East Comp

West Interlocks

Logic Diagrams

West Cooling

---- Level 3 ----

I W-1A W-1B W-1CI IMech Flow Pres

Procedures

Sequence Overlay

Startup Overlay

L2 Recovery

PURGEWESTCOMP

MANUAL ACTIONS

NO

NO

NO

Total Flow is Low

West Compressor Interlock Actions

W. Comp Shutdown

Inlet Block Valve

Outlet Block Valve

Flow Cascade to Manual

2 Stg High Press

W. Suction Pres Low

E. Comp Override to 100%

NO

OPEN

OPEN

NO

NO

OK

OK

OKTotal Flow LowNO

NO

Inter-cooler

---- Level 4 ----


For example, Section 8.2.4 states “Color should not be the only indication for information.

That is, pertinent information should also be available from some other cue in addition to color

such as a symbol or piece of text.”

Yet throughout the remainder of the document, examples are shown that routinely violate

this principle. Figure 19 shows only a few of the “recommended practice examples” from API-

RP-1165. In many of these examples, only subtle color differences, not distinguishable by a

substantial fraction of the operator population, are the only means to distinguish a significant

status difference.

In one table, API-1165 recommends color coding alarms by type. The well-known best practice

is that they are redundantly coded by priority, not type.

Users of API-1165 are therefore advised to pay more attention to the principles it contains than

to the example depictions.

Figure 19: Sub-optimal Examples from API-RP-1165


A Real-World Case Study and Test ofHP HMI ConceptsThe following section is taken from a study conducted by the Electric Power research Institute

(EPRI).

Operator HMI Case Study: The Evaluation ofExisting “Traditional” Operator Graphics vs. High Performance Graphics in a Coal Fired Power PlantSimulator , Product ID 1017637The EPRI study tested these HP HMI concepts at a large, coal-fired power plant. The plant

had a full simulator used for operator training. The existing graphics on the simulator (created

in the early 1990s) operated the same as those on the actual control system. PAS was

retained to prepare several high performance graphics for the simulator. Several operators

were then put through several abnormal situations using both the existing and the new high

performance graphics.

Four examples of the existing graphics are in Figure 20. They have the following

characteristics:

• No graphic hierarchy

• No Overview

• Many controller elements are not shown on any of the existing graphics

• Numbers and digital states are presented inconsistently

• Poor graphic space utilization

• Inconsistent selectability of numbers and elements

• Poor color choices, overuse, and inconsistencies. Bright red and yellow are used for

normal conditions

• Poor interlock depiction

• No trends are implemented, “trend-on-demand” is not used

• Alarm conditions are generally not indicated on graphics – even if the value is a

precursor to an automated action


The operators used dozens of such graphics to control the process.

PAS prepared the following high performance graphics:

• Power Plant Overview (Level 1) – see figure 16.

• Pulverizer Overview Graphic (Level “1.5”) – see Figure 21

• Individual Pulverizer Level 2 Control Graphic – see Figure 22

• Runback 1 and 2: Special Abnormal Situation Graphics – see Figure 23

The Level 1 Overview (see prior Figure 16)

The Overview graphic shows the key performance indicators of the entire system under the

operator’s control. The most important parameters incorporate trends. It is easy to scan

these at a glance and detect any non-normal conditions. Status of major equipment is shown.

Alarms are easily detected.

The operators found the overview display to be far more useful than the existing graphics in

providing overall situation awareness and useful in detecting burgeoning abnormal situations.

Figure 20: 1990s Graphics from the EPRI HP-HMI Test


The Level “1.5” Pulverizer Overview Graphic (Figure 21)

The operator controls eight identical, heavily instrumented, and complex pieces of equipment

called coal pulverizers. At normal rates, seven are in use and one is on standby in case of a

problem. The seven that are running should be showing almost identical performance. It was

immediately apparent that an “Overview” graphic of just these eight items would be very

useful to the operators, since much of their activity is in monitoring and manipulating them.

Being mechanical, they are subject to a variety of problems and abnormal conditions. There

were three separate existing graphics needed for monitoring and controlling each pulverizer,

24 in total for them all. Monitoring using 24 graphics was difficult for the operators.

The new Pulverizer Overview in Figure 21 depicts more than 160 measurements on a single

graphic! The key to making this understandable is that the devices are supposed to run

alike. Instead of blocks of indicators for each pulverizer being grouped together, the same

measurement from each pulverizer is grouped together. Any individual unit operating

differently than the others stands out. The unit that is in standby service also is easily seen. Air

damper positions, a source of problems, are clearly shown.

Note that the trends seemingly violate our recommendation of showing no more than 3 or 4

traces on a single trend. In this case, what the operator is looking for is any trend line that is

not “bunched in” with the others. For such a condition, having these 8 traces was acceptable.

Note that the standby pulverizer’s trace is normally on the bottom.

Even with such a “dense” information depiction and with so many measurements, the

operators found it easy to monitor all eight devices and easily detect burgeoning abnormal

situations. It is easy to scan your eye across the screen and spot any elements that are

inconsistent. Alarm conditions are also easy to spot.

Note that control actions are not taken on this screen, but rather on the eight individual Level

2 graphics, one for each pulverizer.


The Level 2 Pulverizer Control Graphic (Figure 22)

Rather than using three separate graphics to control each pulverizer (24 graphics total), a

single Level 2 graphic for each pulverizer was created with everything needed to accomplish

all typical control manipulations.

08-15-2009 14:22:09PULVERIZER OVERVIEWCoal Flow Trend

Coal Flow

KLB/HR

2 Hrs1

140

PULV A PULV B PULV C PULV D PULV E PULV F PULV G PULV HL1 OVERVIEW RUNBACK 1 RUNBACK 2

EFGH

7574

7845

7674

7574

5150

5050

7565

5551

A A A A M A A A

113113

112113

0113

112112

A B C D E F G H

Mill Amps Trend

A A A A M A A A

4243

4244

043

4343

A B C D E F G H

Diff Pres Trend

A A A A M A A A

8.09.8

10.09.5

0.69.8

9.59.0

A B C D E F G H

Pri. Air Flow Trend

A A A A M A A A

204205

204205

0205

204205

A B C D E F G HA A A A M A A AA B C D E F G H

H H

7574

7877

7878

7574

00

5050

7576

5551

S. Air Flow Trend

A A A A M A A A

204205

204205

0205

204205


Primary Damper

7574

7877

7878

7574

3030

5050

7573

5551

A A A A M A A AA B C D E F G H

North Damper South Damper

7574

7877

7878

7574

4040

5050

7565

5551

C/A Temp Trend

A A A A M A A A

135135

135135

277135

135135


H

7574

7877

7878

7574

5050

5050

7573

5551

A A A A M A A AA B C D E F G H

Hot Damper Cold Damper

ON

Pulverizer Status

G

E

C

AB

D

H

F

ON

ONON

ON

ON

ONON

DiagBurnMaint

ABCD

Diff PresPSI

2 Hrs0

16

EFGH

ABCD

8

OPENOPENSWG Valves

NORMNORMFlm Mnt Mod

OFFOFF

ON

M1Flame

Main Flame

IgniterFlame

Fuel Type Gas -1

Flm Mnt Mod

90 90

30

M2

S

S

OFF

30

ON

SS

L

Figure 21: The Level “1.5” Pulverizer Overview


While complex in appearance to the layman, the trained operator had no difficulty in

understanding and accessing everything they needed for pulverizer startup, shutdown, and

swap situations that arose during the test. Much of the text on the screen has to do with the

status of automated sequences that sometimes require operator intervention. Everything on

the screen is selectable, and when selected the standard faceplate for the element appears

in a reserved “faceplate zone” rather than floating around the screen obscuring the graphic.

Element manipulation is made via the faceplate.

Abnormal Situation Response GraphicsThe operator response for many abnormal plant situations is to cut rates by half, from 700MW

to 350MW. Called a “Runback,” this is a complicated and stressful procedure that takes about

20 minutes to accomplish. If done incorrectly or if important parameters are missed, the plant

can fall to zero output, a very undesirable situation. One of the main purposes of the simulator

was to regularly train the operators for this situation. The operators have to use more than

a dozen of the existing graphics to accomplish the task, involving much navigation, screen

callups and dismissals, and control manipulation.

Figure 22: Level 2 Pulverizer Control

08-15-2009 14:22:09PULVERIZER A

PULV OVRVIEW L1 OVERVIEW PULV B PULV C PULV D PULV E PULV F PULV G PULV H

5.9

AUTOS 9.0O 2%

LOWGAS PRESS

EXECUTE

Open Swg Valves

CANCEL

F. SpeedAUTOBIAS 0.0P 81%S 80%O 80.8%

28

100

40

140“A” Coal Flow

Coal FlowKlb/Hr

113

2 Hrs

“A” C/A Temperature

72.1AirTmpAUTOP 135S 130O 71.2%

50

200

2 Hrs

“A” Primary Air Flow

PA FlowAUTOBIAS 0.0P 92.3%S 92 .3% O 49.0%

100

220

2 Hrs

Klb/Hr200

PA Flow

100

45

Sec Air TotalAUTOBIAS 0.0P 50.1%S 50.0%O 60.0%

“A” Sec. Air Flow (Total)

200

900

2 Hrs

Sec Air Total

Klb/Hr615

86

19

Any Seal Air blower stoppedNo coal on the feeder beltAny pulv grp trip not reset

Lube oil press lowLube oil pump not runningFlame det clg air press lo

Flame detected

Min boiler A.F. requiredLTR atom air press low

Pulv seal air dif f press lowFeeder inlet gate not openLTR oil press low or HDR VLV not open

All PA fans stopped or PAH stopped

Sequence Blocked By:

Diff Press

AUTOS 71.2%O 71.2%

Hot DampMill Amps

AUTOS 49.0%O 49.0%

Damper Pos

AUTOS 28.8%O 28.8%

Start Pulv

Start Feeder

Pulv Tmp to Auto

Rel Sec Air

Stop Ltrs

Rel Pul Dmd

X

Ready

Ready

Open Swg Vlvs

VALVES

CLOSED

OPEN

CLOSED

IG HDR VNT VLV

IG GAS TRP VLV

IG OIL TRP VLV

BRNR VNT VLV CLOSED

NOT OKMN GAS HDR PR

10.7 4549

TRIP “A”Pulv Group

Trip Status Reset

PTR Reset

FAULT

Pulv A DiagH

AUTOBIAS 0.0S 50%O 50%

AUTOS 50%O 50%

N Damp

50

S. Damp

50

STOP“A” PULVERIZER

Ready

Ready

Ready

Ready

DoneStatus

Start Sequence

Sec Air to L.O.

Start Ltrs.

Pulv Grp Dmd L.O.

Start PA Flow

START

OPENOPENOPENOPENOPENOPENSWG Valves

MNT-BMNT-BMNT-BNRMLNRMLNRMLFlm Mnt Mod

OFFOFFOFFOFFOFFOFF

ONONONONONON

A7A6A5A3A2A1Flame

Main Flame

IgniterFlame

Fuel Type Gas

Flm Mnt Mod

90 90 90 90 90 90

30 30 30 30 30 30

DoneStatus

Sec Air N

312

Sec Air S

312

Ready

Ready

Ready

Ready

Ready

Ready

XXXXX


However, in more than a decade of such training, it had never occurred to anyone to design

special graphics specifically designed to assist in this task! This demonstrates the power of

inertia in dealing with our HMIs. Specific Abnormal Situation Detection and Response graphics

are an important element of an HP HMI.

PAS created two “Runback” graphics designed specifically to assist in this task. Every element

that the operator needed to effectively monitor and control the runback situation was

included on them. In use, the operators placed them on adjacent physical screens. Figure 23

shows “Runback 1,” Runback 2 was similar. The reserved faceplate zone is on the lower right.

As a simple example of providing information rather than data, consider the trend graph at

the upper left of Runback 1. To be successful, the rate of power reduction must not be too

slow or too fast. The existing graphics had no trend of this, simply showing the current power

megawatt number. This new trend graph had the “sloped-line” element placed next to it,

indicating the ideal rate of power reduction, the full load zone, and the target half-rate zone.

On the figure, the actual rate of drop is exceeding the desired rate, and that condition is easily

seen. (Note: It would have been more desirable to have the sloped lines on the background of

the trend area itself, but the DCS could not accomplish such a depiction. This is a compromise,

but one the operators still found very useful.)

Figure 23: Abnormal Situation Graphics – Runback 1


The TestingEight Operators, averaging 8 years console operating experience each, were used in the test.

They received only one hour’s familiarity with the new graphics prior to the start of testing.

They were tested on four increasingly complex situations, lasting about 20 minutes each.

• Coal Pulverizer Swap Under Load

• Pulverizer Trip and Load Reduction

• Manual Load Drop with Malfunctions

• Total Plant Load Runback

All operators did all scenarios twice, using the old graphics alone, and the HP HMI graphics.

Half used the old graphics first, and half used the HP HMI graphics first).

Quantitative and qualitative measurements were made on the performance of each scenario

(e.g. detection of the abnormal condition, time to respond, correct and successful response).

The ResultsThe high performance graphics were significantly better in assisting the operator in:

• Maintaining situational awareness

• Recognizing abnormal situations

• Recognizing equipment malfunctions

• Dealing with abnormal situations

• Embedding knowledge into the control system

Operators highly rated the Overview screen, agreeing that it provided highly useful “big picture”

situation awareness. Even with only a few hours total with the new graphics, operators had no difficulties

in operating the unit. The High Performance graphics are designed to have intuitive depictions.

Very positive Operator comments were received on the analog depictions, alarm depictions,

and embedded trends. There were consistent positive comments on how “obvious” the HP

HMI made the various process situations. Values moving towards a unit trip were clearly

shown and noticed by the operators.

The operators commented that HP HMI would enable faster and more effective training of

new operations personnel. The negative operator comments generally had to do with lack of

familiarity with the graphics prior to the test.

The best summary quote was this one:

The effect of inertia being the controlling factor for HMI change was once more confirmed. The

existing HMI had been in use since the early 1990s, with simulator training for more than a decade.

Despite clear deficiencies, almost no change to the existing HMI had been made since inception.

“Once you got used to these new graphics,

going back to the old ones would be hell.”


Operators using the existing graphics first in the test were then asked “What improvements

would you make on the existing graphics to help in these situations? In response, there were

very few or no suggestions!

However, operators using the existing graphics after they used the HP HMI graphics had

many suggestions for improvement, namely analog depictions, embedded trends, consistent

navigation, etc!

So, people get “used to” what they have – and do not complain or know what they are missing

if they are unfamiliar with these HP HMI concepts.

The Seven Step Work ProcessThere is a seven step methodology for the development of a high performance HMI.

Step 1: Adopt a high performance HMI philosophy and style guide. You must have a written

set of principles detailing the proper way to construct and implement a high performance

HMI.

Step 2: Assess and benchmark existing graphics against the HMI philosophy. It is necessary to

know your starting point and have a gap analysis.

Step 3: Determine specific performance and goal objectives for the control of the operation

and for all modes of operation. These are such factors as:

• Safety parameters/limits

• Production rate

• Run length

• Equipment health

• Environmental (i.e. Emission control)

• Production cost

• Quality

• Reliability

It is important to document these, along with their goals and targets. This is rarely done and is

one reason for the current state of most HMIs.

Step 4: Perform task analysis to determine the control manipulations needed to achieve the

performance and goal objectives. This is a simple step involving the determination of which

specific controls and measurements are needed to accomplish the operation’s goal objectives.

The answer determines the content of each Level 2, 3, and 4 graphic.

Step 5: Design high performance graphics, using the design principles in the HMI philosophy

and elements from the style guide to address the identified tasks.

A lack of complaints does not indicate that you have a good HMI!


Step 6: Install, commission, and provide training on the new HMI.

Step 7: Control, maintain, and periodically reassess the HMI performance.

HP HMI Implementation on a BudgetWhile desirable, it is not necessary to replace all of your existing graphics to obtain much of

the benefit of HP HMI. A partial implementation can provide most of the benefits. A partial

implementation involves:

• Design new Level 1, Level 2, and Abnormal Situation Management graphics

• Existing graphics can be designated as Level 3 and navigation paths to them altered

• Improvements to those Level 3s (correcting color choices, adding status indications,

adding embedded trends, and providing proper context.) can be made over time

Any facility can afford about twenty new graphics! A High Performance HMI is affordable.

Conclusion

Sophisticated, capable, computer-based control systems are currently operated via ineffective

and problematic HMIs, which were created without adequate knowledge. In many cases,

guidelines did not exist at the time of graphic creation and the resistance to change has kept

those graphics in commission for two or more decades.

The functionality and effectiveness of these systems can be greatly enhanced if redesigned in

accordance with proper principles. A High Performance HMI is practical and achievable.

References

Hollifield, B., Oliver, D., Habibi, E., & Nimmo, I. (2008). The High Performance HMI Handbook.

Operator HMI Case Study: The Evaluation of Existing “Traditional” Operator Graphics vs. High

Performance Graphics in a Coal Fired Power Plant Simulator , Product ID 1017637


About the Authors

Bill R. Hollifield, PAS Principal Alarm Management and HMI Consultant

Bill is the Principal Consultant responsible for the PAS work processes and intellectual property

in the areas of both Alarm Management and High Performance HMI. He is a member of the

American Petroleum Institute’s API RP-1167 Alarm Management Recommended Practice

committee, the ISA SP-18 Alarm Management committee, the ISA SP101 HMI committee, and the

Engineering Equipment and Materials Users Association (EEMUA) Industry Review Group.

Bill has multi-company, international experience in all aspects of Alarm Management and HMI

development. He has 26 years of experience in the petrochemical industry in engineering and

operations, and an additional 9 years in alarm management and HMI software and services for

the petrochemical, power generation, pipeline, and mining industries.

Bill is co-author of The Alarm Management Handbook, The High Performance HMI Handbook,

and The Electric Power Research Institute (EPRI) guideline on Alarm Management.

Bill has authored several papers on Alarm Management and HMI, and is a regular presenter on

such topics in such venues as API, ISA, and Electric Power symposiums. He has a BSME from

Louisiana Tech University and an MBA from the University of Houston.

Hector R. Perez, PAS High Performance HMI Product Manager

Hector oversees the High Performance HMI business line at PAS. He is a chief designer of high

performance graphics intended to facilitate situation awareness in a variety of industries. At

PAS, Hector oversees PAS software directions to improve product design and capabilities.

Prior to working with PAS, Hector was a senior engineer at Schlumberger. His strength in design

contributed to his success in creating new and improved HMI’s for reservoir evaluation services

and interfaces for business Key Performance Indicator tracking.

Bill Hollifield Hector Perez


In addition to his expertise in High Performance HMI, Hector has widespread experience in all

aspects of Alarm Management. He has facilitated numerous Alarm Management workshops,

conducted alarm rationalization projects, and developed Alarm Philosophy documents for a

wide range of clients in the petrochemical, power generation, pipeline, and mining industries.

Hector has authored technical articles on High Performance HMI. In 2009, he collaborated with

the Electrical Power Research Institute (EPRI) on a comparative research study evaluating high

performance graphics and operator effectiveness.

Hector holds a Bachelor of Science in Chemical Engineering from Rice University.


The High Performance HMI Overview v2.01 (1)

Documents

Transcript of The High Performance HMI Overview v2.01 (1)