SY DE 542

1
SY DE 542
Design Phase 3 Multi-Variate Constraints
Configural Mass Data Displays Feb 7, 2005 R.
Chow Email chow_at_mie.utoronto.ca
2
Multivariate Constraints

• Relationships between 2 or more variables
• May be at same abstraction level
• May be across levels
• Often equations

3
Identifying Multivariate Constraints

• Re-visit Variable List
• For each AH level, look
• within a level
• across levels

4
Example Conservation

• In Out Stored
• Holds for Mass, Energy, Money, Information
• Also People, Aircraft, Requests
• As long as nothing is transformed!
• Which AH level?
• Think laws and principles

5
Example Transformation

• If transformation occurs, identify defining
  relationship
• Example Food manufacturing
• ? Butter ? Sugar ? Flour ? Cookies
• Which AH level?
• Relationships may be identified empirically
  (based on experiments or history)

6
Constraints Across Levels

• Shows how low level elements work towards high
  level purposes
• Examples from DURESS
• (1) Mass from 2 feedwater streams
• (2) Energy leaving reservoir
• (3) Flow split
• (Vicente, 1999)

7
Example (1) Mass from 2 Feedwater Streams

• MI1(t) FA1(t) FB1(t)
• MI1(t) Which AH level?
• FA1(t), FB1(t) Which AH level?
• MI1(t) FI1(t) ??
• Why should we be interested in MI1(t)?

8
Example (2)Energy Leaving Reservoir

• EO1(t) MO1(t) cp T1(t)
• EO1(t) Which AH level?
• MO1(t)?
• Cp?
• T1(t)?
• Why should we be interested in EO1(t)?

9
Example (3)Flow Split

• FA1(t) FVA(t) VA1(t)
• VA1(t) VA2(t)
• FA1(t), FVA(t) Which AH level?
• VA1(t), VA2(t) ?
• Why would FA1(t) not be equal to VA1(t)?

10
Designing for Multivariate Constraints

• Visually show relationships between variables
• Eliminate / reduce need for real-time computation
  by user
• Eliminate / reduce need for real-time lookup (of
  data tables, other documentation)
• Show context for relationships

11
Configural Displays

• Idea is display of information for larger systems
• Individual pieces of data interact in a more
  global relationship - higher order relationship
• Right mapping makes that relationship emerge

12
Definitions

• Low level data usually individual sensor data
• High level relation a more global and general
  display of what the data means
• Emergent property or emergent feature a pattern
  or shape that is created from the low level data,
  is recognizable and has meaning

13
AH -gt Design Phase 3

• Bottom of abstraction hierarchy tells you what
  lower level data should be displayed
• Higher levels of the hierarchy tell you why those
  data are important, what relation has meaning
• Emergent feature must mean something to the task

14
Examples
Network health
Heat transfer efficiency
Network parameters
T1, T2, T3, T4, water flow
15
Configural/Separable/Integral?

• Separable
• show each variable as a single output
• equivalent to single sensor single indicator
  display (SSSI)
• integration or higher level relations must be
  derived

16
Integral Displays

• Show high level information but not low level
  information
• Low level information must be derived.

Normal
Not normal
17
Configural Displays

• Arrange low level data into a meaningful form
• whole is greater than the sum of the parts
• based on principles of gestalt psychology

18
Separable vs Configural

• Separable generally makes it easier to extract
  low level information
• harder to integrate
• Configural makes it harder to extract low level
  information
• easier to integrate

19
Bar Graphs

• Can be configural and separable
• Each element can be separated
• Pattern can be configural

20
Configural Displays

• Configural displays typically form an object
• Sometimes called object displays
• The emergent property is the shape of the object
• Emergent property can be found from your
  abstraction hierarchy

21
Emergent Features

• Example
• two variables, x, y
• could map xy
• only meaningful if area, xy has meaning for the
  task
• good mapping xmass, yvelocity, areamomentum

y
x
22
Visual Mathematics

• Equality
• Does xyz
• horizontal line
• Addition
• Does xyz

x
y
z
x
y
z
23
Visual Mathematics

• Simple average
• Does z(xy)/2
• Multiplication
• Does zxy

x
y
z
z
y
x
24
Visual Mathematics

• Division
• does zx/y
• Mapping
• x - vertical
• y - horizontal
• z - tan(Ø)
• Ø tan-1(z)

x
Ø
y
25
N-gon Feature
Construction Select key variables that measure
overall status. Get normal values. Normalize
x/xnormal. Determine alarm limits, colour
coding. Normalization creates the shape.
26
Another Polar Display
Constant angle Variant radius Not
configural Designed by Florence Nightingale
27
Straight Line Feature
Individual temperatures
Vessel temp profile, vessel state
28
Design Exercise

• You have been hired as an interface designer to
  work for Mrs. Fields cookies. Mrs. Fields
  cookie plant is aging and the company has
  realised that they are losing production and
  potential profits whenever cookies turn out
  flawed. Sugar (S kg), butter (B kg) and flour (F
  kg) are mixed to make dough which is then dropped
  onto a conveyor belt. The conveyor belt runs
  through an oven at temperature T and the finished
  cookies exit the oven.
•  To make the best possible cookie, Mrs. Fields
  cookie research team has determined that the
  dough must consist of a consistent relation
  between the amounts of butter, sugar, and flour..
  This is a general property of cookie dough which
  holds over all different kinds of cookies.
  Precisely,
•   Butter ½ sugar, Sugar 1/3 flour

29
Mass Data Displays

• Basic Idea to show large amounts of data in a
  way that is quickly understood
• to show global patterns in data without hiding
  data
• capitalize on human pattern recognition abilities
  and visual perception
• give an overview, a feeling for the behaviour of
  the process

30
Comparison with Configural Displays

• Both are suited for overviews
• both show global relations
• both can make it hard to separate data, get
  individual values

31
Comparison with Configural Displays

• MDD typically handle larger amounts of data
• dont form an object so much as a pattern
• both show elemental data and dont hide data

32
MDDs

• Are somewhat under-used in computer displays
• have been used for years in paper based displays
• similar to the idea of alarm lights in power
  plants
• get a feeling of system state
• analogous to sounds, e.g. hums

33
General Principles

• Show each piece of data as a simple mark on the
  screen (graphic atom)
• Establish the mapping of the dynamics
• what changes?
• how does the mark change?
• graphic atom level changes such as size, shape,
  colour

34
General Principles

• Determine the arrangement of marks
• what is the organization?
• what is the mapping to location in the display
  space?
• Possibilities
• Topological - follow system connections
• Type of Data - organize temps, pressures, etc.
• Frame of Reference, scaling.

35
General Principles

• What is the pattern that should emerge?
• What does the display look like under different
  conditions?
• Separability To what extent must the operator be
  able to extract the individual value?

36
ABB displays
Polar Star
Mass Data Display
Plant mimic
Plant graphs
37
ABB MDD
Data are normalised so normalhorizontal
Data are normal
Mark is line Change is in angle Organisation is
plant topology
Data are deviating
38
ABB MDD

• Normalisation adds context
• Not normal is more salient
• Faults cascade through plant
• Experimental results
• fault detection 20 times faster

39
ABB MDD

• Other marks they considered
• Circle
• Lines change in thickness

40
The Daisy Wheel
Website use (access and errors) Mark is the line
between elements Clustering of lines shows
information
41
www.smartmoney.com/marketmap
Mass Data Display for Financial Market Mark is
rectangular shape, Tile map Varies in Colour to
show Gains and Losses
42
Ozone levels in LA, 10 years
Technique Coloured tiles
43
Scatterplots
44
Design Exercise

• It is estimated that Mrs. Field's produces
  500,000 cookies a day.
• Each cookie is inspected for size, shape, and
  baking quality (undercooked, cooked, and
  overcooked). Design a mass data display for this
  situation. What sort of dimensions could you
  organize your display with?
•  (Note you don't have to show all 500,000
  cookies)

45
Next Week

• Guest Lecturer Prof. Greg Jamieson
• EID for Petrochemical Processing
• Work Domain Task Analysis
• Design and Evaluation
• No slides will be posted
• Submit final checkpoint to Munira by email
• Before Wed. Feb. 16, 5pm