Created at

1

• Program for North American Mobility in Higher
  Education (NAMP)
• Introducing Process Integration for Environmental
  Control in Engineering Curricula (PIECE)

Module 8 Introduction to Process Integration
Tier 2
2
Project Summary

• Objectives
• Create web-based modules to assist universities
  to address the introduction to Process
  Integration into engineering curricula
• Make these modules widely available in each of
  the participating countries
• Participating institutions
• Two universities in each of the three countries
  (Canada, Mexico and the USA)
• Two research institutes in different industry
  sectors petroleum (Mexico) and pulp and paper
  (Canada)
• Each of the six universities has sponsored 7
  exchange students during the period of the grant
  subsidised in part by each of the three
  countries governments

3
Structure of Module 8

• What is the structure of this module?
• All Modules are divided into 3 tiers, each with a
  specific goal
• Tier I Background Information
• Tier II Case Study Applications
• Tier III Open-Ended Design Problem
• These tiers are intended to be completed in that
  particular order. Students are quizzed at various
  points to measure their degree of understanding,
  before proceeding to the next level.
• Each tier contains a statement of intent at the
  beginning and there is a quiz at the end of Tiers
  I and II.

4
Purpose of Module 8

• What is the purpose of this module?
• It is the intent of this module to cover the
  basic aspects of Process Integration Methods and
  Tools, and to place Process Integration into a
  broad perspective. It is identified as a
  pre-requisite for other modules related to the
  learning of Process Integration.

5
Tier 2 Worked Examples
6
Tier II Objective

• Tier II Statement of intent
• The goal of this tier is to demonstrate various
  concepts and tools of Process Integration using
  real examples. Three examples will be given,
  focusing mainly on three Process Integration
  tools. At the end of Tier II, the student should
  have a general idea of what is
• Data-Driven Modeling - Multivariate Analysis
• Thermal Pinch Analysis
• Integrated Process Control and Design
  Controllability Analysis

7
Tier II Contents
Tier II is broken down into three sections 2.1
Worked example using Data-Driven Modeling, more
specifically Multivariate Analysis 2.2 Worked
example using Thermal Pinch Analysis 2.3 Worked
example using Integrated Process Control and
Design, more specifically Controllability
Analysis A short multiple-choice quiz will
follow at the end of this tier.
8
Outline
2.1 Worked example 1 Data-Driven Modeling
Multivariate Analysis 2.2 Worked example 2
Thermal Pinch Analysis 2.3 Worked example 3
Integrated Process Control and Design
Controllability Analysis
2.1 Worked example 1 Data-Driven Modeling
Multivariate Analysis 2.2 Worked example 2
Thermal Pinch Analysis 2.3 Worked example 3
Integrated Process Control and Design
Controllability Analysis
9
Worked example 1 Data-Driven Modeling
Multivariate Analysis
10
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis Reminder
Graphical representation of MVA
Statistical Model
(internal to software)
Tmt X1 X4 X5 Rep Y avec Y sans
1 -1 -1 -1 1 2.51 2.74
1 -1 -1 -1 2 2.36 3.22
1 -1 -1 -1 3 2.45 2.56
2 -1 0 1 1 2.63 3.23
2 -1 0 1 2 2.55 2.47
2 -1 0 1 3 2.65 2.31
3 -1 1 0 1 2.45 2.67
3 -1 1 0 2 2.6 2.45
3 -1 1 0 3 2.53 2.98
4 0 -1 1 1 3.02 3.22
4 0 -1 1 2 2.7 2.57
4 0 -1 1 3 2.97 2.63
5 0 0 0 1 2.89 3.16
5 0 0 0 2 2.56 3.32
5 0 0 0 3 2.52 3.26
6 0 1 -1 1 2.44 3.1
6 0 1 -1 2 2.22 2.97
6 0 1 -1 3 2.27 2.92
.
.
.
.
.
.
.
.
.
.
.
.
Raw Data impossible to interpret
hundreds of columns
thousands of rows
11
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Basic Statistics

• It is assumed that the student is familiar with
  the following basic statistical concepts mean,
  median, mode standard deviation, variance
  normality, symmetry degree of association,
  correlation coefficients R2, Q2, F-test
  significance of differences, t-test, Chi-square
  eigen values and vectors
• Statistical tests help characterize an existing
  dataset. They do NOT enable you to make
  predictions about future data. For this we must
  turn to regression techniques

Regression

• Take a set of data points, each described by a
  vector of values (y, x1, x2, xn)
• Find an algebraic equation that best expresses
  the relationship between y and the xis
• Y b1x1 b2x2 bnxn e
• Data Requirements normalized data, errors
  normally distributed with mean zero and
  independent variables uncorrelated

12
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Types of MVA
Types of MVA outputs

• MVA software generates two types of outputs
  results, and diagnostics.
• Results Score Plots, Loadings Plots
• Diagnostics Plot of Residuals, Observed
• vs. Predicted, and many more

Q1
Q2
13
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis PCA
Principal Component Analysis (PCA)
Consider these fish. We could measure, for each
fish, its length and breadth.
Suppose that 50 fish were measured, a plot like
the one shown in figure 2 might be obtained.
There is an obvious relationship between length
and breadth as longer fish tend to be broader.
Reference Manchester Metropolitan University
14
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis PCA

• Move the axes so that their origins are now
  centered on the cloud of points this is a
  change in the measurement scale. In this case the
  relevant means were subtracted from each value.

• In effect the major axis is a new variable,
  size. At its simplest, size length breadth
• ? linear combination of the two existing
  variables, which are given equal weighting
• Suppose that we consider length to be more
  important than breadth in the determination of
  size. In this case we could use weights or
  coefficients to introduce differential
  contributions size 0.75 x length 0.25 x
  breadth
• For convenience, we would normally plot the
  graph with the X axis horizontal, this would give
  the appearance of rotating the points rather than
  the axes.

Reference Manchester Metropolitan University
15
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis PCA

• A criterion for the second axis is that it
  should account for as much of the remaining
  variation as possible. However, it must also be
  uncorrelated (orthogonal) with the first.

• In this example the lengths and orientations of
  these axes are given by the eigen values and
  eigen vectors of the correlation matrix. If we
  retain only the 'size' variable we would retain
  1.75/2.00 x 100 (87.5) of the original
  variation. Thus, if we discard the second axis we
  would lose 12.5 of the original information.

Reference Manchester Metropolitan University
16
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Projection to Latent Structures (PLS)

• PLS finds a set of orthogonal components that
• maximize the level of explanation of both X and Y
• provide a predictive equation for Y in terms of
  the Xs
• This is done by
• fitting a set of components to X (as in PCA)
• similarly fitting a set of components to Y
• reconciling the two sets of components so as to
  maximize explanation of X and Y
• Interpretation of the PLS results has all the
  difficulties of PCA, plus another one making
  sense of the individual components in both X and
  Y space. In other words, for the results to make
  sense, the first component in X must be related
  somehow to the first component in Y

17
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Problem Statement
Lets look at a typical integrated
thermomechanical pulp (TMP) newsprint mill in
North America. The mill manager of that
particular plant recognizes that there is too
much data to deal with and that there is a need
to estimate the quality of their final product,
i.e. paper. He decides to use Multivariate
Analysis to derive as much information as
possible from the data set and try to determine
the most important variables that could have an
impact on paper quality in order to be able to
classify final product quality. The mill manager
decides to first look at the refining portion of
the pulping process.
18
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
X and Y Variables

• X variables
• Incoming chips size distribution, bulk density,
  humidity
• Refiner operating data throughput energy split
  between the primary and secondary refiner
  dilution rates levels, pressures and
  temperatures in various units immediately
  connected to the refiners voltage at chip screw
  conveyors refiner body temperature
• Season, represented by the average monthly
  temperature measured at a nearby meteorological
  station

• Y variables
• Pulp quality data after the latency chest
  (automated, on-line analysis of grab samples)
  standard industry parameters including fibre
  length distribution, freeness, consistency, and
  brightness

19
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Results
This is the R2 and Q2 plot for the model. The R2
values tell us that the first component explains
32 of the variability in the original data, the
second another 7 and the third another 6. The
Q2 values are lower. This means that the
predictive power of the model is around 40 when
using all three components. This may seem low,
but is normal for real process data.
20
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Interpretation of the results Score Plot
Autumn Winter Spring Summer
21
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Interpretation of the results Loadings plot
Figure 13
22
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Interpretation of the results

• First Component
• The first component corresponds to throughput
  many process variables are related either
  directly or indirectly to throughput. Remember we
  said that the 1st component was something that
  varied within an individual season?
• Second Component
• The 2nd component explains only 7 of the total
  variability. It is therefore messier than the
  first component, and will be less easy to
  interpret. It is also possible to note that the
  three years were separated with respect to this
  second component
• A major clue occurs in the prominence of two
  important and related tags bleach consumption
  and pulp brightness. This would suggest that
  perhaps the brightness of the incoming wood chips
  was different from year to year, requiring more
  bleaching to get a less white pulp
• Note also that Season is prominent. This can be
  seen with the obvious separation of the seasons
  on the score plot. This suggests that winter
  chips are less bright than summer chips
• Third Component
• The 3rd component explains only 6 of the total
  variability
• The 3rd component is related to the time of year.
  A reasonable interpretation would be that summer
  chips differ from winter chips in some way other
  than brightness, which was already covered by the
  second component. This could be, for instance,
  the ease with which the wood fibres can be
  separated from each other

23
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Summary of the PCA results
Using PCA, we have determined that 45 of the
variability in the original 130 variables can be
represented by using just 3 new variables or
components. These three components are
orthogonal, meaning that the variation within
each one occurs independently of the others. In
other words, the new components are uncorrelated
with each other.
24
2.1 Worked example 1 Data-Driven
Modeling Multivariate Analysis
Quality reference map
X
X
X
Figure 14
25
Outline
2.1 Worked example 1 Data-Driven Modeling
Multivariate Analysis 2.2 Worked example 2
Thermal Pinch Analysis 2.3 Worked example 3
Integrated Process Control and Design
Controllability Analysis
2.1 Worked example 1 Data-Driven Modeling
Multivariate Analysis 2.2 Worked example 2
Thermal Pinch Analysis 2.3 Worked example 3
Integrated Process Control and Design
Controllability Analysis
26
Worked example 2 Thermal Pinch Analysis
27
2.2 Worked example 2 Thermal Pinch Analysis
Reminder
What is Thermal Pinch Analysis?
Utility Usage
Utility costs go down
Costs related to exchange area go up
Internal Exchanges
From 100 utility...
... to 100 internal exchanges
28
2.2 Worked example 2 Thermal Pinch Analysis
What about an entire site ?
At least 40 streams to heat and cool

• Example Recovery Boiler
• Obvious solution preheat entering fresh water
  with hot condensate leaving boiler

Figure 15
29
2.2 Worked example 2 Thermal Pinch Analysis
Simulation
?Tmin
30
2.2 Worked example 2 Thermal Pinch Analysis
Composite Curves
31
2.2 Worked example 2 Thermal Pinch Analysis
Mass Integration Composite Curves for pollution
prevention
Figure 18
Figure 17
32
2.2 Worked example 2 Thermal Pinch Analysis
Problem Statement
A process engineer in a consulting firm is hired
by an oil refinery to design the Conventional
Atmospheric Crude Fractionation Units section of
the refinery facility, as shown in figure 17. The
main objective of this project is to minimize the
energy consumption by using Thermal Pinch
Analysis. The plant is currently using 75000 kW
in hot utilities. In this example, stress will be
put on the construction of the composite curves
with the objective of identifying energy savings
opportunities.
33
2.2 Worked example 2 Thermal Pinch Analysis
Data Extraction
34
2.2 Worked example 2 Thermal Pinch
Analysis Composite Curves
1. Sort in ascending order the hot streams
temperatures, omitting common temperatures
Using the data above, we form temperature
intervals for the process
35
2.2 Worked example 2 Thermal Pinch
Analysis Composite Curves
2. Sum up the CP of every stream present in each
temperature interval
36
2.2 Worked example 2 Thermal Pinch
Analysis Composite Curves
3. Calculate the net enthalpy for each
temperature interval
37
2.2 Worked example 2 Thermal Pinch
Analysis Composite Curves
4. Obtain the accumulated enthalpy for each
temperature interval
38
2.2 Worked example 2 Thermal Pinch
Analysis Composite Curves
5. Plot temperature on the Y axis versus
accumulated enthalpy on the X axis
39
2.2 Worked example 2 Thermal Pinch
Analysis Composite Curves
The construction of the Cold Composite Curve is
similar to that of the Hot Composite Curve.
40
2.2 Worked example 2 Thermal Pinch
Analysis Composite Curves

• This representation reduces the entire process
  into one combined hot and cold stream

• The heat recovery between the composite curves
  can be increased until we reach DTmin. Composite
  curves, just like individual streams can be
  shifted horizontally on the T-H diagram without
  causing changes to the process because H is a
  state function

• This sets the minimum hot (QHmin) and cold
  (QCmin) utilities requirements for the entire
  process and the maximum possible process-process
  heat recovery

41
2.2 Worked example 2 Thermal Pinch Analysis
Summary of results

• As seen in the previous slides, from the
  temperature-enthalpy plot, we can determine three
  useful pieces of information
• Amount of possible process-process heat recovery
  represented by the area between the two
  composites curves
• Hot Utility requirement or target 57668 kW
• Cold Utility requirement or target 30784 kW

Composite curves are excellent tools for learning
the methods and understanding the overall energy
situation, but minimum energy consumption and the
heat recovery Pinch are more often obtained by
numerical procedures. This method is called the
Problem Table Algorithm. Typically, it is based
on notions of Heat Cascade.
Q5
Q6
42
Outline
2.1 Worked example 1 Data-Driven Modeling
Multivariate Analysis 2.2 Worked example 2
Thermal Pinch Analysis 2.3 Worked example 3
Integrated Process Control and Design
Controllability Analysis
2.1 Worked example 1 Data-Driven Modeling
Multivariate Analysis 2.2 Worked example 2
Thermal Pinch Analysis 2.3 Worked example 3
Integrated Process Control and Design
Controllability Analysis
43
Worked example 3 Integrated Process Control and
Design Controllability Analysis
44
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis Reminder
Fundamentals
PROCESS RESILIENCY
Input Variables
Control Loop
Disturbances
Input Variables (manipulated)
Process
Internal interactions
Output Variables (controlled and measured)
Uncertainties
PROCESS FLEXIBILITY
Figure 25
45
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Water F1,C1
CC
FC
C, F
Pulp F2,C2
Figure 26
INPUTS (manipulated variables or disturbances)
OUTPUTS (Best Selection by Controllability
analysis)
EFFECTS
46
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
_
u1
y1


F11
C1
y1
y1sp

F21
F12

y2
y2sp
u2
y2

F22
C2

_
Figure 27
47
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Experiment 1 Step Change in u1 with all loops
open
ss
Du1
Figure 28
Main Effect
48
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Experiment 2 Step Change in u1 with all loops
closed
Du1
ss
Figure 29
Main Effect
Total Effect
Interactive Effect
49
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Relative Gain and Relative Gain Array (RGA)
?11 measure of the extent of steady state
interaction in using u1 to control y1, while
using u2 to control y2
Main Effect (1st Experiment)
Total Effect (2nd Experiment)
Relative Gain
Relative Gain Array
y1
yi
u1
uj
50
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Selection of Loops using RGA How to select the
configuration with minimum interaction
Table 8
yi Controlled variable uj Manipulated variable
51
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Other Controllability Indexes

• Niederlinski (NI) system stability index
• Condition Number (CN) and Disturbance Condition
  Number (DCN) sensibility measure
• Relative Disturbance Gain (RDG) index that
  gives an idea of the influence of internal
  interactions on the effect of disturbances
• Others Singular Value Decomposition (SVD)

52
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Problem Statement
In this case-study, a process control engineer is
asked to create a model of the thermomechanical
pulping process to find the best process control
selection and variable pairing for a plant that
has not been built yet. Consider the simplified
newsprint paper machine short loop configuration
shown in figure 30. Variable pairing techniques
will be applied as well as the use of
controllability indexes.
Figure 30
53
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
controlled
Problem Statement
manipulated
disturbances
Pfin Fines retained
54
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Disturbances
C
Fines
Controlled
BR
Manipulated
Figure 31
55
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Process Gain Matrices and Steady-State
Controllability


Gd
Gp
Manipulated
Controlled
Disturbances
?
56
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Figure 32
57
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Controllability Indexes (1)

• Niederlinski Index (NI) ? Stability
  considerations
• NI lt 0. System will be unstable under closed-loop
  conditions
• NI gt 0. System is stabilizable (function of
  controller parameters)
• Condition number (CN) ? Sensitivity to model
  uncertainty
• CN lt 2. Multivariable effects of uncertainty are
  not likely to be serious
• CN gt 10. ILL-CONDITIONED process

NI0.73
CN713
58
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Controllability Indexes (2)

• Disturbance Condition Number (DCN) ? Is the
  action taken by the manipulated variable large or
  small?
• 1 DCN CN
• Relative Disturbance Gain (RDG) ? Internal
  interaction among the loops is favorable or
  unfavorable to reject disturbances?
• RDG lt2 . Internal interactions reduce the effect
  of the disturbance

DCN for Cfresh pulp 9.2 DCN for finesfresh
pulp 4.6
? It is harder to reject a sudden change in fresh
pulp consistency
The effect of both disturbances, C and fines in
FRESH PULP, is reduced by internal interactions.
All RDGs are lt2
59
2.3 Worked example 3 Integrated Process Control
and Design Controllability Analysis
Conclusion

• Control structure configuration RGA results
  confirmed current implementation in newsprint
  mills
• Internal interactions of the aforementioned
  configuration reduce the effect of disturbances
  on output variables
• The process is ill-conditioned. Model
  uncertainty may be highly amplified
• Resiliency Indexes, DCN and RDG, can be used to
  account for disturbance rejection in newsprint
  processes

60
End of Tier II

• This is the end of Tier II. At this point, we
  assume that you have done all the reading. You
  should have a pretty good idea of what Process
  Integration is as well as basic knowledge in
  regards to Multivariate Analysis, Thermal Pinch
  Analysis and Controllability Analysis. For
  further information on the tools presented in
  Tier II as well as on other Process Integration
  tools introduced in Tier I, please consult the
  references slides in Tiers I and II.
• Prior to advancing to Tier III, a short multiple
  choice quiz will follow.

61
QUIZ
62
TIER II - QUIZ
Question 1

• What is Principal Components Analysis used for?
• Understand relations between the variables of a
  system
• Identify the components having an influence on
  one or many outputs
• Predict certain outputs
• Maximize the covariance of a set of variables

2 and 3
1 and 3
1
3
1 and 2
1,2 and 3
63
TIER II - QUIZ
Question 2
Associate each Multivariate Analysis output with
the kind of information it provides the user
with. 1. Residuals plot A. Shows all the
original data points in a new set of
coordinates or components 2. Score plot B.
Shows the distance between each real
observation in the initial dataset and the
predicted value based on the model 3.
Observed vs. Predicted C. Shows the accuracy of
prediction 4. Loadings plot D. Shows how
strongly each variable is associated with
each new component
1B, 2A, 3C, 4D
1D, 2B, 3A, 4C
1C, 2D, 3A, 4B
1B, 2C, 3D, 4A
1A, 2D, 3B, 4C
1B, 2D, 3C, 4A
64
TIER II - QUIZ
Question 3
The lengths and orientations of the axes obtained
with a PCA are given by the eigen values and
eigen vectors of the correlation matrix. Let's
say the length and breadth variables have a lower
correlation coefficient than in the example given
in slide 13 and that we obtain the eigen values
shown in the figure below. If we discard the
second axis, what percentage of the original
information would we lose?
12,5
75
25
62,5
37,5
0
65
TIER II - QUIZ
Question 4
In the context of a Thermal Pinch Analysis, what
is a hot stream? 1. A process stream that needs
to be heated 2. A process stream with a very high
temperature 3. A process stream that is used to
generate steam 4. A process stream that needs to
be cooled
1
3
2
4
66
TIER II - QUIZ
Question 5
A Thermal Pinch Analysis has been performed at a
plant and the DTmin was set at 40ºC. If another
plant was to be built with a lower DTmin, how
would the corresponding energy costs be in
comparison to the first plant?
Higher
Lower
Would stay the same
67
TIER II - QUIZ
Question 6

• Which of the following statements are true?
• Minimum energy consumption and the heat recovery
  Pinch are more often obtained by Composite Curves
• Composite curves, just like individual streams,
  can be shifted horizontally on the T-H diagram
  without causing changes to the process
• Heat can sometimes be transferred across the
  Pinch
• With the help of ?Tmin and the thermal data,
  Pinch Analysis provides a target for the minimum
  energy consumption

2 and 3
2 and 4
1 and 3
3 and 4
1 and 2
All of the above
68
TIER II - QUIZ
Question 7
Associate each controllability tool or index with
the kind of information it provides the user
with. 1. Niederlinski Index A. Shows the
importance of interactions in a system 2.
Relative Disturbance Gain B. Estimates the
sensitivity of the problem's answer to error
in the input 3. Condition Number C. Includes
disturbances in interactions analysis 4.
Relative Gain Array D. Discusses the stability
of a closed-loop control configuration
1B, 2A, 3C, 4D
1D, 2B, 3A, 4C
1C, 2D, 3A, 4B
1B, 2C, 3D, 4A
1A, 2D, 3B, 4C
1D, 2C, 3B, 4A
69
TIER II - QUIZ
Question 8
In the Relative Gain Array shown in slide 54,
what do the values 1.566 and 1.603 for the
pairing of F40 and C34, and Pfin and Ret, tell
you? 1. There is no interaction with other
control loops 2. The interactive effect is
more important than the main effect 3. The
manipulated input has no effect on output 4. The
interactions from the other loops are opposite in
direction but smaller in magnitude than the
effect of the main loop 5. Pairing is
recommended 6. Pairing is not recommended
1 and 5
4 and 5
3 and 6
2 and 5
2 and 6
4 and 6
70
TIER II - QUIZ
Question 9

• Which of the following statements are false?
• Feedforward control compensates for immeasurable
  disturbances
• Feedback control compensates for measurable
  disturbances
• Resiliency is the degree to which a processing
  system can meet its design objectives despite
  uncertainties in its design parameters
• Flexibility is the degree to which a processing
  system can meet its design objectives despite
  external disturbances

2 and 3
2 and 4
1 and 3
3 and 4
1 and 2
All of the above
71
TIER II - QUIZ
Answers
Question 1 1 and 2 Question 2 1B, 2A, 3C,
4D Question 3 37,5 Question 4 4 Question
5 Lower Question 6 2 and 4 Question 7 1D, 2C, 3B,
4A Question 8 4 and 5 Question 9 All of the above