MONITORING AND DIAGNOSIS

1
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
MONITORING AND DIAGNOSIS
To perform monitoring and diagnosis, people need
information.
Sensor Issues (a little review)

• Real-time Decisions
• Trouble shoot incidents using proven method
• Information to support decisions required quickly
• Process Performance Improvement
• Longer term performance indicators based on data
  and calculations
• Usually identifying slow trends

Real-time sensors Fast lab analyses
Sensors for calculations Lab analysis Data for
statistical analysis
See trouble shooting lesson for more on
strategy and examples
2
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Sensors
MONITORING AND DIAGNOSIS Sensors
Selecting sensor technology

• Achieve required accuracy and reproducibility
• Achieve required reliability
• - Functions well for range of process
  environments
• - Protect from damage (e.g., thermowell) or
  sample
• from process and measure in less harsh
  conditions
• - Redundancy, if required
• - Redundancy with diversity, if required
• Provide means for calibration and maintenance
• Balance cost (capital,maintenance and operating)
  with benefit to achieve economic return

http//www.pc-education.mcmaster.ca/instrumentatio
n/go_inst.htm (Select 2.0 Sensors)
3
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Sensors
MONITORING AND DIAGNOSIS Sensors
Sensors used for plant control
Sensors used for plant display and monitoring
Higham, E.H., A Route to Better Process
Measurements, I.Chem.E. Hazards X - Process
Safety in Fine Chemical and Specialty Plants,
1989. (Current practice might include more
on-stream analysis.)
4
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Sensors
MONITORING AND DIAGNOSIS Sensors
Class exercise on sensors You have been asked to
evaluate the performance of a process. You will
use all forms of data, from trend to historical
reports. What are some of the questions that you
might to ensure that you interpret the sensor
data correctly?
Why should I ask questions? Every data value is
exact, at least to the number of significant
figures on the plot or in the numerical display!
5
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Sensors
MONITORING AND DIAGNOSIS Process Performance
Improvements
Class exercise a. Sensor range

• Most sensor accuracies depend on the range (or
  span). A sensor with a range of 1000 K has a
  poorer accuracy than one with a range of 100 K.
• Some sensor accuracies depend on the position in
  the range. For example, an orifice meter has a
  very poor accuracy at 10 of maximum range.
• Also, a flow meter can report a very small value
  when the flow is actually zero.
• 3. If the measured value equals the maximum
  (minimum) of the range, the plant value could be
  much higher (lower) than the reported value.

6
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Sensors
MONITORING AND DIAGNOSIS Process Performance
Improvements
Class exercise b. Sensor Technology

• Different sensors for the same process variable
  (P, T, F, L, etc) can have very different
  accuracies. For example, a RTD is much more
  accurate that a thermocouple.
• Different sensor technologies are robust
  (sensitive) to changes in process conditions,such
  as suspended solids in a flowing fluid. This
  knowledge helps in anticipating potential sensor
  problems, sensor loss of accuracy and in trouble
  shooting process operation.

7
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Sensors
MONITORING AND DIAGNOSIS Process Performance
Improvements
Class exercise c. Measurement compensation.
Several measurements can be used to provide a
more accurate value for a single process variable.
Orifice flow meters relate the pressure change
across the orifice to the flow rate. The
relationship depends on the fluid density.
Either the density or gas pressure and
temperature can be used to correct for changes
for an assumed (design) value. You need to know
whether or not this correction has been performed.
Is this measured and used in the calculation, or
is the value assumed constant (at the design
value?)
8
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
We have learned the importance of measurements in
trouble shooting.

• During process design, we need to brainstorm a
  list of likely faults that will require trouble
  shooting
• Faults in sensors are expected and can easily
  mislead personnel. Select reliable sensors that
  match process environment and add redundancy
  where appropriate.
• For each, the diagnosis procedure can be
  documented (and used for training and manual
  writing). Fishbone diagrams aid the thought
  process and documentation.
• We need to ensure that the required measurements
  (sensors and laboratory) are provided to support
  successful trouble shooting!

9
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
We select measurements based on an analysis of
design goals and possible incidents
Root cause 1

• Significant Incident 3
• Safety
• Equipment damage
• Environmental impact
• Product quality
• Production rate
• Profitability

Root cause 2

Root cause n
Select measurements that enable operations
personnel to uniquely determine the most likely
root causes of every significant incident.
10
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions

• Balancing Function and Cost Location of sensors
• Control room Displayed Automatic Control
• The variable is required to achieve one of the
  control objectives for the process.
• Necessary information for the controller can be
  measured. Variable should be monitored reliably
  and rapidly.
• Rapid and reliable feedback is required.
• Automated action may be discrete or continuous.
• Manipulated variable can be adjusted
  automatically
• Stored in computer history and available on trend
  plot
• All of above applies to safety shutdown as well
  as modulating control.

11
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions

• Balancing Function and Cost Locations of
  sensors
• Control room Displayed
• Variable need not be controlled automatically
  can be used to monitor process/equipment
  performance
• If controlled, slowly by operator action is
  acceptable
• If controlled, manipulated variable adjusted from
  control room (manual station to change valve
  open or motor on/off)
• Stored in computer history and available on trend
  plot

12
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions

• Balancing Function and Cost Location of sensors
• Locally Displayed
• Variable need not be controlled can be used to
  monitor process/equipment performance
• Operator must travel to unit to observe the
  display
• Used to monitor slowly changing process/equipment
  performance, for example, heat exchanger fouling
  and pressure drops in packed beds
• Can be used for trouble shooting when rapid
  response is not required
• Not stored in computer history - can be recorded
  on written log very infrequently (1/shift or day)

T
3
13
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
Real-time data is used to determine the health
of the process. Use your process knowledge!
For a chemical reaction with a significant heat
of reaction (endo- or exothermic), the
temperature change across the reactor provides an
inference of the extent of reaction. Very useful,
especially when on-stream analysis is not
practical
Hydrocracker reactions are highly
exothermic Runaway is possible.
14
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
Real-time data is used to determine the health
of the process. Use your process knowledge!
For semi-batch (batch-fed) bio-reactor, the
off-gas flow rate and composition can be used to
monitor the health of the reaction system
should the batch continue to completion or be
aborted?
PC
TC
Gas produced by the biological process
FC
fo
fc
L
fo
CW
fo
15
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
Real-time data is used to determine the health
of process equipment. Use your knowledge of
equipment.

• Used for monitoring, display, and alarms to
  personnel
• Used for SIS for safety and equipment protection
• Equipment manufacturers often provide monitoring

• Monitoring rotating equipment can include
• Vibration
• Temperature
• Flow
• Lubrication pressure
• Power consumption

Turbine
Compressor
T. Reeves, EPTQ, Q3, 2005 (www.eptq.com)
16
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
Real-time data is used to determine the health
of process equipment. Use your knowledge of
equipment.

• Used for monitoring, display, and alarms to
  personnel
• Used for SIS for safety and equipment protection

Figure shows schematic of monitor for pump and
motor by T. Reeves, EPTQ, Q3, 2005 (www.eptq.com)
17
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
Real-time data is used to determine the health
of process equipment. Use your knowledge of
equipment.
In a chemical reactor, poor flow distribution can
lead to hot spots that can damage catalyst or
even the reactor vessel. Locating many
temperature sensors at various locations in the
bed provides monitoring for poor flow
distribution. Used for monitoring, alarms and
control.
18
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
We need information for process troubleshooting.
Remember that people are the ultimate backup
protection they have to correct for equipment
malfunctions.
Class Workshop Add the sensors required to
monitor this distillation tower in real time.

• Hints
• What limits must not be violated?
• What incidents must we diagnose?
• What equipment can fail?
• What redundancy is needed?
• Where should display be located?

19
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
Class Workshop Add the sensors required to
monitor this distillation tower in real time.

• Some thoughts
• pressure of closed vessel is very important
• levels in the accumulators are unstable
• trays can leak liquid to lower trays
• trays can become blocked
• internal flows can exceed hydraulic limits
• utility streams are important (e.g., steam)
• the composition of trays should have an expected
  profile
• pumps can malfunction
• cooling water temperature should be below maximum
  limit

20
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Real-time
MONITORING AND DIAGNOSIS Real-time Decisions
YES !!
21
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements

• Many features of a process change slowly, over
  days, weeks or months
• The performance of complex systems is often not
  obvious from direct observation of the data
• Engineers can identify key process performance
  measures that can be calculated automatically and
  stored in history
• Actions are based on careful analysis of the data
  and might require either minor changes or
  extensive plant changes, during shutdowns

22
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Often, several measurements are used to calculate
a key process variable from a complex process.
Examples of performance measures are

• Yields from a reactor
• Electricity consumption per kg feed
• Total effluent (of water, sulfur, etc.)
• Efficiency of equipment (turbine, compressor,
  fired heater, etc.)
• Operating conditions for successful and
  unsuccessful batches
• Inventory in plant(work in progress, feed, and
  finished products)

23
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Engineers must understand the sensors used and
the methods for data storage before deciding how
to use the measurement data.

• Each sensor has a physical principle affecting
  its accuracy and reproducibility
• Data is stored in a history data base, but some
  information is lost
• Some actions by people are also recorded
• Some times, measurement and computing equipment
  fail

24
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements

• The historical data base typically contains
• Sensor measured values
• Events Alarms, set point changes, controller
  mode and tuning changes, SIS activation
• Calculated variables (defined and build by
  engineer)
• Cause of SIS activation
• Laboratory analysis

25
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements

• Some typical questions answered using the
  historical data base
• How much feed did we process last week?
• What was the yield of vinyl chloride monomer?
• How much energy was consumed per 1000 kg of
  product?
• What was the total release of sulfur from the
  plant last month?
• What was the distribution of product quality,
  displayed as a histogram?
• How much valuable hydrogen was diverted to fuel
  gas last week?

26
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Storing ½ second data for 1000 measurements for
30 years is a lot of data!
1000 x 120 x 60 x 24 x 365 x 30 ? 2 x 1012
Therefore, the data base has several layers with
data aggregated to reduce storage. While storage
capacity and computer speeds will continue to
increase, the basic design will likely persist
for a long time because we dont need every data
point for many analyses.
27
MONITORING AND DIAGNOSIS Process Performance
Improvements
Typical structure
Sampling periods and storage duration tailored to
need. For example, SIS period is very fast
(milliseconds) to diagnose fault.
Special purpose systems
Aggregation Large amounts of data stored for
long times. Data is taken periodically with long
sampling periods, e.g., 5 minutes.
Historian
Data is stored with decreasing resolution, e.g.,
3 days of 1/min, 7 days of 1/hr, 30 days 1/day
DCS control system
Trend plots are updated 1/sec for new data, but
this high-frequency data is not stored for later
recall.
28
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements

• Some typical ways to aggregate a variable
• Average (shift, daily, weekly, monthly)
• Integration, e.g., total flow per day (for
  material or energy)
• Maximum and minimum over a period
• Standard deviation (or histogram)
• Analysis can have multiple values

• To aggregate several variables, use process
  insight
• Calculate key process performance measures
  (efficiencies, yields, etc.)

29
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Class Exercise Occasionally, a sensor fails or
the history storage fails for a period of time.
How does the aggregation method handle these
situations?
30
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Class exercise Periods when a reliable value is
not available is marked questionable.

• Reports should provide information on whether any
  data within the aggregation period was
  questionable. (Best if of data that is
  questionable reported)
• You need to determine how the algorithm deals
  with missing data (ignore, interpolate, use last
  good before, use first good after, etc.) when
  calculating results, such as average or total.

31
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Plant personnel also check local sensors
periodically (per shift or day) and record
values. This data can be stored on paper, or the
values can be entered into a hand-held computer
and transferred to the history data base.
32
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Plant personnel extract samples of material for
laboratory analysis. This data can be stored on
paper, or the values can be entered into the
history data base. The time the sample was taken
is essential information.
33
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Class Exercise The data changes from sample to
sample. How do we determine when a significant
change has occurred?
34
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
We can use principles of statistical process
control (SPC) to monitor and decide when a
significant change has occurred. Process
personnel trouble shoot, diagnose and eliminate
the cause
Shewhart Chart
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
35
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Class Workshop In your first job, you are
responsible for the heat exchangers and fired
heater in this process. Design a monitoring
procedure (with sensors and lab analyses).
Heat exchange with process streams that must be
cooled.
feed
tank
fuel

• Hints
• What can change?
• Is it important for plant performance?
• How does it affect measurements?
• Define measurements and calculations.

product
36
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Process Performance
Improvements
Class Workshop In your first job, you are
responsible for the reactor (and catalyst
regenerator) in the process shown on the next
slide. Design a monitoring procedure (with
sensors and lab analyses).

• Hints
• What can change?
• Is it important for plant performance?
• How does it affect measurements? Include
  laboratory samples for variables that might not
  be measured on-stream.
• Define measurements and calculations.

37
Longer term

• Hints
• What can change?
• Is it important for plant performance?
• How does it affect measurements?
• Define measurements and calculations.

38
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
Longer term
MONITORING AND DIAGNOSIS Real-Time Process
Performance Improvements

• INDUSTRIAL PRACTICE
• Standard measurement choices have been developed
  for most unit operations.
• Some issues require advanced analysis, for
  example, pipeline leak detection, rotating
  machinery vibration, pump alignment/lubrication,
  compressor surge, etc.
• Process performance monitoring has not been
  comprehensively studied. Opportunity exists for
  innovation, especially using statistical
  correlation to distinguish good/bad.

39
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
MONITORING AND DIAGNOSIS
To perform monitoring and diagnosis, people need
information.
Sensor Issues (a little review)

• Real-time Decisions
• Trouble shoot incidents using proven method
• Information to support decisions required quickly
• Process Performance Improvement
• Longer term performance indicators based on data
  and calculations
• Usually identifying slow trends

Real-time sensors Fast lab analyses
Sensors for calculations Lab analysis Data for
statistical analysis
See trouble shooting lesson for more on
strategy and examples
40
Key Operability issues 1. Operating window 2.
Flexibility/ controllability 3.
Reliability 4. Safety equipment protection 5.
Efficiency profitability 6. Operation during
transitions 7. Dynamic Performance 8.
Monitoring diagnosis
PROCESS OPERABILITY Achieving a Robust Design

• The job is not done when the design functions for
  (only) a base case operation
• Process structure and equipment must be designed
  to provide good operability
• The engineer must understand process and
  equipment behavior to ensure operability.
• The lowest initial cost design is not necessarily
  the design that gives the best economic return
  over the life of the project.