Process Monitoring

• Data acquistion systems,

• Sampling criteria for continuous process

variables,

• Effects of very low and very high sampling

rates,

• Noise in measurement systems and noise

characteristics.
Data Acquisition System
Fundamentals
 Data Acquisition System
Introduction I
A data acquisition system consists of
many components that are integrated to:

• Sense physical variables (use of

transducers)
• Condition the electrical signal to make it
readable by an A/D board
 Data Acquisition System
Introduction II

• Convert the signal into a digital format

acceptable by a computer
• Process, analyze, store, and display the
acquired data with the help of software
Data Acquisition System
Block Diagram
 Transducers
•SenseDisplacement
physical phenomena
•andLevel
translate it into electric
•signals.
Electric signals
• ON/OFF switch
Examples:

• Temperature
• Pressure
• Light
• Force
 Signal Conditioning
Electrical signals are conditioned so they
can be
used by an analog input board. The
following
features may be available:
– Amplification
– Isolation
– Filtering
– Linearization
 Analog to Digital (A/D)
Converter

• Input signal • Resolution

• Sampling rate • Range
• Throughput • Gain
 A/D Converter: Input Signal
• Analog
– Signal is continuous
– Example: strain gage. Most transducers
produce analog signals

• Digital
– Signal is either ON or OFF
– Example: light switch.
A/D Converter: Sampling Rate
• Determines how often conversions take
place.
• The higher the sampling rate, the better.
Analog
Input

4 Samples/cycle 16 Samples/cycle

8 Samples/cycle
 A/D Converter: Sampling Rate
• Aliasing
– Acquired signal gets distorted if sampling rate
is too small.
 A/D Converter: Throughput
• Effective rate of each individual channel is
inversely proportional to the number of
channels sampled.

• Example:
– 100 KHz maximum.
– 16 channels.
100 KHz/16 = 6.25 KHz per channel.
 A/D Converter: Range
• Minimum and maximum voltage levels that
the A/D converter can quantize

– Ranges are selectable (either hardware or

software) to accurately measure the signal
A/D Converter: Resolution
 Data Acquisition Software
• It can be the most critical factor in
obtaining reliable, high performance
operation.
• Transforms the PC and DAQ hardware
into a complete DAQ, analysis, and
display system.
• Different alternatives:
– Programmable software.
– Data acquisition software packages.
 Programmable Software

• Involves the use of a programming

language, such as:
– C++, Visual C++
– BASIC, Visual Basic + Add-on tools (such as
VisuaLab with VTX)
– Fortran
– C#
• Advantage: flexibility
• Disadvantages: complexity and steep
learning curve
 Data Acquisition Software
• Does not require programming.
• Enables developers to design the custom
instrument best suited to their application.

• Examples: TestPoint, SnapMaster,

LabView, DADISP, DASYLAB, etc.
 Designing a DAS: Factors to
Consider
• Is it a fixed or a mobile application?
• Type of input/output signal: digital or
analog?
• Frequency of input signal ?
• Resolution, range, and gain?
• Continuous operation?
• Compatibility between hardware and
software. Are the drivers available?
• Overall price.
Data Acquisition and Analysis
• Data acquisition
– – equipment
• Data validity
– reproducibility
• Data errors
– – error types and causes
• Data analysis
– – traceability
• Data presentation
– – types
 Data Acquisition (1)
• Appropriate equipment is essential for
effective data acquisition
– pencil and paper ?
– computer based data acquisition system ?
 Data Acquisition (2)
• Can we acquire (record) the required data
– sufficiently quickly ?
– with sufficient accuracy and/or resolution ?
 Data Acquisition (3)
• Example 1
– We need to record the number of cars, buses
and lorries which cross a concrete bridge:
• if the bridge is on a single track road, a single
person could undertake the data acquisition with a
pencil and paper;
• however if the bridge was on an eight-lane
motorway either more people would be required, or
another measurement technique would have to be
used.
 Data Acquisition (4)
• Example 2
– We need to record pressure fluctuations in a
fluid as the fluctuations propagate along a
pipeline:
• we will need a computer based data acquisition
system along with appropriate pressure
transducers installed on the pipeline
• will the computer based system be able to record
and store multiple readings sufficiently fast and
with sufficient accuracy ?
 Data Acquisition (5)
• Example 3
– We need to record the number of bytes of
data being transmitted along a computer
network cable:
• we will need a computer connected to the cable,
equipped with appropriate hardware and software.
 Data Acquisition (6)
• Many research projects fail because
insufficient thought was given to the
requirements of the data acquisition
system
– don’t let yours fail for the same reason !
 Data Validity (1)

• It is important that we can demonstrate the

validity of research data
– Is our data reproducible ?
– If we are using a sampling technique, is the
sample of sufficient size ?
Data Validity (2)
 Data Validity (3)
• The validity of data can be verified
– by repeating experiments and comparing
results
– By comparing experimental data with modeled
data or data produced by others
Data Validity (4)
 Data Errors
• •Error types
– Systematic errors
• AKA calibration and bias errors
– Blunders
• AKA human error
– Random errors
• AKA precision errors
Systematic Errors (1)
Systematic Errors (2)
 Systematic Errors (3)
• Predictable errors
• Can be compensated for in calibration
• Generally easy to deal with
– once identified
Blunders (1)
 Blunders (2)
• Almost always due to human error
• Many are obvious and can be easily
identified
• Some are less obvious and can only be
identified by statistical analysis
– or sometimes not at all
Random Errors (1)
Random + Systematic Errors
 Random Errors (2)
• Most experimental data sets contain
random errors
• Random errors can be analysed using
statistical methods
 Analysis of Random Errors
(1)
• In order to analyse the nature of random
errors, it is useful to take a given reading a
number of times
• Lets consider applying a 50kg weight to a
set of scales 120 times and plot the
readings we obtain
Analysis of Random Errors
(2)
Analysis of Random Errors
(2)
Analysis of Random Errors
(2)
Analysis of Random Errors
(2)
Analysis of Random Errors
(2)
Analysis of Random Errors
(3)
 Analysis of Random Errors
(4)
• This is known as a Normal or Gaussian
distribution
• It can be characterised by the mean value
and the standard deviation
• Analysis of Random Errors (5)
– The mean value
Analysis of Random Errors
(6)
 Analysis of Random Errors
(7)
• For our data set
– Mean = 50kg
– Standard deviation = 0.28kg
 Analysis of Random Errors
(8)
• For a gaussian or normal distribution
– approximately 68% of data points lie within Sx
of the mean
– approximately 95% of data points lie within
2Sx of the mean
– approximately 99.7% of data points lie within
3Sx of the mean
Analysis of Random Errors
(9)
Analysis of Random Errors (10)
Analysis of Random Errors (11)
 Analysis of Random Errors (12)

• This type of data analysis allows us to

quantify random errors
– and often to spot blunders or instruments
which need re-calibration
 Data Analysis
• We must be certain that any data analysis
methods which we apply do not introduce
data errors
– We must therefore “trace through” some
samples of our data to ensure the integrity of
our data analysis
 Data Presentation
• Quantitative data are usually presented in:
– Tabular form
– Graphically
– Both