lesson 2 digital data acquisition and data processing
TRANSCRIPT
Digital Data
Acquisition and Processing for
Nondestructive Evaluation
Prof. Krishnan Balasubramaniam
Professor of Mechanical Engineering and
Head of Centre for Nondestructive Evaluation
Department of Mechanical Engineering,
IIT Chennai 600 036
Lesson 2: Advanced Nondestructive Evaluation
Why Data Acquistion ?
• Complex specifications leads to difficulty in decision making.
• Field environment requires un-biased, quick decisions.
• Seeing is sometimes deceiving.
• Quantitative information is essential.
• Degree of automation is process and product dependent.
Why Data Analysis ?
• Reduce Data Volume
• Noise Reduction
• Feature Extraction for Discrimination
Analysis
• Multi-facet approach to data interpretation
(time, frequency,…domains)
Nondestructive Evaluation
Excitation
Source
Test
Specimen
Signal
Conditioning
Inverse
Model
Input Transducer
Output Transducer
Systems Based Approaches
Signal
Conditioning
Output
Transducer
Response
Characterization
Results
Signal Conditioning
Sampling and Quantization
Signal Enhancement
Signal Restoration
Inverse
Model
Signal Classification
• Analog Digital
-4
-3
-2
-1
0
1
2
3
4
1 2 3 4 5 6 7 8 9
-4
-3
-2
-1
0
1
2
3
4
1 2 3 4 5 6 7 8 9
Introduction To Linear Systems
Signals
x(t)
t Continuous-Time Signal
Discrete-Time Signal
x[n]
x(t)
Input
y(t)
Output x[n]
Input
y[n]
Output
1
2
1
3
1.5
n 0 1 2 3 4
Continuous-Time System Discrete-Time System
If x1[n] y1[n]
x2[n] y2[n]
Linearity ax1[n] + bx2[n] ay1[n] + by2[n]
Similarly, ax1(t) + bx2(t) ay1(t) + by2(t)
Properties of Systems
Linearity
x[n]
x(t)
y[n]
y(t)
Types of Data
• Transient 1 dimensional Signals
• Spatially Distributed-Time Signals
• Static Images (photos)
• Dynamic Images (movies)
Signal Domains
• Time/Spatial Domain
• Frequency Domain (Amplitude)
• Frequency Domain (Phase)
• Time (Space)-Frequency Domain (Wavelets,
STFT)
• Transfer Function Domains
Digital Signal Analysis
• Acquire digital data using DAQ.
• Process Data for Noise Reduction.
• Extract Features
• Multi-parameter Discrimination.
• Damage Estimation.
• Damage Reporting
DAQ System
Analog Digital
DAQ Hardware
• Analog to Digital Convertors
– Ni 5102/5112 – up to 2.5 GHz. Data Aq Boards
– DAQ Boards – Slow (1-1 MHz)
– DSO – Agilent, Tektronics
• Frame Grabbers
Image Grabbers
Triggering
• Allows you to efficiently capture short-
duration and high-speed events by
eliminating the need to continuously acquire
data while waiting for the event to occur
• Usually done using a voltage signal input.
• The Trigger level and +/- values are key.
• Both analog and digital triggers are
possible.
Speed of Acquisition
• Images of a moving toy motorcycle taken
with a progressive scan camera using
exposure times of 33 ms, 10 ms, and 1 ms,
respectively
Nyquist Criterion SR at least = 2*fmax
Digitization Basics
• Horizontal Resolution (Sampling 100 kHz.)
• Vertical Resolution (Bits =16 bits)
• Amplitude Range (+10V to –10V)
• Gain (10 dB)
• Multiplexing
• Effective bits
Rv = 20/216 = 3.05* 10-4V
Rh = 1/100*103=10-5s
Delays and Offsets and …..
• DC Bias Offset
• Trigger Delay
• Pre-trigger
• Equivalent Time Sampling for stationary
signals
Interfaces
• RS-232
• RS-170
• IEEE 488.2 GPIB
• IEEE 1394
• PCI bus Analog
Cameras
Parallel Digital
Cameras
Camera Link
Cameras
IEEE-1394
Cameras
Data Rate Slow Fast Fast Slow
Spatial
Resolution Low High High Medium
Functionality Simple and
easy Advanced Advanced
Simple and
easy
Pixel Depth 8-bit to 10-bit Up to 16-bit Up to 16-bit Typically 8-bit
Cabling Simple BNC
cabling
Thicker, custom
cabling
Simple,
standard
cabling
Simple,
standard
cabling
USB 2.0
RJ 45
PCIe
Frequency Bandwidth
Time Frequency • Fourier Transform
F(t) = aicos(wi)
• Fast Fourier Transform (FFT)
– Complex Operation
• Discrete Fast Fourier Transform (DFFT)
Time-Freq. Paradox
• A classic trade-off.
• High resolution in Time – Poor resolution in
Freq. – vice versa.
• New time-freq. Analysis such as wavelets
allow a more balanced approach –
SIMULTANEOUSLY.
EXAMPLE
Data Reduction
• Fourier Domain
• JPEG
• MPEG
• CODEC
• WT
• …….
EXAMPLE
NOISE Reduction
• SNR improvement
• Filtering
• Time Averaging
• Spatial Averaging
• Split Spectrum Processing
• Wide Spectrum Processing
EXAMPLE
Median Filtering
RAW DATA MEDIAN FILTERED
Filte
Feature Extraction
• Time Domain Features
– Peak to Peak Amplitude, +Peak/-Peak, Pulse Duration, ………..
• Envelope Domain Features
– Energy, Skewness, Kurtosis ,…….
• Frequency Domain Features
– Peak Frequency, Mag of Peak Freq., Slope of Phase Spectrum…. EXAMPLE
Multiple Domains The figure below shows the RF signal and the Amplitude Spectrum. Controls such as
Zeropadding, Undersampling, phase wrap/unwrap, window type, are provided. The
signal can be windowed by dragging the ends of the red bar over the RF signal.
Multiple Domains The figure below shows the Envelope of the RF signal and the Phase Spectrum. It is possible to
analyze several signal simultaneously by using data name patterns using wildcard (*).
Feature Extraction The extracted features are imaged here using a intensity plot. Each
sample is displayed separately and the selector can be used to
dynamically change the feature type. The histogram is also provided for
the 5 samples. The color bar and the # of bins can also be dynamically
changed. Mean and Standard deviation are provided for each sample.
Discrimination Analysis It can be observed by looking at the two feature scatter plot, that the samples 1 and 2
are clustered near the origin (0,0) location while the Samples 3 and 4 are at relatively
large distance from the origin.
Y-Feature
X-Feature
Some Applications
• Find the hidden echo and reduce dead zone effect. `
• Find thickness of thin coatings/structures in frequency domain
• Find Grain Characteristics from noise data analysis.
• Discrimination between defect types.
• Computer Aided Tomography
EXAMPLE
EXAMPLE
EXAMPLE
Defect Discrimination Extraction of Signal
Shape will allow for
discrimination between
defect types.
Digital Industrial Laminography Ewert, Redmer, BAM-VIII.3
Camera trip through the object:
3D- reconstruction
of the crack range after
numeric filtering.
Reconstruction from the
line scan measurements
at 180 kV
Mechanized Weld Inspection
by Tomosynthesis
from 5 Projections
3D Representation
Summary • Data Acquisition is the single most important step
in an NDE procedure.
• Good Data Acquisition, Data Archiving, Digital Signal Processing, Discrimination Algorithms are all now feasible with low cost hardware and software.
• Good DAQ is a combination of hardware, software, and settings, and the data is only as good as the weakest link in this chain.
• Signal Analysis solves problems, previously declared un-solvable.
• Signal Analysis will not fully compensate for poor data acquisition since new information cannot be created.