Application of ICAMM to Impact-Echo Testing - On Statistical Pattern Recognition in Independent Component Analysis Mixture Modelling - page 107

same material and rather similar forms would be the only constraint to take into
account.
Elements with dimensions of about 0.01 m. were used in the finite element
models. This size can accurately capture the frequency response up to 40 kHz.
Surface displacement waveforms were taken from the simulation results at 8 nodes
in different locations on the specimen surface. This would be equivalent to the
signals that could be measured by sensors in a real experiment. The signals con-
sisted of 5000 samples obtained with a simulation step size of 1e-5 s. (sampling
frequency of 100 kHz). To compare simulations with experiments, the second
derivative of the displacement was calculated to work with accelerations since the
sensors available for experiments were mono-axial accelerometers. These accel-
erations were measured in the normal direction to the plane of the material surface
where the sensors were hypothetically to be located.
5.2.2 Experimental Signals
To perform impact-echo measurements on the test blocks, the equipment used in
measuring was composed of five main components: an impact hammer, sensors, a
data acquisition module, a signal conditioner, and a notebook (see Fig. 5.1 ).
• Instrumented impacts hammer 084A14 PCB
• Sensor accelerometers PCB model 353B17. Bandwidth: 0.7-20 kHz. Sensitivity
10 mV./g. and a weight of 1.7 gr. They are fixed to the test specimen for
measurements.
• ICP signal conditioner model F482A18. Up to 8 channels with 91, 910 and
9100 independent gains.
• Data acquisition module 6067E with BNC connectors, which is able to digitalize
up to 16 channels with a total maximum sampling frequency of 1.2 M samples
per second. It works in single sampling mode or continuous sampling mode and
connects to a PC through a fire wire bus.
• Notebook. This controls the data acquisition module and signal storing.
The following parameters were used for measuring: (i) 100 kHz of sampling
frequency per channel; (ii) 5,000 samples acquired per channel; (iii) 16 bits of
vertical resolution; (iv) 10 dB of conditioning gain per channel; (v) single
acquisition mode; (vi) trigger level = 1 V.; and (vii) trigger channel = channel 0.
Figure 5.2 shows the first 2,000 samples of some of the signals collected. Note
the half-sine signal of the impact in channel 1 (positions of the sensors are depicted
in Fig. 5.3 ). The waveform of the measured signals depends on several variables,
for instance, impact location, shape of the defect, and the relation between the
sensor and the defect location. In the case of Fig. 5.2 , the material contained a
fairly symmetric defect to the longitudinal axis that yielded similar waveforms in
sensors 4 and 5 that were located on opposite faces of the material, parallel to the