Partial Compression.
The sensor scale was 0–100 N, corresponding to
the initial stage of the compression curve, for values of measured force lower
than 100 N (Fig. 2). The crosshead speed was set to 10 mm/min. Ten trials
were performed for each sample. The experimental curve was fitted with a straight line and (1
-
r
2
) used as an indicator of jaggedness,
r
2
being the
regression coefficient.
From experimental curves, different values were derived:
F
max
, the force
at peak;
Sl
max
, the maximum slope before peak of the complete compression
curve;
S
, the surface area under curve before peak, homogeneous to an energy
and
Sl
min
, the slope of the partial compression curve.
Acoustic Recordings and Signal Processing
Sound was recorded during the partial compression step with an NT3
Rode microphone (Silverwater, Australia) put against the top of the immobile
part of the Kramer cell and connected to an acquisition card at a 44.1-kHz
sampling rate. The acoustic signal was composed by peaks of variable amplitude,
coded in the 16 bits signed numeric format: values in range from
-
2
15
to 2
15
(as shown for standard cornflakes, Fig. 3A). The simultaneous acquisition
of force and sound was monitored by Labview 5.1 software from the
National Instruments Corporation. A blank trial, without any flakes sample,
is performed to assess the background noise and its frequencies of emission.
This noise is due to the electrical engine of the Instron testing machine, and
more than 90% of its power spectrum is in the range 0–1 kHz. The acoustic
intensity analysis of recorded temporal signals was performed for the last 10 s
of recordings for each sample. The FFT linear power spectra were calculated
with the Matlab 6.1 Signal Processing Toolbox software (Mathworks Inc.,
Natick, MA). The calculated power spectra were normalized with the same
area under curves and studied between 1 and 10 kHz (Fig. 3B). From the
temporal signal, i.e., amplitude–time recordings, among the numerous possibilities
offered by data treatment, the average amplitude of the signal AAS
and the fraction of peaks of higher amplitude (>5000, experimentally defined
signal threshold) named “Rp+” were derived.
In spite of the undoubted interest on this approach, no fractal analysis of
the signals, either acoustic or mechanic, was carried out at this stage, since
we focused on the setup and validation of the experimental design aimed to
distinguish between various brands of commercial cornflakes.