8 Industry 4.0 in Welding
289
Fig. 8.15 Signal decomposition by using DWT
testing machine, a direct monitoring technique. Figure 8.14 shows a surface plot of the
obtained tensile strength values with respect to the selected parametric combinations.
While the base AA6061 had a tensile strength of 270 MPa, a wide range of strength
values, i.e. from 146 MPa to 249.5 MPa, were recorded. The parametric combinations
yielding the best and worst tensile strength values can be referred from the colour
map. The tensile strength can be observed to be lower with high and low values
of welding speed and rotational speed, respectively. With these values, sufficient
amount of heat is not available to plastically deform the workpieces. This in turn
leads to formation of poor welds because of improper flow of the material, and also
welding defects such as voids are found in the weld. Thus, it is always essential
to identify the optimum combination of parameters for welding. For this case, the
weld fabricated with 1000 rpm and 200 mm/min have the highest tensile strength
value, i.e. 249.5 MPa. This tensile strength map, obtained with the wider range of
parametric combinations, was utilized to come up with an efficient predictive model
having enough knowledge of the parametric combinations.
Data of force variation, rotational speed and welding speed, for all the parametric
combinations were also recorded during the welding. An initial database was created
consisting of these parameters, force and the obtained tensile strength values.
8.4.4 Signal Processing, Feature Extraction and Building
of a Database
The acquired force signature was studied in the time–frequency domain by applying
DWT. Wavelets can be defined as a wave which exists for a limited duration of time,
have an irregular shape and a zero mean value. Usually, in order to transform a signal
from one domain to another domain, the signal is multiplied with the basis vectors
to determine the similarity index in the transform domain. As the sinusoids are the
basis vectors in Fourier transform, mother wavelets are the basis vectors in case of
wavelet transform [118].
Figure 8.15 shows the process of signal decomposition in DWT, in which the
original signal is being divided into various frequency bands. These frequency bands
289
Fig. 8.15 Signal decomposition by using DWT
testing machine, a direct monitoring technique. Figure 8.14 shows a surface plot of the
obtained tensile strength values with respect to the selected parametric combinations.
While the base AA6061 had a tensile strength of 270 MPa, a wide range of strength
values, i.e. from 146 MPa to 249.5 MPa, were recorded. The parametric combinations
yielding the best and worst tensile strength values can be referred from the colour
map. The tensile strength can be observed to be lower with high and low values
of welding speed and rotational speed, respectively. With these values, sufficient
amount of heat is not available to plastically deform the workpieces. This in turn
leads to formation of poor welds because of improper flow of the material, and also
welding defects such as voids are found in the weld. Thus, it is always essential
to identify the optimum combination of parameters for welding. For this case, the
weld fabricated with 1000 rpm and 200 mm/min have the highest tensile strength
value, i.e. 249.5 MPa. This tensile strength map, obtained with the wider range of
parametric combinations, was utilized to come up with an efficient predictive model
having enough knowledge of the parametric combinations.
Data of force variation, rotational speed and welding speed, for all the parametric
combinations were also recorded during the welding. An initial database was created
consisting of these parameters, force and the obtained tensile strength values.
8.4.4 Signal Processing, Feature Extraction and Building
of a Database
The acquired force signature was studied in the time–frequency domain by applying
DWT. Wavelets can be defined as a wave which exists for a limited duration of time,
have an irregular shape and a zero mean value. Usually, in order to transform a signal
from one domain to another domain, the signal is multiplied with the basis vectors
to determine the similarity index in the transform domain. As the sinusoids are the
basis vectors in Fourier transform, mother wavelets are the basis vectors in case of
wavelet transform [118].
Figure 8.15 shows the process of signal decomposition in DWT, in which the
original signal is being divided into various frequency bands. These frequency bands