270
D. Mishra et al.
packets extracted from the current signal were considered for predicting the weld
strength. Similarly, the geometrical parameters of the weld, such as height, width
of the bead and reinforcement, have been predicted using features extracted from
the current and voltage signals [44]. The task of classifying welds into defective
and defect-free has also been carried out by using these electrical signatures [45].
Various time-domain features have been noted from the signals for classification.
Other findings by using these electrical parameters include quality of arc at the start
of weld, variation in the distance between the electrode tip and job, problems in
feeding of electrode, problems in joint fit-up, mode of metal transfer [46]. A recent
study demonstrated the usage of these electrical parameters for process monitoring
of arc welding by introducing certain anomalies such as cutting notches in workpiece
for degrading the surface preparation, insufficient flow of shielding gas by reducing
the flow rate and contaminating the job surface by applying grease [47]. Several
time-domain features were found to be correlated with these disturbances. Further
on the utility of the electrical parameters in monitoring of the arc welding technique
and usage of various signal processing techniques and ML models for prediction of
the weld quality, readers can refer to Ref. [41, 48].
The electrical parameters, namely electrode current and tip voltage, have also been
utilized for online monitoring of the RSW technique [49, 50]. These parameters have
been utilized to obtain information about the dynamic resistance and input impedance
in the welding process [51]. While the dynamic resistance refers to the quantity with
respect to change in the current or voltage during the welding, the input impedance
is a measure of the opposition to the current in the circuit. The dynamic resistance
has been found to be sensitive to the phases of the welding, and thus, forms a crucial
quantity for monitoring the process.
Similarly, the electrical parameters have also been proven to be useful for monitoring FSW technique [52]. FSW is usually performed in dedicated machines which
enable the variation of parameters through the use of motors [53]. As the welding
technique involves distinct stages starting from tool rotation to plunging and movement of machine bed to retraction of the tool, the power consumption by the machine
varies with these events. Any sort of anomaly occurring during the welding, hits the
power signature [54]. The power signal has been utilized to identify the defects in
a weldment by analysing it through the time–frequency signal processing technique
[54, 55]. The current and voltage signals of the feed and spindle motors have also been
utilized to predict the joint strength of the weldment by using a ML model, where
the features were extracted from time-domain and time–frequency domain [56, 57].
Other than these, both power and force signals were utilized in FSW for monitoring
the tool quality [1]. The research has been carried out by considering tools of varying
health conditions, and the signals were studied in the time–frequency domain. Later,
a ML model was utilized for classifying those tool conditions in real time.
An advantage of the electrical signatures is their inherent availability in the system,
and non-interference with the manufacturing process; i.e. they do not create any
disturbance. During sampling at large frequencies, very minute fluctuations occurring
in the process can also be captured.
D. Mishra et al.
packets extracted from the current signal were considered for predicting the weld
strength. Similarly, the geometrical parameters of the weld, such as height, width
of the bead and reinforcement, have been predicted using features extracted from
the current and voltage signals [44]. The task of classifying welds into defective
and defect-free has also been carried out by using these electrical signatures [45].
Various time-domain features have been noted from the signals for classification.
Other findings by using these electrical parameters include quality of arc at the start
of weld, variation in the distance between the electrode tip and job, problems in
feeding of electrode, problems in joint fit-up, mode of metal transfer [46]. A recent
study demonstrated the usage of these electrical parameters for process monitoring
of arc welding by introducing certain anomalies such as cutting notches in workpiece
for degrading the surface preparation, insufficient flow of shielding gas by reducing
the flow rate and contaminating the job surface by applying grease [47]. Several
time-domain features were found to be correlated with these disturbances. Further
on the utility of the electrical parameters in monitoring of the arc welding technique
and usage of various signal processing techniques and ML models for prediction of
the weld quality, readers can refer to Ref. [41, 48].
The electrical parameters, namely electrode current and tip voltage, have also been
utilized for online monitoring of the RSW technique [49, 50]. These parameters have
been utilized to obtain information about the dynamic resistance and input impedance
in the welding process [51]. While the dynamic resistance refers to the quantity with
respect to change in the current or voltage during the welding, the input impedance
is a measure of the opposition to the current in the circuit. The dynamic resistance
has been found to be sensitive to the phases of the welding, and thus, forms a crucial
quantity for monitoring the process.
Similarly, the electrical parameters have also been proven to be useful for monitoring FSW technique [52]. FSW is usually performed in dedicated machines which
enable the variation of parameters through the use of motors [53]. As the welding
technique involves distinct stages starting from tool rotation to plunging and movement of machine bed to retraction of the tool, the power consumption by the machine
varies with these events. Any sort of anomaly occurring during the welding, hits the
power signature [54]. The power signal has been utilized to identify the defects in
a weldment by analysing it through the time–frequency signal processing technique
[54, 55]. The current and voltage signals of the feed and spindle motors have also been
utilized to predict the joint strength of the weldment by using a ML model, where
the features were extracted from time-domain and time–frequency domain [56, 57].
Other than these, both power and force signals were utilized in FSW for monitoring
the tool quality [1]. The research has been carried out by considering tools of varying
health conditions, and the signals were studied in the time–frequency domain. Later,
a ML model was utilized for classifying those tool conditions in real time.
An advantage of the electrical signatures is their inherent availability in the system,
and non-interference with the manufacturing process; i.e. they do not create any
disturbance. During sampling at large frequencies, very minute fluctuations occurring
in the process can also be captured.