274
D. Mishra et al.
DWT and “Continuous Wavelet Transform” (CWT) for classification of welds [87,
88]. Feature extracted from the wavelet coefficients of the signals were the “mean
of square of errors” and “variance". The frequency spectrum of the force signal
has also been studied for classification of the welds [89]. A variation in the amplitude spectrums of the defective and defect-free welds was found. Further to assist
in automating the FSW technique, the quality of the tool has been monitored in real
time by using force and power signals [1]. The study considered tools with certain
induced defects. The wavelet coefficients extracted from the said two signals were
utilized to train a ML model in order to classify the tool conditions in real time.
In addition to monitoring, these signals have also been utilized to control the
FSW technique. These studies have designed proportional integral derivative (PID)
controllers [90, 91]. After acquiring the actual force signal, the process was controlled
by calculating the error signal from the reference signal. The obtained result was
used to vary a suitable input variable. A recent study has proposed the usage of
multiple sensors, i.e. force torque and power, for monitoring and controlling the
welding process [54]. The signals collected have been analysed in a cloud server
which included: (a) predicting the ultimate tensile strength (UTS) of the weld joint
in real time by using a ML model, (b) comparing the obtained UTS with a reference
UTS value and (c) predicting new input parameters upon finding deviation between
the predicted and fixed UTS values. The study has also shown the usage of multiple
sensors resulting in higher accuracy over a single sensor.
8.3.3 Context Setting—A Summary
The previous sections have stated the use of direct and indirect methods for monitoring and control of the welding techniques. The direct monitoring methods produce
accurate results, but work offline. Thus, they do not help in improving the process
quality in real time. The indirect means of monitoring are more practical and can
help in fulfilling the real-time needs. It can also be noted that a single sensor cannot
address all the issues in a welding technique. However, among all the sensors that
have been discussed, current/power is one such physical quantity which is common
to all welding techniques, and is available inherently. For instance, this may refer to
the measurement of arc current in arc welding, the power (intensity) of laser beam
in laser welding, electrode current in spot welding or the current consumption by
the motors in case of FSW. Thus, the variation in the electrical parameters may be
considered as a reliable source for monitoring of the welding techniques.
As automation is a core part of Industry 4.0, the use of sensors in manufacturing
becomes essential. The following paragraph discusses the principles of Industry 4.0,
and their application in welding. Later, the digital tools required for implementation
of these principles have been elaborated.
D. Mishra et al.
DWT and “Continuous Wavelet Transform” (CWT) for classification of welds [87,
88]. Feature extracted from the wavelet coefficients of the signals were the “mean
of square of errors” and “variance". The frequency spectrum of the force signal
has also been studied for classification of the welds [89]. A variation in the amplitude spectrums of the defective and defect-free welds was found. Further to assist
in automating the FSW technique, the quality of the tool has been monitored in real
time by using force and power signals [1]. The study considered tools with certain
induced defects. The wavelet coefficients extracted from the said two signals were
utilized to train a ML model in order to classify the tool conditions in real time.
In addition to monitoring, these signals have also been utilized to control the
FSW technique. These studies have designed proportional integral derivative (PID)
controllers [90, 91]. After acquiring the actual force signal, the process was controlled
by calculating the error signal from the reference signal. The obtained result was
used to vary a suitable input variable. A recent study has proposed the usage of
multiple sensors, i.e. force torque and power, for monitoring and controlling the
welding process [54]. The signals collected have been analysed in a cloud server
which included: (a) predicting the ultimate tensile strength (UTS) of the weld joint
in real time by using a ML model, (b) comparing the obtained UTS with a reference
UTS value and (c) predicting new input parameters upon finding deviation between
the predicted and fixed UTS values. The study has also shown the usage of multiple
sensors resulting in higher accuracy over a single sensor.
8.3.3 Context Setting—A Summary
The previous sections have stated the use of direct and indirect methods for monitoring and control of the welding techniques. The direct monitoring methods produce
accurate results, but work offline. Thus, they do not help in improving the process
quality in real time. The indirect means of monitoring are more practical and can
help in fulfilling the real-time needs. It can also be noted that a single sensor cannot
address all the issues in a welding technique. However, among all the sensors that
have been discussed, current/power is one such physical quantity which is common
to all welding techniques, and is available inherently. For instance, this may refer to
the measurement of arc current in arc welding, the power (intensity) of laser beam
in laser welding, electrode current in spot welding or the current consumption by
the motors in case of FSW. Thus, the variation in the electrical parameters may be
considered as a reliable source for monitoring of the welding techniques.
As automation is a core part of Industry 4.0, the use of sensors in manufacturing
becomes essential. The following paragraph discusses the principles of Industry 4.0,
and their application in welding. Later, the digital tools required for implementation
of these principles have been elaborated.