*2.3.5 Spattering index*

According to Moinuddin and Sharma [24], using the cyclograms it is possible to represent characteristics of droplet detachment and arc burning stage. The authors also carried out an analysis of probability density distribution of arc voltage, weld bead, and microstructure analysis for various welding conditions, allowing to extend the stability study to spray transfer mode. The study showed that there is a strong correlation between the microstructure and the stability of the arc. Besides, the different types of electrodes and their electrical conductivity capacity also has influence on the resulting microstructure in a welded bead. A stable arc produces greater penetration and improves melting efficiency. The authors mention that the study can be expanded taking into account other parameters such as electrode type, electrode extension, shield protection gas, welding speed, and other current modes

Cayo [25] uses the cyclograms to detect defects in the weld reflected in the arc and current–voltage signals. The cyclograms allowed to identify three types of disturbances, a variation of the stand of, presence of grease and absence of protection gas. Each type of defect showed changes in the cyclograms, allowing to analyze the changes in voltage and current. One of the advantages of the cyclograms is that it provides a visual result that allows a quick analysis of the values obtained in the process. Again a powerful stability indicator is shown, but it has been oriented only

Suban [26] uses this index to determine a more stable short-circuit material transfer. As a result, open arc, short-circuit, and spray transfer moments are identified depending on the type of gas used. In addition, the author performs an analysis of the probability distribution of voltage and current using Fourier analysis. Among the conclusions, the authors emphasize that with pure CO2, more stability is

The control of droplet size ensures transfer stability. For measuring this variable, image processing, laser shadowing, and sound processing techniques are generally used. The appropriate control ensures proper transfer mode; increases the quality of welding, and decreases the number of defects. Large drops do not represent a

The transfer of the drop is dependent on welding current and arc voltage waveforms influenced by gravity force, electromagnetic force, plasma drag force, and surface tension. Suban [26] ensures that to maximize stability, the time between the

Mousavi and Kulkarni [27, 28] demonstrate that a relationship between droplet detachment and statistical parameters of current exists, assuring that lesser standard deviation and coefficient to variation was considered to be of uniform droplet

Then it can be concluded that there is a correlation between the waveform of the

current and the detachment of the drop. A lower coefficient of variation in the mean of the welding current represents uniformity in the detachment frequency. Additionally, for variable transfer time, the welding arc tends to be unstable and the current signals exhibit irregular behavior. In the case of short-circuit transfer mode,

Soderstrom and Mendez [29] use high-speed laser shadowgraphs and fast Fourier transform (FFT) of the voltage signal for droplet diameter and detachment frequency measurement. It has been found that a relationship between average droplet diameter and current for the different diameter electrodes exists. In addition, it states that the increase in CO2 above normal standards causes an erratic

achieved. This method is simple and can be implemented in real time.

transfers of two subsequent drops should always be the same.

detachment and arc length uniformity.

such as pulsed.

*Welding - Modern Topics*

*2.3.4 Control of droplet size*

suitable condition.

detachment.

**14**

to the analysis of the short-circuit transfer mode.

The amount of spatters generated during the welding process has been another indicator widely used; the spatters are a product of instability in the arc and should be minimized. The largest amount of study is developed in the short-circuit area. The moment when the short circuit occurs and the arc is reset is when the largest number of spatters is produced. Also, if the mean of the short-circuit time is irregular, more spatters will be generated.

Silva et al. [31] propose a criterion for the spattering index correlating spattering rate (S—Eq. 1, **Table 2**) and the deposition rate (D—Eq. 2, **Table 1**). The purpose was to demonstrate that the correct control of these indicators allows to choose appropriate parameters for any specific welding application.

On the other hand, Kang and Rhee [32] develop statistical regression models to predict the amount of spatter in the short-circuit transfer for GMAW. It is shown, in the same way, that voltage and welding current waveforms can be satisfactorily used to predict the presence of spatters. Kang et al. [33] in a similar work use four different linear and nonlinear regression models composed of the waveform factors to develop the spatter prediction model. Proving that the amount of spatter depends on the number of arc extinctions, arc extinctions occur when the welding voltage is


**Table 2.** *Summary of transfer stability indexes.*

below the optimum. In another study, models were developed for evaluating the spatter rate based on the conventional feed-forward multilayer perceptrons with the error back-propagation as the learning algorithm to estimated spatter rate.

transfer mode, the drops are small and practically imperceptible during the acoustic

Roca et al. [39] also applied acoustic monitoring, and the results obtained were used for the training of a neural network. To perform the analysis, they obtain the standard deviations of the peak amplitudes of the sound at the moment in which the short circuit is made, and they use as stability indicator. In Eq. 12, **Table 1** shows the stability index previously established. The combination of statistical technique, acoustic monitoring, and artificial intelligence allowed to use online monitoring,

It can be summarized that the electrical and acoustic signals are correlated mainly in the short-circuit transfer mode where it is possible to identify the detachment of the drop and the arc reignition. In addition, it is possible through sound monitoring to identify the transference modes. It is a method that is not expensive and that is feasible to implement in the industry. The combination of this method with machine learning techniques that allow prediction and classification is open for

To synthesize the study, an analysis of the documentation was obtained, the metadata of the document collection was exported in Information Systems Research (RIS) format, and a bibliometric analysis was performed using the VOSViewer software. A graph with groups of the main authors and their relationship of coauthorship (taking five as a frequency of occurrence of the author's surname) is presented in **Figure 11**. It is possible to identify as the largest cluster the Chinese authors, followed by smaller groups of Brazilian and Indian authors, highlighting

**Figure 12** shows the most used terms in the area that can be defined as

*Authors and their relationship of co-authorship (two as a frequency of occurrence of the author's surname).*

analysis. Already in the case of the short circuit, it is possible to monitor the

occurrence of each short and the reignition of the arc.

*Stability on the GMAW Process*

*DOI: http://dx.doi.org/10.5772/intechopen.90386*

considering it an efficient and non-destructive technique.

**3. Synthesis of the study and future research directions**

that there is little cooperation between those groups.

future works.

keywords.

**Figure 11.**

**17**

Lastly, Fernandes et al. [34] propose a spatter index (Eq. 7, **Table 2**) relating in a mathematical equation of the weight of the spatter collected in the box and weight of the weld bead. Using the calculated value of the spatter index, they propose a new index of stability (Eq. 8, **Table 1**) that enhances the electrical stability of the process and the weight of spatter generated during welding. The proposed method is efficient as soon as the collection of spatters is carried out correctly. It is suitable for a laboratory environment but can hardly be implemented in the industry since it depends on the collection device. However, the results obtained can be generalized in an automatic learning model and implemented for the control of spatters.

#### *2.3.6 Acoustic monitoring*

According to Grad et al. [35], the acoustic signal contains information about the transfer mode and the behavior of the arc. It is also possible to identify changes in arc dimensions and geometry; changes in arc intensity; and metal transfer and oscillations of the molten pool.

Even according to Mota et al. [36], it is possible to observe that the sound signal accompanies the electrical signal, specifically the voltage, in relation to the moments of extinction and ignition of the arc. It is easy to see in **Figure 10** the sound pulses from the moments of the abrupt change in the voltage of the electric arc, and the time intervals between them follow the same pattern observed in the electric signal.

Grum et al. [37] use the sound signal and the light signal to detect even the smallest deviations of arc behavior, as well as large deviations due to the material transfer mode and excessive/inadequate weld penetration. They propose a mathematical model using sound and light values. The authors demonstrated the existence of a correlation between light signals and the energy provided to the system. With the monitoring of sound, it was possible to identify oscillations in the arc that indicated instability. The model was developed for the short-circuit transfer mode but was generalized for the spray transfer mode.

Cayo and Alfaro [38] use the sound to define the difference between the transfer modes on the GMAW process. They use sound pressure and current signals to identify changes in the transfer mode and identify defects. In the case of the spray

**Figure 10.** *Comparison between sound and current signals (modified from [36]).*

#### *Stability on the GMAW Process DOI: http://dx.doi.org/10.5772/intechopen.90386*

below the optimum. In another study, models were developed for evaluating the spatter rate based on the conventional feed-forward multilayer perceptrons with the error back-propagation as the learning algorithm to estimated spatter rate.

*2.3.6 Acoustic monitoring*

*Welding - Modern Topics*

electric signal.

**Figure 10.**

**16**

oscillations of the molten pool.

but was generalized for the spray transfer mode.

*Comparison between sound and current signals (modified from [36]).*

Lastly, Fernandes et al. [34] propose a spatter index (Eq. 7, **Table 2**) relating in a mathematical equation of the weight of the spatter collected in the box and weight of the weld bead. Using the calculated value of the spatter index, they propose a new index of stability (Eq. 8, **Table 1**) that enhances the electrical stability of the process and the weight of spatter generated during welding. The proposed method is efficient as soon as the collection of spatters is carried out correctly. It is suitable for a laboratory environment but can hardly be implemented in the industry since it depends on the collection device. However, the results obtained can be generalized in an automatic learning model and implemented for the control of spatters.

According to Grad et al. [35], the acoustic signal contains information about the transfer mode and the behavior of the arc. It is also possible to identify changes in arc dimensions and geometry; changes in arc intensity; and metal transfer and

Even according to Mota et al. [36], it is possible to observe that the sound signal

Grum et al. [37] use the sound signal and the light signal to detect even the smallest deviations of arc behavior, as well as large deviations due to the material transfer mode and excessive/inadequate weld penetration. They propose a mathematical model using sound and light values. The authors demonstrated the existence of a correlation between light signals and the energy provided to the system. With the monitoring of sound, it was possible to identify oscillations in the arc that indicated instability. The model was developed for the short-circuit transfer mode

Cayo and Alfaro [38] use the sound to define the difference between the transfer

modes on the GMAW process. They use sound pressure and current signals to identify changes in the transfer mode and identify defects. In the case of the spray

accompanies the electrical signal, specifically the voltage, in relation to the moments of extinction and ignition of the arc. It is easy to see in **Figure 10** the sound pulses from the moments of the abrupt change in the voltage of the electric arc, and the time intervals between them follow the same pattern observed in the

transfer mode, the drops are small and practically imperceptible during the acoustic analysis. Already in the case of the short circuit, it is possible to monitor the occurrence of each short and the reignition of the arc.

Roca et al. [39] also applied acoustic monitoring, and the results obtained were used for the training of a neural network. To perform the analysis, they obtain the standard deviations of the peak amplitudes of the sound at the moment in which the short circuit is made, and they use as stability indicator. In Eq. 12, **Table 1** shows the stability index previously established. The combination of statistical technique, acoustic monitoring, and artificial intelligence allowed to use online monitoring, considering it an efficient and non-destructive technique.

It can be summarized that the electrical and acoustic signals are correlated mainly in the short-circuit transfer mode where it is possible to identify the detachment of the drop and the arc reignition. In addition, it is possible through sound monitoring to identify the transference modes. It is a method that is not expensive and that is feasible to implement in the industry. The combination of this method with machine learning techniques that allow prediction and classification is open for future works.
