The light emissions of the plasma plume in the deep penetration laser welding of metals typically have the character of irregular short-time pulses. Their nature indicates that the size of the plume significantly fluctuates and the plasma flows out of the keyhole in the form of short bursts rather than a continuous flow. In this paper, the authors study the plasma plume light emissions using an autocorrelation analysis. The authors show that it is an efficient tool for the detection of the plasma bursts period which is typically in the order of milliseconds. The authors compare the autocorrelation characteristics and the geometry of the welds made on a 2 kW ytterbium-doped yttrium aluminium garnet (Yb:YAG) fiber laser welding machine for the X5CrNi 18-10 stainless steel and the S235JR carbon steel. The welding parameters settings is varied over a range of laser power (1–2 kW) and welding speed (10–30 mm/s) usually used in industry. As a result, the authors identify a linear dependence between the plasma bursts period and the weld depth.

The deep penetration laser welding of metals is accompanied by the formation of the keyhole filled with the laser induced plasma and the bright plasma plume above the keyhole opening.1 The light emissions generated by the plasma plume have pulsing character that indicates continuous fluctuations of its brightness and size.2 

In recent years, the light and acoustic emissions accompanying high power CO2 and fiber laser welding of steel have been investigated by various methods with regard to possible applications in welding process diagnostics. The frequency analysis of the light emissions has shown that the power spectrum has a noise character and the ratio of the mean power in well-chosen frequency bands indicates the transition between the deep penetration and conduction welding mode.3 The wavelet analysis of the acoustic emissions also has allowed for the detection of the welding mode and the welding defects.4 The tools of the chaos theory have revealed that the time-varying intensity of the acoustic and light emissions correspond to the deterministic chaos.5 The nature of the light emissions has also been studied in connection with the dynamics of the keyhole, the weld pool and the plasma plume.6,7 The method of high-speed imaging has visualized the fluctuations of the plasma plume and has revealed that it is repeatedly generated from a keyhole in the form of short bursts with the period in the order of milliseconds.8 The spectroscopy analysis has revealed the difference in properties of plasma induced by CO2 and fiber lasers, which has indicated that the laser absorption in plasma and the keyhole oscillations have different degree of influence in the plasma bursts generated by different laser sources.9,10

In our previous papers,11,12 we have studied the light emissions using the short-time frequency analysis. We have detected the presence of the plasma bursts through a significant peak in the frequency spectrum that corresponds to the frequency of the short-time pulses observed in the waveforms of the light emissions. However, the complex time series of the plasma bursts limit the use of the Fourier transform suitable rather for the periodic signals.

In this paper, we study the plasma plume light emissions using an autocorrelation analysis that is more feasible for the recognition of the repetitive patterns (such as the pulses corresponding to the plasma bursts) in the noise-obscured signal. We show that it is an efficient tool for the detection of the plasma bursts period that is typically in the order of milliseconds. We compare the autocorrelation characteristics and the geometry of the welds made on a 2 kW Yb:YAG fiber laser welding machine for the X5CrNi 18-10 stainless steel and the S235JR carbon steel. The welding parameters settings is varied over a range of laser power (1–2 kW) and welding speed (10–30 mm/s) usually used in industry. As a result, we identify a direct relationship between the plasma bursts period and the weld depth.

The experimental observations of the plasma bursts during the laser welding of the steel or aluminum alloys have been described by other authors8,9 as well as in our previous papers.11–13 In Fig. 1, we show the typical time series of the plasma plume images acquired by a high-speed camera at the frame rate of 2000 fps during the bead-on-plate welding of the S235JR carbon steel with the laser power 1.0 and 2.0 kW. Although the frame rate is relatively low, it is evident that both the size of the plasma plume and the intensity of the emitted light significantly fluctuate due to the temporal evolution of the plasma bursts. The maximum size of the plasma plume is about 10 mm for the laser power 1.0 kW and reaches about 15 mm for 2.0 kW. The higher laser power also causes the liquid metal spatters that correspond to the small bright spots observable in the camera images. In Fig. 2, we show the typical waveforms of the plasma plume light emissions that reveal the short-time pulses corresponding to the plasma bursts observed during the bead-on plate welding of the X5CrNi 18-10 stainless steel with the laser power 1.0 and 2.0 kW. The recognizability of the individual pulses is better for the higher laser power and the period of the plasma bursts can be roughly estimated at about 2 ms for 1.0 kW and 5 ms for 2.0 kW. These observations indicate the dependence between the plasma burst period and the welding parameters settings. To further study this topic, we describe the physical model used in our considerations and the properties of the autocorrelation function employed for the signal analysis.

FIG. 1.

The images of the plasma plume acquired during the deep penetration laser welding of the S235JR carbon steel. Welding conditions: plate thickness 6 mm, argon shielding gas 18 l/min, laser power 1.0 and 2.0 kW, and welding speed 30 mm/s. The relative time in milliseconds is given in the upper right corner.

FIG. 1.

The images of the plasma plume acquired during the deep penetration laser welding of the S235JR carbon steel. Welding conditions: plate thickness 6 mm, argon shielding gas 18 l/min, laser power 1.0 and 2.0 kW, and welding speed 30 mm/s. The relative time in milliseconds is given in the upper right corner.

Close modal
FIG. 2.

Plasma bursts observed as the short-time pulses of the light emissions. Welding conditions: X5CrNi 18-10 stainless steel, plate thickness 6 mm, argon 18 l/min, laser power 1.0 and 2.0 kW, and welding speed 10 mm/s.

FIG. 2.

Plasma bursts observed as the short-time pulses of the light emissions. Welding conditions: X5CrNi 18-10 stainless steel, plate thickness 6 mm, argon 18 l/min, laser power 1.0 and 2.0 kW, and welding speed 10 mm/s.

Close modal

The theoretical model of the plasma bursts is based on a repetitive process that takes place in the plasma-filled keyhole surrounded by the weld pool and comprises two principal physical phenomena: the accumulation of laser induced plasma and the keyhole/weld pool oscillations.

1. Laser induced plasma

The laser induced plasma which fills the keyhole is formed during the interaction of the high power laser beam with the welded metal and its vapors. The keyhole walls are heated by the laser beam due to the Fresnel absorption and its multiple reflections. The evaporation of the molten metal, caused by the laser beam, generates the stream of vapor particles that passes through a thin Knudsen layer where the collisions among the particles cause the transformation of an anisotropic velocity distribution into the isotropic distribution. Beyond the Knudsen layer, i.e., in the vast majority of the keyhole volume, part of the metal vapors is ionized and the properties of such a laser induced plasma are described to a good approximation by the theory of plasma in local thermodynamic equilibrium (LTE).14 The necessary condition for the LTE is that the collisional processes in plasma dominate over the radiative ones. The assumption of the LTE means that the dynamic properties of the plasma particles, such as electron and ion velocities, population partition among the excited atomic states and ionization densities, follow Maxwell–Boltzmann distributions. The ionization degree of the plasma in the LTE can be calculated by Saha equation and the electron temperature can be determined from relative intensities of two spectral lines of the same element under same ionization stage. The laser absorption in plasma is dominantly driven by the process of inverse bremsstrahlung in which an electron absorbs a photon while colliding with an ion or with another electron. The absorption coefficient α is given by the formula

α=z2e6nenilnΛ3ω2cε03(2πmekBTe)3/21(ωp/ω)2,
(1)

where ne is the electron number density, ni is the ion number density, z is the charge number, e is the electron charge, c is the speed of light, ε0 is the dielectric constant, me is the electron mass, kB is the Boltzmann constant, Te is the electron temperature, ω is the angular frequency of the incident wave, ωp is the angular frequency of the plasma oscillations, and lnΛ is the Coulomb logarithm. Using the relation ω=2πc/λ, Eq. (1) shows that the absorption coefficient α is proportional to the square of the laser wavelength λ.

The dependence between α and λ implies that the plasma absorption significantly differs for various laser sources. Considering the same plasma, the α is approximately 100 times higher for the CO2 laser (λ = 10.6 μm) than for the fiber laser (λ = 1.07 μm) due to the ten times longer wavelength. The difference implies that the plasma absorption plays more important role in the explanation of the plasma bursts induced by the CO2 laser than by the fiber laser. In fact, it is well known that the strong plasma absorption during the high power CO2 laser welding of steel can totally shield the incident wave. This results in significant reduction of the total power coupled to the workpiece which may cause periodic generation of the plasma plume.10 In contrast, Kawahito's study8 on fiber laser welding of type 304 stainless steel (20 mm thick plate) has shown that the plume consists of weakly ionized plasma in the whole range of the laser power 1–10 kW where the maximum attenuation reaches only 4%. However, it is necessary to keep in mind that this result was obtained for the plasma outside the keyhole where the conditions are different from its interior. Since the pressure and the plasma density are higher inside the keyhole, the higher absorption is also expected. Although this issue still awaits its experimental verification and proper evaluation through the optical emission spectroscopy,15 we consider the plasma absorption as a partial contribution to the accumulation of plasma as described below in the model of the plasma bursts induced in fiber laser welding.

2. Keyhole and weld pool dynamics

The keyhole and weld pool oscillations are inherent to the laser welding process even in the case of constant laser power and continuous wave operational mode. Among the possible causes of the oscillations, we consider two principal factors: the pressure imbalance at the keyhole/weld pool boundary and the Marangoni flow due to the surface tension gradient.

The keyhole oscillations are caused by the imbalance between the ablation pressure that keeps the keyhole opened and the pressure due to the surface tension that tends to close the capillary. The oscillations have been analyzed for a simplified model of an ideal cylindrical keyhole under the assumption that the amplitude of the deformations is small compared to the keyhole radius. Using the Lagrangian mechanics, the oscillation frequency has been calculated for various radial, azimuthal, and axial modes.16 Among them, the most important is the ground radial mode that corresponds to the oscillations of the keyhole radius with the angular frequency ω0 given by the formula

ω02=r(pγpablδpg)|r=aϱaln(am/a),
(2)

where pγ is the pressure due to the surface tension, pabl is the ablation pressure, δpg is the excess pressure which builds up due to the gas flow along the keyhole axis on its way out, ϱ is the average mass density of the liquid phase, and a and am represent the equilibrium radius of the keyhole and the melt pool, respectively. In the model of the plasma bursts described below, we consider this ground radial oscillatory mode as the one that match most closely the oscillations of the keyhole opening. It is also worth to note that through the individual terms, Eq. (2) involves the dependence of the oscillation frequency on the welding parameters settings. The theoretical analysis has predicted the increase of the frequency with the increasing welding speed and the decreasing laser power.16 

The dynamics of the weld pool is significantly influenced by the Marangoni flow due to the gradient of the surface tension coefficient γ(T) that is generally temperature dependent.17 In the surface layer of the weld pool, the liquid melt flows toward the region with the higher surface tension. The gradient dγ/dT is closely related to the chemical composition of the welded steel and is primarily determined by the concentration of the surface active elements (surfactants) such as sulfur, phosphorus, or oxygen. Although this dependence is generally very complex, it is possible to distinguish between two basic cases. The first case corresponds to the low concentration or even the total absence of the surfactants which makes the gradient dγ/dT negative. This causes the surface melt flow that is directed from the keyhole region, characterized by the high temperature, toward the outer weld pool boundary where the temperature is relatively lower. Below the surface layer, the liquid melt flows in the opposite direction which results in the melt circulation in the weld pool. Due to this configuration, the keyhole has relatively higher radius and the temperature gradients within the weld pool are relatively low. In contrast, the second case is characterized by the high concentration of the surfactants which makes the gradient dγ/dT positive. This causes the surface flow that is directed toward the keyhole region where the melt flows down along the keyhole wall and then continues back to the weld pool outer boundary. Such a configuration results in the keyhole with the relatively smaller radius and the high temperature gradients in the keyhole region causing a higher tendency to the spattering of the molten metal. In general, for the moderate concentration of the surfactants, the sign of the gradient dγ/dT can be temperature dependent which results in a bifurcated melt flow. These facts indicate that the chemical composition of the welded steel significantly influences the dynamics of the weld pool and the keyhole. Depending on the orientation of the Marangoni flow, the weld pool has a different volume, and the keyhole has a different radius which affects its oscillation period and equivalently the plasma bursts period.

3. Plasma bursts

The model of the periodic plasma bursts generation is considered for the continuous wave fiber laser welding with the bottom closed keyhole. As mentioned above, we involve the accumulation of the laser induced plasma and the continuous keyhole/weld pool oscillations as the principal physical phenomena that are closely synchronized. One period of the repetitive process can be divided into two consecutive stages—the accumulation of plasma in the keyhole and the plasma burst itself. The accumulation of plasma begins in the keyhole where the pressure, the electron temperature, and the plasma density are relatively low. The keyhole volume and the size of its opening are relatively small because the pressure due to the surface tension dominates over the sum of the ablation pressure and the plasma flow pressure (in fact, the latter is nearly zero). The plasma is mostly enclosed inside the keyhole, and the observed light emissions have minimal intensity. The conditions in the keyhole are gradually changed due to the Fresnel absorption of the laser beam at the keyhole walls, the continuous generation of plasma through the ionization of the metal vapors, and also by the laser absorption in plasma through the process of inverse bremsstrahlung. We come out of the fact that the absorption coefficient given by Eq. (1) is a positively dependent function of electron temperature and pressure in the keyhole, so the increase in both these quantities caused by the laser absorption leads to the further increase of the absorption itself. When the electron temperature and pressure in the keyhole exceed a critical limit, the first stage of the plasma accumulation is finished, and the second stage of the plasma burst begins. The plasma is ejected out of the keyhole as a short burst and forms a bright plasma plume above the workpiece. During the plasma burst, the observable light emissions have maximum intensity and the keyhole opening reaches the maximum size. This is given by the fact that the pressure due to the surface tension is smaller than the sum of the ablation pressure and the plasma flow pressure which is nonzero and reaches its maximum. The plasma plume gradually expands until it disappears, the keyhole opening shrinks, and the whole process starts again. When observing the interaction zone by a high-speed camera, we detect the significant fluctuations of the plasma plume corresponding to the temporal evolution of the plasma bursts (see Fig. 1). Similarly, when observing the process by a photodetector, we detect the short-time pulses corresponding to the plasma bursts (see Fig. 2). To further study the pulsing character of the light emissions, we employ the autocorrelation analysis.

The discrete normalized autocorrelation function R(τ) with the time lag τ for a real discrete signal I(t) with the mean value μ is defined as

R(τ)=t[ I(t)μ ][ I(tτ)μ ]t[ I(t)μ ]2.
(3)

In Fig. 3, we show the comparison of the signal I(t) and the corresponding autocorrelation R(τ) for three different signal types. If the signal I(t) represents a Gaussian noise, the autocorrelation R(τ) has a sharp maximum for τ = 0. For a periodic signal I(t), we obtain the function R(τ) that oscillates with the same period and has linearly decreasing envelope due to the finite signal length. Finally, for a noisy periodic signal I(t) we get the autocorrelation R(τ) that has a sharp maximum at τ = 0 and also periodically oscillates with decreasing amplitude. Since the last example is similar to the autocorrelations that we meet in our experiments, in Fig. 3(c), we indicate several autocorrelation characteristics used in the experimental data evaluation. The nonzero time lags τmin and τmax correspond to the first local minimum and maximum of R(τ), respectively, and a parameter Rm=R(τmax)R(τmin) represents the difference of R(τ) in these two points. The time lag τmax can be interpreted as the oscillation period of the original signal I(t). The parameter Rm can be used as the quantitative measure for the recognizability of the signal oscillations. The interpretation of the time lag τmax and the parameter Rm indicates that they are closely related to the period and the recognizability of the plasma bursts, respectively, as described in Sec. III.

FIG. 3.

Autocorrelation function R(τ) for the signal I(t) representing (a) Gaussian noise, (b) periodic signal and (c) noisy periodic signal.

FIG. 3.

Autocorrelation function R(τ) for the signal I(t) representing (a) Gaussian noise, (b) periodic signal and (c) noisy periodic signal.

Close modal

The experiments were carried out on a laser welding machine equipped with a 2 kW Yb:YAG fiber laser operating in the continuous wave mode at the wavelength of 1.07 μm. The experimental set-up is shown in Fig. 4. The laser beam was delivered by a 200 μm core diameter fiber, collimated by a 100 mm lens, and focused by a 200 mm lens to get a focus diameter of 0.4 mm. The intensity of the light emissions produced during the bead-on-plate welding was continuously detected by a photodetector pointed at the plasma plume (the viewing angle ±3°, the wavelength range 350–1100 nm), sampled by a data acquisition device at the 40 kHz sample rate, and stored in a personal computer (PC) for further analysis.

FIG. 4.

Experimental set-up of the laser welding machine: 1—laser beam, 2—coaxial nozzle, 3—shielding gas, 4—plasma plume, 5—workpiece, 6—neutral density absorptive (gray) filter, 7—photodetector, 8—photodetector mounting tube, 9—amplifier with adjustable gain, 10—data acquisition device, and 11—PC with control software.

FIG. 4.

Experimental set-up of the laser welding machine: 1—laser beam, 2—coaxial nozzle, 3—shielding gas, 4—plasma plume, 5—workpiece, 6—neutral density absorptive (gray) filter, 7—photodetector, 8—photodetector mounting tube, 9—amplifier with adjustable gain, 10—data acquisition device, and 11—PC with control software.

Close modal

The primary set of experiments was aimed to study in detail the relations among the welding parameters settings, the plasma bursts characteristics, and the weld geometry. It is represented by the partial penetration welds carried out under the welding parameters chosen within the ranges usually used in industry. As a workpiece material, we used the 6 mm thick plates of the X5CrNi 18-10 stainless steel and the S235JR carbon steel (the chemical composition is specified in Table I). The shielding gas was argon with the volumetric flow rate of 18 l/min. The welding speed v was 10, 20, and 30 mm/s, and the focus position was 1 mm below the workpiece surface. The laser power P was changed in 0.2 kW steps between 1.0 and 2.0 kW along each weld of the length 83 mm.

TABLE I.

Chemical composition of the welded materials (wt. %).

MaterialCMnPSCrNiSi
S235JR carbon steel <0.17 <1.4 <0.045 <0.045 — — — 
X5CrNi 18-10 stainless steel <0.07 <2.0 <0.045 <0.015 17.5–19.5 8.0–10.5 <1.0 
MaterialCMnPSCrNiSi
S235JR carbon steel <0.17 <1.4 <0.045 <0.045 — — — 
X5CrNi 18-10 stainless steel <0.07 <2.0 <0.045 <0.015 17.5–19.5 8.0–10.5 <1.0 

The secondary set of experiments was made in order to reveal the basic differences in the plasma bursts characteristics for the partial and full penetration welds. It is represented by the full penetration welds carried out under the same conditions as the primary set with two exceptions—we used the 3 mm thick plates of the X5CrNi 18-10 stainless steel and the S235JR carbon steel and the welding speed was limited to 10 mm/s. As an additional detector, we used a digital camera providing a coaxial top view of the keyhole area at the frame rate of 30 fps (it was used also for the 6 mm plates and the welding speed 10 mm/s).

The evaluation of the weld segments corresponding to the steps of the laser power was based on mutual comparison of several process characteristics. The signal I(t) sampled by the photodetector was displayed in a short time scale to reveal the nature of the plasma plume fluctuations. The autocorrelation R(τ) of the signal I(t) was calculated using Eq. (3) where the signal length was 38 000, 19 000, and 13 000 samples that are equivalent to 0.95, 0.48, and 0.33 s for the welding speeds 10, 20, and 30 mm/s, respectively. The function R(τ) was filtered using a low pass filter in order to determine the nonzero time lags τmin and τmax corresponding to the first local minimum and maximum, respectively. The time lag τmax was identified as the period of the plasma bursts T. The parameter Rm=R(τmax)R(τmin) was used as the quantitative measure for the recognizability of the plasma bursts. Finally, the weld depth d was measured from the scanned image of the weld cross section obtained by metallographic processing.

We start the description with the primary set of experiments represented by the partial penetration welds. The waveforms of the light emissions revealed the plasma bursts in the form of the repetitive short-time pulses with the period in the order of milliseconds (see Fig. 2). Using the autocorrelation function, the plasma bursts were detected for all chosen welding parameters settings except for the combination of the S235JR carbon steel, the laser power of 1.0 kW, and the welding speed of 20 and 30 mm/s. In Fig. 5, we show the plots of the parameter Rm that indicate the recognizability of the plasma bursts. In the case of the X5CrNi 18-10 stainless steel, the parameter Rm ranges between 0.04 and 0.41 and mostly monotonously increases with increasing laser power. However, in the case of the S235JR carbon steel, the parameter Rm ranges between 0.03 and 0.33 but significantly increases only for 1.8 and 2.0 kW. This difference can be explained by the different thermo-mechanical properties of the stainless steel and the carbon steel characterized by the different chemical composition (see Table I). Primarily due to the content of nickel, the stainless steel has a lower thermal conductivity and a higher viscosity in the liquid state.18 This implies that the volume of the weld pool and the tendency of the liquid material toward the collective oscillations are higher for the stainless steel compared to the carbon steel. So, the keyhole/weld pool oscillations and the plasma bursts are more significant and more recognizable in the case of the stainless steel.

FIG. 5.

Relations between the laser power P, the welding speed v, and the parameter Rm that indicates the recognizability of the plasma bursts.

FIG. 5.

Relations between the laser power P, the welding speed v, and the parameter Rm that indicates the recognizability of the plasma bursts.

Close modal

In Fig. 6, we show the examples of the autocorrelations R(τ) with indicated first local maximum and the images of the weld cross sections corresponding to the X5CrNi 18-10 stainless steel, the laser power 1.0–2.0 kW, and the welding speed 30 mm/s. The time lag of the first maximum that is equivalent to the period of the plasma bursts ranges between 1.0 and 3.4 ms, the weld depth ranges between 1.6 and 3.1 mm. The period and recognizability of the plasma bursts as well as the weld depth increase with increasing laser power, which can be justified by two facts. First, the weld pool has larger volume and mass due to the higher laser power, so the system oscillations are more significant but slower. Second, it takes longer time to reach a critical temperature in the keyhole that has a larger volume due to the higher laser power.

FIG. 6.

Autocorrelation functions R(τ) with the position of the first local maximum indicated by the dashed line and the images of the weld cross sections. Welding conditions: X5CrNi 18-10 stainless steel, plate thickness 6 mm, argon 18 l/min, laser power 1.0–2.0 kW, and welding speed 30 mm/s.

FIG. 6.

Autocorrelation functions R(τ) with the position of the first local maximum indicated by the dashed line and the images of the weld cross sections. Welding conditions: X5CrNi 18-10 stainless steel, plate thickness 6 mm, argon 18 l/min, laser power 1.0–2.0 kW, and welding speed 30 mm/s.

Close modal

In Fig. 7, we present the overall plots for the X5CrNi 18-10 stainless steel and the S235JR carbon steel showing the relations between the laser power P, the welding speed v, the plasma bursts period T, and the weld depth d. The upper and middle plots show that both T and d depend linearly on P for the constant v and both T and d decrease with increasing v for the constant P for both stainless steel (left) and carbon steel (right). The comparison of the stainless and carbon steel indicates that the maximum difference in d for a given P and v reaches only 15% but T differs up to 50%. The bottom plots reveal a linear dependence between the weld depth d and the burst period T that is unique for all chosen combinations of P and v and is characterized by the rates 0.62 and 1.07 mm/ms for the stainless steel and the carbon steel, respectively. These results can be interpreted and justified as follows. The linear dependence between P and d is consistent with the previous theoretical and experimental studies1,8,19 showing that the weld depth is proportional to the heat input given by the ratio P/v. The linear dependence between the quantities P, T and d, T implies that the burst period is mainly given by the keyhole depth that determines the keyhole volume and consequently the time necessary to reach the critical temperature and pressure of the plasma inside the keyhole. Moreover, the relation between P and T is in qualitative agreement with the theoretical predictions for the frequency of the radial oscillatory mode of an ideal cylindrical keyhole16 which matches most closely to the oscillations of the keyhole opening. The higher plasma bursts period T for the stainless steel compared to the carbon steel can be explained by the different thermomechanical properties due to the different chemical composition (see Table I). The X5CrNi 18-10 stainless steel has relatively low content of sulfur which implies that the surface tension gradient dγ/dT can be considered negative. The Marangoni flow in the surface layer is directed from the keyhole region to the outer weld pool boundary and the keyhole of the given depth has relatively higher radius and also the volume. The higher keyhole radius implies the higher period of the keyhole oscillations (see Eq. (2)), the higher volume requires longer time to reach the critical temperature and pressure in the keyhole. Moreover, due to the lower thermal conductivity of the stainless steel, the weld pool has larger volume and its oscillations are relatively slower (this is also due to the higher viscosity given by a nickel content). In contrast, the S235JR carbon steel has relatively high content of sulfur which makes the surface tension gradient dγ/dT rather positive. The surface layer flow is directed toward the keyhole region and the keyhole of the given depth has relatively smaller radius and volume. Therefore, the period of the keyhole oscillations is lower as well as is the time required to reach the critical conditions in the keyhole. Finally, due to the higher thermal conductivity of the carbon steel, the weld pool has smaller volume and its oscillations are relatively faster (this is also due to the lower viscosity).

FIG. 7.

Relations between the laser power P, the welding speed v, the burst period T, and the weld depth d. The dashed lines represent the linear fit of the experimentally measured data.

FIG. 7.

Relations between the laser power P, the welding speed v, the burst period T, and the weld depth d. The dashed lines represent the linear fit of the experimentally measured data.

Close modal

The comparison of the plasma bursts characteristics for the partial and full penetration welds was made for the 3 and 6 mm plates of the X5CrNi 18–10 stainless steel and the S235JR carbon steel, the laser power 1.0–2.0 kW, and the welding speed 10 mm/s. For both types of steel, the partial penetration was observed for all welds in the 6 mm plates and the full penetration was achieved for all welds in the 3 mm plates. The plasma bursts were detected for all chosen welding parameters settings, but their characteristics proved to be dependent on the welding mode. When dealing with the partial and full penetration welding, it is necessary to distinguish between three stable modes and one unstable mode.20 The first stable mode corresponds to the partial penetration characterized by the bottom closed keyhole and the absence of the weld root. The second and the third stable mode are characterized by the presence of the weld root and represent the full penetration with the bottom closed and bottom opened keyhole, respectively. The unstable mode occurs when the keyhole randomly alters between the bottom opened and bottom closed state which results in irregular weld root. To experimentally distinguish the welding modes, we used the images from the digital camera providing a top view of the keyhole area. Since the full penetration was achieved for all welds in the 3 mm plates, it is meaningless to study the relation between the laser power P and the weld depth d. Therefore, we primarily investigated the dependence between the laser power P and the plasma burst period T.

In Fig. 8, we show the examples of the autocorrelations R(τ) with indicated first local maximum and the top view images of the keyhole area corresponding to the 3 and 6 mm plates of the X5CrNi 18-10 stainless steel and the S235JR carbon steel, the laser power 2.0 kW, and the welding speed 10 mm/s. In the case of the stainless steel, the autocorrelation characteristics are very similar for both plate thicknesses—the detected plasma burst period T is 4.9 ms for the 6 mm plate and 5.1 ms for the 3 mm plate. The results indicate that the welding was conducted with the bottom closed keyhole for both plate thicknesses. This is consistent with the similarity of the camera images showing the keyhole area—the bright spots have almost uniform irradiance profile due to the signal saturation and the diameter of the saturated area is 0.55 mm for the 6 mm plate and 0.52 mm for the 3 mm plate. In contrast, the autocorrelation characteristics for the 6 and 3 mm plates of the carbon steel are significantly different. The plasma burst period T is 3.3 ms for the 6 mm plate but increases to 8.1 ms for the 3 mm plate. These results indicate that the welding was conducted with the bottom opened keyhole for the 3 mm plate. This is confirmed by the camera images showing the keyhole area. For the 6 mm plate, we observe a bright spot with almost uniform irradiance due to the signal saturation. This is similar to the case of the stainless steel, but the diameter of the saturated area is 0.47 mm. However, for the 3 mm plate, we clearly observe a central drop in the irradiance profile of the bright spot that indicates the bottom side keyhole opening. The outer diameter of the saturated area is reduced only to 0.37 mm which relates to the fact that the part of the plasma flows out of the bottom keyhole opening.

FIG. 8.

Autocorrelation functions R(τ) and the top-view images of the keyhole opening for the partial and full penetration welds made in the 3 and 6 mm plates of the X5CrNi 18-10 stainless steel and the S235JR carbon steel. The first local maximum of the R(τ) is indicated by the dashed line. Welding conditions: laser power 2.0 kW, welding speed 10 mm/s, argon 18 l/min.

FIG. 8.

Autocorrelation functions R(τ) and the top-view images of the keyhole opening for the partial and full penetration welds made in the 3 and 6 mm plates of the X5CrNi 18-10 stainless steel and the S235JR carbon steel. The first local maximum of the R(τ) is indicated by the dashed line. Welding conditions: laser power 2.0 kW, welding speed 10 mm/s, argon 18 l/min.

Close modal

In Fig. 9, we present the overall plots for the 3 and 6 mm plates of the X5CrNi 18-10 stainless steel and the S235JR carbon steel showing the relations between the laser power P and the plasma bursts period T. In the case of the stainless steel, we observe a linear dependence between P and T for both plate thicknesses. The values of T for the 3 mm plate are only 13% higher in average compared to the 6 mm plate. This similarity indicates that the full penetration welding of the 3 mm plate is characterized by the bottom closed keyhole within the whole range of the laser power 1.0–2.0 kW. In contrast, the 3 and 6 mm plates of the carbon steel are characterized by similar values of T only for the laser power of 1.0 kW (the relative difference is 12%). For higher laser power, the relations between P and T are also linear but the values of T for the 3 mm plate are about 2.7 times higher in average compared to the 6 mm plate. These results show that the welding of the 3 mm plate is characterized by the bottom closed keyhole only for the laser power of 1.0 kW; otherwise, the bottom opened keyhole occurs. The significant difference in the plasma burst characteristics for the full penetration welds in the X5CrNi 18–10 stainless steel and the S235JR carbon steel can be explained by the different thermomechanical properties. The viscosity of the molten stainless steel is sufficiently high to keep the keyhole closed at the bottom. However, the low viscosity of the molten carbon steel causes its low cohesion which results in the bottom opened keyhole. The difference in the keyhole geometry has a significant influence on the plasma bursts period. This is due to the fact that the pressure in the bottom opened keyhole is generally lower compared to the bottom closed keyhole. Therefore, the bottom opened keyhole has higher radius than the bottom closed keyhole which implies higher period of the keyhole oscillations and of the corresponding plasma bursts. This is also related to the fact that the lower pressure in the bottom opened keyhole requires longer time to the accumulation of the laser induced plasma required for the plasma burst.

FIG. 9.

Relations between the laser power P and the burst period T for the partial and full penetration welds. The dashed lines represent the linear fit of the experimentally measured data.

FIG. 9.

Relations between the laser power P and the burst period T for the partial and full penetration welds. The dashed lines represent the linear fit of the experimentally measured data.

Close modal

The autocorrelation analysis has proven to be an efficient tool for studying the light emissions generated by the plasma plume. Due to its fundamental simplicity and versatility, the method enables to detect the plasma bursts period that is typically in the range of several milliseconds. The experimental data acquired for the partial penetration welds in the X5CrNi 18-10 stainless steel and the S235JR carbon steel revealed the linear relationship between the burst period and the weld depth that is unique for all chosen combinations of the laser power and welding speed. This represents an important result since we have found the direct connection between the weld geometry and the physical quantity characterizing the welding process that can be inferred from the observable light emissions. The autocorrelation analysis has also revealed a significant difference in the plasma bursts period for the full penetration welds with the bottom closed and bottom opened keyhole. This finding provides an interesting topic for further investigations aimed to the detection of the transition between the different welding modes.

However, there are limitations of the autocorrelation analysis that should be mentioned. Its applicability to the welding process diagnostics is limited to the cases when the plasma bursts are recognizable. This is primarily given by the welding parameters settings and the thermomechanical parameters of the welded materials. The comparison of the X5CrNi 18-10 stainless steel and the S235JR carbon steel has shown that the plasma bursts are more significant for the higher laser power, the higher welding speed and the materials characterized by the lower thermal conductivity, the higher viscosity and the lower content of the surfactants. The recognizability of the plasma bursts is also influenced by the parameters used in the numerical signal processing, especially by the length of the signal I(t) taken for the calculation of the autocorrelation R(τ). As described in Sec. III, we used the signal length of 38 000, 19 000, and 13 000 samples that are equivalent to 0.95, 0.48, and 0.33 s for the welding speeds 10, 20, and 30 mm/s, respectively. These reference signal lengths approximately correspond to the 10 mm of the weld length. To find the lower limit of the signal length that will be sufficient for a reliable detection of the plasma bursts period, we numerically tested the acquired data. We investigated the relation between the plasma burst period T and the signal length N (given in samples). As expected, the values of the period T significantly oscillated for small N but converged to a stable value with increasing N. The results showed that the relative change of the plasma bursts period T decreases below 5% for the reduced signal length that represents only 30%–50% of the reference signal length depending on the recognizability of the plasma bursts. It means that the reliable values of the plasma burst period T can be obtained from the signal of the reduced length corresponding to the 3–5 mm of the weld length. Since more detailed analysis of this issue is desirable for practical use, we intend to study the waterfall plots of the autocorrelation function that are analogous to the spectrograms used in the short-time frequency analysis. It is also highly desirable to develop more rigorous physical model of the plasma bursts that will provide further insight into this interesting topic.

We presented the autocorrelation analysis of the plasma plume light emissions in the deep penetration laser welding of steel. We showed that the autocorrelation function is an efficient tool to detect the period of the plasma bursts. As a result, we identified the linear dependence between the burst period and the weld depth that is valid for both the X5CrNi 18-10 stainless steel and the S235JR carbon steel. This is an important finding that represents motivation for further study of the plasma bursts with regard to possible applications in welding process diagnostics.

The research was supported by MEYS CR (LO1212), its infrastructure by MEYS CR and EC (CZ.1.05/2.1.00/01.0017) and by ASCR (RVO:68081731).

1.
A.
Kaplan
, “
A model of deep penetration laser welding based on calculation of the keyhole profile
,”
J. Phys. D: Appl. Phys.
27
,
1805
1814
(
1994
).
2.
D. F.
Farson
and
K. R.
Kim
, “
Generation of optical and acoustic emissions in laser weld plumes
,”
J. Appl. Phys.
85
,
1329
1336
(
1999
).
3.
S.
Nakamura
,
M.
Sakurai
,
K.
Kamimuki
,
T.
Inoue
, and
Y.
Ito
, “
Detection technique for transition between deep penetration mode and shallow penetration mode in CO2 laser welding of metals
,”
J. Phys. D: Appl. Phys.
33
,
2941
2948
(
2000
).
4.
H.
Zeng
,
Z.
Zhou
,
Y.
Chen
,
H.
Luo
, and
L.
Hu
, “
Wavelet analysis of acoustic emission signals and quality control in laser welding
,”
J. Laser Appl.
13
,
167
173
(
2001
).
5.
Z.
Szymanski
,
J.
Hoffman
, and
J.
Kurzyna
, “
Plasma plume oscillations during welding of thin metal sheets with a CW CO2 laser
,”
J. Phys. D: Appl. Phys.
34
,
189
199
(
2001
).
6.
J.
Zou
,
W.
Yang
,
S.
Wu
,
Y.
He
, and
R.
Xiao
, “
Effect of plume on weld penetration during high-power fiber laser welding
,”
J. Laser Appl.
28
,
022003
(
2016
).
7.
A.
Matsunawa
,
J.
Kim
,
N.
Seto
,
M.
Mizutani
, and
S.
Katayama
, “
Dynamics of keyhole and molten pool in laser welding
,”
J. Laser Appl.
10
,
247
254
(
1998
).
8.
Y.
Kawahito
,
N.
Matsumoto
,
M.
Mizutani
, and
S.
Katayama
, “
Characterisation of plasma induced during high power fibre laser welding of stainless steel
,”
Sci. Technol. Weld. Joining
13
,
744
748
(
2008
).
9.
J.
Wang
,
Ch.
Wang
,
X.
Meng
,
X.
Hu
,
Y.
Yu
, and
S.
Yu
, “
Study on the periodic oscillation of plasma/vapour induced during high power fiber laser penetration welding
,”
Opt. Laser Technol.
44
,
67
70
(
2012
).
10.
E.
Beyer
,
L.
Bakowsky
,
P.
Loosen
,
R.
Poprawe
, and
G.
Herziger
, “
Development and optical absorption properties of a laser induced plasma during CO2-laser processing
,”
Proc. SPIE
0455
,
75
80
(
1984
).
11.
L.
Mrna
,
M.
Sarbort
,
S.
Rerucha
, and
P.
Jedlicka
, “
Correlation between the keyhole depth and the frequency characteristics of light emissions in laser welding
,”
Phys. Proc.
41
,
469
477
(
2013
).
12.
L.
Mrna
and
M.
Sarbort
, “
Plasma bursts in deep penetration laser welding
,”
Phys. Proc.
56
,
1261
1267
(
2014
).
13.
L.
Mrna
,
M.
Sarbort
, and
S.
Rerucha
, “
Autocorrelation analysis of plasma plume oscillations in deep penetration laser welding
,” in
Lasers in Manufacturing Conference
, Munich, Germany (
2015
), Paper No. 255.
14.
S.
Eliezer
,
The Interaction of High-Power Lasers with Plasmas
(
Institute of Physics Publishing
,
Bristol
,
2002
).
15.
M.
Gao
,
C.
Chen
,
M.
Hu
,
L.
Guo
,
Z.
Wang
, and
X.
Zeng
, “
Characteristics of plasma plume in fiber laser welding of aluminum alloy
,”
Appl. Surf. Sci.
326
,
181
186
(
2015
).
16.
T.
Klein
,
M.
Vicanek
,
J.
Kroos
,
I.
Decker
, and
G.
Simon
, “
Oscillations of the keyhole in penetration laser beam welding
,”
J. Phys. D: Appl. Phys.
27
,
2023
2030
(
1994
).
17.
T.
Fuhrich
,
P.
Berger
, and
H.
Hügel
, “
Marangoni effect in laser deep penetration welding of steel
,”
J. Laser Appl.
13
,
178
186
(
2001
).
18.
E.
Kannatey-Asibu
,
Principles of Laser Material Processing
(
Wiley
,
New Jersey
,
2009
).
19.
J.
Dowden
and
P.
Kapadia
, “
A mathematical investigation of the penetration depth in keyhole welding with continuous CO2 lasers
,”
J. Phys. D: Appl. Phys.
28
,
2252
2261
(
1995
).
20.
M. Y.
Krasnoperov
,
R. R. G. M.
Pieters
, and
I. M.
Richardson
, “
Weld pool geometry during keyhole laser welding of thin steel sheets
,”
Sci. Technol. Weld. Joining
9
,
501
506
(
2004
).