- Email: [email protected]
Influence of weld imperfection on short circuit GMA welding arc stability
Journal of Materials Processing Technology 175 (2006) 285–290
Influence of weld imperfection on short circuit GMA welding arc stability K. Luksa Welding Department, Faculty of Mechanical Engineering, Silesian University of Technology, ul. Konarskiego 18a, 44-100 Gliwice, Poland
Abstract Monitoring of welding process can play important role in on-line quality control of the GMA welding process. It is very important to find a quick and safe method of analysing data recorded during monitoring. Two signals: momentary arc power and momentary arc resistance were included in the analysis and the results of analysis were compared with the results of statistic analysis of welding current and welding voltage. Sensitivity of mean value and variance of signals to artificial disturbances of welding arc was tested. The results show that it is possible to detect imperfections related to phenomena that take place in welding arc and affect the gas shield of the welding arc or the length of wire extension. © 2005 Elsevier B.V. All rights reserved. Keywords: Welding; GMA welding; Welding monitoring
1. Introduction The welding process can be characterised by four sets of parameters: those that describe the property of base metals and consumables, those that describe the burning of welding arc, those that describe the configuration of the welding electrode–base metal system and those that describe property the of welding machine. Some of welding parameters, that are thought to be input parameters for the welding process, as a result of interaction with the welding process, receive new features and can be treated as signals carrying information about the welding process. Welding voltage, welding current and welding wire speed, in the case of short circuit metal transfer, are examples of those parameters. Those parameters are modulated by the arc metal transfer during short circuit GMA welding process. The simplicity of welding voltage and current measurement is the reason why those parameters are the most frequently used for monitoring and evaluation of welding process quality. Recording and analysing of welding parameters let us test directly the conditions of welding arc burning and also the process of molten metal transfer from electrode to weld pool. We can assume that properties of base metal, consumables and also properties of welding machine do not change during the execution of welding, whereas changes in geometrical configuration between an 0924-0136/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2005.04.053
electrode and welded metal influence the condition of the welding arc. In that case, we can simplify the welding process to a model in which the quality of a welded joint first of all depends on the stability of arc.
2. Stability of GMA welding arc A welding arc can be thought of as a cylinder-like shape bounded by a barrier of temperature gradient. It is kept in equilibrium if all acting forces are in equilibrium [1]. Three main forces which control the equilibrium in the arc region are: surface tension force, electromagnetic force and gravitational force. Losing the state of equilibrium can be caused by liquid metal transport through the welding arc or disturbance of the gaseous shield of welding arc. At the beginning of metal transfer the surface tension force resists detachment of the molten droplet hanging at the welding wire and when the droplet touches the welding pool the force changes its direction. The electromagnetic power results in flow and rotation of the medium that conducts the electric current. The influence of electromagnetic force in the area around the molten electrode tip is known as the Pinch effect that supports detachment of liquid droplet from wire.
286
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
Intensive spatter, frequent arc stubbing and resistantexplosive melting of wire are the external symptoms of welding arc stability loss. The energy balance during gas metal arc welding with short circuit metal transfer is negative. The temperature of the molten welding pool and surrounding material goes down during short circuit. When the short circuit is too long, the temperature of arc area is too low to support the emission of electrons and the following welding arc ignition. Conditions of arc ignition are strongly influenced by the chemical composition of the shielding gas and presence of the oxide film on welded metal surface. The areas rich in oxide films have low emission work. Research of welding process stability are carried on by testing characteristics of signals recorded during welding. Testing of mean value and standard deviation of welding voltage and welding current is in common use [2,3]. Some researchers [4] test statistical characteristics of short circuit and arc burning time, for example, standard deviation of instantaneous short circuit frequency can be used as a measure of welding arc stability. The authors [4] admit that expressed in that way the criterion of arc stability depends on other welding parameters. Another criterion of GMA arc stability was presented in [5]. At first the authors present results of earlier research in which were proposed as a criterion of arc stability the variation of time intervals between transfer of succeeding molten droplets from a wire to welding pool [6], or standard deviation of welding current maximum values [7]. Other research results are also cited [8], in which probability density plots of welding parameters are used to estimate welding process stability. Similar criterions of arc stability are presented in [9]. For analysis of metal transportation through the welding arc, the theory of welding pool oscillation was exploited [5]. The theory was presented in [10] and also exploited in [11]. The most important conclusion resulting from that theory is that the most favourable condition for welding can be observed when the standard deviation of short circuit frequency reaches the minimum value [10]. In the same article, two conditions of stable arc burning are put into words: wire velocity equal to wire feed velocity and transport of molten metal from electrode to welding pool should not disturb the burning of welding arc. In this work [12] time of short circuit and time of arc burning were investigated by calculation of their mean values and standard deviations. The coefficient of variation was defined as the quotient of standard deviation value and mean value. The range of welding parameters in which the coefficient of variation reaches its minimum value is recognised as the range that corresponds to the most stable arc. In the following paper [13], the authors changed their opinion, and proposed [13,14] a different approach to this problem. The standard deviation of each loop at U–I plot that corresponds to the cycle of each droplet transfer from electrode to welding pool was proposed as a measure of welding arc stability. A plot of defined in this way stability factor against welding parameters have a minimum value. This helps to evaluate the state of the welding
arc and the distance from actual parameters to optimal parameters.
3. Experiments The estimation of the gas metal arc stability and concluding on this base about the stability of the welding process requires creations (for research) circumstances in which the unstable burning of the welding arc can be observe. In the experiments disturbances which were found characteristic for the GMA welding process were taken into account. Five artificial disturbances were tested: -
a lack of shielding gas, a layer of grease on the top surface of joint, a layer of paint on the top surface of joint, too narrow root gap, too wide root gap.
The disturbances were applied two or three times at the same joint and the length of disturbance was 30 mm each time. The distance between disturbances was 40 mm. During the welding process execution, the values of selected features of recorded voltage and welding currents signal were recorded in sections of the joint, where did not appear the disturbance and in sections with introduced disturbances. Tests were made on butt weld joints made of S235 structure steel. Each test joint was 250 mm long, 90 mm wide and 5 mm thick. The angle of the V bevel was 45◦ and the width of root gap was 1.2 mm. Joints were welded by gas metal arc welding with shielding gas M21 and G382CG3Si1 wire of diameter 1.2 mm. Welding parameters were the same for all joints: welding current, 120–130 A; welding voltage, 18 V; welding speed, 266–275 mm/min; wire extension, 15 mm. Welding current and welding voltage were recorded with a speed of 4000 samples per second to detect the influence of disturbances on welding parameters. The law pass filter was applied.
4. Results and discussion Values of the features of signals are dependent on the number of measurements in the sample. In the case of the GMA welding process with short circuit metal transport through the arc these features should represent elementary phenomena happening in the arc: the creation of droplet (the time of arc burning) and the passage of the droplet from molten end of wire to welding pool (the time of the welding arc short circuit). We can assume that the minimum number of the arc burnings and short circuits which should be taken into account to calculate the features of the signal is 10. We also know that the range of short circuit time in examined files is 2.4–10.2 ms, the range of arc burning time in examined files is 4.60–45.0 ms. In this case, the size of the time window should be longer than more or less 550 ms, what
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
287
Fig. 1. Influence of time window length on values of selected features of recorded welding current and welding voltage signals. I: welding current, U: welding voltage; TJ: time of arc burning; TZ: time of short circuit; UMI = U × I; UDI = U/I; m: mean value; v: variance.
at the sampling rate 4000 samples per the second is equal to 2200 samples. Experimental results show however, that the values of the features stabilize after the slightly longer time (Fig. 1). It is also important to know what happed when the time window in which we calculate the features of signals is not filled completely by disturbance (Fig. 2). These research were executed for the lack of the shielding gas and the burn through because these defects affect the recorded signals in different way. The lack of shielding gas changes the character the welding arc burning and influences recorded signals along the whole length of disturbance. The burn through defect can affect recorded signals only at the beginning and final section of the defect. In this case, changes in recorded signals result from the abrupt change of the wire extension and connected with this transition of the arc to new equilibrium conditions at the longer wire extension. Results show that even if the time window is only partially fulfilled by disturbance, the changes of tested features are big enough to notice them (Fig. 2). For each test joint four signals were analysed: welding current (I), welding voltage (U), momentary power of arc (UMI) and momentary arc resistance (UDI). For our case, “momentary” means that values of UDI and UMI were also calculated 4000 times per second and later averaged, if necessary. Welding voltage signal was used to calculate time of short circuit
(TZ), time of arc burning (TJ) and frequency of short circuit (FZW). Statistical features of each signal (labelled in figures as m – mean value and v – variance) calculated separately for the area of weld with disturbances and the area of weld free of disturbances. Two coefficients were calculated. The first coefficient (WWBW) expresses the quotient of the mean value of statistical features in the area of weld where disturbance was applied and the mean value of the statistical features in the area of weld free of disturbance. The second coefficient (WWC) expresses the quotient of the mean value of statistical features in the area of weld where disturbance was applied and the mean value of statistical features calculated for the full length of weld. Results of that calculation are presented in Table 1. GMA is a welding process in which heat for melting the base metal and wire is transferred from an arc where electrical energy is changed to thermal energy. Monitoring of welding current and welding voltage does not completely represent the thermal processes which are essential for evaluation of welding process stability. UMI and UDI signals, which represent “the thermal element” of welding process, are included in the analysis to determine if they are more suitable for finding areas of a joint where imperfections are expected. Analysis of results presented in Table 1 shows that in most cases signals of UMI and UDI better indicate areas where disturbances were applied, especially when the WWBW coefficient is investigated.
Fig. 2. Influence of time window length and location on values of selected features of recorded welding current and welding voltage signals. v TJw: variance of time of arc burning in area of disturbance; v TJok: variance of time of arc burning in area without of disturbance; v UDIw: variance of UDI in area of disturbance; v UDIok: variance of time of arc burning in area without disturbance.
288
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
Table 1 Relative change of statistical features of tested signals in the weld area with disturbances Relative values Variance of welding voltage
Variance of welding current
Variance of arc burning time
Variance of arc short circuit time
Frequency of short circuit
Variance of UMI
Variance of UDI
Disturbance: lack of gas shield WWCa 1.85 WWBWb 4.09
1.64 2.88
0.15 0.18
1.38 4.15
1.55 2.40
1.50 2.78
2.12 307.90
Disturbance: layer of paint WWCa 1.53 WWBWb 2.04
1.39 2.96
0.66 7.63
3.25 130.80
1.06 1.01
1.00 1.91
1.52 1065.76
Disturbance: layer of grease WWCa 1.22 WWBWb 1.35
1.48 2.90
4.29 35.95
4.57 81.48
1.03 0.96
1.18 2.15
0.93 634.51
Disturbance: too narrow root gap WWCa 1.00 WWBWb 0.97
1.01 1.04
1.11 1.65
0.86 0.81
0.96 0,89
1.01 1.00
1.02 1.01
Disturbance: too wide root gap WWCa 1.37 WWBWb 1.49
2.13 4.97
4.27 70.22
6.25 326.70
0.63 0.51
1.63 2.99
2.62 1274.82
UMI: momentary power of welding arc; UDI: momentary resistance of welding arc. a WWC expresses quotient of mean value of statistical feature in the area where disturbance was applied and mean value of statistical feature calculated for all length of joint. b WWMW expresses quotient of mean value of statistical feature in the area where disturbance was applied and mean value of statistical feature in the area free of disturbance.
A different approach to analysis of signals emitted by the welding process is presented in Figs. 4–7. In this case, recorded signals are divided into samples. Each sample represents 500 recorded values (0.125 s). Mean value of statistical features of samples were smoothed by moving average method (Fig. 3). The fifth applied disturbance (too narrow root gap—lack of penetration) is not presented because no changes in tested signals were observed. That kind of imperfection probably influence neither welding arc stability nor welding parameters. It is also possible that changes in welding parameter were to small and to short in time to notice them. The four remaining disturbance influence the run of recorded signals in different way. Lack of shielding gas influences almost all calculated features of signals (Fig. 4). Layer of grease
Fig. 3. Mean value ( m) and variance ( v) of welding current (I) computed for each 1/8 of second and smoothened using moving average method; f: smoothened.
was detected but not in all three areas where grease was applied. Thickness of grease layer was gradually increased and only the thicker layer of grease was clearly revealed in plots (Fig. 5). Layer of paint, and too wide root gap similar to lack of gas are clearly seen in plots (Figs. 6 and 7). For all applied disturbances traditional signals (welding current and welding voltage) features are compared with features of UMI and UDI signals. In many cases, UMI and UDI plots clearly reveal places where disturbances were applied. UMI and UDI signals derive from welding current and welding voltage so they cannot contain more information than the source signals, but some features of signals can be amplified. It happened when both welding current and welding voltage are influenced by introduced disturbance to limited degree. No changes in UMI or UDI signals were observed when neither welding current nor welding voltage is influenced by applied disturbance. Grease and paint were applied with layers of gradually increased thickness. At the beginning of the plates one layer was applied, in the middle two layer were applied and at the and three layer were applied on surface of plates (Figs. 5 and 6). In the case of the introduction of the layer of paint as the disturbance, the changes in the course of investigated signals response to the relative thickness of the layer of paint. Investigated signals otherwise answer to the change of the thickness of the layer of grease. Only the most thicker layer of grease causes distinct changes in investigated signals. Comparing these two cases one ought to take into account that paint has a solid consistency whereas grease has a semi-fluid consistency, changing in liquid at the elevated temperature.
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
289
Fig. 4. Mean value ( m), variance ( v) of welding current (I), welding voltage (U), momentary power of arc (UMI) and momentary arc resistance (UDI). Disturbance: lack of shielding gas.
Fig. 5. Mean value ( m), variance ( v) of welding current (I), welding voltage (U), momentary power of arc (UMI) and momentary arc resistance (UDI). Disturbance: layer of grease.
Fig. 6. Mean value ( m) and variance ( v) of welding voltage (U), welding current (I), momentary power of arc (UMI) and momentary arc resistance (UDI). Disturbance: layer of paint.
Fig. 7. Mean value( m) and variance ( v) of welding voltage (U), welding current (I), momentary power of arc (UMI) and momentary arc resistance (UDI). Disturbance: too wide root opening.
290
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
This causes that majority of burning products of paint probably enters the atmosphere of the welding arc; however, only the part of products of burning of grease enters the atmosphere of the arc. The rest is blown away by the pressure of the arc what can be seen, observing the process of welding.
5. Conclusions The monitoring of the GMA welding process by digital recording and analysis of welding parameters enables us to detect imperfections related to phenomena that take place in welding arc and influence the stability of welding arc. There are two types of disturbances that meet the mentioned requirements: imperfections that affect the gas shield of the welding arc and disturbances that affect the length of wire extension. Lack of shielding gas, a layer of grease, a layer of paint affect the gas shield of welding arc. Oxygen, nitrogen and hydrogen as components of the air or the products of decomposition of grease and paint. They change the welding arc shield properties, influence the ionisation potential, the surface tension of molten metal, the condition of welding arc ignition, the power of welding arc and other essential conditions of arc burning. All mentioned above disturbances change the condition of arc burning as long as they are applied. Never the less some additional conditions, like liquid grease blow out should be considered. Disturbances that influence the wire extension are related to the electrode–base metal system. They change the resistance of the welding circuit. The defects caused by this type of disturbances can be unrecognised in courses of investigated parameters if the degree of the change of the wire extension is not large.
Two methods of detection of both types of imperfections were tested for this paper. Signals of welding current, welding voltage, arc power and arc resistance were tested. Artificial disturbances of welding arc stability were detected in the tested signal. Not all signals react on arc disturbance in the same manner and intensity. This means that it is better for full detection to analyse all signals for proper detection of welding imperfection, but there is a chance of identifying the imperfections by using all signals simultaneously.
References [1] [2] [3] [4] [5]
[6] [7] [8] [9]
[10] [11] [12] [13] [14]
J.F. Lancaster (Ed.), The Physics of Welding, Pergamon Press, 1984. T.B. Quinn, C. Smith, et al., Weld. J. 9 (1999) 322-s. S. Adolfsson, A. Bahrami, et al., Weld. J. 2 (1999) 59-s. M.J. Hermans, G. Den Quden, Weld. J. 4 (1999) 137-s. M.J. Hermans, G. Den Quden, Proceedings of International Conference on the Joining of Materials, JOM-8, Helsingor, Denmark, 1997, p. 312. S. Liu, T.A. Siewert, Weld. J. 2 (1989) 52-s. W. Lucas, Met. Constr. 7 (1985) 431. S.R. Gupta, P.C. Gupta, D. Rehfeldt, Weld. Rev. 4 (1966) 232. U. Dilthey, T. Reihel, W. Scheller, Proceedings of Internatonal Conference on the Joining of Materials, JOM 7, Helsingor, Denmark, 1995, p. 410. P.J. Modensi, J.H. Nixon, Weld. J. 9 (1994) 219. D. Rehfeldt, T. Schmitz, IIW-DOC (1995) 212-882-95. T. Shinoda, H. Kaneda, Y. Takeuchi, Weld. Met. Fabrication 12 (1989) 522. T. Shinoda, H. Kaneda, Y. Takeuchi, IIW-DOC (1991) XII-1214-91. T. Shinoda, H. Nishikawa, Proceedings of International Conference on the Joining of Materials, JOM 7, Helsingor, Denmark, 1995, p. 558.
Influence of weld imperfection on short circuit GMA welding arc stability K. Luksa Welding Department, Faculty of Mechanical Engineering, Silesian University of Technology, ul. Konarskiego 18a, 44-100 Gliwice, Poland
Abstract Monitoring of welding process can play important role in on-line quality control of the GMA welding process. It is very important to find a quick and safe method of analysing data recorded during monitoring. Two signals: momentary arc power and momentary arc resistance were included in the analysis and the results of analysis were compared with the results of statistic analysis of welding current and welding voltage. Sensitivity of mean value and variance of signals to artificial disturbances of welding arc was tested. The results show that it is possible to detect imperfections related to phenomena that take place in welding arc and affect the gas shield of the welding arc or the length of wire extension. © 2005 Elsevier B.V. All rights reserved. Keywords: Welding; GMA welding; Welding monitoring
1. Introduction The welding process can be characterised by four sets of parameters: those that describe the property of base metals and consumables, those that describe the burning of welding arc, those that describe the configuration of the welding electrode–base metal system and those that describe property the of welding machine. Some of welding parameters, that are thought to be input parameters for the welding process, as a result of interaction with the welding process, receive new features and can be treated as signals carrying information about the welding process. Welding voltage, welding current and welding wire speed, in the case of short circuit metal transfer, are examples of those parameters. Those parameters are modulated by the arc metal transfer during short circuit GMA welding process. The simplicity of welding voltage and current measurement is the reason why those parameters are the most frequently used for monitoring and evaluation of welding process quality. Recording and analysing of welding parameters let us test directly the conditions of welding arc burning and also the process of molten metal transfer from electrode to weld pool. We can assume that properties of base metal, consumables and also properties of welding machine do not change during the execution of welding, whereas changes in geometrical configuration between an 0924-0136/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2005.04.053
electrode and welded metal influence the condition of the welding arc. In that case, we can simplify the welding process to a model in which the quality of a welded joint first of all depends on the stability of arc.
2. Stability of GMA welding arc A welding arc can be thought of as a cylinder-like shape bounded by a barrier of temperature gradient. It is kept in equilibrium if all acting forces are in equilibrium [1]. Three main forces which control the equilibrium in the arc region are: surface tension force, electromagnetic force and gravitational force. Losing the state of equilibrium can be caused by liquid metal transport through the welding arc or disturbance of the gaseous shield of welding arc. At the beginning of metal transfer the surface tension force resists detachment of the molten droplet hanging at the welding wire and when the droplet touches the welding pool the force changes its direction. The electromagnetic power results in flow and rotation of the medium that conducts the electric current. The influence of electromagnetic force in the area around the molten electrode tip is known as the Pinch effect that supports detachment of liquid droplet from wire.
286
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
Intensive spatter, frequent arc stubbing and resistantexplosive melting of wire are the external symptoms of welding arc stability loss. The energy balance during gas metal arc welding with short circuit metal transfer is negative. The temperature of the molten welding pool and surrounding material goes down during short circuit. When the short circuit is too long, the temperature of arc area is too low to support the emission of electrons and the following welding arc ignition. Conditions of arc ignition are strongly influenced by the chemical composition of the shielding gas and presence of the oxide film on welded metal surface. The areas rich in oxide films have low emission work. Research of welding process stability are carried on by testing characteristics of signals recorded during welding. Testing of mean value and standard deviation of welding voltage and welding current is in common use [2,3]. Some researchers [4] test statistical characteristics of short circuit and arc burning time, for example, standard deviation of instantaneous short circuit frequency can be used as a measure of welding arc stability. The authors [4] admit that expressed in that way the criterion of arc stability depends on other welding parameters. Another criterion of GMA arc stability was presented in [5]. At first the authors present results of earlier research in which were proposed as a criterion of arc stability the variation of time intervals between transfer of succeeding molten droplets from a wire to welding pool [6], or standard deviation of welding current maximum values [7]. Other research results are also cited [8], in which probability density plots of welding parameters are used to estimate welding process stability. Similar criterions of arc stability are presented in [9]. For analysis of metal transportation through the welding arc, the theory of welding pool oscillation was exploited [5]. The theory was presented in [10] and also exploited in [11]. The most important conclusion resulting from that theory is that the most favourable condition for welding can be observed when the standard deviation of short circuit frequency reaches the minimum value [10]. In the same article, two conditions of stable arc burning are put into words: wire velocity equal to wire feed velocity and transport of molten metal from electrode to welding pool should not disturb the burning of welding arc. In this work [12] time of short circuit and time of arc burning were investigated by calculation of their mean values and standard deviations. The coefficient of variation was defined as the quotient of standard deviation value and mean value. The range of welding parameters in which the coefficient of variation reaches its minimum value is recognised as the range that corresponds to the most stable arc. In the following paper [13], the authors changed their opinion, and proposed [13,14] a different approach to this problem. The standard deviation of each loop at U–I plot that corresponds to the cycle of each droplet transfer from electrode to welding pool was proposed as a measure of welding arc stability. A plot of defined in this way stability factor against welding parameters have a minimum value. This helps to evaluate the state of the welding
arc and the distance from actual parameters to optimal parameters.
3. Experiments The estimation of the gas metal arc stability and concluding on this base about the stability of the welding process requires creations (for research) circumstances in which the unstable burning of the welding arc can be observe. In the experiments disturbances which were found characteristic for the GMA welding process were taken into account. Five artificial disturbances were tested: -
a lack of shielding gas, a layer of grease on the top surface of joint, a layer of paint on the top surface of joint, too narrow root gap, too wide root gap.
The disturbances were applied two or three times at the same joint and the length of disturbance was 30 mm each time. The distance between disturbances was 40 mm. During the welding process execution, the values of selected features of recorded voltage and welding currents signal were recorded in sections of the joint, where did not appear the disturbance and in sections with introduced disturbances. Tests were made on butt weld joints made of S235 structure steel. Each test joint was 250 mm long, 90 mm wide and 5 mm thick. The angle of the V bevel was 45◦ and the width of root gap was 1.2 mm. Joints were welded by gas metal arc welding with shielding gas M21 and G382CG3Si1 wire of diameter 1.2 mm. Welding parameters were the same for all joints: welding current, 120–130 A; welding voltage, 18 V; welding speed, 266–275 mm/min; wire extension, 15 mm. Welding current and welding voltage were recorded with a speed of 4000 samples per second to detect the influence of disturbances on welding parameters. The law pass filter was applied.
4. Results and discussion Values of the features of signals are dependent on the number of measurements in the sample. In the case of the GMA welding process with short circuit metal transport through the arc these features should represent elementary phenomena happening in the arc: the creation of droplet (the time of arc burning) and the passage of the droplet from molten end of wire to welding pool (the time of the welding arc short circuit). We can assume that the minimum number of the arc burnings and short circuits which should be taken into account to calculate the features of the signal is 10. We also know that the range of short circuit time in examined files is 2.4–10.2 ms, the range of arc burning time in examined files is 4.60–45.0 ms. In this case, the size of the time window should be longer than more or less 550 ms, what
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
287
Fig. 1. Influence of time window length on values of selected features of recorded welding current and welding voltage signals. I: welding current, U: welding voltage; TJ: time of arc burning; TZ: time of short circuit; UMI = U × I; UDI = U/I; m: mean value; v: variance.
at the sampling rate 4000 samples per the second is equal to 2200 samples. Experimental results show however, that the values of the features stabilize after the slightly longer time (Fig. 1). It is also important to know what happed when the time window in which we calculate the features of signals is not filled completely by disturbance (Fig. 2). These research were executed for the lack of the shielding gas and the burn through because these defects affect the recorded signals in different way. The lack of shielding gas changes the character the welding arc burning and influences recorded signals along the whole length of disturbance. The burn through defect can affect recorded signals only at the beginning and final section of the defect. In this case, changes in recorded signals result from the abrupt change of the wire extension and connected with this transition of the arc to new equilibrium conditions at the longer wire extension. Results show that even if the time window is only partially fulfilled by disturbance, the changes of tested features are big enough to notice them (Fig. 2). For each test joint four signals were analysed: welding current (I), welding voltage (U), momentary power of arc (UMI) and momentary arc resistance (UDI). For our case, “momentary” means that values of UDI and UMI were also calculated 4000 times per second and later averaged, if necessary. Welding voltage signal was used to calculate time of short circuit
(TZ), time of arc burning (TJ) and frequency of short circuit (FZW). Statistical features of each signal (labelled in figures as m – mean value and v – variance) calculated separately for the area of weld with disturbances and the area of weld free of disturbances. Two coefficients were calculated. The first coefficient (WWBW) expresses the quotient of the mean value of statistical features in the area of weld where disturbance was applied and the mean value of the statistical features in the area of weld free of disturbance. The second coefficient (WWC) expresses the quotient of the mean value of statistical features in the area of weld where disturbance was applied and the mean value of statistical features calculated for the full length of weld. Results of that calculation are presented in Table 1. GMA is a welding process in which heat for melting the base metal and wire is transferred from an arc where electrical energy is changed to thermal energy. Monitoring of welding current and welding voltage does not completely represent the thermal processes which are essential for evaluation of welding process stability. UMI and UDI signals, which represent “the thermal element” of welding process, are included in the analysis to determine if they are more suitable for finding areas of a joint where imperfections are expected. Analysis of results presented in Table 1 shows that in most cases signals of UMI and UDI better indicate areas where disturbances were applied, especially when the WWBW coefficient is investigated.
Fig. 2. Influence of time window length and location on values of selected features of recorded welding current and welding voltage signals. v TJw: variance of time of arc burning in area of disturbance; v TJok: variance of time of arc burning in area without of disturbance; v UDIw: variance of UDI in area of disturbance; v UDIok: variance of time of arc burning in area without disturbance.
288
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
Table 1 Relative change of statistical features of tested signals in the weld area with disturbances Relative values Variance of welding voltage
Variance of welding current
Variance of arc burning time
Variance of arc short circuit time
Frequency of short circuit
Variance of UMI
Variance of UDI
Disturbance: lack of gas shield WWCa 1.85 WWBWb 4.09
1.64 2.88
0.15 0.18
1.38 4.15
1.55 2.40
1.50 2.78
2.12 307.90
Disturbance: layer of paint WWCa 1.53 WWBWb 2.04
1.39 2.96
0.66 7.63
3.25 130.80
1.06 1.01
1.00 1.91
1.52 1065.76
Disturbance: layer of grease WWCa 1.22 WWBWb 1.35
1.48 2.90
4.29 35.95
4.57 81.48
1.03 0.96
1.18 2.15
0.93 634.51
Disturbance: too narrow root gap WWCa 1.00 WWBWb 0.97
1.01 1.04
1.11 1.65
0.86 0.81
0.96 0,89
1.01 1.00
1.02 1.01
Disturbance: too wide root gap WWCa 1.37 WWBWb 1.49
2.13 4.97
4.27 70.22
6.25 326.70
0.63 0.51
1.63 2.99
2.62 1274.82
UMI: momentary power of welding arc; UDI: momentary resistance of welding arc. a WWC expresses quotient of mean value of statistical feature in the area where disturbance was applied and mean value of statistical feature calculated for all length of joint. b WWMW expresses quotient of mean value of statistical feature in the area where disturbance was applied and mean value of statistical feature in the area free of disturbance.
A different approach to analysis of signals emitted by the welding process is presented in Figs. 4–7. In this case, recorded signals are divided into samples. Each sample represents 500 recorded values (0.125 s). Mean value of statistical features of samples were smoothed by moving average method (Fig. 3). The fifth applied disturbance (too narrow root gap—lack of penetration) is not presented because no changes in tested signals were observed. That kind of imperfection probably influence neither welding arc stability nor welding parameters. It is also possible that changes in welding parameter were to small and to short in time to notice them. The four remaining disturbance influence the run of recorded signals in different way. Lack of shielding gas influences almost all calculated features of signals (Fig. 4). Layer of grease
Fig. 3. Mean value ( m) and variance ( v) of welding current (I) computed for each 1/8 of second and smoothened using moving average method; f: smoothened.
was detected but not in all three areas where grease was applied. Thickness of grease layer was gradually increased and only the thicker layer of grease was clearly revealed in plots (Fig. 5). Layer of paint, and too wide root gap similar to lack of gas are clearly seen in plots (Figs. 6 and 7). For all applied disturbances traditional signals (welding current and welding voltage) features are compared with features of UMI and UDI signals. In many cases, UMI and UDI plots clearly reveal places where disturbances were applied. UMI and UDI signals derive from welding current and welding voltage so they cannot contain more information than the source signals, but some features of signals can be amplified. It happened when both welding current and welding voltage are influenced by introduced disturbance to limited degree. No changes in UMI or UDI signals were observed when neither welding current nor welding voltage is influenced by applied disturbance. Grease and paint were applied with layers of gradually increased thickness. At the beginning of the plates one layer was applied, in the middle two layer were applied and at the and three layer were applied on surface of plates (Figs. 5 and 6). In the case of the introduction of the layer of paint as the disturbance, the changes in the course of investigated signals response to the relative thickness of the layer of paint. Investigated signals otherwise answer to the change of the thickness of the layer of grease. Only the most thicker layer of grease causes distinct changes in investigated signals. Comparing these two cases one ought to take into account that paint has a solid consistency whereas grease has a semi-fluid consistency, changing in liquid at the elevated temperature.
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
289
Fig. 4. Mean value ( m), variance ( v) of welding current (I), welding voltage (U), momentary power of arc (UMI) and momentary arc resistance (UDI). Disturbance: lack of shielding gas.
Fig. 5. Mean value ( m), variance ( v) of welding current (I), welding voltage (U), momentary power of arc (UMI) and momentary arc resistance (UDI). Disturbance: layer of grease.
Fig. 6. Mean value ( m) and variance ( v) of welding voltage (U), welding current (I), momentary power of arc (UMI) and momentary arc resistance (UDI). Disturbance: layer of paint.
Fig. 7. Mean value( m) and variance ( v) of welding voltage (U), welding current (I), momentary power of arc (UMI) and momentary arc resistance (UDI). Disturbance: too wide root opening.
290
K. Luksa / Journal of Materials Processing Technology 175 (2006) 285–290
This causes that majority of burning products of paint probably enters the atmosphere of the welding arc; however, only the part of products of burning of grease enters the atmosphere of the arc. The rest is blown away by the pressure of the arc what can be seen, observing the process of welding.
5. Conclusions The monitoring of the GMA welding process by digital recording and analysis of welding parameters enables us to detect imperfections related to phenomena that take place in welding arc and influence the stability of welding arc. There are two types of disturbances that meet the mentioned requirements: imperfections that affect the gas shield of the welding arc and disturbances that affect the length of wire extension. Lack of shielding gas, a layer of grease, a layer of paint affect the gas shield of welding arc. Oxygen, nitrogen and hydrogen as components of the air or the products of decomposition of grease and paint. They change the welding arc shield properties, influence the ionisation potential, the surface tension of molten metal, the condition of welding arc ignition, the power of welding arc and other essential conditions of arc burning. All mentioned above disturbances change the condition of arc burning as long as they are applied. Never the less some additional conditions, like liquid grease blow out should be considered. Disturbances that influence the wire extension are related to the electrode–base metal system. They change the resistance of the welding circuit. The defects caused by this type of disturbances can be unrecognised in courses of investigated parameters if the degree of the change of the wire extension is not large.
Two methods of detection of both types of imperfections were tested for this paper. Signals of welding current, welding voltage, arc power and arc resistance were tested. Artificial disturbances of welding arc stability were detected in the tested signal. Not all signals react on arc disturbance in the same manner and intensity. This means that it is better for full detection to analyse all signals for proper detection of welding imperfection, but there is a chance of identifying the imperfections by using all signals simultaneously.
References [1] [2] [3] [4] [5]
[6] [7] [8] [9]
[10] [11] [12] [13] [14]
J.F. Lancaster (Ed.), The Physics of Welding, Pergamon Press, 1984. T.B. Quinn, C. Smith, et al., Weld. J. 9 (1999) 322-s. S. Adolfsson, A. Bahrami, et al., Weld. J. 2 (1999) 59-s. M.J. Hermans, G. Den Quden, Weld. J. 4 (1999) 137-s. M.J. Hermans, G. Den Quden, Proceedings of International Conference on the Joining of Materials, JOM-8, Helsingor, Denmark, 1997, p. 312. S. Liu, T.A. Siewert, Weld. J. 2 (1989) 52-s. W. Lucas, Met. Constr. 7 (1985) 431. S.R. Gupta, P.C. Gupta, D. Rehfeldt, Weld. Rev. 4 (1966) 232. U. Dilthey, T. Reihel, W. Scheller, Proceedings of Internatonal Conference on the Joining of Materials, JOM 7, Helsingor, Denmark, 1995, p. 410. P.J. Modensi, J.H. Nixon, Weld. J. 9 (1994) 219. D. Rehfeldt, T. Schmitz, IIW-DOC (1995) 212-882-95. T. Shinoda, H. Kaneda, Y. Takeuchi, Weld. Met. Fabrication 12 (1989) 522. T. Shinoda, H. Kaneda, Y. Takeuchi, IIW-DOC (1991) XII-1214-91. T. Shinoda, H. Nishikawa, Proceedings of International Conference on the Joining of Materials, JOM 7, Helsingor, Denmark, 1995, p. 558.