Journal of Materials Processing Technology 89±90 (1999) 254±259
A sensing system for weld process control H.C. Wikle III, R.H. Zee, B.A. Chin* Materials Research and Education Center, Auburn University, 201 Ross Hall, Auburn, AL 36830, USA
Abstract The development of a point infrared sensor was undertaken in an effort to reduce the size and cost of an infrared sensing system for welding process monitoring and control. Numerical modeling, sensor development and welding process control were integral phases in the course of this effort. A numerical model of the heat transfer during autogenous arc welding was used to estimate the net heat exchange between a weldment surface and a point infrared detector as a function of sensor position about the welding arc. The gas tungsten arc welding process was monitored with a point infrared sensor. Feedback control of the infrared exchange between the weldment and the point infrared sensor was accomplished by adjusting the welding current. Welding process control was demonstrated on both constant thickness plates and plates with a step change in thickness. # 1999 Elsevier Science S.A. All rights reserved. Keywords: Welding; Penetration control; Infrared sensor; Arc; Gap
1. Introduction The development of arc welding procedures for gas tungsten arc (GTA) welding involves the speci®cation of the process parameters such as current, voltage, welding speed and shielding gas. These process parameters strongly in¯uence the ®nal weld bead geometry, microstructure, and mechanical properties of the weldment. Typically, empirical relationships between the welding parameters and the weldment properties are developed for use as guides in the selection of the process parameters necessary to obtain the desired properties. These relationships, however, are strictly valid only for the speci®c conditions under which they were determined. Unexpected perturbations encountered during the welding process can seriously affect the ®nal properties of the weldment. These considerations tend to limit the effectiveness of the empirical guidelines. In order to overcome these limitations, the implementation of sensing and control techniques into the welding process can provide means of welding under dynamic conditions. In this manner, outside perturbations can be accommodated and the desired weld properties maintained throughout the welding process. In order to consistently produce high quality welds, the desired weld bead geometry must be maintained throughout the welding pass. Accurate determination of the weld *Corresponding author. E-mail:
[email protected]
penetration depth during the welding process is dif®cult in that it cannot be easily observed. Previous investigations have shown that the weld pool penetration depth affects the surface temperature distribution of the base metal [1,2]. Various infrared sensors have been used to measure the surface temperature distribution during arc welding [3±5]. Whilst these infrared sensors have shown promise as a tool for measuring penetration depth, they have not been successfully deployed in a commercial penetration control system. High initial cost, delicate optical components and arc radiation interference have contributed to the slow implementation of this class of sensors. Recent advances in infrared detector technology, however, have brought about low cost and rugged devices that are not as susceptible to arc radiation effects. The work described within, investigates the viability of a point infrared sensor for use in controlling the GTA welding process. A numerical model of the heat transfer that occurs during arc welding was used to calculate the surface temperature ®elds of a weldment and to predict the strength of the radiative heat exchange between the weldment and a detector as a function of position about the welding arc. A point infrared sensor, based upon a thermopile detector, was used to monitor the infrared emissions at a particular location with reference to the welding torch. The net heat exchange was directly controlled, thereby indirectly controlling the weld bead penetration depth. Whilst GTA welding was the focus of this research, the results of this work are equally applicable to the other arc welding processes.
0924-0136/99/$ ± see front matter # 1999 Elsevier Science S.A. All rights reserved. PII: S 0 9 2 4 - 0 1 3 6 ( 9 9 ) 0 0 0 4 4 - 8
H.C. Wikle III et al. / Journal of Materials Processing Technology 89±90 (1999) 254±259
255
Table 1 Numerical model welding conditions Parameter
Value
Current DC voltage Welding speed
120, 150, 180 A 12 V 2.54 10ÿ3 m/sec
2. Modeling The net heat exchange between the plate surface and the sensing element was calculated by combining a numerical steady state solution to the convection±diffusion equation with the equations for thermal radiative exchange. The net exchange was calculated for the cases of variable: (1) radial position from the arc center; (2) angular displacement from the welding direction (polar angle), and (3) distance from the plate surface to the detector sensing element (target distance). The geometry and physical size of the point infrared sensor were taken into account in these calculations. The calculations were performed for the welding conditions given in Table 1. A projection of the ®eld-of-view of the point IR sensor onto the calculated surface temperature distribution is shown in Fig. 1. The sensor's viewing area could easily be repositioned about the welding arc and the corresponding infrared thermal exchange calculated for the new position. The results of these calculations indicate that the greatest exchange of heat occurs behind the welding arc, for polar
Fig. 1. Projection of the field of view of the IR sensor onto a contour plot of a solution to the convection±diffusion equation.
Fig. 2. Polar plot of net heat exchange for a sensing height L20.6 mm. Greatest exchange occurs for the smallest radial distance.
angles in the range 90 < < 270 . A polar plot of the case where the target distance is 20.6 mm is shown in Fig. 2. The outermost curve corresponds to a radial position of 12.5 mm progressing inward, 17.5, 25.0, 37.5 and 50 mm radial position. This polar representation of the data clearly shows the angular dependence of the heat exchange between the weldment and the sensor. The forward directions show very little heat exchange with the sensing element. The results of these calculations indicate that the greatest heat exchange between the weldment surface and the thermopile detector occurs nearest to the welding arc in the rearwards direction. Whilst this is the intuitive choice for the greatest sensor response, positioning the sensor in a rearwards location is counter-productive for use in a welding process control system. As the sensor is positioned increasingly further behind the arc, the time lag between the phenomena that are occurring in the weld pool and the recognition of that phenomena increases. Fig. 3 shows the calculated time required to reach 95% of steady state net heat exchange as a function of polar angle for different radial positions. The shortest times are in front of the arc center whilst the longest are directly behind the arc center. Close inspection of all of the data shows that at a polar angle of 908 for small radial displacements, the net heat exchange and thus the sensor signal, begins to rapidly increase. This position is more acceptable for sensor placement during welding process control since it is directly perpendicular to the welding direction through the center of the welding arc. This eliminates any time lag due to positioning in the welding direction (x-direction with reference to the welding model) and leaves only the unavoidable lag due to displacement in the perpendicular direction (ydirection).
256
H.C. Wikle III et al. / Journal of Materials Processing Technology 89±90 (1999) 254±259
Fig. 3. Time required for heat exchange to reach 95% steady-state after a step change in heat input of 1.4±2.1 kW.
3. Control Bead-on-plate GTA welds were performed on 6.4 mm thick plain carbon steel plate using the welding parameters shown in Table 2. A test ®xture, shown in Fig. 4, was constructed to form the welding platform and support for the welding torch. Arc initiation, torch travel, sensor data acquisition and welding process control were managed from a personal computer. Control of the infrared exchange between the weldment and the sensor was accomplished by adjusting the welding current during welding on plates of constant thickness and plates with a step change in thickness. A step change in heat input was used to investigate the response of the point IR sensor as a function of radial position and polar angle about the welding arc. The signal gain was measured for each step change in heat input. A step change in the welding current from 131 amps (35%) to 188 amps (50%) was performed and the sensor response measured. The step change in welding current took place 100 s after the start of welding, long enough to ensure that steady state had been achieved after the start of the weld. The welding process was allowed to continue until well after the sensor signal had attained a new steady-state value at the increased welding current. The sensor output measurements
Fig. 4. The test fixture for welding process control. The fixture is capable of two axes of motion, one vertical and one horizontal.
were then taken from the average steady-state values before and after the step, as shown in Fig. 5. This experiment was performed for a series of polar angles about the welding arc for a speci®ed radial distance. The polar angle of 08 was in the welding direction. This experiment was repeated for a series of sensing element radial displacements of 17.5, 20.7, and 23.9 mm. The results of these experiments are shown in Fig. 6. The trend shown in the experimental results of Fig. 6 agrees with the calculated results from the heat transfer
Table 2 Experimental welding conditions Parameter
Value
Current DC voltage Welding speed Shielding gas Base metal
120±200 A approximately 12 V 2.54 10ÿ3 m/sec Ar, 25 cfm AISI 1020
Fig. 5. Step change in heat input during the welding process. The step occurs at 100 s after the start of welding.
H.C. Wikle III et al. / Journal of Materials Processing Technology 89±90 (1999) 254±259
257
Fig. 6. Measured signal gain as a function of the polar angle displacement of the sensor about the welding arc.
Fig. 7. Set-point (SP), sensor signal (PV), and welding current (MV) for a weld made on a constant thickness plate under computer control.
model. For all cases, the closest radial displacement, 17.5 mm, exhibits the largest signal gain. The polar angle of the sensor placement, however, is seen to have the greatest effect on the measured sensor signal gain. For polar angles of less than 908, the increase in signal gain is negligible. At the polar angle of 908, the gain begins to increase slightly. For polar angles that give positions behind the welding arc, large changes in the signal gain are seen to occur. The measured results for a polar angle of 1808 were not plotted in Fig. 6. This is because of the intense heat coming from the plate saturated the sensor signal in every case and no useful information was obtained at these positions. Both the modeling efforts and the experimental results indicate that the placement of the sensor behind the welding arc shows the greatest sensitivity to process variations. The sensitivity of the sensor is very poor for sensor positions ahead of the welding arc. For radial positions near to the welding arc, the placement of the sensor at a polar angle of 908 shows a slight increase in the radiative heat exchange, see Fig. 2, and the sensor signal gain, see Fig. 6. Whilst this position does not appear to give the greatest sensitivity to changes that occur in the weld pool, it does appear to offer the best compromise between sensitivity and response time. Referring back to Fig. 3, the time required to reach steady state is expected to increase with the polar angle. Close inspection of Fig. 3 though, shows that an in¯ection point occurs near to the polar angle of 908. Based on these results, a polar angle of 908 and a radial displacement of 17.5 mm was selected as the position of the detector for all further experiments. Feedback control of the infrared thermal exchange was implemented using the developed point infrared sensor to monitor the welding process. The control system was tested on two different plate geometries, plates of constant thick-
ness and plates with a step change in thickness. A proportional±integral (PI) controller was used to maintain the net heat exchange about a speci®ed set-point. The welding current was adjusted automatically by the computer according to the output of a PI controller algorithm. Initial attempts at controlling the welding process were made with plates of constant 6.4 mm thickness. A step change in the set point was used to induce a change in the welding conditions. An initial welding current of 150 amps and a travel speed of 6.35 10ÿ3 m/s were used. Automatic control of the welding current was typically started after 76 mm of torch travel. This allowed the temperature distribution to reach steady state before automatic control was initiated. Automatic control would then remain in effect until the end of the weld. The set point (SP), sensor signal (PV) and the control signal (MV) are shown for a weld on a constant thickness plate in Fig. 7. A step change in SP was made at 152 mm. Oscillations in PV and MV can be seen at the initiation of control and again at the step change in SP. A photograph of the plate is shown in Fig. 8. The temper lines on the surface of the plate show that the controller increased the welding current to bring PV up to SP. Step changes in plate thickness were used to induce disturbances to the welding process. A 3.2 mm step change in plate thickness, 76.2 mm inches in length, was milled into the bottom surface of a 6.4 mm thick mild steel plate approximately 165 mm from the end of the plate. This con®guration gave two step changes in thickness, 6.4± 3.2 mm and then back to 6.4 mm thick. GTA welds were made with an initial welding current of 150 amps and a travel speed of 6.35 10ÿ3 m/s. Automatic computer control of the welding current was typically started after 76 mm of torch travel. This allowed the temperature distribution to reach steady state before automatic control was initiated.
258
H.C. Wikle III et al. / Journal of Materials Processing Technology 89±90 (1999) 254±259
Fig. 8. Photograph of a controlled weld made on a plate with constant thickness: (a) top side of plate; (b) bottom side of plate.
Automatic control would then remain in effect until the end of the weld. The set-point, the sensor signal and the control signal are shown for a weld on a step plate in Fig. 9. At the ®rst step
Fig. 9. Set-point (SP), sensor signal (PV), and welding current (MV) during the control of the heat exchange for a plate with a 3.2 mm (0.125") step change in plate thickness.
Fig. 10. Photograph of the front and back sides of a step plate in which control action was implemented. The welding progressed from left to right in the photograph.
change in thickness, the controller reduced the welding current in order to maintain a constant sensor signal. At the second step change, the controller increased the welding current. A damped oscillation can be seen in both the PV and MV curves at the point where automatic control was initiated and again at both of the step changes in thickness. A photograph of the plate with a 3.2 mm step change is shown in Fig. 10. Note that the controller prevented the weld bead from burning through the plate in the thin section. In addition, the width of the temper lines on the surface of the plate was maintained at a constant width throughout the welding process, indicating that the surface temperature distribution was maintained constant. For all possible positions of the infrared sensor, time lags are present in the measurement system. This was because the sensor viewed an area that was not exactly at the arc center. The radial displacement of the sensor was never equal to zero. This was due to the physical size limitations that were present in the welding and measurement system. The tungsten electrode where the arc is established is surrounded by a gas cup, which measured 19.1 mm diameter. Due to the intense heat surrounding the welding arc, a cooling system
H.C. Wikle III et al. / Journal of Materials Processing Technology 89±90 (1999) 254±259
was required to maintain the thermopile detector at a constant temperature. The diameter of the thermopile detector was 8.2 mm whilst the outside diameter of the cooling jacket was 15.9 mm in diameter. These dimensions allow a minimum possible radial displacement of 17.5 mm. 4. Conclusions The heat-transfer model was capable of predicting the trends in the surface temperature ®eld and net heat exchange between a differential surface element (detector) and a circular area on the surface of the weldment. The results of these predictions were used to pre-determine the placement of an infrared sensor near to the welding arc and give an estimation of the sensor response. Polar angles 908 were calculated to give the greatest sensor response. A feed-back control system was developed which maintained a constant heat exchange between the weldment surface and the point
259
infrared sensor during a welding pass. The system was demonstrated on both plates of constant thickness and plates with a step change in thickness. References [1] N. Christensen, V de L. Davies, K. Gjermundsen, The distribution of temperature in arc welding, Brit. Welding J. 12 (1965) 54±75. [2] B.A. Chin, N.H. Madsen, J.S. Goodling, Infrared thermography for sensing the arc welding process, Welding J. 62(9) (1983) 227±234. [3] E.R. Bangs, Infrared signature analysis: real time monitoring of manufacturing processes, in: R.D. Lucier (Ed.), Thermosense X. Proc. SPIE 934 Int. Conf. on Thermal Infrared Sensing for Diagnostics and Control, 1988, pp. 111±119. [4] P.W. Ramsey, J.J. Chyle, J.N. Kuhr, P.S. Myers, M. Weiss, W. Groth, Infrared temperature sensing systems for automatic fusion welding, Welding J. 42(8), 337±346. [5] J.-B. Song, D.E. Hardt. 1990. Estimation of weld bead depth for inprocess control, in: K. Danai, S. Malkin (Eds.), Automation of Manufacturing Processes, New York, ASME, 1990, pp. 39±45.