Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 48

Penetration Depth Monitoring and Control in Submerged Arc

The partial penetration depth in the submerged process is modeled and feedback controlled based on the base metal current

BY X. R. LI, Y. M. ZHANG, AND L. KVIDAHL

monitoring and control in SAW for ship- ABSTRACT building can be appreciated and demon- strated through its capability in facilitating Submerged arc welding (SAW) is known for its high productivity. However, there is the so-called “no backgouge” method. a lack of capability to monitor and control weld penetration. Because penetration is be- This applies to the automated, two-sided lieved to be primarily determined by base metal current, a butt joint welding processes such as panel- (GMAW) gun is added into the SAW process to bypass part of the total current. The line assemblies where plates are welded base metal current that controls weld penetration is directly reduced, and the ability to using SAW tractors from both sides to en- adjust the base metal current to control weld penetration without reducing deposition sure complete joint penetration and ac- rate is introduced into SAW. To conveniently monitor weld penetration and acquire the ceptable weld quality. Two-sided welding WELDING RESEARCH needed feedback for weld penetration control, welding parameters and conditions af- in shipbuilding applications involves weld- fecting weld penetration were analyzed and specific variables subject to variation and ing from one side, flipping the plate, and fluctuation were identified. Experiments were conducted to see what parameters affect then completing the weld from the other the weld penetration and what their significances are. It was found that the base metal side. current is the dominant parameter that determines weld penetration with a sufficient For applications that require complete accuracy when other major parameters are in their stated ranges. A control system has joint penetration, the key to two-sided been established to monitor and control weld penetration using a proportional integral welding is ensuring that the “backside” derivative (PID) control algorithm. This algorithm is based on penetration feedback weld completely penetrates the joint and provided by the penetration model that correlates weld penetration depth to base metal fuses into the weld deposited on the op- current. Experiments on DH36 butt joints verified the effectiveness of the pro- posite side. In most cases, complete joint posed method. penetration is assured by backgouging prior to welding of the opposite side (or stant voltage (CV) mode to balance the backside weld) (Refs. 3, 4). The elimina- Introduction melting of the wire with its given feeding tion of backgouging could significantly speed. While the deposition rate and wire save time and cost through the following: Submerged arc welding (SAW) is a balance are controlled, the melting cur- 1) eliminating the steps associated with major process used to join ship structures, rent is subject to change. Because the the backgouging process, and 2) reducing and its efficiency plays a significant role in melting current is the same as the base the volume of weld metal needed to com- determining the total ownership costs of metal current, which controls weld pene- plete the backside weld. Unfortunately, ships. Unlike other arc welding processes, tration, the resultant weld penetration can this so-called no backgouge method is ap- in SAW the arc and molten weld metal are vary with welding conditions that may af- proved in limited cases, and the major shielded by a covering envelope of molten fect the melting-feeding balance of the issue limiting its use is due to the inability and a layer of unfused granular flux wire causing the CV power supply to to reliably control weld penetration. If one particles (Refs. 1, 2). This unique ap- change the current. Efforts are needed to could accurately control weld penetration proach for shielding allows the use of large control welding conditions such as root from the two sides, one would be able to welding currents without spatter and high opening, joint geometry, and contact tube- extend the use of no backgouge proce- travel speeds without exposing high-tem- to-work distance (CTWD) within certain dures to greater component thicknesses. perature liquid metal to the surrounding ranges. However, the resultant weld pen- Sensing and control of weld penetra- atmosphere. Thus, quality welds can be etration is still not ensured. tion are critical variables for the competi- made at high travel speeds. The significance of weld penetration tive next-generation manufacturing indus- However, in current practices with try. Unfortunately, the penetration depth SAW, the power supply is set to the con- in partial penetration weld applications is not visible, and the weld bead on the back- X. R. LI is with Adaptive Intelligent Systems, KEYWORDS side of the workpiece is not visible from LLC, Lexington, Ky. Y. M. ZHANG the front side. Monitoring weld penetra- ([email protected];yuming.zhang@uk Submerged Arc Welding (SAW) tion, either the penetration depth or weld y.edu) is with Adaptive Intelligent Systems, LLC, Weld Penetration and the University of Kentucky Institute for Sus- bead on the backside of the workpiece, is Modeling challenging. tainable Manufacturing and Department of Elec- Control trical and Computer Engineering. L. KVIDAHL Despite the difficulties, a number of is with Huntington Ingalls Industries, Pascagoula, Proportional Integral methods have been proposed to detect Miss. Derivative (PID) weld penetration. For SAW, methods that DH36 are based on direct observation of the Presented during the AWS Professional Program electric arc or weld pool such as camera at FABTECH 2012, Las Vegas, Nev.

48-s FEBRUARY 2013, VOL. 92 Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 49

Fig. 1 — Bypass SAW process block diagram. Fig. 2 — Block diagram of automatic welding system.

Fig. 3 — Bypass SAW platform. Fig. 4 — Installation of bypass GMAW gun.

and pool oscillation are apparently not ing weld penetration without reducing verify the effectiveness of the proposed suitable. Other methods that have been deposition rate. In-situ testing and data method and system. WELDING RESEARCH explored for SAW, which include ultra- analysis were performed to model the sonic penetration sensors (Refs. 5, 6), in- depth of the weld penetration and corre- Process and System frared camera-based sensors (Refs. 7–9), late the penetration depth to the arc sig- and numerical analysis-based methods nals. Weld penetration monitoring and Principle (Refs. 10, 11). However, to facilitate a subsequent control system were estab- method that is more suitable for a manu- lished with feasibility verified by prelimi- A modified SAW process that allows facturing environment, it may be pre- nary experiments. Finally, closed-loop the base metal current to be adjusted with- ferred if only arc signals can be conve- control experiments were conducted to out reducing the deposition rate is intro- niently measured. This paper proposes a method to mon- itor and control the depth of weld pene- tration using SAW, which can be conve- Table 1 — Welding Conditions for Modeling Experiments niently implemented in a manufacturing Parameters Unit Value environment. At first, the double-elec- trode bypass method that has previously Material AISI 1018 been studied for gas metal arc welding Butt joint root opening in. 0 3 (GMAW) with an added gas tungsten arc CTWD in. ⁄4 welding (GTAW) bypass torch (Ref. 12) Travel speed in./min 20 and GMAW bypass welding gun (Refs. 13, Main voltage V 30.0 1 14), respectively, is introduced to the SAW Main wire ⁄8 in. (3.2 mm) AWS A5.17 to provide the ability to change the base Bypass voltage V 30.0 metal current. It is believed the base metal Bypass wire 0.045 in. (1.14 mm) AWS A5.18 current is a major parameter in determin-

WELDING JOURNAL 49-s Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 50

Fig. 5 — Bypass SAW control system (left: control box; right: initial parameter input screen on HMI terminal touchscreen).

Fig. 6 — Model accuracy for group 1.

welding processes, ship becomes as follows: including SAW with a consumable It = Ibm + Ibp (1) wire as a terminal WELDING RESEARCH of the arc, the total As a result, the total current can be set current It that melts large enough to achieve the needed depo- the main wire is the sition rate, while the bypass current can be same as the base well controlled to achieve the desired base metal current Ibm metal current to produce the required that determines the penetration. As a result, two control vari- penetration on the ables, the wire feed speeds of the main Fig. 7 — Model accuracy for group 2. workpiece. The wire and bypass wire, can be adjusted to deposition rate is produce the two outputs, weld penetration proportional to and deposition rate, at their desired val- duced by adding a GMAW gun into a con- Ibm, since Ibm = It. If a partial penetration ues. The required controllability for weld ventional SAW process as illustrated in of specific depth is needed, the deposition penetration without reducing the deposi- Fig. 1. Two power supplies are used to rate will have to change accordingly. With tion rate is provided. power the SAW gun and added GMAW the introduction of the bypass GMAW The principle of the system established gun, separately. In all consumable arc gun and bypass current Ibm, the relation- in this study is to implement the proposed

Table 2 — Measurements from Identification Experiments

No. IIbm Ibp P WFSm Wm WFSbp Wbp Wt (A) (A) (A) (in.) (in./min) (lb/h) (in./min) (lb/h) (lb/h)

1 466.23 301 165.23 0.157 91 18.63 322 8.55 27.18 2 464.23 279 185.23 0.111 91 18.63 371 9.84 28.47 3 464.23 264 200.23 0.084 91 18.63 408 10.81 29.44 4 464.23 251 213.23 0.074 91 18.63 439 11.65 30.28 5 464.23 237 227.23 0.071 91 18.63 473 12.55 31.18 6 462.23 245 217.23 0.058 91 18.63 448 11.91 30.53 7 586.23 431 155.23 0.222 101 20.67 298 7.91 28.58 8 584.23 406 178.23 0.208 101 20.67 354 9.39 30.06 9 597.23 397 200.23 0.208 101 20.67 407 10.81 31.48 10 600.23 414 186.23 0.207 101 20.67 373 9.91 30.58 11 611.23 393 218.23 0.199 101 20.67 451 11.97 32.67 12 621.23 395 226.23 0.181 101 20.67 470 12.49 33.16 13 692.23 577 115.23 0.346 112 22.9 201 5.33 28.23 14 647.23 491 156.23 0.280 112 22.9 301 7.97 30.87 15 638.23 458 180.23 0.265 112 22.9 359 9.52 32.42 16 632.23 430 202.23 0.222 112 22.9 412 10.94 33.84 17 626.23 399 227.23 0.214 112 22.9 473 12.55 35.45 18 632.23 388 244.23 0.208 112 22.9 514 13.65 36.55

50-s FEBRUARY 2013, VOL. 92 Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 51

Fig. 9 — Block diagram of control system.

Fig. 8 — Model accuracy for group 3. Fig. 10 — Weld bead made from experiment for ¼-in. penetration depth.

process and control shown in Fig. 2. The protection. 4. The bypass welding gun was powered by embedded controller is the core of the The bypass GMAW gun was fixed to a Thermal Arc Powermaster 500 CC/CV control system. It is interfaced with the the tractor with an approximate 45-deg welding machine, which is also operated process through a number of isolation angle with the SAW gun, as shown in Fig. under CV mode. The AWS A5.18 ER70S- modules. A human machine interface (HMI) terminal is used by the operator to input initial welding parameters. The op- tional data acquisition module is used only Table 3 — RMSEs for the Four Models when it is necessary to record the on-line Model RMSE (in.) Regression variables/number of parameters measurements. All the components of the control system are installed in a portable Model 1 0.0112 base metal current and deposition rates/4 control enclosure. Both power supplies Model 2 0.0124 base metal current/2 are operated under CV mode. For each Model 3 0.0206 deposition rates/3 wire feeder, the wire feed speed command Model 4 0.0122 base metal current, square of base metal current/3 signal is provided by the embedded con- troller via the output isolation module. Two current sensors are used to measure 1 the base metal current and bypass current. Table 4 — Welding Parameters for ⁄4-in. Penetration The measurements from sensors are di- rectly connected to the input isolation Parameters Unit Parameter Value modules. Travel speed in./min 20 WELDING RESEARCH 3 CTWD in. ⁄4 Experimental Setup Main voltage V 30 Total current Follow main wire speed An LT-7 tractor from Lincoln Electric Main wire 0.125 in. (3.175 mm) AWS A5.17 was used to perform SAW in this study and Main wire feed speed in./min 70 is shown together with the bypass GMAW Bypass voltage V 30 gun in Fig. 3. A Miller Deltaweld 652 Bypass current Follow bypass wire feeding speed CV/DC welding machine was used to Bypass wire 0.045 in. (1.14 mm) AWS A5.18 power the SAW gun under constant volt- Initial bypass wire feed speed in./min 200 age (CV) mode. With a preset arc voltage, Desired base metal current A 436 the welding current can be changed with the wire feeding speed setting. The AWS 1 8 A5.17 EM12K wire with ⁄ -in. (3.18-mm) 1 Table 5 — Experimental Results for ⁄4-in. (0.25-in.) Penetration Depth diameter from Lincoln Electric was cho- sen as the consumable wire for SAW. It is Point # I I I Penetration Error fed by the LT-7 tractor, with adjustable bm bp A A A in. % wire feeding speed denoted as WFSm. Lin- colnweld 882 flux was used to protect the 1 438 117 555 0.2449 –2.05 submerged arc and weld pool. The lower 2 436 123 559 0.2500 0 end of the flux hopper merges with the tip 3 432 119 551 0.2571 2.83 of the SAW gun in order to supply flux 4 434 113 547 0.2421 –3.15 when the gun is moving and make sure 5 438 112 550 0.2457 –1.73 that the electric arc is protected from sur- 6 434 116 550 0.2512 0.47 rounding atmosphere with the flux 7 435 116 551 0.2520 0.79

WELDING JOURNAL 51-s Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 52

Fig. 12 — Completed weld bead for ¼-in. penetration depth on DH36.

3 Fig. 13 — Completed weld bead for ⁄8-in. penetration test on DH36.

6 wire with 0.045-in. be determined by the based metal current (1.14-mm) diameter with sufficient accuracy. The ability to from Kobelco was se- control the weld penetration is established lected as the bypass wire. using the bypass SAW. To this end, all A Miller S-74D wire major parameters affecting penetration feeder was used to supply depth would have to be taken into consid-

WELDING RESEARCH the bypass wire at a wire eration first. feeding speed of WFSbp. A number of studies have been de- Fig. 11 — Measured welding parameters with ¼-in. penetration control. A bypass SAW process voted to modeling the SAW process (Refs. controller was installed 16–20). Based on these studies, a compre- based on the universal hensive model is proposed to correlate the welding process control depth of the weld penetration to a number system developed by Adaptive Intelligent of welding parameters as the regression Systems, LLC, Lexington, Ky. (Ref. 15) The variables: Table 6 — Alloy Composition (wt-%) of DH36 controller was used to monitor and control Steel (Ref. 23) various welding parameters, such as weld- P = f(Ibm ,Wt ,G, CTWD, S) (2) ing current, arc voltage, wire feeding speed, Alloy Material Wt-% etc. A HMI terminal was used to enhance Here, P is the depth of partial penetra- C 0.18 the communication between and tion weld (in.), Ibm the base metal current Mn 0.90–1.60 process. The controller and HMI screen are (A), Wt the total deposition rate (lb/h), G Si 0.10–0.50 shown in Fig. 5. the root opening (in.), CTWD the contact S 0.035 tip-to-work distance (in.), and S the travel P 0.035 Modeling speed (in./min). Al 0.015 min In the welding parameters included in Nb 0.02–0.05 Regression Variables Equation 2, CTWD and S can be accu- V 0.05–0.10 rately controlled, and G may be controlled Ti 0.02 As discussed above, the bypass SAW with a certain range. The approach is to Cu 0.35 introduced an ability to control both the simplify the model into the following Cr 0.20 base metal current and deposition rate. In form: Ni 0.40 this section, the authors will demonstrate Mo 0.08 that the depth of the weld penetration can P = f(Ibm ,Wt), (3)

identify the model under given CTWD 1 Table 7 — Welding Parameters for ⁄4-in. Penetration Test on DH36 and S and nominal G, and then examine how the accuracy may be affected by G Parameters Unit Value when it is in the tolerant range. It is ap- parent that Material DH 36 Root opening in. 0 W = W + W (4) Travel speed in./min 20 t m bp 3 CTWD in. ⁄4 Here, W and W are the deposition Main voltage V 30.0 m bp Total current A Follow main wire speed rates from the main (SAW) wire and by- 1 pass wire, respectively. In general, a depo- Main wire ⁄8-in. (3.17-mm) AWS A5.17 Main wire speed in./min 70 sition rate W can be calculated from its Bypass voltage V 30.0 wire feed speed WFS for a steel wire in the Bypass current A Follow bypass wire speed following expression: Bypass wire 0.045-in. (1.14-mm) AWS A5.18 Initial bypass wire speed in./min 200 W = 13.1 × D2 × WFS × EE (5) Desired base metal current A 470

52-s FEBRUARY 2013, VOL. 92 Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 53

Here, D is the diameter of the wire (in.), WFS the wire feed speed (in./min), and EE the efficiency that is considered 100% for a solid wire. The con- stant of 13.1 is due to the density and units used. Hence,

Wt = Wm + Wbp = 13.1 2 2 × (D m × WFSm + D bp × WFSbp) (6)

The proposed model becomes

P = f(Ibm, Wm, Wbp)(7) Fig. 14 — Weld bead profiles for ¼-in. partial penetration test etchants. Identification Experiments

To provide sufficient variations for the regression parameters in the model, three groups of butt joint welding experiments were conducted on ½-in.-thick AISI 1018 plates under conditions in Table 1. In the butt joint welding experiments, no root openings were set intentionally. For each experiment, WFSm was manually set as a constant on the panel of the SAW tractor, and there were variations in the total current from experiment to experi-

3 ment with the same nominal WFSm. Fig. 15 — Weld bead profiles for ⁄8-in. partial penetration test etchants. Within each group, WFSbp was changed in a relatively large range and bypass current changed accordingly. After experiments, spectively. Index i indicates the ith sample specimens were cut in 1-in. intervals to in the given experiment group, and n is the measure the depth of the weld penetra- 2 n ⎛  ⎞ number of samples from the given experi- tion. Table 2 lists all the measured experi- ∑ ⎜ pP− ⎟ i=1⎝ li⎠ ment group. The resultant RMSEs are mental data, including the regression vari- RMSE =(9)listed in Table 3. ables/model inputs and model output. ∧ n When the number of parameters is the where p and p denote the measured and same, a model with a smaller RMSE Model Identification model-estimated penetration depths, re- should be selected. Model 3 is thus elimi- Four tentative model structures were

proposed in Equation 8. 1 Table 8 — Measured Welding Current and Penetration Depth for ⁄4-in. Penetration Test on DH36 ⎧ Pa=+ aI + aWaWModel + (1) ⎪ 01bm 2 m 3 bp Point # Ibm Ibp It Penetration Error ⎪ ⎪Pa=+ aI (2)Model A A A in. % WELDING RESEARCH ⎨ 01bm = ++ ⎪Pa aW aW (3)(Model 88) 1 480 63 543 0.2547 1.89 ⎪ 00 1mbp 2 2 470 66 536 0.2516 0.63 ⎪Pa=+ aI + aI2 (Modell 4) ⎩ 01bm 2 bm 3 470 63 533 0.2551 2.05 4 465 63 528 0.2539 1.57 Model 1 includes the base metal current 5 469 63 532 0.2539 1.57 and deposition rates as regression parame- ters/model inputs; Models 2 and 3 only con- 3 sider either the base metal current or depo- Table 9 — Welding Parameters for ⁄8-in. Penetration Test on DH36 sition rate, respectively; and Model 4 contains linear and quadratic equations Parameters Unit Value representing the base metal current. The Least Squares method (Ref. 21) Material DH 36 Root opening in. 0 was used to estimate model coefficients Travel speed in./min 20 for each given model structure in Equa- 3 CTWD in. ⁄4 tion 8 using experimental data in Table 2. Main voltage V 30.0 The measured output was compared with Total current A Follow main wire speed 1 the model fitted penetration depth in Figs. Main wire ⁄8-in. (3.17-mm) AWS A5.17 6–8 for each of the three experiment Main wire speed in./min 90 groups. Bypass voltage V 30.0 To evaluate model performances, the Bypass current A Follow bypass wire speed root mean square error (RMSE) was used Bypass wire 0.045-in. (1.14-mm) AWS A5.18 for each model: Initial bypass wire speed in./min 200 Desired base metal current A 570

WELDING JOURNAL 53-s Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 54

plies, wire feeders, and resultant arcs. The 3 Table 10 — Measured Welding Currents and Penetration Depth for ⁄8-in. Penetration Test on travel speed, main wire feed speed, and DH36 other constant parameters are also applied to the process. However, only the bypass Point # Ibm Ibp I Penetration Error wire feed speed is manipulated as the con- A A A in. % trol variable of the process. The outputs of the process contain all variables from the 1 563 106 669 0.3772 0.58 bypass SAW process and can be recorded 2 568 104 672 0.3709 –1.10 for off-line analysis, but only the base metal 3 575 102 677 0.3752 0.05 current is used as the feedback. 4 576 107 683 0.3728 –0.58 The PID controller was implemented by 5 561 102 663 0.3748 –0.05 the embedded control system introduced in the process and system section. The A/D port of controller samples the current sig- Table 11 — Experimental Results from the Joint Root Opening Test nals from the current sensors at 100 Hz (100 samples per second per channel). The con- trol period was empirically selected to be 0.5 Point # Ibm Ibp I Penetration Error s. The average of the measurements of a A A A in. % particular signal in a control period was used as the measurement/feedback of this 1 434 78 512 0.2512 0.47 signal in this control period. 2 438 78 516 0.2539 1.57 To demonstrate the effectiveness of the 3 448 80 528 0.2547 1.89 proposed monitoring and control method, 4 436 89 525 0.2528 1.10 5 443 92 535 0.2449 –2.05 closed-loop control verification experi- ments were conducted by butt joint weld-

WELDING RESEARCH ing ½-in.-thick AISI 1018 carbon steel

5 Table 12 — Experimental Results from ⁄8-in. CTWD Test plates with the targeted depth of penetra- tion at ¼ in. (50% of whole thickness). AISI 1018 carbon steel was selected be- Point # Ibm Ibp I Penetration Error A A A in. % cause of its combination of typical traits of steel: strength, ductility, and comparative 1 432 110 542 0.2472 –1.10 ease of . Chemically, it is very 2 438 105 543 0.2504 0.16 similar to A36 hot rolled steel, but the cold 3 438 112 550 0.2480 –0.79 rolling process creates a better surface fin- 4 434 109 543 0.2449 –2.05 ish and better properties. Its good weld- ability and low cost especially make it an appropriate choice for extensive experi- nated. As the number of parameters in- mined and then set at a constant level (30 ments based on process and method de- creases, the resultant reduction in RMSE V, for example); and 3) main wire feed velopments, as in this paper. must be significant. The significance may speed appropriate to the needed penetra- In all the preliminary experiments for be examined using F-test (Ref. 22) with a tion level should be determined and then verification, two ½-in.-thick by 2-in.-wide given confidence level α. For α = 0.05, F set at a constant. Reduced ranges of these by 24-in.-long AISI 1018 plates were = 0.5 and 1.6 from Models 2 to 3 and from parameters/variables in general should tacked with no intentional root opening. Models 2 to 1, respectively. Both these F improve the accuracy of local models over Square butt joints were welded with no values are far from being significant. the global model established on wide grooves at the flat position. The SAW trac- Model 2 is selected. Because of the small ranges of these parameters/variables. tor traveled along the experimental track RMSE, the resultant model (model 2) with the welding gun in line with the joint. Feedback Control and The optimal values of constant param- P = –0.1258 + 0.0008Ibm (10) Verification Experiment eters were determined based on prelimi- nary experiments. For ¼-in. (50%) partial is considered to exhibit the best accuracy. The principle of the proposed penetra- penetration control, multiple tests with tion control system is illustrated in Fig. 9. the optimal values for the constant pa- Local Modeling In this system, only the bypass wire feed rameters were conducted. These tests not speed is adjusted to control the penetra- only produced more samples for further The model identified above is a global tion depth/base metal current. If the dep- analysis, but also proved the stability and model. The modeling accuracy may be osition rate and weld penetration both consistency of the proposed bypass SAW better assured if models are identified for need to be accurately controlled, the main method. The optimal values for constant different penetration levels of concern, wire speed should be adjusted together welding parameters for targeted ¼-in. such as 50% and 75% of the full plate with the bypass wire feed speed. penetration are listed in Table 4. thickness. This is because a specific pene- In Fig. 9, the penetration model calcu- The travel speed of 20 in./min is accept- tration level can narrow down many pa- lates/estimates the depth of weld penetra- able for butt joint welding in productivity rameters/variables: 1) Root opening, tion P as the feedback. Its difference with lines, but not too fast to add difficulties for CTWD, and travel speed appropriate to the desired depth of weld penetration P* is penetration. The voltage settings and diam- the needed penetration level should be de- used as the input of the proportional inte- eter of the main wire (0.125 in.) were deter- termined and then set to the correspon- gral derivative (PID) control algorithm. mined based on typical shipbuilding appli- ding constants; 2) arc voltage for main and The output of the PID control algorithm is cations. The bypass wire diameter (0.045 bypass power supply appropriate to the the change of the bypass wire feed speed in.) is typical for GMAW applications/guns. Δ needed penetration level should be deter- WFSbp. The process includes power sup- The initial value for the bypass wire feed

54-s FEBRUARY 2013, VOL. 92 Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 55

speed was used before the base metal cur- DH-36 Experiment Results section sample are listed in Table 8. rent feedback became available. and Analysis From the measurement of partial pen- Experiments have been conducted to etration depth, it can be seen that the max- identify the local model corresponding to In order to prove the effectiveness of imum error for ¼-in. penetration is 0.051 the 50% penetration level using the ex- the partial penetration control method, ½- in. (2.05% of desired penetration depth), perimental conditions given in Table 5. in.-thick DH36 plates were welded to which is far less than the maximum error 1 The resultant local model is achieve a desired partial penetration of ¼ of ⁄16 in. acceptable to shipyards. 3 in. (50%) and ⁄8 in. (75%) targeting ship- 3 P = –0.0361 + 0.000656Ibm (11) building applications. The two penetra- Partial Penetration at ⁄8 in. (75%) tion depths are sufficient to avoid back Based on this local model, the desired gouging. If specific penetration depths are Following the same procedure, the fol- 3 base metal current is 436 A in order to required for future applications, another lowing local model for ⁄8-in. partial pene- produce ¼-in. penetration depth on the set of welding and control parameters can tration was obtained: ½-in.-thick plates. Figure 10 shows the be given to achieve those levels. resultant butt joint weld. As can be ob- P = –0.4942 + 0.001312 × Ibm (13) served, the weld made is consistent and Material 3 smooth after brushing off the solidified By similar calculation, for ⁄8-in. pene- flux. The recorded currents are plotted As extensively used in ship structures, tration depth, the required base metal cur- in Fig. 11. DH36 plates were selected to demonstrate rent should be 570 A. Other welding pa- Figure 11 includes the bypass wire the feasibility of the proposed method in rameters are listed in Table 9. With feed speed command signal. Before t = practical applications. The chemical com- increased weld penetration, the main wire 3 s approximately, the bypass wire feed position of DH36 is listed in Table 6. feeding speed was increased from 70 to 90 speed command signal is 0. This is be- All the welding experiments were car- in./min, in order to provide larger welding cause the bypass wire feed system was ried out on square butt joints in the flat po- current and higher deposition rate. switched on manually after the main arc sition. Two pieces of DH36 plates with ½ in. Similarly consistent weld and results was established. The bypass wire feed thickness × 3 in. width × 24 in. length were were obtained as shown in Fig. 13 and speed command signal was then adjusted butt joint welded with zero root opening. Table 10. from its initial value of 200 in./min as de- Although the root opening may be slightly By examining the results, the maximum 3 termined by the PID control algorithm. increased during the welding process, it error for ⁄8-in. partial penetration is 0.0041 As a result, the base metal current is con- won’t affect the control accuracy as learned in. (1.10% of required penetration depth), trolled at its desired level, 436 A, ap- in previous experiments and will be demon- which is also inside the tolerated error of 1 proximately. strated in this section. The SAW tractor ⁄16 in. The completed weldment was then traveled along the experimental track with cut into 1-in. segments to measure the re- the GMAW gun in line with the joint. Weld Profile sultant depth of the weld penetration along the coupon length. The results to- Partial Penetration at ¼ in. (50%) The obtained weld bead specimens gether with the measured arc signals are from the partial penetration test were given in Table 5 in 1-in. intervals. For To achieve ¼-in. penetration on ½-in.- later processed by SECAT, Inc., Lexing- partial penetration weld joints, it was thick DH36 square butt joints, a series of ton, Ky., using standard preparation tech- easy to identify the weld interface and preliminary experiments were conducted niques for metallographic examination. penetration depth from the weld profile to obtain the following local model: By polishing and etching those specimens, of each small sample. In comparison with the penetration depth could be more the welding parameters measured in Fig. P = –0.3633 + 0.001312 × Ibm (12) clearly identified and measured. Some se- 11, the measured penetration depths lected weld bead profiles were shown corresponded to the welding parameters This local model calls for 470-A base below in Figs. 14 and 15. WELDING RESEARCH around each point. The information of metal current to produce ¼-in. penetra- Based on the etching results, it can be weld samples is listed in Table 5. tion depth. The welding parameters used observed that in certain cases, the main As can be seen, the base metal current to conduct these experiments and result- and bypass welding guns may deviate from was closed-loop controlled accurately ant base metal current from this local the joint. This may have been caused by an around the desired level of 436 A, with a model are listed in Table 7. irregular edge or deformation of the weld maximum error of 4 A. The average weld Again, the DH36 square butt joints joint, or the travel trajectory of the SAW penetration from the experiment was were tacked with zero root opening before tractor may not have been exactly in line 0.2484 in. The maximum error was welding. The completed weld bead is il- with the weld joint. However, the weld in- 0.0079 in., only 3.15% of the desired pen- lustrated in Fig. 12. Similarly, as in the terface of all specimens reached the de- etration depth. conventional SAW process, the weld bead sired penetration depth. These weld bead In further tests, it was also found that in Fig. 12 is smooth without spatter. In ad- profiles further proved the effectiveness 1 the root opening of less than ⁄16 in. will dition, due to the closed-loop control ef- of the proposed partial penetration con- not have a noticeable influence on the fect, the weld bead is also consistent in trol method by the bypass SAW process. penetration control accuracy. Actually, width and height. To accurately measure even though the weld joint was prepared partial penetration depth, the specimen Variation Experiments to have a zero root opening, the root was cut into several small sections along opening may become larger during weld- the weld bead with 1-in. interval. On the In ideal cases with all the welding pa- ing because of heat distortion of the base profile of weld bead cross section, the pen- rameters and conditions kept constant, an material. Therefore, if the weld joint is etration depth can be measured by caliper open-loop SAW process using predeter- prepared with a root opening smaller with accepted accuracy. The welding cur- mined welding parameters (wire feed 1 than ⁄16 in., the monitoring and control rent and partial penetration depth meas- speed, voltage, CTWD, root opening, etc.) accuracy will not be affected visibly. urements corresponding to each cross- would produce consistent welds and weld-

WELDING JOURNAL 55-s Li et al Supplement February 2013_Layout 1 1/16/13 10:05 AM Page 56

penetration depth as typically used in in- in. CTWD under the same total WFS and 6. Hopko, S. N., and Ume, I. C. 1999. Laser dustrial practices. However, the base power supply voltage settings. Through generated ultrasound by material ablation metal current that controls the weld pen- feedback control, the base metal current, using fiber optic delivery. Ultrasonics 37: 1–7. etration depth does subject to the effects as well as the penetration depth, will be 7. Chin, B. A., Madsen, N. H., and from possible variations in the CTWD and stable at the set points. The obtained weld Goodling, J. S. 1983. Infrared thermography for sensing the arc welding process. Welding Journal root opening. A closed-loop control would bead was processed in the same way, and 62(9): 227-s to 234-s. overcome the effects of these variations the measurements are listed in Table 12. 8. Chen, W. H., and Chin, B. A. 1990. Moni- on the base metal current and weld pene- As can be seen, the maximum error was toring joint penetration using infrared sensing tration depth. 2.05%. The closed-loop control still guar- techniques. Welding Journal 69(5): 181-s to Further, while the introduction of the anteed consistent weld penetration depth 185-s. bypass arc provides advantages, the arcing at the desired level. 9. Wikle, H. C., Kottilingam, S., Zee, R. H., process becomes more complicated and is It is apparent that consistent weld pen- and Chin, B. A. 2001. Infrared sensing tech- determined by more parameters including etration depths were achieved with ac- niques for penetration depth control of the sub- bypass wire position/angle. Ideally, setting ceptable accuracies despite the merged arc welding process. Journal of Materi- als Processing Technology 133: 228–233. all parameters at their nominal values be- changes/variations in the CTWD and root 10. Murugan, N., Parmer, R. S., and Sud, S. comes challenging. A closed-loop control opening. Further verification experiments K. 1993. Effect of submerged arc process vari- would overcome the effects of possible are needed in order to confirm the effec- ables on dilution and bead geometry in single setting inaccuracy on the base metal cur- tiveness of the closed-loop control system wire surfacing. Journal of Materials Processing rent and weld penetration depth. In addi- under other different changes/variations Technology 37: 767–780. tion, when an open-loop method is used, a in welding conditions. The constraints on 11. Murugan, N., and Gunaraj, V. 2005. Pre- few experiments are needed to determine and the design of the closed-loop control diction and control of weld bead geometry and the values for the welding parameters that system may be subject to changes. shape relationship in submerged arc welding of produce the desired base metal current pipes. Journal of Materials Processing Technol- ogy 168: 478–487. and weld penetration depth. With a Conclusions 12. Li, K., Chen, J., and Zhang, Y. 2007. closed-loop control, such experiments be- Double-electrode GMAW process and control. WELDING RESEARCH come unnecessary. 1) Bypass SAW provided an effective Welding Journal 86(8): 231-s to 237-s. To further demonstrate the effective- method to adjust the weld penetration for 13. Li, K., and Zhang, Y. M. 2008. Consum- ness of the closed-loop controlled bypass SAW without reducing the deposition able double-electrode GMAW Part I: The SAW process, a series of experiments were rate. process. Welding Journal 87(1): 11-s to 17-s. designed and conducted with varying root 2) A controlled bypass SAW system 14. Li, K., and Zhang, Y. M. 2008. Consum- openings and CTWDs. was established with shipbuilding welding able double-electrode GMAW Part II: Moni- as the target application. toring, modeling, and control. Welding Journal 87(2): 44-s to 50-s. Varying Root Opening 3) The penetration depth in welding 15. Li, X., Heusman, J., Kvidahl, L., Hoyt, square butt joints can be determined by P., and Zhang, Y. 2011. Manual keyhole PAW In previous experiments, butt joint the base metal current with an acceptable with application. Welding Journal 90(12): 258-s welds were made without intentional root accuracy. to 264-s. openings. To demonstrate the effective- 4) Local models provided a method to 16. Gupta, V. K., and Parmar, R. S. 1989. ness of the closed-loop control, the same obtain more accurate weld penetration es- Fractional factorial technique to predict di- welding parameters from the feedback timates for specific applications. mensions of the weld bead in automatic sub- control and verification experiment sec- 5) The effectiveness of the proposed merged arc welding. Journal of Institute of Engi- tion (Table 4) were used to weld the same penetration estimation/modeling and con- neering (India), Part ME 70: 67–75. AISI 1018 plates but with an intentional 17. Chandel, R. S., and Bala, S. 1998. Rela- trol method for DH-36 was experimen- tionship between submerged arc welding pa- variation in the root opening that in- tally demonstrated. rameters and weld bead size. Schweissen 1 creased from 0 to ⁄16 in. The obtained weld Schneiden 40: 28–31. bead was processed in the same way, and Acknowledgment 18. Chan, B., Chandel, R. S., Yang, L. J., and the measurements are listed in Table 11. Bibly, M. J. 1994. Software system for antici- 1 As can be seen, even with a ⁄16-in. root This work was funded by the Navy pating the size and shape of submerged arc opening, the base metal current was closed- SBIR Program under contract N65538-10- welds. Journal of Materials Processing Technol- loop controlled accurately around the de- M-0110. The technical guidance and assis- ogy 40: 249–262. sired level of 436 A. The average weld pen- tance from the technical point of contact, 19. Tarng, Y. S., Yang, W. H., and Juang, S. C. 2000. The use of fuzzy logic in the Taguchi etration from the experiment was 0.2515 in. Jonnie Deloach, is greatly appreciated. method for the optimization of the submerged The maximum error was 0.0066 in., only arc welding process. International Journal of 2.05% of the desired penetration depth. References Advanced Manufacturing Technology 16: Therefore, for the root opening varying be- 688–694. 1 tween 0 and ⁄16 in., the closed-loop control 1. Messler, R. W. 1999. Principles of welding: 20. Tarng, Y. S., Yang, W. H., and Juang, S. still guaranteed consistent weld penetration processes, Physics, Chemistry, and . C. 2002. The use of grey-based Taguchi meth- depth at the desired level. John Wiley. ods to determine submerged arc welding 2. AWS Welding Handbook, 8th ed., Vol. 2. process parameters in . Journal of Varying CTWD Test 1991. Miami, Fla.: American Welding Society. Materials Processing Technology 128: 1–6. 3. Bennett, A. E., and Siy, L. J. 2008. Blue- 21. Haber, R., and Keviczky, L. 1999. Non- print Reading for , 8th ed. Florence, Ky.: linear System Identification: Nonlinear System For the varying CTWD test, the same Parameter Identification, Vol. 1. Dordrecht, experiment parameters and weld joint Delmar Cengage Learning. 4. Jeffus, L., and Bower, L. 2009. Welding Netherlands: Kluwer Academic Publishers. were used as described in Table 4 and the Skills, Processes and Practices for Entry-Level 22. Ramachandran, K. M., and Tsokos, C. P. feedback control and verification experi- Welders. Florence, Ky.: Delmar Cengage 2009. Mathematical Statistics with Applications. ment section. For this test, the CTWD was Learning. Academic Press. 1 3 set ⁄8 in. lower than the standard ⁄4-in. 5. Graham, G. M., and Ume, I. C. 1997. Au- 23. , Standard Specification for Structural CTWD. The total current is expected to tomated system for laser ultrasonic sensing of Steel for Ships. 2004. West Conshohocken, Pa.: 3 change from the case with the standard ⁄4- weld penetration. Mechatronics 7: 711–721. ASTM International.

56-s FEBRUARY 2013, VOL. 92