Detection of short circuit in pulse gas metal arc welding process

Transcription

1 of Achevements n Materals and Manufacturng Engneerng VOLUME 4 ISSUE 1 September 007 Detecton of short crcut n pulse gas metal arc weldng process.k.d.v. Yarlagadda a, * co-operatng wth. raveen a, *, V.K. Madasu a, S. Rhee b a School of Engneerng Systems, Queensland Unversty of Technology, George Street, rsbane, Qld 4001, Australa b Department of Mechancal Engneerng, Hanyang Unversty, 17 Haegdang-dong, Seongdong-ku, Seoul , Korea * Correspondng author: E-mal address: p.posnaseet@qut.edu.au Receved ; publshed n revsed form Abstract urpose: The paper dscusses several methods of detectng occurrence of short crcut and short crcut severty n pulse gas metal arc weldng process (GMAW-). Desgn/methodology/approach: Weldng experments wth dfferent values of pulsng parameter and smultaneous recordng of hgh speed camera pctures and weldng sgnals (such as current and voltage) were used to dentfy the occurrence of short crcut and ts severty n GMAW- process. The nvestgaton s based on the measurement of weldng sgnals specfcally current and voltage sgnals and ther synchronzaton wth hgh speed camera to nvestgate the short crcut phenomenon n GMAW- process. Fndngs: The results reveal that short crcut can be detected usng sgnal processng technques and ts severty can be predcted by usng statstcal models and artfcal ntellgence technques n GMAW- process. Research lmtatons/mplcatons: Several factors are responsble for short crcut occurrence n GMAW- process. The results show that voltage and current sgnal carry rch nformaton about the metal transfer and especally short crcut occurrence n GMAW- process. Hence t s possble to detect short crcut occurrence n GMAW- process. Future work should concentrate on development of advance technques to mprove relablty of technques mentoned n ths paper for short crcut detecton and predcton n GMAW- process. Orgnalty/value: For achevng atomzaton of the weldng processes, mplementaton of real tme montorng of weld qualty s essental. Specfcally for GMAW- process whch s wdely used for lght weght metal whch s wdely ganng popularty n manufacturng ndustry. However, n case of GMAW- process hardly any attempt s made to analyse technques to detect and predct occurrence of short crcut. Ths paper analyses dfferent technques that can be employed for real tme montorng and predcton of short crcut and ts severty n the GMAW- process. Keywords: Short crcut; ulse gas metal arc weldng 1. Introducton Achevng good qualty wth lght weght metals lke Alumnum s a challengng job [1]. GMAW- process s wdely used n weldng of lght metal alloys lke Alumnum etc. Thus ndustry has great research nterest n automzaton and robotzaton of weldng processes though mplementaton of realtme montorng and control of weldng processes especally for lght weght metal alloy. 38 Occasonal paper Copyrght by Internatonal OCSCO World ress. All rghts reserved. 007

2 Several sensors lke vsual magng, nfrared camera, ultrasonc sensng and radographc sensng methods have been employed for acheng automaton []. In addton external sensors capable of attachng to weldng torch has also been develped. The drawbacks of these sensors are: expensve, relablty, and may cause practcal lmtaton by restrctng the moton flexblty and accessblty of the weldng torch. To avod the above-mentoned drawback, through-the-arc sensng (a sensor that uses the weldng voltage and current alone) has been developed and successfully appled n real tme montorng of weldng processes. In the expermental work descrbed, several technques for short crcut analyss based on captured sgnals of current and voltage have been dscussed. To ths purpose, a data acquston system ntegrated wth personal computer was used. The results from sgnal processng was verfed by usng the hgh speed camera pctures. Experments were carred out usng the prncple of back-lght hgh speed xenon lamp cnematography whch was synchronzed wth data acquston system. The arc voltage was measured between the contact tp of the weldng gun and the fxture. The weldng current was measured wth a Hall sensor, whch was attached to the earth cable. A data acquston card n combnaton wth a laptop personal computer was used to acqure voltage and current sgnals alsong wth hgh speed camera pctures. The samplng rate was 10 khz. The nose on the sgnals was removed by a dgtal low pass flter wth a 00 Hz cut-off frequency. The waveform sgnals were collected durng a s perod after 10 s elapsed from the start of weldng.. Materal Transfer transfer and Weldng rocess weldng Stablty process stablty ulse mode (or projected spray or drop spray mode), s characterzed by pulsng of current between low-level background current and hgh-level peak current n such a way that mean current s always below the threshold level of spray transfer. It has advantages over the other transfer modes whch ncludes: unform droplet sze closer to electrode dameter, regular detachment, drectonal droplet transfer, and lttle spatter. However, ths transfer mode can be obtaned only n very narrow ranges of current and t s also dependent on other weldng parameters such as wre dameter, electrode extenson etc. In order to mantan a stable GMAW- process followng crtera s must be fulflled: (1) Acheve stable arc, () Spray type metal arc transfer (Avod short crcut), (3) Constant arc length must be mantaned, and (4) No spatter (Avod short crcut) Occurrence of short crcut n GMAW- process means arc length does not reman contact and metal transfer through contact wth weld pool. erhaps the most unwanted result of unstable arc n GMAW- process s the producton of large amount of spatter whch s unpleasant aesthetcally. Hence avodng short crcut s the most essental condton for achevng the stable weldng arc. The easest method of avodng short crcut and ensurng stablty of GMAW- process s through control and analyss of captured sgnals lke current and voltage. Ths s generally followed by statstcal and soft computng technques for extracton of useful nformaton from these sgnals. 3. Experments 3.1. Expermental Setup setup The expermental set-up used n ths study s shown n fgure 1. Through-the-arc sensng of the weldng current, arc voltage and hgh speed magng of the droplet transfer are assembled n conjuncton wth each other to study the nfluence of the pulse current waveform on the metal transfer process durng GMAW-. Fg. 1. Schematc dagram of expermental setup 3.. Weldng Condtons condtons ead on plate welds were made n the flat poston usng an nverter controlled GMAW power source. Welds were made n the constant current mode. The fller materal used was a 4047 alumnum alloy weldng wre wth a 1.-mm dameter. The base materal was 6061 alumnum alloy wth a thckness of 6 mm. The sheldng gas used throughout the experments was pure argon wth a gas flow rate of 0 L/mn. Contact tp to workpece dstance (CTWD) and speed was kept constant through out the experments. The values of CTWD and speed used for all the experments were 0 mm and 4 mm/s. Table 1 shows the settng condtons of pulsng parameters used for the expermentaton. Table 1. Expermental desgn plan Levels I p (A) I b (A) T p (ms) T b (ms) WFR (m/mn) READING DIRECT: 39

3 Journal of Achevements n Materals and Manufacturng Engneerng Volume 4 Issue 1 September On-lne On-Lne detecton of Short of short Crcut crcut Fgure shows the weldng current and voltage sgnal correspondng to short crcut and drop transfer mode respectvely. Short crcut waveform s characterzed by three regons namely A, and C whch are shown n Fgure (a). a) A C Threshold Voltage 5. arametrc arameterc Approach approach for Short for Crut short Severty crcut Estmaton severty estmaton Usng the sgnal processng and mage processng technques, short crcut data was separated from the condtons under whch drop transfer mode was observed. Total 187 experments resulted n short crcut out of 786 experments performed. Further analyss nvolved quantfcaton of short crcut severty based on sgnal processng and mage processng technques. Dependng upon the pulsng parameters, short crcut occurs wth dfferent severtes whch have been quantfed n ths work on the bass of number of tmes short crcut occurs n a sngle pulse. Several waveform factors can be extracted from the wre meltng rate equaton. Waveform factors used n ths study are: (1) ulse characterstcs shape factors: eak Current (I ), ase Current (I ), eak Tme (T ), ase Tme (T ), Unt eak Current (I F), and Unt ase Current (I F); () ulse area characterstcs shape factors: b) D Threshold Voltage eak Tme Arc Heatng Fracton I T ase Tme Arc Heatng Fracton I T Unt eak Tme Arc Heatng Fracton I T F, Unt ase Tme Arc Heatng Fracton I T F Square of eak Tme Arc Heatng Fracton I T, Square of ase Tme Arc Heatng Fracton I T, Square of Unt eak Tme Arc Heatng Fracton I T F Square of Unt ase Tme Arc Heatng Fracton I T F I T F eak Tme Resstve Heatng Fracton ase Tme Resstve Heatng Fracton I T F. Fg.. Typcal weldng current (top) and voltage (bottom) as functons of tme for dfferent metal transfer modes, namely (a) Short crcut, and (b) Drop Transfer Regon A and C s characterzed by the drop n voltage below the threshold level (here 5 Volts) where short crcut occurs. Whle Regon s characterzed by voltage well above the threshold voltage when there s no contact between the electrode and workpece. On the other hand, drop trasnfer mode s characterzed by pulses whch occur repettvely throughout the voltage waveform and voltage value s well above the threshold value as shown n Fgure (b). Hence a smple test to detect short crcut s based on voltage magntude whch s shown n equaton [1]. V V mn Short Crcut (1) 5.1. Statstcal statstcal Modellng modellng A lnear regresson model was composed as shown n Eq., usng the 16 factors obtaned durng two seconds of weldng. A correlaton and multple regresson analyss were performed to nvestgate how much nfluence these factors had on the number of shorts per pulse (Y). The resultant equaton can be wrtten as: Y WFR I I I T I T F I T F pi T F T The multple correlaton coeffcent and standard error between the estmated results and the number of shorts per pulse was found to be 83.9% and 0.35 respectvely. Ths proves that nput parameters greatly affect the output and confrms the adequacy of the model n accurately predctng the short crcut severty. () 330 Occasonal paper.k.d.v. Yarlagadda,. raveen, V.K. Madasu, S. Rhee

4 a) b) Fg. 3. Average rms errors between the predcted results and the number of shorts per pulse at dfferent wre feed rates for: (a) Expermental data from whch model s made, and (b) Test data To ensure the accuracy of the developed model descrbed above and to survey the spread of the values, model results were compared wth expermental results. Comparson of average rms error between the expermental observaton and results predcted from the model for dfferent wre feed rates are shown n Fg. 3 (a) and (b) respectvely. For all shorts per pulses, over estmaton of the number of shorts per pulse was found and as number of shorts per pulse ncreased over estmaton decreased. It s also evdent from the Fg. 3 (a) and (b) that reasonable agreement between expermental and predcted number of shorts per pulse s shown. resence of error suggests that developed model should be used wth care. ' where, (1 t) t x (1 s) s x ' x x m 5.. Fuzzy Fuzzy Modelng modellng In ths secton, we brefly explan the newly developed fuzzy modelng method for estmatng the short crcut severty for GMAW- process. Sxteen factors representng the characterstcs of the pulse waveforms are employed as predctor varables and the short crcut severty (or number of shorts per pulse) s predcted on the bass of a modfed exponental membershp functon ftted to the fuzzy sets derved from predctor varables. The system s traned wth means and varances computed from each one of the fuzzy sets and then tested usng untraned nput predctor varables. The membershp functon charactersng a fuzzy set s chosen as, x m x e (4) where: x s the th feature of the unknown state, m represents means and stands for varances for each of the fuzzy sets It s observed that some of the fuzzy sets have a very small varance and others, a large varance. Ths spurred the choce of a new membershp functon n [3] nvolvng the structural parameters s and t gven by, ' ' x / x e (5) Fg. 4. Convergence of parameter s and t wth number of teratons on X-axs and tme taken (secs) on Y-axs. The new mean and the new varance are functons of the mean and varance of the reference fuzzy set. Thus the structural parameters- s, t model the varatons n the mean and varance. The choce of these parameters has reasonng. That s, f s=1, ' x = x. Thus, s would be perturbed around 1 to reflect changes n the means. Smlarly, f t=-1, then ' = thus, t would Detecton of short crcut n pulse gas metal arc weldng process 331

5 Journal of Achevements n Materals and Manufacturng Engneerng Volume 4 Issue 1 September 007 reflect the changes n the varances. The ntal values of s and t are taken as 11 and 6 respectvely and are later updated usng the gradent descent technque. The above mdoelng strategy was appled on a dataset consstng on a subset of 56 experments conducted at a wre feed rate of 4 m/mn (as shwon n Table 1). The convergence of structural parameters s and t for constant s shown n below gven fgure. The fnal values of s & t were found to be and respectvely. We obtaned a recognton rate of 100% wth a smaller dataset of 64 experments and 94.5% wth a second test data set consstng of 56 experments Conclusons conclusons Ths paper presents varous methods by whch short crcut occurrence and ts severty can be predcted n GMAW- durng real tme operaton. To ths end varous experments were performed to fnd the pulsng condtons under whch short crcut occurs and ther characterstc nfluence on current and voltage sgnal was characterzed for detectng short crcut n GMAW- process. Further work nvolved dentfcaton of waveform factors and ther nfluence on short crcut severty. Through arc sensng was found to be relable for detecton of short crcut and for predcton of short crcut severty. References [1]. raveen, M.J. Kang,. K. D. V.Yarlagadda, Arc voltage behavor of one drop per pulse mode n GMAW-, Journal of Achevements n Materals and Manufacturng Engneerng, 17 (006) []. raveen, M.J. Kang,.K.D.V. Yarlagadda, Characterzaton of dynamc behavor of short crcut n pulse gas metal arc weldng of alumnum, Journal of Achevements n Materals and Manufacturng Engneerng 14 (006) [3] M. Hanmandlu, M. H. M. Yusof, V.K. Madasu, Off-lne sgnature verfcaton and forgery detecton usng fuzzy modelng, attern Recognton 38/3 (005) Occasonal paper.k.d.v. Yarlagadda,. raveen, V.K. Madasu, S. Rhee