18.06.2013 Views

LIBRARY ı6ıul 0) - Cranfield University

LIBRARY ı6ıul 0) - Cranfield University

LIBRARY ı6ıul 0) - Cranfield University

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Appendix J), it was observed that the proportionality constant 0 used in equation 2.22<br />

and calculated using equation 6.5 differs from the measured value (see Figure 6.9).<br />

The value of the difference observed was found to be dependent on the wire feed<br />

speed and the mode of metal transfer.<br />

This observed inaccuracy in the estimation of 0 was possibly caused by the<br />

fact that the model does not fully describe the non-linearities present in the process.<br />

Also, the model (see equation 2.21) is not a theoretical model but an empirical one<br />

and the model inversion carried out could be a possible source of error.<br />

While adequate for indicating changes in stand-off, Ogunbiyi's model does not<br />

provide the accuracy and robustness necessary for control purposes, especially when<br />

using low wire feed speed dip transfer (see Figure 6.9).<br />

Previous work by Philpott [ref. 131] has shown that the dip-resistance can be<br />

used for stand-off monitoring. Based on this, a method for monitoring the dip<br />

resistance was developed and the dip resistance measurements were correlated to<br />

stand-off variation. A stand-off estimation model based on the dip resistance was<br />

developed and used in conjunction with the windowing technique developed by<br />

Chawla [ref. 161] (see section 2.6.3).<br />

The dip resistance based stand-off estimation model was found to be less<br />

sensitive to process instabilities in dip transfer because it uses the resistance measured<br />

during the short circuiting phase only. Hence, voltage and current spikes do not affect<br />

the prediction as was the case in the Ogunbiyi's model (see Figures 6.15 to 6.24).<br />

It should be noted that the model proposed by Ogunbiyi used an integral<br />

approach' to reduce the effect of random variation commonly observed in the welding<br />

current. However, the use of the integral approach makes the model very sensitive to<br />

the stability of the process at the start of the welding. Also, Ogunbiyi's model only<br />

outputs estimated changes in the stand-off while the dip resistance based model<br />

produces an absolute stand-off estimation.<br />

The dip resistance model was also extended to the spray transfer mode, the<br />

only difference being that, in this mode, the measured resistance also included a<br />

component due to the welding arc (see section 6.2 and Figures 6.13,6.14 and 6.25 to<br />

6.27).<br />

The stand-off models for dip and spray transfer were validated successfully for<br />

bead-on-plate welding trials. However, when transferred to fillet joints in the flat<br />

position, it was observed that the models would produce a prediction smaller than the<br />

actual stand-off by a constant average value. This was hypothesised to be caused by<br />

the weld pool build-up under the welding arc which occurs due to the restricted flow<br />

of the molten metal resulting from the geometry of the fillet joint. This was confirmed<br />

by changing the deposition rate and comparing the resulting estimation errors (see<br />

Figure 6.29). In order to compensate for the weld pool constriction effect found in<br />

fillet joints, the offset in the stand-off prediction (illustrated in Figure 6.29) was<br />

estimated at the start of the weld and its value was added to the subsequent<br />

estimations (see section 6.2). It should be noted, however, that this approach can only<br />

be applied if the initial stand-off is known, as was the case in this work. Good results<br />

were obtained for the fillet weld trials (see Figures 7.6 to 7.14).<br />

7 Characterised by the use of a cumulative summation of the differences between successive welding<br />

current values.<br />

193

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!