LIBRARY ı6ıul 0) - Cranfield University

stable arc condition for dip mode metal transfer by automatically adjusting the
welding voltage until stability was achieved. Overall, the technique presented by the
authors [ref. 201] was computationally time consuming; they reported that it takes
about 15 cycles (90 seconds) to obtain a stable welding condition, whereas the
controller developed in this work takes less than 6 seconds (see Section 7.1).
In a similar work, Mita et al. [ref. 200] used the fuzzy logic method to
automatically set the welding voltage in CO2 gas metal arc welding. The fuzzy rules
used were developed from the standard deviation of short circuiting and arcing times
and were distributed into three groups, dependending on the welding current (low
range: 80-200A; medium range: 210-290A; and high range: >300A). This was due to
the fact that the correlation between the stability of the welding arc and the standard
deviation of these parameters becomes poor as the welding current increases
[ref. 37].
In each group, different rules were used for assessing the stability of the welding arc
resulting in a fast but complicated method (for example, 20 fuzzy production rules
were used for the low current range and 25 rules for the medium range). Although
this is a simpler approach to automatically tuning the voltage compared to the work
carried out by Won and Cho [ref. 201], Mita et al. [ref. 200] work was found to use
too many rules. The algorithm presented in this section uses only 17 rules accross the
whole range of welding currents.
4.2.2 Data acquisition and processing
In order to implement the control algorithms presented in section 4.2.1, a
monitoring system was developed for acquiring and digitising the welding current and
voltage transient waveforms and extracting the statistical features from which the
monitoring indices are calculated. The description of the hardware involved can be
found in section 5.3 and the description of the software can be found in Appendix E.
Basically, the monitoring system acquires a fixed number of data points (512)
at a fixed sampling rate (2.5 kHz) for both the welding current and the welding
voltage. The basic statistical features of both signals are extracted using equations
(2.26) to (2.33). Then the monitoring indices are calculated using equations (2.3) to
(2.6) and filtered using a moving average filter, as shown in equation (4.7).
where
Si
Si filtered
Si-I. filtered
a
Sl. iltered - \i - aý Sý -F a' Sý-1 filtered
is the current calculated value of the variable
is the current filtered value of the variable
is the previous filtered value of the variable
smoothing factor (0.3)
This filter is used to reduce the effect of random variation in the monitoring
indices. The filtered values are then applied to the control rules to generate the
required voltage correction.
At the sampling rate used, 2.5 kHz, the acquisition boards take approximately
205 milliseconds to acquire 512 data points. This time was found sufficient to allow
109
(4.7)