Performance analysis of arc welding parameters using self organizing maps and probability density distributions

During welding, random variations in current and voltage occur, which cannot be recorded with ordinary ammeter and voltmeter. Acquisition of voltage and current signals while welding is in progress at a very high speed using digital storage oscilloscope (DSO) and subsequent analysis of the stored data can be very useful to understand the arc welding process. In the present study, welding data were acquired for two inverter and two generator power sources while welding with two different electrodes using a DSO at the sampling rate of 100000 samples/s. This data was filtered using the Fast Fourier Transform (FFT) low pass filter and subjected to time domain and statistical analysis. Probability Density Distributions (PDDs) and artificial neural network comprising of Self Organizing Maps (SOM) were used to evaluate performance of power sources and welders. This paper explores the use of self-organizing maps as a mechanism for performing unsupervised learning for comparing performance characteristics of various welding parameters which includes welding power supplies and welders. Results obtained using SOM has been compared with the PDDs obtained during statistical analysis. Finally it is shown that in addition to PDD, analysis of voltage and current data using SOM technique can also be used to evaluate the arc welding process.