Abstract: An approximate entropy (ApEn)-support vector machine (SVM) method of arc voltage was proposed to evaluate the stability of aluminum alloy twin-wire pulse metal insert gas (PMIG) welding process. A set of welding experiments were carried out and the ApEn of welding current and voltage signals was calculated. The results showed that the smaller the ApEn of current and voltage signals is the more stable, the welding process is. The application of ApEn on the welding current and the welding voltage was compared. It showed that the voltage based ApEn is sounder in measuring the stability of aluminum alloy twin-wire PMIG welding. Then a support vector machine (SVM) algorithm based on approximate entropy (ApEn) has been developed on voltage signals. And the results of the classification showed that the SVM algorithm based on ApEn can mark off the stable processes from the unstable ones. When the training data is more than 20%, the classification accuracy is more than 90%.