The document compares different ANN algorithms for detecting the saturation level in the magnetic core of a welding transformer. Four algorithms are evaluated: Resilient Backpropagation, Gradient Descent, Levenberg-Marquardt, and Bayesian Regularization. The algorithms are assessed based on computational time, error, gradient, and complexity. Detecting saturation is important to prevent current spikes that could shut down the welding system. An ANN uses the primary current as input to identify spikes and control the flux density to prevent saturation and overcurrent.