Monitoring is a crucial aspect of modern production systems, especially in additive manufacturing, where instabilities and defects can lead to significant economic losses due to defective components. Consequently, artificial intelligence is increasingly used to monitor processes, enabling machines with self-analysis capabilities to generate stops or provide automatic feedback to operators. In the Wire Arc Additive Manufacturing (WAAM) process, frequency analysis of voltage signals offers an additional method to study signal characteristics, enabling the extraction of features that describe the process state. This study conducted deposition tests of Inconel 718 using the Pulsed Gas Metal Arc Welding process with pre-optimized parameters. Features were extracted by analysing the time-frequency behaviour of welding voltage signals using wavelet decomposition. Subsequently, a Gaussian Mixture Model was employed to identify clusters that define the process state. By utilizing the centroids of these clusters, the process was monitored online by assigning new samples arriving online from the real deposition process to the nearest centroid. This enabled alerts to be generated for an operator or an autonomous decision-making module regarding current state of the WAAM system.