<p>Vacuum-assisted composite manufacturing techniques utilize atmospheric pressure to consolidate fabric components. However, the presence of leakages in the vacuum bag can lead to air bubbles, resin traps, non-uniform surface finishes, and subpar mechanical properties. Therefore, identifying and repairing leakages prior to the curing stage is essential. This paper presents an intelligent, machine learning-based approach to leakage detection. Initially, we introduce an electric circuit analogy for simulating the vacuum process. Our proposed method is compared and validated against experimental configurations. The simulation serves as a rapid and reliable alternative for generating data for the machine learning agent, obviating the need for complex analytical simulations and extensive physical experiments. In our proposed framework, a classification model is trained to categorize the number of leakages. Afterwards, various regression machine learning models have been trained for leakage localization. Our models predict leakage locations with acceptable errors and demonstrate robust performance across diverse configurations.</p>