An experimental study of the droplet sizes generated by a liquid ligament emerging from a rotary atomizer is described. A high-speed-shutter CCD camera was used to observe the ligament path, the ligament thinning, and the breakup. Droplet sizes were measured with laser diffraction technique. The atomizer consisted of a rotating drum with a small orifice at the lateral surface. Due to the experimental setup, the radial and tangential velocity components of the emerging ligament can he varied independently. The experimental work was carried out with variation of rotational speed, mass flow rate, and fluid composition (mixtures of water, glycerol, and surfactant). A neural network (NN) was applied to model and simulate the droplet diameter for different operating conditions, showing good agreement with the measured values. Based on the NN predictions of the system behavior, the model was used as a tool to detect outliers in the experimental data.