Dataset for Data-Driven Assessment of Fan Noise in the Presence of Inflow Distortion: A First Demonstration

Systematic measurements on a subsonic axial-flow fan rig investigated how major noise sources scale with inflow distortion parameters. Common in compact nacelles for next-generation UHBR engines, these distortions significantly amplify fan noise and present an aeroacoustic design challenge. The Co/Counter Rotating Acoustic Fan Test Rig (CRAFT) was developed by the German Aerospace Center (DLR) in Berlin specifically for the investigation of sound sources in ducted, subsonic fan stages. A diverse set of distortion conditions was created, varying distortion type and intensity through scaled and modified distortion screens, resulting in a rich dataset of simultaneous acoustic and aerodynamic measurements. Advanced signal-processing techniques are employed to isolate the distortion-induced noise contributions and to quantify their dependence on key flow features. The resulting data are used to calibrate and validate data-driven models that predict noise generation in the presence of distorted inflows to illustrate the potential of the dataset. This work combines advanced aeroacoustic signal processing with interpretable machine learning analysis. The generated dataset allows a better understanding of the impact of inflow distortions on the sound sources at the rotor and stator. Shapley value analysis shows that low-order circumferential inflow harmonics, rig power setting, and rotor–stator distance significantly affect most outputs. In this way, prediction methods can be optimized in the long term. This dataset enables data-driven modelling of aeroacoustic dependencies. Due to the nonlinear relationship between aerodynamics, operating conditions, and radiated noise, it can be used to evaluate machine learning approaches.