Synthetic End-Milled Sidewall Topography Dataset for Surface Recovery under Controlled Stochastic Distortions

This self-contained synthetic numerical dataset was developed as an educational assignment and controlled benchmark for the course Methods of Analysis and Forecasting of Manufacturing Processes (Методи аналізу та прогнозування виробничих процесів). Its primary purpose is to support reconstruction of a clean modelled milled sidewall surface from one, two, or three distorted observations and to enable analysis of related manufacturing-process responses. The dataset contains 48 deterministic variants generated for 12 radial depths of cut and four processing stages, together with 48 stochastic variants. Each stochastic variant includes three distorted observations, resulting in 144 noisy observations. 3 noise levels are provided: 20 dB 12,5 and 5 dB. Each native surface field has a resolution of 25 × 5000 points. For educational use, the dataset is divided into 40 primary and 8 reserve assignment variants, with 20 primary and 4 reserve variants provided for each noise level. The dataset is fully synthetic and is intended for numerical analysis, signal reconstruction, model evaluation, and educational benchmarking. It should not be interpreted as experimental measurement data. Some parameter combinations intentionally extend beyond the locally supported modelling interval and are included as controlled synthetic extrapolations. The archive includes clean target fields, noisy observations, process-related diagnostic data, model parameters, schemas, generation code, validation materials, a public benchmark protocol, and standalone HTML viewers in English and Ukrainian. Creator: Artur V. MyhovychNational Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", Kyiv, UkraineORCID: 0000-0001-8687-6879ROR: https://ror.org/00syn5v21