Steady state time series data from a continuous distillation plant with and without anomalies for developing machine learning anomaly detection methods

The dataset contains multivariate steady-state time-series data collected from a continuous distillation mini-plant operated under normal and anomalous conditions. The data comprises 28 experiments spanning 630 hours of steady-state data conducted across three process scenarios:
•    Scenario A: Single-component water distillation
•    Scenario B: Binary heteroazeotropic n-butanol/water separation
•    Scenario C: Reactive distillation for polyoxymethylene ether (OME) synthesis
The time-series data include temperatures, pressures, flow rates, pressure differences, valve positions, and other sensor or actuator data. For scenario B, besides time-series data, we also provide concentration data analyzed by techniques such as gas chromatography and Karl Fischer titration. Each experiment is supplemented by a structured metadata file documenting the operating conditions and detailed information on anomalies. This dataset supports a broad range of machine learning research tasks, including anomaly detection, synthetic data generation, and interactive visualization and exploratory analysis of real chemical process dynamics. It is easy to download and use. An online previewer of the datasets is available under: https://continuousdistillationtum.streamlit.app/.