AI-Ready Simulation Dataset for Chemical Process Safety and Accident Prevention

This dataset comprises 500 time-ordered records generated within a high-fidelity digital twin of a hazardous chemical process unit. It integrates process variables, control actions, and multi-domain safety indicators to support the development and training of artificial intelligence models for industrial accident prevention. The dataset includes thermodynamic and kinetic process measurements (temperature, pressure, flow rate, and reactant concentration), manipulated variables associated with cooling, agitation, and ventilation, and condition monitoring signals related to mechanical integrity, dust explosibility, and fire precursors. Controlled fault injection scenarios, including cooling system degradation, pressure sensor bias, and dust handling upsets, are embedded to represent realistic accident precursor dynamics. Each record is annotated with a continuous risk score derived from physics-consistent safety indicators and discrete labels identifying near-miss and hazardous operating states. The dataset is specifically designed for supervised, unsupervised, reinforcement learning, and physics-informed machine learning applications, enabling the study of early fault detection, anomaly identification, and safety-oriented decision support under severe class imbalance.