Elevator 820-Trip Multisensor Dataset for Predictive Maintenance Research

This dataset contains 820 multisensor elevator trips collected from an electromechanical traction elevator equipped with a worm-gear transmission. The dataset was prepared for research on IoT-based condition monitoring and unsupervised anomaly detection for predictive maintenance applications. The deposited package includes selected trip-level Parquet files, processed trip-level Parquet files, feature tables, model outputs, ISO 20816 reference calculations, validation summaries, metadata files and SHA256 checksums. The acquisition system included vibration, electrical current, temperature, operational state and travel direction signals. The dataset supports the reproducibility of the associated 820-trip study, including the extraction of global trip-level features and operational-state features used with Isolation Forest models. The model outputs identify statistically atypical trips with respect to the learned operational baseline. These events should be interpreted as candidates for technical inspection or further analysis, not as confirmed mechanical failures. The dataset does not include credentials, IP addresses, building-identifying information or private configuration files.