The ALERT Dataset: Benchmarking Anomaly Detection of Non-Stationary Vibrational Signals.

Summary In recent years, automatic audio anomaly detection has gained considerable attention. However, most existing methods and benchmarks assume stationary or periodic signals, limiting their applicability to industrial environments characterized by non-stationary operating regimes such as speed ramps and transient load variations. We introduce the ALERT Dataset, a large-scale collection of non-stationary vibration recordings from electric powertrains acquired on an industrial end-of-line test bench. Each recording captures ramp-up and ramp-down phases with continuously varying rotational speed and includes synchronized speed measurements to enable explicit conditioning on operational dynamics. The dataset comprises 224 healthy training recordings and 80 healthy test recordings, along with an additional 80-sample hold-out set reserved for anomaly generation. From this hold-out set, multiple anomalous test suites (80 samples each) are constructed via expert-designed amplitude-based degradations and structured noise perturbations at varying signal-to-noise ratios, simulating realistic fault scenarios. Models are evaluated by discriminating these anomalies from the 80 healthy test recordings under a one-class learning paradigm. The benchmark further supports diverse protocols, including zero-shot cross-phase testing. To our knowledge, the ALERT Dataset is the first large-scale collection of non-stationary industrial vibration signals with synchronized speed references, addressing a critical gap in existing benchmarks. Data Acquisition Parameter Value System Electric powertrains Environment Industrial end-of-line production test bench Operating profile Ramp-up (~18 s) and ramp-down (~10 s), near piece-wise linear speed trajectory Raw sampling rate 100 kHz Low-pass filter 10 kHz Recommended working sampling rate 10 kHz (downsampled) Additional signal Synchronized rotational speed measurement Dataset Structure Training set: 224 healthy recordings Normal test set: 80 healthy recordings Hold-out set: 80 healthy recordings (used for anomaly generation) From the hold-out set, multiple anomalous test sets (80 samples each) are generated. Base sets: Training set (224 samples) Normal test set (80 samples) Hold-out set (80 samples) Anomalous test sets: Amplitude shift (1 set) White noise (7 sets) Variance-dependent noise (7 sets) TVAR(1) noise (8 sets) Alpha-stable noise (7 sets) Band-limited noise (8 sets) Data Format Each sample consists of four .npz files: Vibration signal (ramp-up) Rotational speed (ramp-up) Vibration signal (ramp-down) Rotational speed (ramp-down) File naming convention: {dataset_name}_{snr}_sample_{sample_id} Evaluation Protocol One-class learning: training on healthy data only Intra-phase evaluation: train and test on the same phase (e.g., ramp-up) Cross-phase evaluation: train on one phase, test on another (e.g., ramp-up → ramp-down) Multi-SNR robustness evaluation: for noise-based anomaly scenarios Intended Use Time-series anomaly detection Vibro-acoustic condition monitoring Non-stationary signal analysis Benchmarking under dynamic operating conditions Citation If you use this dataset, please cite: The ALERT Dataset: Benchmarking Anomaly Detection of Non-Stationary Vibrational Signals ALERT: Anomaly Detection in Vibrational Data Using a Linear Autoregressive Latent Space License The ALERT Dataset: Benchmarking Anomaly Detection of Non-Stationary Vibrational Signals © 2026 by Anton EMELCHENKOV is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0).To view a copy of this license, visit: https://creativecommons.org/licenses/by-nc-sa/4.0/ Notes Released with permission from the industrial partner. No confidential or proprietary information is included.