Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper investigates unsupervised anomaly detection in complex industrial time series data, highlighting the limitations of methods validated on simpler benchmarks. The authors evaluate Isolation Forest and several autoencoder architectures on a novel dataset from operational industrial machinery with process-induced variability. Results demonstrate that temporal convolutional autoencoders (TCAEs) outperform Isolation Forest, recurrent, and variational autoencoders in capturing non-periodic, multi-scale dynamics, though the latter require more hyperparameter tuning.
Turns out, your fancy anomaly detection algorithm probably won't work on real-world industrial data unless it can handle the chaos of multi-stage processes.
Industrial time-series data from real production environments exhibits substantially higher complexity than commonly used benchmark datasets, primarily due to heterogeneous, multi-stage operational processes. As a result, anomaly detection methods validated under simplified conditions often fail to generalize to industrial settings. This work presents an empirical study on a unique dataset collected from fully operational industrial machinery, explicitly capturing pronounced process-induced variability. We evaluate which model classes are capable of capturing this complexity, starting with a classical Isolation Forest baseline and extending to multiple autoencoder architectures. Experimental results show that Isolation Forest is insufficient for modeling the non-periodic, multi-scale dynamics present in the data, whereas autoencoders consistently perform better. Among them, temporal convolutional autoencoders achieve the most robust performance, while recurrent and variational variants require more careful tuning.