Distributed memory banks improve factory anomaly detection performance
Federated Coreset Subsampling for Distributed Memory Banks
Distributed, Parallel, and Cluster ComputingComputer Vision and Pattern Recognition
Summary
Factories often produce different types of data because they use different machines or sensors. This makes it hard to catch unusual problems by sharing data since the data is not the same everywhere. The authors created two new ways to combine and make small summaries of data from different places without sharing all the data. They adjusted an existing method to work better in these situations, which helps spots problems faster and handles changes in sensors or missing data from some locations.
What this means in practice
- •For industrial process engineers: Improve early detection of equipment or process anomalies across multiple factory sites with varying sensor data.
- •For manufacturing quality control teams: Use federated coreset methods to maintain reliable anomaly detection despite sensor inconsistencies or partial data sharing.
Authors
Jason Witzemann, Daniel Pröll, Sebastian Hegenbart
Abstract
In real-world scenarios, industrial processes can span multiple factories and goods are often produced at multiple sites, leading to a shared yet non-identical distribution of data across locations. To address this, we propose two federated approaches for i) combining existing coresets (Double Subsampling) and ii) creating new coresets (Federated Subsampling) in a distributed anomaly detection setting. In particular, we adapt the memory bank based PatchCore algorithm to a federated setting, resulting in increased cold-start performance, resilience to domain- and sensor-shift, and resilience to unreliable actor participation, a common problem in classical, network based federated learning strategies.