Compact model improves machine fault detection with few labels
When Labels Are Scarce: An Oscillatory State Space Model for Vibration Diagnosis
Machine Learning
Summary
Detecting problems in machines from vibration sounds usually needs lots of labeled examples, which are hard to get. The authors created a small computer model called DualRes that learns from just a few labeled seconds of vibration data by looking at sound in two different ways. Their model works well on multiple machine datasets and runs faster and uses less memory than similar methods. This approach helps diagnose machine faults efficiently even when labeled data is scarce.
What this means in practice
- •For maintenance engineers: Identify faults in industrial bearings and gearboxes quickly using very limited labeled vibration data on-site.
- •For embedded systems developers: Build efficient vibration fault diagnosis features into resource-constrained edge devices for real-time local inference.
- •For automotive diagnostic service providers: Offer diagnosis of gearbox faults to customers with minimal training data by deploying compact models on diagnostic tools.$Commercial implications: Enables compact and fast diagnostic software for gearbox fault detection deployable in commercial automotive service tools.
Authors
Mainak Mallick, Seung-Kyum Choi
Abstract
Machine fault diagnosis from vibration requires learning from scarce labelled fault recordings while meeting the computational constraints of edge devices for local inference. We introduce DualRes, a compact oscillatory state-space model that combines two complementary spectral views of vibration, capturing rapid changes and fine frequency structure. Time-aligned views are processed by selective oscillatory memory, which learns how long to retain temporal patterns. The encoder contains 39,528 parameters. We evaluate supervised learning across six bearing datasets and a gearbox benchmark, with an additional gearbox pilot. Recording-level splits and explicit accounting of labelled duration distinguish data efficiency from repeated exposure to correlated samples. On the main gearbox benchmark, DualRes achieves state-of-the-art performance among the nine evaluated methods at six of seven label budgets. With about six labelled seconds per class, it improves macro-F1 by 16.1 percentage points over the next strongest comparator. On the same benchmark, DualRes achieves a 1.44-fold recording-level speedup and a 24.8-fold reduction in checkpoint storage relative to a selective state-space baseline under matched hardware and runtime conditions. Bearing results reveal task-dependent trade-offs. These findings support oscillatory memory as a compact approach to vibration diagnosis under limited labelled exposure.