Robust Multimodal Fusion for Audio–Vibration–Current Signals Using Cross-Attention with Missing-Modality Handling
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
In the domain of Industry 4.0 and the Industrial Internet of Things (IIoT), the reliability of predictive maintenance systems relies heavily on the accurate interpretation of multisensor data. While conventional fault diagnosis frameworks often utilize single-modality signals, the fusion of heterogeneous data sources—specifically Audio, Vibration, and Current (AVC)—offers a more comprehensive representation of machinery health states. However, the practical deployment of such multimodal systems is frequently hindered by sensor malfunctions, transmission errors, or environmental noise, leading to missing or corrupted modalities. This paper presents a novel deep learning architecture, the AVC-FusionNet, which employs a robust cross-attention mechanism designed to dynamically weigh the importance of each modality while explicitly handling missing data scenarios. By integrating Motor Current Signature Analysis (MCSA) with acoustic and vibrational features, our approach captures complex inter-signal correlations that unimodal systems overlook. We introduce a specialized Modality Dropout training strategy that simulates sensor failure, forcing the network to learn resilient representations. Extensive experiments on a rigorous synthetic and real-world industrial dataset demonstrate that the proposed method outperforms state-of-the-art fusion techniques, maintaining high classification accuracy even when one or more modalities are entirely absent.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it