Detector —
Anomaly Detection Autoencoder
Detector performs continuous health monitoring across all mechanical systems via reconstruction-error object detection. Named for the watchful guard. Unlike the other fleet models which predict optimal control actions, Detector learns the normal operating envelope and flags deviations — catching failures that site-specific models are not explicitly trained to detect (novel fault modes, sensor drift, control valve failures).
Abstract
Background
Detector performs continuous health monitoring across all mechanical systems via reconstruction-error object detection. Named for the watchful guard. Unlike the other fleet models which predict optimal control actions, Detector learns the normal operating envelope and flags deviations — catching failures that site-specific models are not explicitly trained to detect (novel fault modes, sensor drift, control valve failures).
Approach
The model was trained in AxonML (a pure-Rust deep learning framework) and compiled through the Hailo Dataflow Compiler (DFC 3.33.1) targeting Hailo-8 silicon. Post-training INT8 quantization was applied during the DFC compilation pass with production telemetry calibration data. The resulting Hailo Executable Format (HEF) binary executes on Hailo’s fixed-function dataflow architecture with deterministic latency and zero framework overhead at the edge.
Results
On production hardware (Hailo-8 M.2 (P/N: HM218B1C2FAE, S/N: HLDDM2A234600289)), Detector achieves 30,762 FPS (hw_only) with 0.005 ms hardware latency at 0.90 W average power draw.
Conclusion
Detector is production-ready as a single HEF binary deployed to edge devices with no external dependencies beyond the HailoRT vendor runtime. The model meets real-time latency requirements for its target hvac monitoring application.
| Model | Detector |
| Domain | HVAC Monitoring |
| Architecture | Anomaly Detection Autoencoder |
| Target silicon | Hailo-8 |
| Measured on | Hailo-8 M.2 (P/N: HM218B1C2FAE, S/N: HLDDM2A234600289) |
| DFC compiler | 3.33.1 |
| Framework | AxonML v0.6 (pure-Rust, CUDA + CPU backends) |
| Author | Andrew Jewell Sr. · ORCID 0009-0005-2158-7060 |
| Organization | AutomataNexus LLC · Fort Wayne, Indiana |
Executive overview
Detector performs continuous health monitoring across all mechanical systems via reconstruction-error object detection. Named for the watchful guard. Unlike the other fleet models which predict optimal control actions, Detector learns the normal operating envelope and flags deviations — catching failures that site-specific models are not explicitly trained to detect (novel fault modes, sensor drift, control valve failures).
Network I/O
Input: 57-channel sensor state. Output: 57-channel reconstruction (anomaly = high MSE between input and output).
Architecture
Anomaly Detection Autoencoder
Encoder-decoder autoencoder architecture (not a TCN predictor). The encoder compresses the full 57-channel sensor state into a 16-dimensional latent code via 4 progressively narrowing convolution layers. The decoder reconstructs the input from the latent code. Anomaly score = reconstruction MSE. High reconstruction error indicates the system is in a state not seen during normal operation training data. Extremely fast inference (sub-microsecond HW latency) due to the compact architecture.
Compilation constraints
All AxonML models targeting Hailo silicon are compiled under the fixed-function dataflow constraints: no dynamic control flow, no variable-length dimensions, all activations representable in INT8 after calibration, and no operations requiring dedicated softmax hardware (replaced with ReLU gating or depthwise convolution equivalents where necessary).
Silicon performance
Measured on production hardware via hailortcli benchmark with 5-second sustained inference. Device: Hailo-8 M.2 (P/N: HM218B1C2FAE, S/N: HLDDM2A234600289).
| Metric | Measured Value |
|---|---|
| FPS (hw_only) | 30,761.80 |
| FPS (streaming) | 20,425.80 |
| HW Latency | 0.005000 ms |
| Power (streaming avg) | 0.89900 W |
| Power (streaming max) | 0.90300 W |
| Power (idle) | 0.74979 W |
| Quantization | INT8 (post-training, DFC calibration) |
| DFC Compiler | 3.33.1 |
| HailoRT | 4.20.0 |
| Measured On | Hailo-8 M.2 (P/N: HM218B1C2FAE, S/N: HLDDM2A234600289) |
Deployment
Deployed as a single HEF binary. No ONNX runtime, TensorFlow Lite, or Python inference stack required at the edge.
| Target silicon | Hailo-8 |
| Measured on | Hailo-8 M.2 (P/N: HM218B1C2FAE, S/N: HLDDM2A234600289) |
| DFC compiler | 3.33.1 |
| Quantization | INT8 (post-training, production telemetry calibration) |
| Runtime | HailoRT (vendor runtime) |
| Edge platform | Raspberry Pi 5 + Hailo AI HAT+ (M.2 Key M) |
Deployment procedure
Copy the .hef binary to the target device. hailortcli run loads the HEF directly into the Hailo-8 dataflow engine over PCIe. Inference begins immediately with deterministic per-frame latency. No model conversion, graph optimization, or warmup phase required.
References
- Jewell, A. (2026). AxonML: A Pure-Rust Deep Learning Framework for Edge Inference. AutomataNexus LLC. Technical whitepaper.
- Hailo Technologies Ltd. (2024). Hailo Dataflow Compiler User Guide. DFC v3.33.1.
- Hailo Technologies Ltd. (2024). Hailo-8 Product Datasheet.
AutomataNexus LLC · Fort Wayne, Indiana · andrew.jewellsr@automatanexus.com
Andrew Jewell Sr. · ORCID 0009-0005-2158-7060
May 2026 · All rights reserved.