axonml-hvac

Overview

axonml-hvac is a domain-specific sub-crate of the AxonML deep-learning framework providing nine specialized neural-network models for HVAC fault detection, predictive maintenance, and facility-wide diagnostic reasoning. It was extracted from the axonml umbrella crate in 0.6.1 (April 2026) to keep the framework itself domain-agnostic.

The models were designed for the TMC HVAC controller project and trained on physics-informed synthetic data derived from real building-automation-system (BAS) control logic.

Last updated: 2026-04-16 — version 0.6.1.

Models

Model	File	Purpose
Apollo	`apollo.rs`	Primary fault classifier (top-level diagnostic head)
Aquilo	`aquilo.rs`	Airflow anomaly detector (AHU/VAV blockage, dirty filter)
Boreas	`boreas.rs`	Cold-side (cooling) specialist — chilled-water / DX cooling faults
Colossus	`colossus.rs`	Large transformer diagnostician — multi-equipment reasoning
Gaia	`gaia.rs`	Environmental context encoder (OAT, humidity, enthalpy)
Naiad	`naiad.rs`	Water-side (hydronic) specialist — loop pressures, flows, pump behavior
Panoptes	`panoptes.rs`	Observability / multi-signal fusion across a full facility
Vulcan	`vulcan.rs`	Heat-side specialist — boilers, hot-water loops, reheat
Zephyrus	`zephyrus.rs`	Temporal predictor + autoencoder for drift detection

Supporting Modules

Module	Purpose
`data`	`HvacSensorData`, `HvacLabels`, `PipelineOutput`, `SyntheticHvacGenerator`
`panoptes_datagen`	Warren HVAC simulator (`PanoptesTrainingData`, `WarrenSimulator`) — physics-informed training data
`pipeline`	End-to-end multi-model diagnostic pipeline (`HvacPipeline`)

Public Re-exports

lib.rs re-exports the full public surface: Apollo, Aquilo, Boreas, Colossus, Gaia, Naiad, Panoptes, Vulcan, Zephyrus, HvacPipeline, HvacSensorData, HvacLabels, PipelineOutput, SyntheticHvacGenerator, PanoptesTrainingData, and WarrenSimulator.

Usage

use axonml_hvac::{Apollo, Boreas, HvacPipeline, Panoptes};

fn main() {
    // Full diagnostic pipeline
    let pipeline = HvacPipeline::new();

    // Or individual specialists
    let apollo = Apollo::new();
    let boreas = Boreas::new();
    let panoptes = Panoptes::new();
    let _ = (pipeline, apollo, boreas, panoptes);
}

Examples

The crate ships four runnable examples under examples/:

# Train Panoptes (facility-wide anomaly detection)
cargo run --release -p axonml-hvac --example train_panoptes

# HVAC inference demo
cargo run --release -p axonml-hvac --example hvac_inference

# HVAC model scaffolding test
cargo run --release -p axonml-hvac --example hvac_model

# HVAC training loop reference
cargo run --release -p axonml-hvac --example hvac_training

Dependencies

Crate	Purpose
`axonml-core`	Device, DType, error types
`axonml-tensor`	Tensor operations
`axonml-autograd`	Automatic differentiation
`axonml-nn`	Neural network layers
`rand`	Random initialization / synthetic data generation

Dev-dependencies add axonml-optim, axonml-serialize, and the axonml umbrella crate for the examples.

Features

Feature	Description
`default`	(none)
`cuda`	Enables CUDA on `axonml-core`, `axonml-tensor`, `axonml-nn`, and `axonml-autograd`

GPU acceleration:

cargo build --release -p axonml-hvac --features cuda

License

Licensed under either of Apache License 2.0 or MIT at your option.

Part of AxonML — a complete ML/AI framework in pure Rust.

axonml-hvac 0.6.2