https://github.com/kretski/orac-nt-ssd-thermal-sdk

Deterministic Physics-based Thermal Control SDK for NVMe Controllers. Extending NAND lifetime by 31.6% using ORAC-NT Vitality modeling and Arrhenius-based BER optimization.
https://github.com/kretski/orac-nt-ssd-thermal-sdk

nand-flash-memory nvme reliability-engineering ssd-controller thermal-management

Last synced: 2 months ago
JSON representation

Deterministic Physics-based Thermal Control SDK for NVMe Controllers. Extending NAND lifetime by 31.6% using ORAC-NT Vitality modeling and Arrhenius-based BER optimization.

• Host: GitHub
• URL: https://github.com/kretski/orac-nt-ssd-thermal-sdk
• Owner: Kretski
• License: other
• Created: 2026-03-28T01:21:37.000Z (2 months ago)
• Default Branch: main
• Last Pushed: 2026-03-28T13:06:42.000Z (2 months ago)
• Last Synced: 2026-03-28T16:27:01.031Z (2 months ago)
• Topics: nand-flash-memory, nvme, reliability-engineering, ssd-controller, thermal-management
• Language: Python
• Homepage:
• Size: 1.29 MB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# ORAC-NT v5.5

### Orbital Resilience Adaptive Controller — Neural-Thermal Edition

> **Deterministic Physical AI for Ultra-Low-Power Embedded Systems**  

> *Replacing probabilistic neural networks with provably stable physics.*

[![License: Proprietary](https://img.shields.io/badge/License-Proprietary-red.svg)](LICENSE)

[![Platform: STM32F4](https://img.shields.io/badge/Platform-STM32F4-blue.svg)]()

[![Validated: Hardware](https://img.shields.io/badge/Validated-Hardware%20%2B%20Simulation-green.svg)]()

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.18898599.svg)](https://doi.org/10.5281/zenodo.18898599)

![ORAC-NT Proof](https://github.com/user-attachments/assets/7b293f4f-ac91-4de4-9fbe-544050d0dee7)

---

## The Problem

Modern AI controllers for space, drones and industrial systems rely on neural networks that:

- Require **megabytes** of RAM and flash

- Consume **milliwatts** continuously — catastrophic for CubeSats

- Are **probabilistic** — no formal stability guarantees

- Show up to **27% disruption rates** in critical regimes (plasma, hypersonic)

## The Hidden Cost of Neural AI

Many companies have already spent hundreds of thousands of euros (some millions) developing neural controllers that:

- Never reach production due to power budget

- Require constant retuning when conditions change

- Provide no formal stability guarantees

ORAC-NT is not “just another optimizer”. It is a **physically deterministic** replacement that makes a large part of existing AI efforts economically unviable.

---

## The Solution: ORAC-NT

ORAC-NT replaces the neural stack with a **single deterministic physics equation**:

```c

W(t) = Q(t) · D(t) − T(t)

Symbol	Meaning

Q(t)	Signal quality from Byzantine-aware dual-sensor fusion

D(t)	Decision capability (available compute & power budget)

T(t)	Entropy/thermal cost (temperature + computational stress)

W(t)	Vitality index — positive = healthy operation

When W ≤ 0, the system automatically enters survival mode (adaptive freeze). This check can even be done with a simple analog comparator — zero CPU cycles.

________________________________________

Hardware Benchmark (STM32F401 @ 84 MHz, 3.3V)

Metric	ORAC-NT	TinyML (16-node)	Improvement

RAM	16 bytes	62,500 bytes	~4,000× smaller

Latency	~35 ns	—	Sub-microsecond

Power	9.90 μW	33.0 mW	~3,334× lower

Energy saved	Adaptive freeze	0%	Up to 51.5%

Measured on real hardware using DWT cycle counter.

________________________________________

Simulation Results (9,000 missions)

Test	Detection Rate	False Alarms	Avg Latency

Silent Drift	100%	0	~12 steps

Byzantine Adversarial	100%	0	~8 steps

Cascading Failure	100%	0	~6 steps

Final Boss (all faults)	100%	0	~9 steps

________________________________________

Quick Start

C

#include "orac_nt.h"

static OracNode_t node;                    // 16 bytes persistent state

static OracConfig_t cfg = orac_default_config();

void loop() {

    OracSensorInput_t input = { mag_A, mag_B, temp_C };

    

    OracStatus_t status;

    orac_fuse(&input, &status, baseline, range);

    uint32_t cycles = 0;

    bool active = orac_step(&node, &cfg, gradient, status.W, step, &cycles);

    

    // active == false → frozen (energy saved)

}

________________________________________

Repository Structure

text

ORAC-NT-Public/

├── orac_nt.h              ← Public API header

├── liborac_lib_final.a    ← Pre-compiled evaluation library

├── main.c                 ← Example

├── LICENSE

├── README.md

└── results/

    └── orac_v54_proof.png

Note: Core implementation is proprietary (delivered as static library). Full source available under paid licenses.

________________________________________

Licensing & Pricing

Tier	Access	Price

Evaluation	Header + .a library	Free

Developer Kit	Full source + docs + 1 year support	€500

Production	Full source + support + IP rights	€25,000

For licensing: kretski1@gmail.com