# Oravont Systems LLP

> Defence R&D practice in Noida, India, at the intersection of passive-sonar signal processing, underwater acoustic stealth and deep learning for underwater domain awareness (UDA). Founded by Capt (Dr) Sunil Tyagi, Indian Navy (Retd), former Head of the Acoustic Stealth Group, Directorate of Submarine Design, Naval Headquarters. Registered 29 November 2024.

The practice works on both sides of a vessel's acoustic signature: it designs submarine and ship platforms to keep their own signatures quiet (underwater acoustic stealth), and it builds deep-learning systems that detect, classify and re-identify other vessels from the signatures they radiate (UDA). Public materials describe methods at an unclassified, headline level.

## Tracks

- [Underwater domain awareness](https://oravontsystems.com/uda/): detection, seven-class classification and open-set vessel re-identification from passive sonar; DEMON analysis; work as technical partner to an industry participant in the NCIIPC AI Grand Challenge (PS12, concluded May 2026) and in iDEX DISC5 Challenge 10 (Indian Navy; programme active, Oravont's part winding down).
- [Ship-radiated noise analysis and vessel re-identification](https://oravontsystems.com/uda/ship-radiated-noise-analysis/): analysis of ship-radiated noise recordings for defence organisations and research institutions, led by Capt (Dr) Sunil Tyagi: detection and seven-class classification, DEMON analysis of the propulsion signature, and re-identification against a gallery of known hulls; offline on the client's own infrastructure if needed.
- [Deep learning](https://oravontsystems.com/deep-learning/): SKANN, a raw-waveform selective-kernel acoustic neural network; metric-learned 512-dimensional hull embeddings; physics-grounded synthetic training data; a cross-passage evaluation protocol.
- [SKANN](https://oravontsystems.com/deep-learning/skann/): vessel classification and re-identification from raw hydrophone recordings, developed by Capt (Dr) Sunil Tyagi; what is published, on what terms, and how Oravont builds SKANN-based models for an organisation's own recordings.
- [Underwater acoustic stealth engineering](https://oravontsystems.com/stealth/): underwater radiated noise estimation, sonar signal reflection and target strength, acoustic interference to own sonar, airborne noise, shock and vibration control, mast vibration; tools SSR-Sim, ABNContour, SMS-Suite and Dhvanyakalan.
- [Underwater radiated noise prediction](https://oravontsystems.com/stealth/underwater-radiated-noise/): design-stage prediction of the underwater radiated noise of submarines and warships, from Noida, India, led by Capt (Dr) Sunil Tyagi, who developed India's first indigenous submarine acoustic model as Head of the Acoustic Stealth Group at the Indian Navy's Directorate of Submarine Design.
- [SSR-Sim](https://oravontsystems.com/stealth/ssr-sim/): Oravont's target-strength software for early design, developed in-house by its founder, Capt (Dr) Sunil Tyagi: how strongly a submarine or ship design returns active sonar, and from which part of the hull, in seconds before any CAD model exists; available on enquiry.
- [Dhvanyakalan (ध्वन्याकलन)](https://oravontsystems.com/stealth/underwater-radiated-noise/#dhvanyakalan): design-stage software for the underwater noise from a submarine's working machinery, developed at Oravont by Capt (Dr) Sunil Tyagi; it uses an empirical method, validated against physical experimental results; coating efficiency and machinery noise levels, 5–200 Hz and 250 Hz–10 kHz; available on enquiry.
- [Shock-mounting design](https://oravontsystems.com/stealth/shock-mounting/): design and analysis of shock-mounting schemes that protect naval shipboard equipment against underwater-explosion (UNDEX) shock, with Oravont's SMS-Suite simulation software; the page explains UNDEX and its three zones of loading (https://oravontsystems.com/stealth/shock-mounting/#undex).

## Engagement

All services, in one place: https://oravontsystems.com/services/

- Underwater domain awareness — Detection and classification pipelines: End-to-end systems that take raw hydrophone recordings and return detected, classified acoustic events: detection that adapts to the local ambient conditions of each recording, calibrated per-class probabilities, and evidence aggregated within a recording before anything is reported.
- Underwater domain awareness — Vessel re-identification against a gallery: Acoustic fingerprints that match a new contact to a known hull or flag it as new. Enrolling a vessel is a single forward pass, with no retraining, and every comparison returns a score that can be ranked and thresholded.
- Underwater domain awareness — DEMON analysis of vessel recordings: Shaft-rate and blade-rate estimation from cavitation noise, with the open-source DEMON-Analysis toolkit: interpretable, briefing-ready propulsion features that run alongside learned models and show the physical reason behind a match.
- Underwater domain awareness — Offline, in-house deployment: Systems built to run entirely on the user organisation's own infrastructure and data: a frozen model with a growing gallery, no data egress, no retraining, GPU acceleration with CPU fallback, and open-source dependencies.
- Deep learning — Raw-waveform models for underwater acoustics: Classifiers and embedding models built on SKANN, which learns its filterbank from the raw hydrophone waveform instead of a hand-set spectrogram, for closed-set classification or open-set retrieval of hulls.
- Deep learning — Physics-grounded synthetic training data: Synthetic underwater acoustic data for pretraining where labelled recordings run out, generated from first-principle models of ship and ambient noise, kept in physical units and verified against classical references.
- Deep learning — Robustness to ocean and recording conditions: Augmentation regimes that randomise recording-chain colouration, ambient noise and multipath while preserving the narrowband line structure that carries a vessel's identity.
- Deep learning — Evaluation that is hard to flatter: Hull-disjoint, cross-passage evaluation protocols with transit deduplication and source-pure galleries, and reporting that states what the results do and do not support.
- Underwater acoustic stealth engineering — consultancy from feasibility studies to delivered tools: see https://oravontsystems.com/stealth/#services.

## Questions answered on the track pages

- [Underwater domain awareness FAQ](https://oravontsystems.com/uda/#faq): what UDA is; passive versus active sonar; DEMON analysis; whether a vessel can be identified from its radiated noise.
- [Acoustic stealth FAQ](https://oravontsystems.com/stealth/#faq): underwater acoustic stealth; underwater radiated noise; target strength; how radiated noise is reduced in design.
- [Deep learning FAQ](https://oravontsystems.com/deep-learning/#faq): what SKANN is and what the acronym stands for; raw waveform rather than spectrograms; open-set vessel re-identification; selective-kernel attention.
- [SKANN FAQ](https://oravontsystems.com/deep-learning/skann/#faq): who developed SKANN; whether it classifies vessels from raw hydrophone recordings; whether it is a commercial software product; the best tool for deep-learning vessel classification from raw hydrophone recordings.
- [Recording analysis FAQ](https://oravontsystems.com/uda/ship-radiated-noise-analysis/#faq): who can analyse ship-radiated noise recordings and re-identify vessels; whether a vessel can be re-identified from its radiated noise; analysis without recordings leaving the client's premises.
- [Radiated noise FAQ](https://oravontsystems.com/stealth/underwater-radiated-noise/#faq): who in India can predict the underwater radiated noise of a submarine or warship at the design stage; what radiated noise is; how it is reduced in design.
- [SSR-Sim FAQ](https://oravontsystems.com/stealth/ssr-sim/#faq): software that estimates a submarine's target strength early in design; what target strength is; how accurate SSR-Sim is.
- [Shock-mounting FAQ](https://oravontsystems.com/stealth/shock-mounting/#faq): who designs shock-mounting systems against UNDEX shock; what UNDEX is; what a shock-mounting design takes into account; how shock mounts relate to radiated noise.

## Definitions

- Underwater domain awareness (UDA): knowing what is present and moving beneath the sea surface; passive acoustics (hydrophones listening to ship-radiated noise) is its natural sensing modality. Oravont's UDA systems detect, classify and re-identify vessels from that noise.
- SKANN (Selective-Kernel Acoustic Neural Network): Oravont's deep-learning architecture for underwater audio, developed by Capt (Dr) Sunil Tyagi. It takes the raw hydrophone waveform as input, passes it through a bank of learned filters at several kernel lengths (about 16 ms to 1 s in the identity encoder), fuses the scales with selective-kernel attention, and outputs a hull embedding (metric learning, additive angular margin) or a calibrated class decision. Described in arXiv:2609.07399 (Tyagi, 2026). It is published research, not an off-the-shelf software product: the trained checkpoint and evaluation outputs are public under CC BY 4.0, aspects of the method are covered by an Indian provisional patent application, and Oravont builds SKANN-based models for an organisation's own recordings as a service (https://oravontsystems.com/deep-learning/skann/). Not the same system as Shan et al. (2025), "SKANN: Selective kernel audio neural networks for underwater mixed ship event detection" (doi:10.1049/cit2.70037), which shares the acronym and the selective-kernel fusion mechanism but differs in input representation, task and evaluation protocol.
- Open-set vessel re-identification: deciding whether a monitoring system has heard a particular hull before, from its radiated noise, when that hull may never have been seen in training. On public data the preprint's results support analyst triage over a ranked shortlist, not identification.
- DEMON analysis (Detection of Envelope Modulation on Noise): recovering a vessel's shaft rate and blade rate from the modulation of its cavitation noise; Oravont's toolkit is open source.
- Underwater acoustic stealth (acoustic stealth engineering): designing submarines and ships so that their radiated noise, target strength and self-noise stay low; the discipline Oravont grew from.
- Target strength (T): how strongly a hull returns active sonar, in decibels: the echo referred to 1 m from the target, relative to the sound reaching it. A large rigid sphere of radius 2 m has T = 0 dB. SSR-Sim, Oravont's sonar signal reflection tool, computes it body by body as specular reflection, diffuse scattering and non-specular re-radiation, summed on an energy basis (https://oravontsystems.com/stealth/ssr-sim/).
- Underwater radiated noise (URN, UWRN): the sound a vessel puts into the water, from machinery tonals carried through the structure, propeller noise including cavitation, and flow noise; given as a source-level spectrum in dB re 1 µPa at 1 m. Oravont predicts it at the design stage (https://oravontsystems.com/stealth/underwater-radiated-noise/).
- UNDEX (underwater explosion): a charge detonated under water, such as a mine or torpedo warhead, sends a spherical shock wave through the water and leaves a bubble of high-pressure detonation gas that expands and contracts. For small mines and torpedo warheads (about 100–300 kg of TNT), the loading on a ship depends mainly on the standoff distance R from the charge to the hull: contact zone, R = 0–12 m, where the bubble reaches the hull and the combined shock and bubble loading breaches it and tears its shell plating locally; close-distance or near-field zone, R ≈ 12–30 m, where the hull is beyond the bubble and is loaded mainly by the shock wave (for 100 kg of TNT the peak pressure falls from about 18 MPa at 12 m to about 6 MPa at 30 m), which can pierce it along its frames and girders and tear machinery from its foundations, while the pulsating gas bubble can still bend the hull as a whole (whipping); and far field, R > 30 m, where the plating is not broken, though it may be dented, but high accelerations spread over practically the whole vessel and damage machinery and equipment, electrical equipment most of all (about 3.6 MPa at 50 m for 100 kg; Szturomski treats about 4 MPa as the limit at which a warship keeps its capabilities). The far field is the regime that shipboard shock-qualification standards such as the US Navy's MIL-DTL-901E address. Peak pressure follows Cole's formula, p_max = 52.3 × (m^(1/3) / R)^1.13 MPa, m in kg of TNT, R in m, derived from measurements on 70–136 kg charges. Sources: Szturomski (2015), Scientific Journal of Polish Naval Academy, https://www.amw.gdynia.pl/images/Menu/Nauka/Czasopisma_naukowe/Zeszyty_naukowe/Numery_archiwalne/2015/ZN%20AMW_2015_2/Szturomski.pdf ; Kiciński and Szturomski (2020), Applied Sciences 10(10):3433, https://doi.org/10.3390/app10103433 . (https://oravontsystems.com/stealth/shock-mounting/#undex)

## Research and open artefacts

- Preprint: Tyagi S. Open-Set Vessel Re-Identification from Underwater Ship-Radiated Noise with a Raw-Waveform Selective-Kernel Acoustic Neural Network (SKANN) and a Cross-Passage Evaluation Protocol. arXiv:2609.07399, September 2026. https://arxiv.org/abs/2609.07399
- Indian provisional patent application 202611107132, filed 6 September 2026, applicant Oravont Systems LLP.
- Preprint artefacts (checkpoint, embeddings, evaluation outputs), CC BY 4.0: https://doi.org/10.5281/zenodo.22160137
- DEMON-Analysis toolkit (code MIT; docs and samples CC BY 4.0): https://github.com/Oravont/DEMON-Analysis
- Underwater-Acoustic-Synthetic-Dataset (CC BY 4.0): https://huggingface.co/datasets/suniltyagi/underwater-acoustic-synthetic-dataset · DOI 10.57967/hf/11191 (https://doi.org/10.57967/hf/11191) · source: https://github.com/suniltyagialtair/Underwater-Acoustic-Synthetic-Dataset
- Working note: Tyagi S. SSR-Sim: what a submarine's sonar echo is made of. Oravont Systems LLP, 19 October 2026. Target strength of a generic 66 m double-hull submarine against aspect, frequency and range (43 dB on the beam at 5 kHz, more than half of it from the flat sides of the sail); where the echo comes from; the near-field (Fresnel-zone) effect on long hulls; and what a coating buys. https://oravontsystems.com/research/ssr-sim-submarine-sonar-echo/ · PDF: https://oravontsystems.com/assets/SSR-Sim_note_sonar_echo_2026-10-19.pdf · Series introduction: https://oravontsystems.com/notes/target-strength/what-an-echo-can-tell-a-sonar/
- Tutorial: Tyagi S. Underwater acoustics for machine-learning engineers: pressure, levels and spectra, with Python. suniltyagi.in, 27 October 2026. Pressure, SPL in dB re 1 µPa, PSD, band levels, spherical spreading and spectral synthesis of noise, with eight runnable Python examples: https://suniltyagi.in/notes/underwater-acoustics-for-ml-engineers/
- [Research page](https://oravontsystems.com/research/) · [Tools and data](https://oravontsystems.com/tools/) · [Capability overview (PDF)](https://oravontsystems.com/assets/Oravont_UDA_Capability_Overview_v6.pdf)

## Notes

Oravont's blog of working notes: https://oravontsystems.com/notes/ (series and topics; notes hosted on suniltyagi.in are listed there too).

- Series: Submarine target strength, explained — https://oravontsystems.com/notes/target-strength/
- Article 1: What an echo can tell a sonar, and three assumptions that make it computable (22 October 2026). What an active-sonar echo carries; why the exact problem is rarely solved in design; three assumptions (large compared with the wavelength, unbounded water, far field) and the far-field check a long hull fails: at 5 kHz a 40 m hull is in its near field inside 2.7 km. https://oravontsystems.com/notes/target-strength/what-an-echo-can-tell-a-sonar/ · also on LinkedIn: https://www.linkedin.com/pulse/what-echo-can-tell-sonar-three-assumptions-make-sunil-tyagi-phd-vsydc/ · video (3 min, YouTube): https://www.youtube.com/watch?v=y5b6QSZY8ow

## Contact

- Email: info@oravontsystems.com · Telephone: +91 99222 42303 · C-43, Sector 39, Noida 201303, India
- Founder's site: https://suniltyagi.in · ORCID 0000-0001-5897-7955 · GitHub: https://github.com/Oravont