AI SENSING FOR PHYSICAL SYSTEMS

Hear what machines cannot tell you.

Resona turns sound and vibration into a real-time model of physical state.

ACQUISITION — CH 01 / 48 kHz● LISTENING
10 kHz5 kHz1 kHz0 Hz
TIME →WINDOW 4096 · HOP 512
SIGNAL ACQUIRED

62.4 dB

RMS · 1 s WINDOW
ROTATIONAL SIGNATURE

24.80 Hz

PHYSICAL STATE INFERENCE
BEARING RESPONSE

0.42 g

PHYSICAL STATE INFERENCE
STRUCTURAL RESONANCE

4.71 kHz

PHYSICAL STATE INFERENCE
02CONCEPT

Every physical system has a voice.

Machines constantly emit acoustic and vibrational signals as their internal components move, interact, heat, deform, wear, and change state.

These signals contain information about the physical system.

Resona learns to decode that information.

PHYSICAL SYSTEM
mechanical movementfrictionresonanceimpactflowdeformation
ACOUSTIC / VIBRATIONAL SIGNAL
RESONA MODEL
PHYSICAL STATE
03THE SIGNAL

The interesting part is usually below the noise.

SPECTROGRAM — 0–20 kHz · LOG SCALE● UPDATING
HARMONIC

integer multiples of a rotational fundamental

RESONANCE

structural modes excited by broadband energy

TRANSIENT

short impulsive events, periodic or sporadic

MODULATION

sidebands around a carrier frequency

ANOMALY

structure absent from the system's own history

RAW SIGNAL
INTERPRETED SIGNAL

Resona does not simply classify sounds. It learns the relationship between acoustic signatures and physical states.

04THE PHYSICAL STATE

From sound to something you can act on.

SIGNAL01
  • Frequency
  • Amplitude
  • Harmonics
  • Transient response
INFERENCE02
  • Rotation
  • Load
  • Friction
  • Vibration mode
  • Structural response
STATE03
  • NORMAL
  • DEGRADED
  • UNSTABLE
  • FAILURE SIGNATURE
STATUS INTERFACEILLUSTRATIVE EXAMPLE
SYSTEM
INDUSTRIAL PUMP 04
STATE
DEGRADING
SIGNATURE
CAVITATION-LIKE
FIRST DETECTED
18:42:07
CONFIDENCE
94.2 %
NOTE

The interface above is an illustrative UI example. It does not represent a real customer result or a validated measurement.

05SENSING MAP

The same microphone can hear more than one failure.

SENSOR NODE

MIC + ACCELEROMETER

48.0 kHz

ONE SENSING MODALITY
MANY PHYSICAL STATES
MOTOR

harmonic series

PUMP

broadband flow noise

BEARING

periodic transients

PIPE

resonant column

TRANSFORMER

line-frequency hum

ROBOT JOINT

actuation envelope

BATTERY

micro-acoustic events

STRUCTURE

modal response

Resona extracts different physical information from the same underlying acoustic and vibrational signal, depending on the system being observed.

06MODEL STACK

Physics leaves structure in the noise.

01RAW
Microphone / accelerometer

Continuous acquisition, no internal access required.

02REPRESENTATION
Time-frequency representation

Signal is resolved in time and frequency simultaneously.

03PHYSICAL FEATURES
HarmonicsResonance modesTransient eventsSpectral shiftsPeriodic structure

Features are physically motivated, not perceptual.

04MODEL
Multimodal / temporal AI

Models temporal relationships, not audio similarity.

05INFERENCE
Physical state

State is inferred, not thresholded.

06PREDICTION
Emerging change

Deviation is evaluated against the system's own history.

07APPLICATIONS
01

ROTATING MACHINERY

Hear changes in bearings, shafts, gears, and motors.

SPECTRAL VIEWILLUSTRATIVE
02

FLUID SYSTEMS

Identify changes in flow, cavitation, pressure-related behavior, and pipe conditions.

SPECTRAL VIEWILLUSTRATIVE
03

STRUCTURES

Detect changes in vibration modes and structural response.

SPECTRAL VIEWILLUSTRATIVE
04

ROBOTICS

Interpret mechanical interactions through acoustic and vibrational feedback.

SPECTRAL VIEWILLUSTRATIVE
05

ENERGY SYSTEMS

Listen for physical signatures inside complex equipment.

SPECTRAL VIEWILLUSTRATIVE
08THE DIFFERENCE

Sensors measure. Resona interprets.

TRADITIONAL SENSING
Sensor
Measurement
Threshold
Alarm
RESONA
Sound / vibration
Representation
Physical inference
State model
Evolving signature
MEASUREMENTbecoming

UNDERSTANDING

Resona does not replace conventional sensors. It adds a layer of interpretation on top of signals that already exist, or that can be collected with low-cost acoustic and vibration hardware.

09LONGITUDINAL MODEL

A machine that listens gets a new kind of memory.

SIGNATURE HISTORY — 147 DAYS● TRACKING
DAY 01

Normal signature

DAY 30

Minor spectral shift

DAY 72

New harmonic

DAY 110

Repeated transient

DAY 147

State transition

SIGNAL HISTORY
BASELINE
DEVIATION
PATTERN
PHYSICAL CHANGE

Instead of treating each measurement independently, Resona builds a longitudinal model. It continuously learns what normal sounds like for each physical system and detects meaningful deviations from that system's own history.

Machines already speak.

We are teaching them to listen to themselves.

RESONA — AI SENSING FOR PHYSICAL SYSTEMS