From signal to decision, in your browser.
This page reproduces the module's real chain: telemetry (vibration, temperature, current) runs through the deterministic rules and comes out as typed decisions — log, monitoring, alert.
Nothing leaves your machine. This computation runs on this page, with no server, no connection. That is the module's principle: the decision is made on site.
The figures shown are simulated, not measured on a real machine. The rules have the same structure as those in the firmware (thresholds, hysteresis, wear, trend), but the parameters shown are illustrative values: the module's actual settings are part of the lab's know-how and are shared only under a confidentiality agreement.
The decision engine
Thresholds, hysteresis, wear calculation, classes and alerts: the demonstration follows the same decision chain as the embedded firmware, with illustrative parameters. The type of reasoning is genuine; the actual settings are not published.
The input signal
Vibration, temperature and power are simulated, with plausible values for a small rotating machine — not yet recorded on a physical test bench.
What this changes, and what it does not. The day the module reads real sensors, only the input changes: the decision is already computed by the same code. That is what makes it useful before the hardware.
Note on the use of the term “artificial intelligence”. This demonstration uses no neural network: it shows the deterministic layer — fixed rules, a given input always produces the same output. The embedded neural network exists (compiled for a microcontroller, validated in tests — not yet flashed to a chip) but remains a complementary vote: it never carries an alert on its own.
What the module measures.
Sensors
Module decision
Local log (excerpt)
| t (s) | class | wear | p | vibration | T °C | W | decision |
|---|---|---|---|---|---|---|---|
| The log fills up as the signal runs… | |||||||
Observe, warn, estimate.
Monitor
The three sensors are recorded continuously. The log stays on the machine — the client retrieves it on a card or over a cable.
Warn
Comparison with normal operation. A drift triggers monitoring, then an alert: the module warns, the staff decides.
Predict
The wear trend is extrapolated: the module estimates the approximate time before the threshold, to plan maintenance.
The module never controls the machine. It observes, it warns, it estimates. The action (slow down, stop) remains with the staff — that is what makes it non-intrusive and risk-free for production.