Silvanet sensors use a gas sensor that is an AI-enabled low-power device that measures air pressure, temperature and humidity and detects gases such as Volatile Organic Compounds (VOCs) and carbon monoxide. This gas sensor uses a Dryad-developed algorithm to detect fires, even in their smoldering phase.
A smoldering fire is defined to be a slow, flameless combustion of a biomass material such as forest floor material, branches, leaves.
Using the Dryad-developed algorithm, the Silvanet sensor calibrates normal air around a Silvanet sensor.
The calibration period is approximately 14 days after deployment. During this period the sensor detects existing VOC compounds (such as diesel exhaust) to form a base "air quality". After this calibration period, the sensor is ready to detect any deviation from the normal air. If a deviation is detected, this triggers a process to determine if the deviation is, in fact, the result of gasses produced by the smoldering phase of a fire.
Verification consists of two phases described below.
Fire detection phases
The Silvanet sensor has two phases for detecting fires:
- Phase 1: Yellow Warning
- Phase 2: Red Alerts
Phase 1 (Yellow) Warnings
During a Phase I fire verification process, the sensors check the air quality index (AQI) once per minute.
If the air quality has significantly deteriorated within a short period of time (using a default sliding time window of 20 minutes), the sensors issue a Phase 1 or Yellow Warning. Phase 1 (yellow) warnings can be triggered by several events including fumes of car exhausts or other environmental triggers.
A Phase 1 alert means that the sensors have "smelled" something unusual. The sensitivity of the Phase 1 algorithm (threshold and sliding time window) can be configured remotely by Dryad’s support team.
In the future, Dryad is planning to enable self-configuration by customers using the Site Management app as well as an adaptive algorithm aiming to reduce the number of Yellow Warnings over time.
Phase 2 (Red) Alerts
When the sensor detects a Phase 1 (Yellow) alert, it automatically starts a gas scan which can take 1-3 minutes to complete.
During the gas scan, the sensor measures hydrogen, carbon monoxide and other volatile organic compounds. It uses a built-in Artificial Intelligence (AI) / Machine Learning (ML) model to compare the measurements with pre-trained models to reliably detect the gas composition typical for a wildfire.
If the AI determines that the gas detected is a fire, the sensor immediately sends a Phase 2 (Red) Alert. Models are pre-trained in the Dryad laboratory and can be fine-tuned / programmed specifically for the species of trees present in a particular deployment. Updated models can be remotely installed in the sensors by Dryad’s support team.
After the sensors have detected a fire, an email alert will be sent to the appropriate users. Additionally, in the Site Management app, a Fire Alert appears.
In the future, Dryad is planning to enable customers to update the ML models using the Site Management app.
For more information about alerts, see Fire alerts.