Understand your machines — without adding more sensors.

Berrytron creates virtual sensors for heavy machinery. We use open data and real machine signals to build continuously calibrated models — generating high-quality synthetic data so OEMs and technology developers can improve AI and prediction models without expensive sensor setups.

Physical sensors cannot measure everything.

Heavy machines operate in complex environments. Many important conditions — such as internal loads, structural stress, wear, fatigue, payload behaviour, and safety limits — are difficult, expensive, or impossible to measure directly with physical sensors.

Adding more sensors increases cost, complexity, installation time, and maintenance effort. Most companies collect machine data, but raw data alone does not explain the full physical reality of the machine.

This is where Berrytron comes in.

Excavator operator's hands on joystick controls inside machinery cabin during construction site operation

How It Works

We build generic models, calibrate them with real machine data, and generate synthetic data to improve AI and prediction models.

Overhead view of autonomous delivery van in testing facility with technicians monitoring sensor calibration and perception systems

Step 1: Build a generic machine model

Instead of relying on expensive proprietary CAD models, we start from generic machine models built from open data, data spaces, and engineering knowledge.

Technician holding tablet displaying energy infrastructure simulation model with interconnected system diagrams and real-time operational data

Step 2: Connect real machine data

The model receives real operating signals from machines or a fleet. The more data it receives, the closer it gets to the physical reality of the machine.

Isometric 3D render of complex prototype component with visible internal cooling channels and optimized geometry under cyan lighting.

Step 3: Calibrate the model toward reality

The model is continuously updated with real operating data so it behaves closer to the actual machine environment — improving accuracy over time.

Step 4: Generate virtual sensors and synthetic data

The calibrated model estimates hidden physical conditions (virtual sensors) and generates realistic synthetic data. OEMs and AI teams use this data to train better prediction models without expensive sensor installations.