StableFlame provides an AI-driven digital twin for municipal solid waste bunkers to optimize thermal waste treatment processes. This self-learning system visualizes waste calorific value and calculates optimal homogenization strategies to maximize energy recovery. The technology reduces operational costs, lowers CO2 emissions, and increases steam throughput in waste-to-energy plants.
Funding
Funding not disclosed
Founders
Product
Problem
Thermal waste‑to‑energy plants often lack precise, real‑time information on the calorific value distribution within their waste bunkers. Operators must rely on manual visual assessments that vary by operator experience, leading to sub‑optimal bunker homogenization, higher auxiliary fuel consumption, increased CO₂ emissions, and elevated operating costs.
Solution
StableFlame delivers an AI‑driven digital twin of the waste bunker that continuously maps the calorific value of the stored waste in three dimensions. By processing images from bunker‑mounted cameras, the system estimates local heating values and tracks material movements as the crane loads and transports waste. The self‑learning model refines its predictions using plant performance data, ensuring accuracy across different waste streams and plant configurations. An optimization engine uses the digital twin data to generate optimal homogenization strategies and feed‑rate plans that keep the furnace operating at a stable, high‑efficiency point. Operators can interact with a visual UI for semi‑automatic guidance, while a fully automated mode can execute crane actions without human intervention. The solution runs on‑premise within the plant’s IT environment and integrates with existing control systems via standard APIs.
Target Audience
Primary customers are operators and engineering teams of thermal waste‑to‑energy plants that use a single waste bunker per boiler and seek to improve combustion efficiency and reduce emissions.
Features
- 3‑D digital twin visualizing calorific value per bunker cell, updated in real time from camera feeds
- AI‑based calorific value estimation using computer‑vision models trained on plant‑specific waste data
- Continuous self‑learning loop that calibrates predictions with combustion performance metrics
- Optimization algorithm that plans crane movements for optimal waste homogenization and trichter (feed) loading, respecting plant‑specific constraints (drying time, feed‑rate limits)
- Semi‑automatic mode with an intuitive UI that advises crane operators on optimal pick‑and‑place actions
- Fully automated mode that autonomously schedules and executes crane actions for complete process automation
- On‑premise deployment on a virtual machine with VPN‑based remote monitoring and maintenance
- Standardized API (REST/OPC‑UA) for integration with existing plant control and data acquisition systems