Self-learning Control

Product · SaaS / cloud software · Operations & supply chain by Gigaton

Self-learning control for energy-intensive industries that reduces cost, complexity, and carbon.

Gigaton's Self-learning Control is a cloud-based AI platform designed for energy-intensive industries such as cement, steel, and glass production. It solves the problem of process degradation in traditional control systems by continuously retraining models to adapt to plant changes. The system works by analyzing sensor and manufacturing data to provide real-time, autonomous set-point recommendations that connect directly to a plant's Distributed Control System (DCS). It features a digital twin for safe simulation and explainable AI that visualizes the reasoning behind every autonomous decision

Deployment Time
8 weeks to go live
Hardware Requirements
No extra hardware needed
Connectivity
Direct connection to DCS
Architecture
Cloud-based platform
AI Method
Deep reinforcement learning

Overview

Gigaton's Self-learning Control is a cloud-based AI platform designed for energy-intensive industries such as cement, steel, and glass production. It solves the problem of process degradation in traditional control systems by continuously retraining models to adapt to plant changes. The system works by analyzing sensor and manufacturing data to provide real-time, autonomous set-point recommendations that connect directly to a plant's Distributed Control System (DCS). It features a digital twin for safe simulation and explainable AI that visualizes the reasoning behind every autonomous decision. The platform is delivered as a software-only solution that requires no extra hardware or plant shutdowns for deployment.

Autonomous Control

  • DCS Integration: The controller acts autonomously, connecting directly to the Distributed Control System (DCS) to execute actions.
  • Flexible Priority Shifting: Plant managers can shift focus from cost to production in a single click to align with changing strategic priorities.
  • Predictive Forecasting: The system acts ahead of problems by forecasting process variables like temperature and gas composition.

Explainability and Simulation

  • Decision Visualisation: The platform visualises the reason behind each decision to help operators understand why a specific control action was made.
  • Set-point Recommendations: The system provides clear set-point recommendations for kiln feed rate, fan speed, and pre-calciner temperature.
  • Human-in-the-loop Design: The solution is designed to augment human expertise while keeping the plant operator in control of production processes.

Industrial Performance

  • 10% Fuel Cost Reduction: The platform targets a 10% reduction in fuel costs and emissions specifically for cement production.
  • Heat Consumption Reduction: The system helps reduce specific heat consumption, with deployments showing a 2.2% reduction.
  • Quality Stability: Deployments have demonstrated a 33% reduction in C3S variability, leading to more stable clinker quality.
  • Clinker Quality Optimization: The platform enables a 25% reduction in clinker-free lime variance and eliminates analysis lag for XRD/XRF measurements.

Specifications

Deployment Time
8 weeks to go live
Hardware Requirements
No extra hardware needed
Connectivity
Direct connection to DCS
Architecture
Cloud-based platform
AI Method
Deep reinforcement learning

Added by AntSpace from public sources (gigaton.co).

Related

More products by Gigaton