How AI can provide smarter glass melting

CelSian outlines its Celfos application, which uses Artificial Intelligence (AI) to predict the future quality of glass.

In traditional glass furnace control systems, the focus has been on stabilising temperature.

However, maintaining a stable temperature doesn’t necessarily result in producing high-quality glass. The true objective of glass production isn’t simply controlling temperature - it’s to achieve optimal glass quality while minimising costs.

The challenge in glass manufacturing arises from the fact that glass quality can be influenced by factors that occur in the furnace over a span of hours, or even days.

Once the glass has passed a certain point in the furnace, there’s no opportunity for immediate corrective action. This makes it difficult for operators to make real-time adjustments to improve quality in the future.

However, what if operators had a tool that can predict glass quality a few days in advance? With this insight, they can proactively adjust furnace settings to ensure better quality outcomes.

Enter Celfos - an application that leverages AI to predict the future quality of glass. Rather than focusing solely on temperature, Celfos enables smarter decision-making by offering a detailed forecast of the future quality of the glass produced.

The core of this solution lies in CFD-powered AI models. Using data from the furnace, such as temperatures and process settings, the model can learn patterns that affect the final product quality – such as the number of bubbles or knots.

This foresight allows operators to anticipate potential issues and make adjustments before quality degradation occurs.

Challenges

The dynamic nature of a furnace complicates the task of accurately predicting quality. Moreover, the various parameters influencing glass quality are often vast and interrelated, making it a complex task to build an AI model that delivers accurate predictions.

To address these challenges, Celfos uses complex neural networks coupled with CelSian’s CFD model (GTM-X).

The CFD model simulates the flow dynamics and thermal behaviour inside the furnace, while the AI model learns from this data to predict future glass quality.

This combination of AI and CFD allows Celfos to account for the furnace’s present state and predict how adjustments to parameters, such as fuel flow or boosting, will impact future quality.

Celfos continually updates its quality predictions by accessing real-time process data. Operators are then presented with up-to-date forecasts of glass quality, enabling them to take proactive actions. The application also provides past predictions for comparison, which builds operator confidence in the system’s accuracy.

The next steps involve refining the system’s recommendations and suggesting specific actions like altering fuel flow or boosting processes to optimise future quality. This level of proactive, AI-driven decision-making can improve both glass quality and production efficiency.

Combining AI with CelSian’s CFD application GTM-X allows operators to look into the future and influence glass quality, enabling smarter, more cost-effective production practices.

The future

Celfos has successfully implemented quality prediction on a float furnace, with a prototype running since Q4 2024. The initial results show good accuracy and ongoing evaluations aim to refine and improve its performance.

The next step is to expand into container glass production, applying and adapting the model to a different process. Development is underway on a recommender system that will assist operators in optimising process settings. This system is set for testing in 2025. It is designed to help operators adjust set points efficiently, factoring in process changes over time to maintain glass quality while managing costs.

To support these advancements, CelSian has received an MIT Grant for a collaborative AI-driven research project. This project, funded by the MIT-R&D programme, aims to enhance the efficiency of glass production by using AI to improve simulation software. The goal is to reduce product losses, which will not only lower costs but also reduce CO₂, NOₓ, and SOₓ emissions.

With this new funding, CelSian is well-positioned to continue advancing its technology and contributing to a more sustainable and efficient glass industry.