Abstract: Computational Fluid Dynamics (CFD) continues to extend its reach in different domains, and with CelSian’s simulation software, GTMX, it is possible to simulate and analyze the physics inside a glass furnace. GTMX has dedicated models that allow furnace designers to predict the factors affecting production time and quality. CelSian uses Cadence® Fidelity™ Pointwise® as the mesh generation software for refinement in locations where the physics is crucial for the glass production process. Using a computer simulation technology wherein the CAD import and meshing are carried out in Fidelity Pointwise, and the simulation is solved on GTMX, different models with multiple design criteria can be tested within the time constraints.

Introduction

The United Nations (UN) has chosen 2022 as the International Year of Glass (IYOG2022) to celebrate the transformative changes that glass has brought into our lives, from glass optical fibers for high-speed internet connectivity to new innovative glass designs for reinforcement structures in buildings and household appliances. The UN has addressed a few target points that they plan to achieve by 2030, which include:

• Organizing events that highlight the link between glass, art, and culture
• Build global alliances focused on science and engineering for young minds
• Promote glass research in academia, industry, and the public domain
• Demonstrate the changing role of glass in advancing civilization

In today’s technological era, a faster turnaround time is essential to satiate the growing demands of customers, whether in communication or healthcare. With sustainable or renewable goals to accomplish, it is necessary to test an enormous number of glass models before the best solution is available for commercialization. This necessitates a computer simulation technology that can predict the life of the glass furnace, the impact of renewable sources of energy on the furnace life, the impact on the burner, and the product quality, instead of relying on experimentation that is both time-consuming and expensive.

The development of computer simulation began in the previous century. The initial goal was to develop software that could understand and predict heat and mass transfer, fluid flow, chemical reactions, and other related processes that occur in engineering equipment, in the natural environment, and living organisms. The processes of heat transfer and fluid flow pervade many aspects of our life.

In real life, the best approach to solve a complex problem is to divide it into smaller pieces and to solve it piece by piece. Some of the first simulation methods were based on this principle and work as follows: a process is cut into many small boxes, and for each box, the laws of physics are applied (Figure 1). There is an exchange of information between the boxes for the entire process. This is the basic principle of computational simulation tools as we know them today, called computational fluid dynamics, or CFD.

Figure 1. Example of an energy flow balance over a control volume 

Computing the Weather

Weather forecasting is a well-known example of the simulation of wind, rain, and temperature that will be undertaken to predict the weather in the near future. Figure 2 is an example of a climate model where the earth’s atmosphere is discretized into several connecting boxes, and in each box, the local situation is studied using a set of equations. The model includes all the basic factors that affect the climate, such as the atmosphere, ocean, land, and sea ice. The boundaries of this model include the incident solar energy and the heat reflected into space.

Weather forecast models are incorporated into our daily lives; we use them to schedule our day-to-day activities, such as our decision on how to commute to the office

Figure 2. U.S. NOAA, a model based on the ocean and atmosphere interactions[2]

CelSian introduced a simulation software for glass furnace simulations, named GTM-X, similar to the software used for weather forecasts or those used in the automotive and airplane industries. The software is equipped with dedicated models that allow it to accurately simulate the phenomena inside a glass furnace. Data from laboratory experiments and many years of field experience are used to define and validate dedicated models like the model for the batch blanket [4] and the model to calculate the dissolution of sand grains [3]. The laboratory at CelSian supports the modeling of cold top furnaces, as there is a sudden surge in demand for full electric furnace technology (Figure 3).

Figure 3. Glass melt in the cold top furnace at CelSian’s laboratory was captured using an infrared camera (left); the top view of the furnace where sand (batch) is used for molding the glass, and heat is applied from the sides of the furnace (right).

GTM-X is used by CelSian’s in-house engineers for process improvement projects and by external teams to analyze the engineering requirements of glass producers and furnace designers. The software helps address current industry requirements such as low energy use, good quality product, low emissions, and longer furnace life. The continuous development of the software is driven by valuable feedback from the industry. One example is implementing a special radiation model for low iron glass melts [6].

To generate the computational grid (i.e., the collection of boxes described above) for the furnace, CelSian uses Fidelity Pointwise. This software directly imports the glass furnace’s shape from CAD drawings. Figure 4 shows an example of a computational grid of a glass furnace generated using Fidelity Pointwise.

Figure 4. Example of a computational grid generated using Cadence Fidelity Pointwise. The lines indicate the outline of each “box” or cell in the grid. The grid adheres to the shape of the furnace, and the refinement of the cells can be controlled in locations where smaller cells are needed to accurately resolve the physics.

Reducing carbon footprint is one of the biggest challenges furnace designers must tackle. Current initiatives focus on reducing energy use and investigating alternative energy sources such as biofuel, electrical energy, and hydrogen. Although the best choice will likely be driven by availability, technical challenges can be solved using computer modeling.

In a conventional furnace, electrical boosting energy should not be increased without carefully inspecting its impact on the product quality. Temperature differences drive the flow of the glass melt, and the wrong location of the electrodes can lead to production issues.

Figure 5. Example of a hybrid furnace design (courtesy of Fives Stein) where the red color denotes big bubbles and the green ones are the tiny bubbles trapped in the glass.

While using an alternative source of energy such as hydrogen, the burner design, the impact on product quality, and the life of the furnace need to be considered during the furnace design. Figure 6 shows the result of a project in which modeling is combined with lab-scale testing to evaluate the impact of hydrogen combustion on heat transfer.

Figure 6. Flame shape and thermal behavior of natural gas (left) and hydrogen gas (right). The intensity of the color hues in the hydrogen flame does not determine its heat capacity.

Defects such as solid inclusions, bubbles, and blisters are investigated in a laboratory, and the source of these defects can be traced via computer modeling. Once the source is known, possible solutions are tested in the computer model, and the most promising solution is applied in the real furnace. This approach has been used in many projects.

For furnaces in which the product color changes, it is profitable to reduce the time by which the product is out of specification. Figure 7 is a model of a float furnace that is progressing towards a color change. The simulation reveals that the changing composition of iron oxide in the glass melt affects the product color and its radiative properties. These properties need to be considered as they affect the crown temperatures in the furnace, and the simulation provides insights into the energy that needs to be corrected to prevent overheating of the crown. 

Figure 7. Time transient modeling of a color change and the impact on temperatures.

Having a time-dependent computer model opens the road to other applications. Since the response to changes in a glass furnace is very slow, a computer model like the one mentioned in this article can help predict how and when the changes can affect the glass production process. In short, the model is a digital twin of the real furnace with many virtual sensors beneath the melt surface.

The simulation, as described above, is almost as fast as in real-time but still too slow to steer the glass furnace directly. For direct control of the furnace, a much faster model is created from the CFD model by performing a series of tests on the digital twin. This method is safe for the furnace, does not impact product quality, and results in a control model without information from unknown disturbances.

Figure 8. Using a detailed furnace simulation to generate a fast and accurate control model.

The reduced model can be directly used in a model-based predictive control (MPC) or for glass melt temperature estimations in the furnace and the setpoints can be modulated to maintain a stable temperature that is within the allowed limits [7]. Using this technique, it has been observed that the standard deviation for average glass melt temperatures reduced from 1° C to 0.25 °C. Further, the controller helped achieve temperature stability, and the unstable period after a large process change was reduced to a minimum.

If you are interested in learning more about GTM-X for glass simulation, contact CelSian at infodesk@celsian.nl.

Regardless of what you manufacture, grid generation from Cadence Fidelity Pointwise can support your computational simulation. Want to find out how? 

Acknowledgment

This article is based on an original written by Andries Habraken from CelSian.

References

[1] Numerical heat transfer and fluid flow, Suhas V. Patankar, 1980, McGraw-Hill Book Company

[2] National Ocean Service, a model based on ocean and atmosphere interactions, January 21, 2021.

[3] R.G.C. Beerkens, H.P.H. Muijsenberg, T. van der Heijden: Modelling of sand grain dissolution in industrial glass melting tanks, Glastech. Ber. 67, 179–188 (1994)

[4] A.J. Faber, R.G.C. Beerkens, H. de Waal: Thermal behaviour of glass batch on batch heating, Glastech. Ber. 65(7), 177–185 (1992)

[5] R.G.C. Beerkens: Analysis of elementary process steps in industrial glass melting tanks – Some ideas on innovations in industrial glass melting, Ceram. Silik. 52(4), 206–217 (2008)

[6] A.M. Lankhorst, L. Thielen, P.J.P.M. Simons, A.F.J.A. Habraken: Proper modeling of radiative heat transfer in clear glass melts. In: 73rd Conference on Glass Problems, ed. by S.K. Sundaram (Wiley, Chichester 2013) pp. 249–258

[7] L. Huisman: Control of Glass Melting Processes Based on Reduced CFD Models, PhD Thesis (Technical University TU Eindhoven, Eindhoven 2005)