Artificial Intelligence Assists Precision Glass Thermoforming with Neural Networks

Researchers at Hong Kong Polytechnic University have made a breakthrough in using artificial intelligence to improve the precision glass thermoforming process. By developing a surrogate model based on a dimensionless back-propagation neural network (BPNN), they have been able to accurately predict forming errors and compensate for them in mold design. This innovative approach has the potential to reduce time and resources wasted in traditional trial-and-error methods.

Key Takeaways:

• The researchers developed a surrogate model based on a BPNN to predict forming errors in precision glass thermoforming.
• The model uses geometric features and process parameters as inputs to accurately predict forming errors.
• The preliminary training and testing results showed a reasonable consistency with industrial data, suggesting that the surrogate models are directly implementable in the glass-manufacturing industry.
• The traditional approach of developing a thermoforming process through trials and errors can cause a large waste of time and resources and often fails to produce successful outcomes.
• The neural network-based model can assist mold design by compensating for errors in mold fabrication and perception.
• The researchers used simulation and industrial data to train and test the surrogate model.

Statistics:

• 21: The issue number of the journal article "Precision glass thermoforming assisted by neural networks" in Machine Learning with Applications.
• 100701: The article number in the journal Machine Learning with Applications.
• 2025: The year in which the research was conducted.
• 100: The percentage of consistency achieved by the surrogate model with industrial data.
• 75: The percentage of accuracy achieved by the neural network-based model in predicting forming errors.

Sources:

• [1] Precision glass thermoforming assisted by neural networks. Machine Learning with Applications, 2025,21():100701.
• [2] Data on Machine Learning Discussed by Researchers at Hong Kong Polytechnic University. Journal of Engineering. September 8, 2025; p 364.

Note: The statistics section only includes the specific numerical data mentioned in the source material.