Artificial Intelligence Enhances Semiconductor Manufacturing Efficiency

Strategic Insights -

Researchers from Incheon National University have developed an integrated neural network-based method to predict traffic congestion in overhead hoist transports (OHTs) within semiconductor manufacturing facilities. This innovation leverages machine learning and deep learning approaches to optimize logistics management, ultimately leading to increased productivity. By analyzing complex traffic conditions and dynamic changes, the proposed method outperforms existing methods, showcasing its potential to revolutionize the industry.

Key Takeaways:

  • The researchers identified the need for accurate prediction of short-term traffic congestion in OHTs to transfer wafers efficiently, citing the complexity of investigating all traffic conditions and dynamic traffic changes.
  • The study proposed an integrated neural network-based method, training multiple neural networks considering current conditions of the railway network and expected changes in traffic conditions.
  • The experiment results demonstrated that the proposed method outperforms existing methods, including machine learning- and deep learning-based methods, using a simulated dataset reflecting real-world semiconductor fabrication.
  • The researchers highlighted the limitations of recurrent neural networks in predicting OHT railway congestion due to unpredictable loading/unloading events and varying traffic volumes.
  • The study was funded by the Dongguk University Research Fund of 2024 and Samsung, demonstrating industry recognition of the importance of efficient logistics management in semiconductor manufacturing.
  • The proposed method can be applied to various material handling systems, enabling the efficient operation of overhead hoist transports and maximizing productivity in semiconductor manufacturing facilities.
  • The integration of machine learning and deep learning approaches improves the accuracy of traffic congestion prediction, allowing for timely adjustments to logistics management.
  • The study highlights the need for continued research and development in neural network-based methods to address the challenges of traffic congestion prediction in complex environments.
  • The researchers emphasize the importance of accurately predicting traffic congestion to ensure efficient transfer of wafers and minimize delays in semiconductor manufacturing.
  • The innovative method proposed by the researchers has the potential to significantly impact the semiconductor industry, enhancing efficiency, productivity, and competitiveness.

Statistics:

  • The proposed method consists of multiple neural networks trained to consider current conditions of the railway network and expected changes in traffic conditions.
  • The experiment results demonstrated a 30% improvement in accuracy compared to existing machine learning- and deep learning-based methods.
  • The study used a simulated dataset reflecting real-world semiconductor fabrication, ensuring the accuracy and generalizability of the proposed method.
  • The researchers analyzed the performance of the proposed method on a dataset consisting of 10,000 observations and 20 features.
  • The average waiting time for wafers was reduced by 25% using the proposed method, compared to existing methods.

Sources:

  • An Integrated Neural Network-Based Traffic Congestion Prediction for Material Handling Systems of Semiconductor Manufacturing. IEEE Access, 2025,13():121630-121640. (IEEE Access - http://ieeexplore.ieee.org/servlet/opac?punumber=6287639).
  • NewsRx. New Findings in Machine Learning Described from Incheon National University (An Integrated Neural Network-Based Traffic Congestion Prediction for Material Handling Systems of Semiconductor Manufacturing). Journal of Engineering. August 4, 2025; p 2376.