Optimizing Temperature Setting for Decomposition Furnace Based on Attention Mechanism and Neural Networks.

Area of Science:

• Industrial Engineering
• Artificial Intelligence
• Materials Science

Background:

• Accurate temperature control in cement decomposition furnaces is crucial for operational stability and high-quality cement production.
• Existing methods may lack the sophistication to handle complex thermal dynamics and optimize settings effectively.

Purpose of the Study:

• To develop and validate a novel deep learning model, CNN-LSTM-A, for optimizing decomposition furnace temperature settings.
• To enhance prediction accuracy and operational efficiency in cement manufacturing processes.

Main Methods:

• Proposed a CNN-LSTM-A model integrating Convolutional Neural Networks (CNNs) for spatial features, Long Short-Term Memory networks (LSTMs) for temporal data, and attention mechanisms for weight optimization.
• Utilized features selected by Least Absolute Shrinkage and Selection Operator (Lasso) alongside expert-defined inputs.
• Collected real-world production data from a cement factory for empirical validation and hyperparameter analysis.

Main Results:

• The CNN-LSTM-A model demonstrated superior prediction accuracy compared to baseline LSTM, deep-convolution-based LSTM, and attention-based LSTM models.
• Experimental results confirmed the model's effectiveness in optimizing temperature settings under real-world conditions.
• Analysis of hyperparameter impacts provided insights for model tuning and deployment.

Conclusions:

• The proposed CNN-LSTM-A model offers a significant advancement in optimizing decomposition furnace temperature control.
• The model shows strong potential for widespread adoption in cement plants to automate and improve furnace operations.
• This approach contributes to enhanced cement quality and manufacturing efficiency through intelligent control.