A soft sensor model based on long&short-term memory dual pathways convolutional gated recurrent unit network for predicting cement specific surface area

Area of Science:

• Materials Science
• Chemical Engineering
• Artificial Intelligence

Background:

• Specific surface area is crucial for cement quality.
• Industrial cement data presents challenges like time delays and non-linearity.
• Accurate online monitoring models are difficult to establish due to data complexity.

Purpose of the Study:

• To develop an accurate soft sensor model for online prediction of cement specific surface area.
• To address the limitations of traditional models in handling complex industrial data.

Main Methods:

• A novel L/S-ConvGRU network is proposed, modifying the gated recurrent unit (GRU) with parameters L and S.
• Dual spatio-temporal feature extraction pathways (long-term memory and short-term dependence) are designed.
• Features are extracted and fused within the L/S-ConvGRU model for prediction.

Main Results:

• The L/S-ConvGRU model demonstrated enhanced feature extraction capabilities.
• The dual pathways effectively captured long-term and short-term time-varying delays.
• The model achieved higher precision and better generalization on cement plant data.

Conclusions:

• The L/S-ConvGRU model offers a robust solution for online specific surface area prediction in cement production.
• The proposed method overcomes limitations of traditional models in complex industrial environments.
• This approach contributes to improved cement quality control through advanced AI techniques.