Encoder-Decoder Network for Sintering Burning Through Point Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Predicting the burning through point (BTP) in the sintering process is challenging due to its lag behind the sintering process, making it difficult to control and affecting the quality and output of sinter ore.
Innovation Solution
A method using an encoder-decoder network with temporal and spatial attention mechanisms is employed to predict BTP by selecting auxiliary variables, preprocessing data, and establishing a prediction model that captures time-series dynamics and correlations, allowing for real-time adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If BTP is detected by traditional methods, then measurement is obtained, but the detection lags behind the sintering process by about 40 minutes making control difficult
Solution Approach 1:
The encoder-decoder prediction model performs preliminary prediction of BTP during the sintering process before the actual burning through occurs. By using real-time process parameters (exhaust gas temperature, airflow rate, material composition) as input features, the model predicts future BTP values, enabling advance control actions rather than reactive adjustments after the 40-minute detection lag.
2Stability of the object's composition
If real-time prediction of BTP is implemented, then control stability is improved, but the model complexity increases requiring encoder-decoder network with attention mechanisms
Solution Approach 1:
The encoder-decoder architecture acts as an intermediary between raw process parameters and BTP prediction. The encoder transforms input sequences into compressed representations, while the decoder generates predictions. Attention mechanisms serve as intermediaries that selectively focus on relevant time steps and features, enabling the model to capture temporal dependencies without requiring excessive complexity in the overall system design.
3Measurement precision
If auxiliary variables are selected for prediction, then prediction accuracy is improved, but data preprocessing and feature selection complexity increases
Solution Approach 1:
The feature selection process is segmented into distinct categories: process parameters (exhaust gas temperature, airflow rate), material parameters (composition, moisture content), and operational parameters (sintering speed, belt conveyor speed). This segmentation allows systematic preprocessing of each category with appropriate techniques, managing complexity by treating different feature types independently rather than as a monolithic preprocessing step.
Data Source
AI summary
A method for predicting burning through point (BTP) based on an encoder-decoder network is provided, which belongs to a field of soft-sensing modeling in an industrial process. A BTP prediction model based on the encoder-decoder network with a temporal attention mechanism and a spatial attention mechanism is developed according to data acquired during an operation of a sintering machine, where the temporal attention mechanism is used to characterize temporal dynamics of samples, and the spatial attention mechanism is used to capture a correlation between an object variable and an advanced feature, to improve accuracy and robustness of the model. With the model, BTP in a sintering process can be predicted in real time, which has great practical significance for on-site process guidance and parameter adjustment.