Dynamic correction of soft measurement model for evaporation process parameters based on ARMA

Area of Science:

• Chemical Engineering
• Process Control
• Data Science

Background:

• Traditional static models for evaporation process parameters suffer from significant soft measurement errors due to continuity and cumulativity.
• Accurate real-time parameter estimation is crucial for optimizing industrial evaporation processes.

Purpose of the Study:

• To develop a dynamic correction method for soft measurement models of evaporation process parameters.
• To improve the prediction accuracy and dynamic estimation capabilities compared to traditional static models.

Main Methods:

• Utilized Powell's directional evolution (Powell-DE) algorithm to identify Autoregressive Moving-Average (ARMA) model orders.
• Employed a mechanism-reduced robust least squares support vector machine ensemble model to predict errors.
• Developed an error time series prediction model for dynamic error estimation.
• Integrated static and dynamic models using an entropy method.

Main Results:

• The proposed dynamic correction method significantly improved prediction accuracy for evaporation process parameters.
• The new model demonstrated enhanced capability for dynamic prediction.
• Validation using production data from an alumina plant confirmed the model's effectiveness.

Conclusions:

• The dynamic correction method effectively addresses soft measurement errors in evaporation processes.
• The integrated model provides a more accurate and dynamic approach to process parameter estimation.
• This approach offers a significant advancement over traditional static soft measurement models.