Design And Realization Of CAPL Steel Quality Prediction And Process Monitoring System Based On PLS

With the development of science and technology, steel is as an important production material in various fields, and the requirements of steel performance have become more sophisticated today. Continuous annealing line is a key process in improving steel's mechanical properties. But in the real production process, the mechanism of continuous annealing is complicated and the detection of steel quality has a long time lag, which brings grate obstacle to improve steel quality.As one of multivariate statistical methods, Partial Least Squares (PLS) modeling method doesn't rely on the mechanism. The model has a clear structure. There is a definite relationship between process and quality variables. This thesis adopts the PLS to build the relationship between process and quality variables. It can guarantee the safety of production process by realizing quality prediction and process monitoring timely.This thesis focus on design and realization of steel quality prediction and process monitoring for continuous annealing line based on PLS algorithm. Firstly, the database of quality prediction and process monitoring system is built after analyzing the input and output of the model. Then PLS offline model is trained by using treated data, at the same time, the control limits of online model are determined. And the precision of online model is tested. At last, the system interface and data preprocessing module are developed according to the output of the model.System database manages complicated data for quality prediction and process monitoring system. Online model predicts steel quality and monitors the process. Data interface and data preprocessing module guarantees data exchange of the whole system. System interface shows the output and provides convenience to workers'operation. These four parts compose the whole quality prediction and process monitoring system.Now the system is successfully running in the site. It proves the validity of PLS algorithm for quality prediction and process monitoring. Meanwhile, it also provides reliable basis for applying multivariate statistical methods to large-scale industrial process.