Soft-sensing Method Based on Lazy Learning Algorithm

In view of lots of unmeasured variables in industrial process,a soft sensing method using lazy learning is presented.The k-vector nearest neighbor is used to generate a neighbor of current regime in order to enhance the predictive capability of the original algorithm.The complexity of the algorithm is decreased by recursive least squares and the optimal solution is addressed by PRESS.Using this method to model ester rate of a local chemical plant,this paper obtains the maximum inaccuracy of 0.874 2%.The simulation results show that the perfect generalization performance can meet highprecision measuring requirements and the method is very understandable and easy to be implemented.