Incorporation of Prior Knowledge in Adaptive Learning for Modeling the Robotic Profile Grinding

When modeling the removal rates of the robotic belt grinding system,it is hard to deal with the affecting factors which changes suddenly.In order to solve this problem,the difference between the belt grinding modeling approach based on machine learning and the statistical learning theory is pointed out firstly.And this difference is formalized.After that, an adaptive learning method incorporating prior knowledge is put forward,which produces virtual examples from semi-empirical formula to remedy the lack of new examples in adaptive learning modeling,and extracts the information from the semi-empirical formula which is imported into the learner then.It is proved by the experiments that this method makes the model adapt to the variances more quickly and with higher precision.Moreover,this method will help to improve the processing efficiency and reduce the reject rate caused by the dynamical factors in the practical use.