Application of Improved BP Neural Network in Controlling the Constant-Force Grinding Feed

摘要

BP neural network is applied to control the amount of feed, which is the key problem during the constant-force grinding. Firstly, BP neural network is constructed. Because its convergence is slow and local minimums often occur, the adaptive learning rate is used and certain momentums are added to improve BP neural network. Then the feature parameters in time and frequency domain are picked up in grinding vibration signals. With these feature parameters BP neural network is trained. As the result it makes the amount of grinding feed recognized precisely. Comparing the practical amount of feed with the set one, the system sends commands to increase or decrease the feed .So the amount of feed regains the set one. This method realizes the auto-control of the grinding feed, and puts forward a new method for constant-force grinding. It combines the features in time domain with those in frequency domain, and overcomes the limitation of the method which picks up feature parameters only in time domain or in frequency domain. At the same time this method provides a new clue of integrating other feature parameters in the grinding. Practice proves its good effect.