RBFNN soft-sensor modeling of pellets sintering permeability based on subtractive clustering algorithm

The nonlinearity, the process complexity, the mathematical modal uncertainty and time-varying characteristics make it very difficult to build a permeability soft-sensor model for pellets sintering process. In order to solve this problem, a RBF (radial basis function) neural network soft-sensing method based on the subtractive clustering algorithm is put forward. Subtractive clustering algorithm is adopted to partition the input space so as to obtain the centers and standardized constants of gauss basis functions of all nodes in hidden layer of neural network. Then the recursive least squares method with forgetting factor is used to update the weights of the output layer. Simulation results show that the proposed model have faster learning ratio and higher predictive accuracy. The predictive accuracy can satisfy the demand of the on-line soft-sensing for controlling the pellets sintering process real-time.