adaptive control and optimization of the machining process:I;clustered neural network

In this paper a newly developed clustered multi-layer preceptron (CMLP)network is applied and compared to the Multi-layer perceptron (MLP)network on the basis of its capacity for non -linear modeling and on-line adappapin.this network has been tested with two simple applications:first whith a simple analytical model and secondly using machinig data acquired experimentally.during on-line training of the analytical model about 12.8% degradation of the data learned a priori was observed using the MLP compared with only 0.3% for the CMLP. in the machining data case the information damage of previously learned model was observed 0.6% and 3.5% respectively during continuous adaptation of the CMLP and MLP networks.