Neural Network Inverse Control Of Two-motor Synchronous System

The paper fouces on the multi-variable decoupling control of synchronizationsystem composed of two AC induction motors and transducers------decoupling thesynchronically modulating velocity system into velocity and tension subsystems by the application of a-th ANN inverse system method.On the basis of theoretic invertibility analysis of that system ,the methods are proposed to fabricate the a -th ANN inverse system of original system.The network gets the initial centers through a k -means algorithm,and gets the weights through recursive least squares(RLS) in off-line training.In on-line training, the network updates the network parameters using a gradient descending error algorithm. The a -th ANN inverse system composed of two AC induction motors and the transducers is constructed .The digital control arithmetic programmed according to the ANN and the synthesis controller of pseudolinear integrated system in PLC,which is designed on the basis of the ANN inverse system ,works well in the controlling experiment on the platform of two-motor synchronization system, which is controlled by PLC and Supervisory Control and Data Acquisition software- WinCC.The online execution of ANN controller after its training convergence is subject to not reference model of controlled object ( the platform of the two-motor synchronization system) or inverse dynamic model .Lots of experiments results manifest that the method designed successfully decouples the system into the velocity and the tension subsystem completely,and the system has good robustance against the disturbance of the load .The dynamic and static has the same effect even if under the condition of alteration of the working methods of system transducer without alteration of other parameters (for example ,the structure of controller and control parameters ). The control method proposed in the paper is promised and can be applied in many industrial control environments.