Research On Intelligent Methods Of The Progress Control Optimization

With the improvement of the control needs and the rapid development of the computer science, the computer technology has penetrated to the control career in-depth step by step, particularly including the intelligent technology, which solves the problems of non-linear model identification and adaptive control, such as fuzzy neural network. The control of heavy manipulator, which is used to make large parts efficiently and precisely, is high non-linear, and the operation progress is ordered by time. This thesis researches the computer intelligent method against the progress identification and control problems of the manipulation progress ordered with time.And it designs the progress identification method based on fuzzy neural network, the control optimization method based on dynamic programming and the coordination control method based on decision tree. These methods provide efficient solution to this kind of engineering problems.The main contents and results obtained by this thesis are summarized as follows:1) We analyze and summarize the intelligent methods applying on the identification and control of non-linear model by the research on relative world-wide theory and reference. By means of study of non-linear model identification, we research and analyze the fuzzy neural network about applying to the non-linear identification and adaptive control against the object features. And we analyze the current control methods and their relative problems of the manipulators, and research the features of relative intelligent methods, including the dynamic programming used to the optimization and the compliance control of mechanical structure. The current method of coordination control of two mechanical structures and the decision tree in data mining career are studied.2) Forging manipulator is one of the most important equipments in the automation of forging work, and plays a great role in the improvement of productivity and the assurance of product quality. To improve the forging manipulator's performance, a similar model of a forging manipulator is designed to facilitate experiments, which is a 3-DOF planar parallel mechanism. The direct kinematics and inverse kinematics are obtained by using geometrical approach and the workspace is analyzed. Based on Jacobian matrix, the singularity loci are obtained while the tongs-carrier orientation is of different angles. Finally, the coupling among its DOFs is analyzed based on direct Jacobian matrix. Analysis proves that the manipulator is free from singular configuration in workspace and the coupling can be ignored while forging in small movements; and these facilitate the kinematic control of the manipulator.The inverse dynamics of this similar model is obtained by using the principal of virtual work; and to decrease the negative effect of the nonlinearity of the mechanism dynamics, the dynamic feed forward control method is adopted, distributing driving forces, which are needed for the tongs carrier of the manipulator to generate desired motions, to each hydraulic cylinder. The tracking performance of the tongs carrier is chosen to be guaranteed by improving the position response performance of each hydraulic cylinder. So, using feedback linearization method and considering the nonlinearities of hydraulic systems, a hydraulic cylinder system is transformed into a second-order linear system which can be easily adjusted by setting four parameters to get desired damp ratio and natural frequency and to improve the response dynamics. Simulation results finally attest the validity of the proposed method. And then the working conditions, principles and control of the manipulators are analyzed for designing suitable intelligent methods to solute this kind of progress control.3) The initiative force control of two heavy manipulators needs a model between the velocity and force in the whole manipulating process, but the high non-linear system made the exact model hard to be established. So this thesis proposes a method of process identification by a new kind of fuzzy neural network with an ordered subsection algorithm, a judgement function to each subsection and an activation function of nerve cells in a layer. Based on these, the simulation results validate the process identification model and hence this method provides a modeling foundation for the coordination control of the heavy manipulators.4) In order to get smooth velocity of the tong-carrier, this thesis constructs a new kind of fuzzy neural network with links between nerve cells in a layer for process identification especially, which also provides the non-linear relation model for dynamic programming, and then the dynamic programming arithmetic with a scale-factor figures out an optimized result to the hydraulic cylinder closed-loop velocity-control. The control optimization for the smooth velocity-output of the tong-carrier is implemented by active compliance, and the simulation proves that this method eliminates the error in contrast with that before control optimization, and so it provides a control foundation of single-manipulator for the coordination control of the heavy manipulators.5) Two heavy manipulators need to be controlled coordinately when they work with a piece of workpiece. The whole coordinate control designed by this thesis concludes decision-making and execution layers. This thesis firstly provides an object model for the decision tree by means of constructing a new fuzzy neural network which especially adapts to process identification, and then a new kind of decision tree with membership function is used to provide upper-layer policy and fuzzy input of each single-controller in the lower-layer, finally the fuzzy neural network controller executes the policy and thus the whole adaptive coordinate control of the two heavy manipulators was completed which is proved by the simulation of manipulators and experiments of hydraulic cylinders in contrast with passive compliance and PID control.The research contents and results of this thesis rich the research work of the fuzzy neural network, dynamic programming and decision tree in data mining career. And they solve the problems of identification and control of ordered model and provide an entire program. This thesis expands the application career from computer to control, which has significant effect to theory and application.