| The parameter selection in Automatic Manufacturing greatly affects the utilization of the machine tool and the manufacturing costs. The amount of metal cutting determined by the method of experience will be greatly different when the worker or the plant is different. The actual effectiveness will be a big difference. And because of new processing materials continue to emerge, and the application of the NC machines, machining centers and flexible manufacturing systems, the method which determine the cutting parameters by experience will be far behind the development of the times.At present, the speed and accuracy of determining the cutting parameters are limited by some factors. So it is difficult for the real-time control of cutting parameters and the realization of intelligent parameter selection. The neural network has ability to process vast amounts of information. So the neural network can be applied to achieve automatic parameter selection in the processing.However, the neural network based on software simulation is actualizing the large mount of parallel calculation by means of the serial method. It lost its basic characteristics. The processing speed is far from adequate to meet the requirements of research and use. Moreover, the frequency of its computational speed is limited by the computer, it was unable to complete the task of real-time control of cutting parameters, online testing and revision process parameters. Although pure hardware neural network can be an effective way to speed up the realization of its computational speed, we should not amend the network structure. It is not the ideal way. In practical engineering, The combination method of software and hardware- using programmable devices as the neural network coprocessor become the focus of the study. DSP processor will be able to meet the design requirements.Taking the BP network for the automatic parameter selection for example, the BP algorithm was actualized successfully on the DSP system in this research. And the special DSP system was designed to choose the parameter automatically. With training data to train this network, it has met the real-time requirement of the processing system. It will make the real-time control of the processing parameter become possible in the NC machine tool.By analyzing the problem of automatic parameter selection, comparing the various characteristics of the neural network, I designed the neural network model to solve the problem of automatic parameter selection in the mechanic processing.In view of this issue, in consideration of all factors, we choose the TMS320C55X chip. TMS320C55X employs variable length instructions to improve code efficiency, and enhances the parallel mechanisms to enhance circulation efficiency. TMS320C55X DSP pipelining was divided into two parts: instruction pipelining and implementation pipelining. Instruction pipelining completes the task of the interview addresses, memory wait to respond, take instruction packs and instruction, and so on. Implementation pipelining completes the task of decoding, reading / revise Register, read operation data and output the results, and so on.Our actualization of the BP algorithm is based on the ICETEK-VC5509-C evaluation board, using C language. The program was debugged in the simulator mode firstly, and then operated on the DSP chip.Based on the ICETEK-VC5509-C evaluation board, I designed the DSP system for the automatic parameter selection, and debugged the hardware and the software system. So it was proved that this method is feasible, and I have made some basic job for the using of the DSP system in the CNC processing to make the real-time modify of the parameters.
|
PDF Full Text Request