Research And Implementation On Dimension Error Prediction And Compensation System For Boring

Dimension error prediction and compensation for boring is a cost effective method to control the dimension distribution and maintain part exchangeability. The research discussed in the thesis, which is supported by the national 863 high-tech scheme(Project No. 2001AA423250), investigates in-depth into the dimension error prediction and compensation system for boring and related techniques. Combining the practical application in transfer line for boring, an on-line dimension error prediction and compensation system for boring in series production is developed and applied. Firstly, the requirements to dimension error prediction and compensation system for boring in series production are identified and described, based on the analysis of transfer lines. Its functional modules are investigated. A set of performance indexes are proposed. The general scheme for such production is developed. The micro compensation device for the boring cutter is the key technique to realize error compensation. Based on the research on micro movement and elastic gemels, a boring cutter micro compensation device amounted on the parallelogram elastic boring rod is developed with self owned intelligence right. The device has reasonable compensation range, compensation accuracy and robust reliability. The dimension error prediction model for boring is proposed based on artificial neural networks. Some issues on modeling are discussed. The standard artificial neural networks have inherent disadvantages. To overcome these disadvantages, an evolved Back Propagation (BP) algorithm is developed by combination of additional inertia and self-adaptation of learning velocity. The results indicate that, when the artificial neural networks is applied to modeling the dimension error prediction for boring, the parts machined have improved dimension distribution and standard deviation. Based on the analysis and research on particle swarm optimization algorithm, the particle swarm optimization algorithm based artificial neural network optimization modeling method is initially proposed and applied to the dimension error prediction for boring. The test indicated that the model established by the method has quicker convergence speed and higher prediction accuracy compared to the improved BP algorithm. The research achievements in the thesis have been applied to the transfer line for support arms in cars in DongFeng Peugeot Citroen Automobile Company Ltd.. The system replaced the existing dimension error compensation system for boring originally imported from abroad. The practical tests indicate that the product approaches the same technique level as the same category of the abroad products, and even achieved better performances on several aspects. The user gives good remarks. The product is pioneer device with such structure in boring in China. "It is leading at home and reaching the frontier abroad in view of technique progress".