The Dynamic Measurement Errors Prediction of Three Coordinates Measuring Machine Based on Hybrid Modeling

基於混合建模的三座標測量機動態誤差預測

馮旭剛(Feng Xugang) ； 徐強(Xu Qiang) ； 杜翠翠(Du Cuicui) ； 章家岩(Zhang Jiayan)

《中國機械工程學刊》 38卷3期 (2017/06) Pp. 261-265

摘要

The dynamic measurement errors prediction of three coordinates measuring machine are influenced by many factors, and many control methods are hard to achieve satisfactory accuracy. In order to present the high precision dynamic error model of the three coordinates measuring machine, apply neural network and fuzzy logic control to study the hybrid modeling methods of dynamic errors, and adopt the steepest descent algorithm to adjust the weights of hybrid model. Application of double-frequency laser interferometer has been used for predicating the dynamic measurement errors on the different positions under different speeds, and the dynamic error of measurement and dynamic error will be one of the points of collection of former 100 values for the establishment and adjustment of hybrid modeling, the later 30 values used for the predication of results. The results showing that the hybrid modeling of dynamic error of three coordinate measuring machine has better predication, and has a value for application. Coordinate Measuring Machine is a kind of new precision measurement instrument in latest 30 years, and playing a vital role in ensuring product quality. The rapid development of modern manufacturing industries put forward higher requirements of CMM. When measuring at higher speed, the moving and measuring speed of probe are also big, and causing larger dynamic error . With the speeding up of the CMM, the dynamic error become the main factors of influencing the accuracy of measurement. Therefore, studying the dynamic errors of CMM in depth , it also has the very vital significance at whatever improving the dynamic error and the design of CMM under high speed. Yang Hongtao adopted the method by combining the least squares with support vector machine(SVM) to model for the dynamic error, and used the double-frequency laser interferometer to verify the effectiveness of the proposed model, the experiment results showing that the accuracy of the model has been proved, the model don’t consider the dynamic characteristics of probe, therefore, it is not the dynamic error model in the true sense. Zhang Mei using three-dimensional coordinates and the parameters of CMM for the original independent variables, attain the dynamic measurement errors model through the three dimensionality of b-spline transform and parameter estimation. Although, the model overcome the problem of multicollinearity, the model doesn’t analyze the dynamic error sources and take their mutual influence into consideration. Fang Chuanzhi put forward the independent component to make decomposition reconstruction for the analysis of the dynamic measurement errors prediction, make full use of the overall sample of error, and provide a reliable basis for the sources of the dynamic error and the establishment of error correction model. Above methods using the internal mechanism analysis of three coordinates measuring machine, establishing the model from the source of the dynamic measurement error, and arising various error sources influenced each other inevitably, causing the restriction for the source of the dynamic error predication. Therefore, this paper establishes the hybrid neural fuzzy system, take measurement for the dynamic error of CMM at the different location, build the external model on the basis of dynamic error, avoid the internal interaction of the various error sources, and expect to realize the accurate predication of the dynamic measurement error.

關鍵字

Three coordinates measuring machine ； The neural fuzzy controlling ； Dynamic error ； Predication of errors

並列摘要

三坐标测量机动态测量误差受到众多因素的影响，许多控制方法都难以达到令人满意的精度要求。为了建立高精度的三坐标测量机动态误差模型，应用神经网络和模糊控制研究动态误差混合建模方法，并采用梯度下降法调整混合模型的权值。应用双频激光干涉仪进行了不同速度下不同位置的动态误差测量实验，并将其中一个位置点的动态误差集合的前100 个值用于混合模型的建立和调整，后30 个值用于结果预测。结果表明，混合建模对三坐标测量机动态误差具有较好的预测精度，具有一定的应用价值。