拉弯成形智能控制关键技术的研究

【作者】 孙凌逸；

【导师】 王俊彪；

【作者基本信息】 西北工业大学 ， 航空宇航制造工程， 2005， 硕士

【摘要】 拉弯成形工艺在航空、航天、武器装备和汽车型材弯曲件的制造中应用非常广泛。但成形过程受工件的塑性变形特性和摩擦条件等诸多非线性因素的影响,回弹等成形缺陷比较严重,极大的影响了制件的成形精度。因此,研究拉弯成形的智能控制技术,提高拉弯成形件精度尤为重要。本文由此入手,对拉弯成形中材料性能参数的在线识别及拉弯成形回弹预测进行了研究。 为保证在材料性能出现变化和波动时,依然保持较高的成形精度,本文对拉弯成形中材料性能参数的在线识别进行了研究,建立了集实时监测和在线识别功能为一体的材料性能参数在线识别系统,通过拉弯变形初始阶段的主缸拉力—位移曲线,利用数字滤波和曲线拟合技术,实现了拉弯成形中材料性能参数的在线识别。对参数识别中存在的误差进行了分析研究,结合拉弯设备的结构特点,建立了零点校正以及变形误差补偿模型,有效地解决了识别误差问题。这项研究为回弹的实时控制及拉弯成形智能化系统的建立奠定了基础。 本文利用神经网络技术对预测拉弯回弹进行了研究,建立了拉弯回弹预测神经网络模型,实现了拉弯过程中的成形条件与拉弯回弹的非线性映射。利用BP神经网络建立了拉弯回弹智能预测系统,根据给定的成形条件,对等边角型材拉弯变形的回弹进行了预测,取得了理想的预测精度,表明了神经网络技术在预测回弹方面的可靠性和优越性。更多还原

【Abstract】 Stretch bending has been widely applied in the manufacture of extrusions forming in the field such as aviation, aerospace, weapon and automobile. The accuracy of the formed part is greatly affected by the forming defects of sheet metals for instance springback. which are affected by many non-linear factors such as the plastic forming characteristic and friction. As a result, it is important that we research the intelligent control technologies to improve the accuracy of the formed part. In this paper, a research was conducted on the on-line identification of material properties in stretch bending process and the prediction of the springback.For ensuring a better forming precision with the changes of material properties, on-line identification of material property parameters was researched. Then an on-line identification system combined with real time detect and on-line identification for material property parameters is established, by which the on-line identification for material property parameters during stretch bending is realized by digital filtering and curve fitting with the curve of loading-main cylinder of primary stretch bending. The error of identification was analyzed, then the model of zero calibrate and error compensation was established, which combined with the structure characteristic of stretch bending machine and resolved the problem of identification error efficiently. The research lays a foundation for further establishment of a system for real-time control of springback and intelligence of stretch bending.In this paper, a neural network model for prediction of forming springback is established by the neural network technology, the non-linear mapping between the deformation condition and springback was realized. An intelligent prediction system of stretch bending springback was build based on BP neural network, the extruded profiles stretch bending is studied respectively. The predicted precision is ideal, as shows the reliability and advantage of use of the neural network technology in forming springback prediction.更多还原