Optimization of correction margin in a CAD/CAM system applying neuro-logic

应用神经逻辑的CDA/CAM系统中校正裕度的优化作为使用光学三维坐标输入设备从同一样品模型获得的冲模和三维点群数据的比较结果，我们了解到，在具有大倾角的轴向表面和尖角处的测量精度降低，并且测量受到形状测量的测量步骤的影响。包括照明和光学接收器装置的光学系统的设备和布置。虽然我们认识到，与移动平均的情况相比，神经逻辑的应用导致在光滑表面的大部分区域上更高的校正误差的能力，但是在连接肩型边缘和轴向表面的截面处以及在从轴向表面到配合表面截面的过渡截面处，两种方法之间没有观察到太大的差异，并且存在许多不能识别通过学习效应进行的校正的情况。另一方面，虽然在移动平均法下在每种情况下都用圆角进行校正，但在该方法下观察到更尖锐的边缘，尽管包括一些误差。在使用H-NeT的曲线插值中，也存在无法观察到部分精度改善的情况，这表明需要通过改变参数进行进一步分析。此外，认为需要改进光学阅读装置和阅读方法。

Optimization of correction margin in a CDA/CAM system applying neuro-logicAs a result of comparing the die and the three dimensional point group data obtained from the same sample model using an optical three dimensional coordinate entering equipment, we learned that the measurement accuracy at axial surfaces with a large inclination and at sharp corners is degraded and that the measurement are affected by the measuring steps of the shape measuring equlpment and arrangement of the optical system including illumination and optical receiver devices. While we earned that the application of neuro-logic resulted in higher ability of correcting errors over most of the areas of a smooth surface comparing to the case of moving average, not much difference was observed between the two methods at a section connecting the shoulder type margin and axial surface and at a transitional section from axial surface to the mating surface section, and there were many cases where correction by learning effect could not be recognized. On the other hand, while correction is made with rounded corner in every case under the moving average method, more sharp edges were observed under this method although some errors were included. In curve interpolation using H-NeT, too, there were cases where partial accuracy improvement could not be observed, which indicated a need for further analysis by changing parameters. Furthermore, it is considered that improvements are required in the optical reading device and the reading method.