論文

Machining data (NC program) is generated by a CAM system, which generates the tool path from the target shape as a plane approximation surface instead of a free-form surface. Owing to this plane approximation, machining accuracy is reduced.In this paper, we propose a method to process the shape with high accuracy by defining the areas where accuracy is not required as a plane approximation surface and defining the part where accuracy is required as free-form surfaces.

Selective laser sintering hybrid milling (SHM) is an advanced manufacturing process that combines the advantages of both metal powder bed fusion and milling. Using this technology, extremely complex products can be fabricated with high accuracy and reasonable surface quality. SHM has enormous potential in the application of mold fabrication; molds with free-form inner cooling channels can be produced with this technology. The key feature of this technology is that milling is conducted not only after, but also during the forming process. However, this leads to the issue that although machined layers of products have a high level of accuracy without thermal deformation, the addition of new layers causes the materials to undergo shrinkage and deformations to occur on their machined layers. The dimensional error due to shrinkage significantly affects to the accuracy of SHM process; therefore, a completely understanding of the formation error due to shrinkage is highly important in developing efficient solutions. In this study, shrinkage deformation is observed in detail through the results of simulation using finite element analysis. Three shrinkage rules and the unclamping effect are proposed to completely explain the shrinkage error shape in SHM. Finally, an experiment is conducted to confirm the final result of shrinkage.

In general, NC programs for machining free-form surfaces using a computer numerical control (CNC) machine tool are generated using a computer-aided manufacturing (CAM) system. The tool paths (CL data) generated by a CAM system are approximated straight-line segments based on tolerance (allowable error). As a result, the tolerance affects the machining accuracy and time. If the tolerance is set to a small value, the lengths of the segments are shortened, and the machining accuracy is improved. The process in which a CNC machine tool reads and analyzes an NC program and controls the motors requires a mini-mum processing time of an NC program block (block-processing time). Therefore, if the lengths of the approximated straight-line segments are too small, it will be impossible to reach the indicated feed speed, and the machining time will be longer. In this study, by identifying the block-processing time of a CNC con-troller and deriving the appropriate length of the approximated straight-line segment based on the block-processing time, a CL data creation method that is ca-pable of high-speed and high-accuracy free-form surface machining is proposed. In addition, experimental verification tests of the method are conducted.

Machining time estimation is essential for the due-date estimation of products as well as for production plan-ning. Conventionally, machining time has been estimated by a computer aided manufacturing (CAM) system, which requires time and effort to create its nu-merical control (NC) program and requires machining expertise to operate it. In addition, among the problems with conventional methods, an error in the estimated machining time arises owing to the machine tool’s control characteristics. In this study, an artificial intelligence (AI)-based system capable of estimating machining time promptly and simply based on shape data without requiring any NC program is developed. The input data to the AI system are color information regarding the machined depths, which are used to estimate the rough-machining time, and color information regarding the machined surface curvature distributions to estimate the finish-machining time. Color information on the machined depths and machined surface curvature distributions is created using three-dimensional computer aided design (3D CAD) data. To build the AI system, the shape data and machining time data accumulated at the machining site are used, so that the machining time estimated reflects the machining method, machining expertise, and the machine tool characteristics employed.

One advantage of a femtosecond laser is that it enables the processing of materials that are otherwise difficult to process via conventional methods such as cutting, grinding, electric discharge machining, etc. However, it is difficult to predict a three dimensional shape processed by the laser. To achieve required three-dimensional shapes using femtosecond lasers with high accuracy and efficiency, it is essential to develop a computer-aided manufacturing (CAM) system that generates processing data (NC data). In this paper, a simulation method is proposed for femtosecond laser processing, which is the basis for the development of a CAM system. A binderless nano-polycrystalline diamond is used as the workpiece material to test the proposed method. Based on a comparison with the experimental results, it is confirmed that the processing shape could be simulated at a high accuracy for fixed-point irradiation as well as scanning irradiation.

Calibration of linear encoders has conventionally been performed using an optical interferometer for reference. Calibration using the optical interferometer, however, cannot fulfill the accuracy and stability requirements when the linear encoders exhibit sub-nanometer resolution. This study describes a new calibration method developed using the picometer resolution laser holographic scale as a reference, employing a range-dependent and minimum error-variance algorithm. The method has been evaluated via comparison tests with conventional optical-interferometer-based methods. Moreover, it has also been verified to achieve higher resolution, accuracy, and higher stability during the measurements, even in the atmospheric environment.