LETR: An End-to-End Detector of Reconstruction Area in Blade’s Adaptive Machining with Transformer

In the leading/trailing edge’s adaptive machining of the near-net-shaped blade, a small portion of the theoretical part is retained for securing aerodynamic performance by manual work. However, this procedure is time-consuming and depends on the human experience. In this paper, we defined retained theoretical leading/trailing edge as the reconstruction area. To accelerate the reconstruction process, an anchor-free neural network model based on Transformer was proposed, named LETR (Leading/trailing Edge Transformer). LETR extracts image features from an aspect of mixed frequency and channel domain. We also integrated LETR with the newest meta-Acon activation function. We tested our model on the self-made dataset LDEG2021 on a single GPU and got an mAP of 91.9\%, which surpassed our baseline model, Deformable DETR by 1.1\%. Furthermore, we modified LETR’s convolution layer and named the new model after GLETR (Ghost Leading/trailing Edge Transformer) as a lightweight model for real-time detection. It is proved that GLETR has fewer weight parameters and converges faster than LETR with an acceptable decrease in mAP (0.1\%) by test results.

Adaptive Machining Scheme for Multi-Hole Part with Multi-Position Accuracy Requirements

The machining of multi-hole parts often has complex correlated position accuracy requirements. When some position accuracies do not meet the requirements, several hole axes need to be adjusted. Previous methods usually correct all deviated axes to their theoretical locations. However, the correction workload is too large and inefficient. This paper proposes an efficient and adaptive hole position correction model for multi-hole part. First, the method establishes the topological relationship of the holes and faces on the part according to the position accuracy requirements of the multi-hole part. Then, the goal is to minimize the number of holes that need to be corrected. In this model, the parallelism of holes, perpendicularity, and other constraints are considered. The simulation and experimental results show that the use of this model can effectively reduce the number of holes that need to be corrected during the compensation of the position error between holes. It improves the efficiency in the subsequent compensation process significantly.

Development of a Novel System for Adaptive Machining of Near-Net-Shape Components

Near-net-shape components are popular among the aerospace industry for low material waste and high manufacturing efficiency. However, it is difficult to machine such components into final shapes because the machining allowance is often distributed unevenly and even insufficient. This paper proposed a novel system for adaptive machining near-net-shape components, which integrates units like on-machine measurement based on probe and ultrasonic-sensor, machining allowance constrained localization, tolerance range constrained shape reconstruction, and TCP (tool cutter position) template-based NC programming. Firstly, localization and free form deformation (FFD)-based shape construction are performed within the tolerance ranges of the component, and an even distribution of the machining allowance can be obtained. Next, the quick NC programming that directly manipulates the TCPs by using spatial deformation is introduced. Last, the data transmission between units is illustrated. A case study of the machining titanium turbine blade is performed, which validates the proposed system.

МЕТОД ВІРТУАЛЬНОГО БАЗУВАННЯ ДЕТАЛЕЙ З ФОРМОЮ, НАБЛИЖЕНОЮ ДО ФОРМИ ЗАГОТОВОК

The subject matter of the article is the processes of virtual localization of near shape parts during adaptive machining. The aim is to develop an effective method for finding the starting location of a CAD model of a part with virtual localization inside a point cloud obtained by laser scanning of a workpiece. The task is to formalize the procedure for starting positioning of the part model as the first stage of the virtual localization process. The second stage for final localization proposed to use iterative algorithms with the objective function which is sensitive to the intersection of the surface parts and the workpiece. In solving the problem the starting position used tools available in today's CAD packages and 3D scanning tools. The methods used are the methods of matrix algebra, in particular, the methods for finding the main central moments of inertia of three-dimensional objects based on the tensor of inertia. The following results were obtained. When calculating the inertia tensor components is proposed to use three-dimensional scanning data of workpiece and geometrical data of part obtained from the CAD system. The result is an algorithm starting location of CAD model in the virtual localization, which in the case of blanks with oversize close to uniform can provide enough current location parts for adaptive machining tasks. It is shown that to minimize computational errors and to ensure satisfactory accuracy of localization proposed algorithm can require several iterations of the shift vector search model. Conclusions. The scientific novelty of the results obtained is as follows: in contrast to the previously used approaches, when solving the problem of virtual localization for the starting position, using the condition of coincidence of the centers of the weight of thin shells coinciding with the surfaces of the workpiece and the part, it was proposed to additionally ensure the alignment of the main central axes of inertia of these shells, which, in the case of near shape blanks, provides a positioning accuracy that may not require additional iterative procedures.