In the modern industry, the parts used in aerospace, defense industry and other applications have high requirements and are mostly expensive and have few spare parts. It is of great economic significance to detect defects and quickly repair them. This paper proposes a method combining 3D scanning and laser cladding to study the repair of profile parts. First, the point cloud of the defective area of the profile parts is obtained by 3D scanning, and then the mathematical model of the defective part is obtained by digital processing. Finally, the regional defects are repaired by industrial robots combined with laser cladding technology. The experimental results show that this method can be used for the precision repair processing of complex profiles.
At present, from the large-span support structure of space satellites to aircraft engine components, new materials and new processes are generally used for research and development, and the production cost is high. However, once the surface of the processed parts is damaged or broken, the service performance or aerodynamic performance of the parts will be affected, especially the surface of complex curved parts, which is difficult to repair. On the basis of ensuring the performance of the parts, how to repair the profile defect area of the damaged parts is a problem we need to solve.
Three-dimensional laser scanning technology is also known as real-scene replication technology. It is a technological revolution in the field of surveying and mapping after GPS technology. It breaks through the traditional single-point measurement method and has the unique advantages of high efficiency and high precision. 3D laser scanning technology can provide 3D point cloud data of the scanned object surface, which can be used for reverse design in the early stage of manufacturing and surface geometry detection in the later stage of manufacturing. Metal laser cladding is an integrated manufacturing technology that cannot be processed by traditional manufacturing technology. It has the advantages of short production cycle, high material utilization rate, strong ability to manufacture complex structures, etc., and can meet the needs of personalized small-batch production. The organic integration of 3D scanning technology and laser cladding technology is an effective exploration of repairing surface defects of curved parts. Figure 1 shows the scene of laser cladding repair of cracks in the casing area of an aircraft engine by the Shenyang Institute of Automation, Chinese Academy of Sciences.
In the manufacturing industry, 3D scanners, as a fast stereo measurement device, are increasingly used due to their advantages such as fast measurement speed, high accuracy, non-contact, and easy use. Compared with general plane scanning or camera photography, 3D scanners are more accurate, fast, multi-angle, and refined, and can collect 3D information of objects in the real world more completely into the computer. Scanning samples and models with a 3D scanner can obtain their three-dimensional size data. These data can be directly interfaced with CAD/CAM software. In the CAD system, the data can be adjusted and repaired, and then sent to the machining center or rapid prototyping equipment for manufacturing, which can greatly shorten the product manufacturing cycle.
Scanning with an industrial robot equipped with a 3D scanner is currently a popular scanning method. The industrial robot is used as a motion platform and equipped with a 3D scanner to complete the scanning operation according to the planned trajectory. It is more stable and efficient than manual handheld scanning. Tao Jingxin et al. proposed a robot-based 3D scanning method for automatic measurement of shell products. According to the shape characteristics of the shell products, especially the position characteristics such as holes, grooves and narrow spaces, the scanning trajectory is set and the robot’s offline programming language is generated. Ai Xiaoxiang et al. generated a 3D scanning path based on the least squares method. In the process of assembling the aircraft wing wall panel, the genetic algorithm was used to optimize the adjustment of the robot’s posture and guide the robot to perform digital scanning. Zhang Zhifeng et al. designed an industrial robot-assisted 3D laser scanning measurement system, which was applied to the assembly platform of complex structural parts, providing a good theoretical reference for the actual assembly of structural parts. Liu Guochao et al. used a 3D laser scanner to obtain a large number of point clouds on the surface of curved parts, and used third-party software to compare the obtained point clouds with the standard model to find product problems. Bian Peiying [5] obtained point clouds of defective parts and complete parts through 3D scanning and formed CAD models. The two models were compared by a secondary development method based on UG to determine model problems and guide the repair of parts. In this study, a 3D scanner, a laser cladding head, and a 6-DOF robot were combined to form a highly flexible part repair system with measurement and 3D printing functions. At the same time, combined with the high flexibility of the robot, a quick-change device was used to realize the switching of measurement and printing functions. The 3D scanning technology was used to detect the geometric morphology of 3D printed parts, and the laser cladding head was used to repair the defective parts. The repair method was studied and simulated and experimentally verified.
1 Curved part repair method
In the process of part repair, the first condition is to obtain the digital model of the compensation area of the part, and the part is scanned by 3D scanning technology to obtain point cloud data. In order to adapt to the measurement range of various parts, we use the measurement method of the robot end clamping scanning system (Figure 2). The measurement system consists of a Staubli robot, a FARO measuring arm, a motion guide rail, a PLC, a workpiece positioning platform and a control system. The robot drives the measuring arm to collect point clouds within a certain range, and the system has sufficient flexibility.
1.1 Defect location acquisition
1.1.1 3D scanning of surface defects
To ensure the quality of point clouds, the scanning head at the end of the robot needs to be calibrated before scanning the part, so that the scanning angle can be controlled along the normal direction of the measuring surface during the measurement process, and the measuring distance of the scanning head can be ensured to be within the measurement range. The flowchart of the part scanning algorithm is shown in Figure 3. Figure 4 shows the point cloud results of the defect area
1.1.2 Part defect model generation
Use third-party software to process point cloud noise, and then perform model matching. In the error analysis process, set the accuracy range requirements, select the point cloud of the defect area, and use the reverse modeling method to obtain the triangulated model, which is output as an stl file, as shown in Figure 5(a). The reconstructed model is matched with the original design model for point cloud, as shown in Figure 5(b).
1.1.3 Boolean difference calculation of model part
The Boolean operation of spatial model mainly refers to the intersection, union and difference of two three-dimensional spatial models. This operation is to obtain a new three-dimensional spatial model. This paper mainly focuses on laser cladding processing, so the Boolean operation is mainly a difference operation.
The required Boolean difference operation is realized by calling the computational geometry algorithm library (CGAL). This library is a geometric data structure and algorithm library based on C++ and is widely used in graphics and other fields. Since the data structure of common mathematical models can only be converted into the Polyhedral format in CGAL, the data structure that can be flexibly operated with Boolean is the Nef-Polyhedral format in CGAL. Therefore, it is necessary to convert the two spatial models into the Polyhedral format, and then convert the CGAL internal format into the Nef-Polyhedral format. After the Boolean difference operation, it is converted into the Polyhedral format, and finally converted into the original data structure of the model for output, as shown in Figure 6.
In practical applications, because there is a certain error between the reconstructed model and the original design model, there will be a certain error after matching. The algorithm shown in Figure 6 is only applicable to the case where the two models are completely inconsistent. The following is the method that needs to be used in the actual application process:
(1) First, obtain the result of the intersection of the above method, as shown in Figure 7(a). It can be seen that there are many scattered grids, some of which are even discrete (that is, these grids constitute patches individually).
(2) Use the maximum clique algorithm in graph theory to obtain the largest complete grid. Figure 7(b) shows the discrete grid in Figure 7(a) (which is not in the main grid body) and the grid in Nef-Polyhedral format (which satisfies the half-edge structure: an edge is occupied by only one patch, that is, this patch only has one half of the edge, and the other half is idle). After screening and reordering the data points of the half-edge structure, the model part that satisfies the half-edge structure can be extracted, that is, the grid in Figure 7(c), which is the maximum clique data of the model that satisfies the half-edge structure.
(3) Extract and sort the maximum clique data, regenerate the grid in Polyhedral format, and obtain the grid shown in Figure 7(d).
1.2 Path Planning
The path planning algorithm for additive manufacturing mainly includes feature recognition of slice contours and path filling. Because the slice contour includes a single polygonal contour, it may also be a nested contour with hollow islands, or even a contour structure with multiple islands. At present, the feature recognition of the contour requires the inner and outer circles of the inclusion relationship to be replaced clockwise and counterclockwise. The additive path is mainly filled in the plane contour after slicing. Because the research in this paper is mainly based on the plane contour, the path filling is relatively simple. You can choose the equidistant contour offset (CPO) or the reciprocating filling method (Zig-Zag) method, as shown in Figure 8 (a) and (b). In order to avoid the problem of insufficient coverage of the contour boundary by the two paths, we also proposed a hybrid path, which combines the above two paths. The outermost layer is 2 or 3 layers of equidistant contours, and the inner layer is a reciprocating path, as shown in Figure 8 (c).
2 Simulation and processing experiments
2.1 Simulation experiments
In order to verify the practicality of the algorithm, the above examples are subjected to Boolean difference operations and maximum cluster extraction, and the obtained model to be filled is shown in Figure 9.
Then, the defective parts are sliced and filled, as shown in Figure 10.
2.2 Application examples
2.2.1 Example 1
In order to verify the practicality of the algorithm, the side wall of the aircraft engine stator is laser repaired, and the repair material is IN625 nickel-based high-temperature alloy. Technical difficulties: The repair area requires high dimensional accuracy, and the thin-walled wall with a forming height of 10mm and a diameter of 150mm is required to reach 1mm. The overall deformation of the part is required to be less than 3%. The process of forming and cooling is adopted to meet the above requirements, with high dimensional accuracy and small deformation. The engine stator repair process is shown in Figure 11.
2.2.2 Example 2
A certain engine end cover part is repaired, and the repair material is stainless steel 316L. Technical difficulties: The repair area requires high dimensional accuracy, and the overall deformation of the part is required to be less than 1%. The repair area is required to meet high flaw detection requirements, and no defects such as cracks and poor fusion are allowed. The overall deformation after repair meets the requirement of less than 1%. The repair process of the end cover part and the finished product after repair are shown in Figure 12.
3 Conclusion
A surface part repair technology is proposed, which combines 3D scanning technology and laser cladding technology to achieve accurate repair of irregular defects of parts, which has certain practical significance. The conclusions are as follows:
(1) The Boolean difference operation is used to solve the original design model and the reconstructed model after scanning. For the discrete grids generated by the Boolean difference operation and the grid data that does not meet the half-edge structure, the maximum clique algorithm in graph theory is used to re-extract the grid surface, and the extracted result meets the half-edge structure.
(2) The algorithm in this paper is used for engineering application, and practice shows that it can be used for high-precision repair of defective parts of thin-walled surfaces.
