1 Introduction With the rapid development of China's aerospace industry and the increasingly fierce competition in the international satellite launch market, higher requirements have been put forward for the production cycle and manufacturing cost of engines. In order to adapt to this situation, it is urgent to improve the quality and stability of welded joints and ensure the production cycle. The flexibility of the robot welding system is a good solution to this contradiction. 2 Problems in Robot Welding To meet the requirements of welding quality, one weld seam needs to be selected for robot programming for each segment. The current programming method is teaching programming. The operator uses the teaching box to control the robot's movement, so that the welding gun reaches the position required to complete the welding operation, and records the position data of each teaching point. Then the robot can complete the welding of this weld seam in the "reproduction" state. According to the previous test, there are currently two problems: (1) The teaching accuracy is unstable, which affects the welding quality. During the teaching process, the programming effect is greatly affected by the operator's level and state. When teaching, the teaching point should be kept on the weld seam trajectory as much as possible, the welding gun height should be kept appropriate, and the welding gun posture should be kept to change continuously. The operator's level is very high. In addition, operators are in a state of high concentration for a long time, making it difficult to ensure the accuracy of each teaching point. This makes the final programming accuracy unstable, and sometimes problems such as the welding torch colliding with the workpiece occur. (2) Long programming time and low welding efficiency. In order to ensure the accuracy of the trajectory, 50 points are usually required for a 100mm weld to ensure the smooth operation of the welding robot and the consistency of the arc termination point. Each segment of online teaching and programming takes 2 hours, that is, the entire product requires 200 hours of teaching and programming, totaling 25 working days, which increases the total welding time of the nozzle extension section. Therefore, how to improve the efficiency and accuracy of programming, shorten the total welding time of the product, and improve the welding quality has become an urgent problem to be solved. 3 Robot Welding Offline Programming Technology Currently, robot programming can be divided into two methods: teaching programming and offline programming. When the task to be completed by the robot is not very complex and the programming time is relatively short compared with the working time, teaching programming is effective and feasible, but it is not satisfactory in many complex operation applications. 3.1 Characteristics of Offline Robot Programming Offline programming and simulation technology for robot welding utilizes computer graphics to create a model of the robot and its working environment on a computer. Through the control and manipulation of the graphics, programming is performed without using the actual robot, thus generating a robot program. Compared with traditional online teaching programming, offline programming has the following advantages: a. Reduces robot downtime. b. Keeps programmers away from hazardous working environments. c. Facilitates modification of robot programs. d. Can be combined with various artificial intelligence technologies to improve programming efficiency. e. Facilitates integration with CADICAM systems, achieving CAD/CAM/Robotics integration. Therefore, offline programming and simulation for robot welding is a key technology for improving the flexibility of robot welding systems and an important development trend in modern robot welding manufacturing. 3.2 Current Status of Offline Robot Programming Technology Currently, there are commercially available robot offline programming software based on ordinary PCs on the international market, such as Workspace, ROBCAD, and IGRIP. Workspace is the first commercially available microcomputer-based robot simulation and offline programming software developed by Robot Simulations. The latest version of this software uses ACIS as its modeling core, achieving excellent data exchange with some computer-based CAD systems such as AutoCAD. ROBCAD is a robot CAD and simulation system launched by Tecnomatix in 1986. Within just a few years, ROBCAD has been widely used in actual industrial systems. Many automotive companies, including Ford, Volkswagen, and Fiat, as well as Lockheed Martin, use ROBCAD for robot production line design, simulation, and offline programming. Another well-known robot offline programming and simulation software package in the United States is IGRIP, an interactive robot graphical programming and simulation software package launched by Deneb Robotics. It is mainly used for robot work cell layout, simulation, and offline programming. IGRIP can run on workstations such as SGI, HP, and SUN. The IGRIP software consists of three parts: IMS, GSL, and GLI. In addition, it provides users with some more advanced functions through a shared library. In China, Harbin Institute of Technology, Beijing University of Technology, Nanjing University of Science and Technology, and other institutions have conducted research on offline programming for robot welding. Harbin Institute of Technology (HIT) began research more than a decade ago, and its research level is among the leading in China. It has successively developed offline programming systems for robot arc welding, such as RAWCAD, which have been applied to some products. 4. Robot Offline Programming and Simulation Solution Develop an offline programming and simulation system for arc welding robots based on the SolidWorks platform to realize offline programming during the welding process of the nozzle extension section. 4.1 Workflow a. Establish CAD models of the nozzle extension section mold, the pipe, and the robot model. b. Divide the weld seam into segments and number them. For each weld seam segment, automatically program it using the offline programming system, including planning the welding torch trajectory and welding torch posture. c. Simulate the programming results and correct the planned pose based on the simulation results. d. Calibrate the robot coordinate system to make it consistent with the coordinate system in the offline programming system. e. Convert the offline programming program into a Motoman robot program and import it into the robot controller via a communication interface or CF card. f. The robot uses the offline-programmed program to complete the welding of the workpiece. 4.2 Three-Point Calibration Method The three-point calibration method uses the spatial coordinates of three feature points of the actual workpiece and the spatial coordinates of three feature points of the virtual workpiece. The label point (X) is a point on the x-axis of the calibration coordinate system, the label point (Y) is a point on the y-axis of the calibration coordinate system, and the label point (o) is the origin of the calibration coordinate system. 4.3 Example 4.3.1 Calibration The robot is taught to the actual robot welding torch at the three feature points of the workpiece. The robot joint angles at these three feature points are recorded and saved to a file. This robot joint angle file is then rewritten into a robot program file. The program upload function of the "Programmer" is then used to upload the program to the offline programming system. The robot is controlled to move step-by-step in the "Programmer," and the position of the robot's end effector is recorded at the corresponding point after each movement, as shown in Figure 1. Figure 1 Workpiece Calibration Coordinate System This provides a calibration function for saddle-shaped weld workpieces. Six label points are required during the calibration process: three on the upper circle and three on the lower circle. The recording method for each point is the same as the previous three-point calibration. Note that the recording order of each point on the circle should be the same, generally counterclockwise. See Figure 2. Figure 2 Simulation results of saddle-shaped weld workpiece 4.3.2 Creating weld joint feature objects Generating weld joint feature objects: a. Perform name check; b. Perform pose calculation. Currently, pose calculation only supports fillet welds. The calculation principle is as follows: the cross section of the fillet weld is approximately an inverted triangle, and the whole is approximately a triangular shape. Welding path points are generated on the bottom edge of the weld where the two weld plates intersect. Spatial position information is extracted from the edge. The tangent direction of the edge at this point is the X-axis direction of the weld point, and the angle between the normal directions of the two sides of the weld is the Z-axis direction of the weld point. The Y-axis is obtained by the cross product of X and Z. See Figure 3. Figure 3 Weld joint features Without generating weld geometry, the calculation principle for generating weld paths and the principle for generating weld points from weld geometry are shown in Figure 4. Figure 4. Simulation results of welding path. 4.3.3 Simulation of welding of nozzle extension section robot. The simulation results of the nozzle extension section robot welding system are shown in Figure 5. 5. Conclusion Because offline programming is used, programming does not affect the normal production of the welding robot. Moreover, the offline programming system can perform automatic programming, and the selection of welding torch position points and the transition of welding torch posture are very smooth, improving programming accuracy. Programmers can visually check the programming results through the simulation system and make manual corrections. Using such an offline programming system can improve programming efficiency, reduce the workload of programmers, and improve product productivity and welding quality.