1. Introduction
In the semiconductor industry, more than 100 kinds of chemicals are used for the purpose of cleaning, etchant, stripper, electroplating, etc. Chemicals in clean drums are delivered to Central Chemical Supply System (CCSS) through dispensing systems. In chemical drum assembly, chemicals or fumes may be exposed to operators. Chemical drum assembly is a process that involves grasping the dispense head, aligning it, mating keycodes, mating threads, screwdriving and fastening. Screw fastening is the most common assembly method in manufacturing, however it is a difficult process to automate. In particular, the complex interactions between the internal and external screws and the various problems caused by cross threading can lead to fastening failure [1]. Additional strategies are needed to improve the safety of automated screw fastening systems and to minimize failures.
Screw fastening is a process in which the threads of two parts are engaged. Improper engagement can damage the threads and cause permanent damage to fastening components. There are three main ways to address cross-threading. The first is to use a mechanical RCC (remote-center-compliance) to reduce the frequency of cross-threading [2]. The second is to use a back-spin first method, in which the fastening components are first rotated backwards before the fastening operation. The third is that the starting points of the two threads are aligned to reduce the frequency of occurrence. This method is slower and requires sensing, but it is advantageous for fastening components with large diameters and small pitches [3,4].
In this paper, the dual-arm collaborative robot system for chemical drum assembly has been developed. The method to increase the precision and stability of chemical drum assembly is proposed through assembly experiments using position/force control [5-10].
2. Chemical Drum Assembly Environment
Fig. 1 shows the chemical drum and dispense head for chemical drum assembly. The chemical drum includes drum cap, dip-tube with female keycode and dispense head with male keycode. The dispense head consists of dispense body, nut and male keycode as shown in Fig. 1(b). For assembly tests, the dispense head are fabricated with a 3D printer. Fig. 2 shows the chemical drum assembly process schematically. Fig. 3(a) shows a commercial screwdriving tool used for small-size screw fastening. However, Fig. 3(b) shows the dispense head with tube mounted on the chemical drum. For the dispense head, a general screwdriving tool can not directly be applied due to a large-size nut of dispense and the interference of the chemical tube. Instead of a screwdriving robot tool, dual-arm robots with compliant F/T sensors and grippers are employed similarly to human hands; one hand aligns the screw axes and grasps the dispense head, while the other hand mates keycodes and fastens the large-size nut.
Fig. 1 Components for chemical drum assembly [11,12]
Fig. 2 Process for chemical drum assembly
Fig. 3 Screw fastening comparison [11-13]
Fig. 4 shows the 3D modeling of dual-arm robot experimental setups for chemical drum assembly. The robot pose for chemical drum assembly is determined by 3D robot vision, and the dispense head can be assembled on the chemical drum using dual-arm collaborative robot system (two sets of UR10e with commercial F/T sensor (Onrobot HEX-E) and gripper (RobotiQ 2F-85). Employing compliant F/T sensors can reduce cross-threading, jamming and dispense head damage in keycode mating and nut fastening.
Fig. 4 3D modeling of the chemical drum assembly system (L and R denote the left and right robot arms)
3. Dual-arm Collaborative Robot System
The dual-arm collaborative robot system for the chemical drum assembly is shown in Fig. 5. Fig. 6 shows a network interface for the Universal Robot e-series. Fig. 7 shows the real-time control configuration of dual-arm robots. Two Universal Robot controllers with the names of L and R are connected and integrated to the target PC. The tool controllers for F/T sensors and external grippers are also connected to the target PC, which performs the integrated and real-time control, and is running on real-time operation system with 1msec loop. The host PC provides graphical user interfaces such as robot simulation and teaching pendant.
Fig. 5 Dual-arm collaborative robot system for the chemical drum assembly
Fig. 6 Network interface for Universal Robot e-series [14]
Fig. 7 Network configuration for the dual-arm collaborative robot system
The robot vision of the dual-arm collaborative robot system for chemical drum assembly is divided into two tasks. The center position of the drum cap is detected using IntelⓇ RealSense depth camera(D455), and the pose of female keycode center at dip-tube is detected by Pickit M-HD camera. For D455 camera, image preprocessing was performed with edge detection of RGB images, and the center point for the drum cap was detected with the CHT (Circular Hough Transform) algorithm [15]. The distance from the depth start point of D455 camera to the front cover glass is 4.55mm. A measurement depth standard for coordinate calibration is defined as 300mm, and the pixel resolution obtained through the experiment is 0.7836mm/pixel. The D455 camera is mounted on the robot R and the M-HD camera is fixed to the frame.
The compliant F/T sensors are mounted between the grippers and robots. The F/T sensors provides enough compliance for force control stability as well as F/T measurement. As shown in Fig. 8, the robot R performs force control when removing the drum cap, grasping the dispense head. The robot L uses position/force control for keycode mating and nut fastening.
Fig. 8 Process of the dual-arm collaborative robot system for chemical drum assembly
4. Assembly Experiments on Chemical Drum
In the assembly experiments, the robot L performs position/force control for keycode mating and nut fastening. First, the dispense head is moved to the pose of female keycode center detected by Pickit M-HD camera, and keycodes are contacted with minimal force. Second, with minimal contact force (1N), the search operation is performed by rotating the male keycode about the axis of dispense head. Finally, when both keycodes are mated, the male keycode is connected to female keycode with pressing force (18N). After the keycode connection is completed, the robot L performs nut fastening. The operations of the robot L can be divided into following steps;
A1: Move to female keycode center
A2: Contact with female keycode
A3: Rotate male keycode in contact with female keycode
A4: Male/female keycodes mating
A5: Press male keycode into female keycode
A6: Keycode assembly completed
The F/T sensor and work frames are defined as shown in Fig. 9. Two frames are parallel and \(\begin{align}\overline {O_SO_D}\end{align}\).
Fig. 9 F/T sensor and work frame definitions
The “force_mode” function in URScript is used to control force. The y-axis is selected as the force control direction (“1” in selection_vector). During the search step (A3), the minimal force of 1N along the y-axis is applied. During the pressing step (A5), the pressing force of 18N along the y-axis is applied and the moment of –3.6Nm about the x-axis is applied to compensate the reaction moment by Fy.
Selection_vector for A3: [1,1,1; 0,0,0]
Selection_vector for A5: [1,1,1; 1,1,1]
Wrench for A3: [0,1,0; 0,0,0]T[N;Nm]
Wrench for A5: [0,18,0; -3.6,0,0]T[N;Nm]
The resulting force/torque at the F/T sensor for keycode mating are shown in Fig. 10. Note that the signs of force/torque values are opposite to the applied ones, since the values are reaction force/torque from F/T sensor. The absolute values of Fy are the range of 2~3N for A3, large overshoot for A5, and 20N for A6. The large force errors result from the low resolution of F/T sensor about 1N, and low force gain for force control stability.
The operations of nut fastening by the robot L can be divided into following steps;
B1: Move to the nut fastening position
B2: Screwdriving on nut
B3: Nut fastening completed
Fig. 10 Resulting force/torque for keycode mating
The resulting force/torque at the F/T sensor for nut fastening are shown in Fig. 11. The reaction forces of Fx and Fz, and torque of My gradually increase as the number of turns increases. At the moment of nut fastening completed, Fx, Fz, and My rise rapidly.
Fig. 11 Resulting force/torque for nut fastening
5. Conclusions
The dual-arm collaborative robot system is developed for chemical drum assembly used in semiconductor industries. The real-time network controller for dual UR10e robots and tools is developed. The object detection and pose estimation algorithms are also developed. The force control experiments for keycode mating and nut fastening are successfully performed. In future research, the dual-arm collaborative robot system will be applied for other assembly tasks in semiconductor industries [16,17].
Acknowledgements
This research was partially supported by Samsung Electronics-University Cooperation Research Project.
참고문헌
- Edward, J. N. & Ronald S. F. "Dynamic modeling of a part mating problem: threaded fastener insertion," Proc. IEEE RSJ Int. Workshop Intell. Robots Syst., pp. 30-37, (1992).
- Edward, J. N. & Ronald S. F. "Compliant Control of Threaded Fastener Insertion," ICRA, pp. 484-490, (1993).
- Dift ler, M. A. & Walker, I. D. "Experiment s in aligning threaded parts using a robot hand," IEEE Trans. Robot. & Autom, vol. 15, pp. 858-868, (1999). https://doi.org/10.1109/70.795791
- Jia, Z. et al. "A Survey of Automated Threaded Fastening," IEEE Trans. Autom. Sci. & Eng., vol. 16, pp. 298-310, (2018).
- Kim, H. S. "Kinestatic Control using Six-axis Parallel-type Compliant Device," KSMTE, vol. 23, pp. 421-427, (2014). https://doi.org/10.7735/ksmte.2014.23.5.421
- Kim, H. S. "Design of a 6-axis Compliance Device with F/T Sensing for Position/Force Control," KSIC, vol. 21, no. 2, pp. 63-70, (2018).
- Park, S. B. Kim, H. S. "Position/force Control using 6-axis Compliance Device for Chemical Coupler Assembly," KSIC, vol. 25, no. 5, pp. 909-915, (2022).
- Jeong, S. H. Kim, G. S. Park, S. B. Kim, H. S. "Development of Robotic Tools for Chemical Coupler Assembly," KSIC, vol. 25, no. 6, pp. 953-959, (2023).
- Raibert, M. J. & Craig, J. J. "Hybrid Position/force Control of Manipulators," ASME J. Dyn. Syst. Meas. Contr., vol. 103, pp. 126-133, (1981). https://doi.org/10.1115/1.3139652
- Duffy, J. "The fallacy of modern hybrid control theory that is based on orthogonal complements of twist and wrench spaces," J. Robot. Syst., vol. 7, pp. 139-144, (1990). https://doi.org/10.1002/rob.4620070202
- www.dusemund.com
- www.aicellomilim.com
- www.staubli.com
- Overview of Client Interfaces, Universal Robots, Denmark, (2023).
- Mohamed R. et al. "Object Detection using Circular Hough Transform," AJAS, vol. 2, no. 12, pp. 1606-1609, (2005). https://doi.org/10.3844/ajassp.2005.1606.1609
- Jeong, S. H. Kim, H. S. "Development of a 6-axis Robotic Base Platform with Force/Moment Sensing," KSIC, vol. 22, no. 3, pp. 315-324, (2019).
- Kim, G. S. Jeong, S. H. Kim, H. S. "Study on the Stability of Force Control using a 6-axis Compliance Device with F/T Sensing," KSIC, vol. 26, no. 1, pp. 211-215, (2023).