1 Object and significance

1.1 Background and significance

After the car is finished welding, it is necessary to polish the weld seam, remove the burrs before painting, remove burrs and further to improve the surface flatness and smoothness. At present, for the case where has high requirements on the grinding accuracy, the main processing method is mainly based on the manual grinding. On the other hand, for some works with high reproducible and relative lower accuracy requirement, the use of industrial robot cooperating with suitable grinding mechanism can improve the manufacturing efficiency and automation of the manufacturing progress while avoid the noise and dust on the worker's body. Meanwhile, it also helps to improve the flatness and roughness of the polished surface when machining complex parts with curved surfaces.

With the development of automotive industry, the industrial robot grinding system has been widely used^[1]. The current industrial robot grinding system integrates robots, grinding mechanisms, force monitoring and feedback, robot control and grinding trajectory planning, and grinding result evaluation system. It can realize trajectory planning, off-line programming and movement simulation of the grinding progress, realize the constant force control of the grinding process and improve the grinding precision. Some robot systems use neural network to transform human experience and the grinding results measured by the laser sensor to further improve the efficiency and precision of the grinding system^[2].

Under the background of “Made in China 2025” strategy, intelligent manufacturing is one of the key projects that China has vigorously promoted^[3][4]. The application of industrial robot technology in the automotive industry and the integration of new generation information technology is an important part of this strategy. At present, foreign industrial robot research has been carried out since the 1980s, and related research on robots has achieved a series of results, relevant industry standards and management methods also have been launched^[5].However, there are still some gaps in the technical and application of industrial robots. The industrial robot industry is facing huge development opportunities and challenges^[6].

1.2 Research brief

1.2.1 Research brief of deburring robot control system

After the welding of the car body, there are burrs and flash edges in the weld seam, which need to be cleaned before painting. Because the deburring robot has the advantages of high automation, high grinding efficiency, high production flexibility while saving labor cost, it has been widely developed and applied both in China and abroad.

Foreign robot deburring technology has been developed for more than 30 years. In 1985, Stepien started the research on the contact force control of deburring robots^[7], Based on the microprocessor, they developed a control system to amend the position error of each axis of the robot. Kazerooni proposed the contact force impedance control of the deburring robot, and used the impedance method to design an adaptive mechanism to compensate the uncertain vibration during the deburring process from the perspective of the frequency domain control problem^[8]. Asada proposed to use the teaching playback method to transfer human experience to the robot, developed a highly user-friendly robot teaching system, and established the teaching and feedback rules, so that users can complete the teaching and control of the robot without learning the robot programming^[9]. Suh proposed fuzzy adaptive control^[10], using position servo drive and acceleration signal filter to avoid vibration during robot deburring progress and improve accuracy of deburring. Lee proposed using a laser sensor to detect the shape of the burr and established a detection and modeling algorithm for the irregular burr. The detected model was used to adjust and supplement to the original deburring plan to solve the problem of irregular burr parts’ deburring, improve both speed and accuracy^[11].

In recent years, intelligent control methods such as computer vision and neural networks have been developed, providing new ways to solve the problem of burr shape uncertainty and further more improve the deburring accuracy. Leo proposed using a computer vision-assisted robot to complete the edge deburring work, and developed a control system that automatically controls and programs the grinding force based on the detected burr shape, further more improve the deburring effect^[12]. Dornfeld proposed using the grinding sound as a feedback parameter to improve the precision of the deburring force control, and use time and frequency domain analysis of the signals collected by the vibration and sound sensors to further judge the working state of the tools, thus improve the accuracy of the control progress ^[13] . Pandiyan proposed using support vector machine and genetic algorithm to judge the working state of the tools. Use existing data to model the abrasive belt wear process based on genetic algorithm and support vector machine, and the signals from the acoustic sensor, force sensor and acceleration sensor to estimate the residual life of the tools^[14]. Caesarendra proposed an adaptive neural network and fuzzy algorithm to monitor the quality of the grinding results, and perform wavelet decomposition and Welch spectrum estimation on the signals collected by the acceleration sensors to obtain characteristic information. Based on those information, an adaptive neural network and fuzzy algorithm are to monitor the deburring effect in real time^[15].

With the transformation and upgrading of China's manufacturing industry and the accelerated application of industrial robots, the domestic research work of industrial robot deburring technology has been carried out in a large number and has achieved a series of results.

In terms of the design of robot deburring tools, Chen Guanlin of Guangdong University of Technology designed a grinding wheel deburring system for complex surface machining. The key components of the belt machine structure, such as the tensioning wheel, the adjusting wheel and the auxiliary wheel were designed. The pneumatic circuit and the linear fixture were analyzed designed and proposed improvements^[16]. Zhang Mengyuan of Huazhong University of Science and Technology has designed a pre-deform grinding tool suitable for large-curve arc-face machining. It transforms the part into a suitable profile by pre-deformation method. The limited element method is used to optimize the structure of the tool, and the deformation measurement method is designed to improve the deformation accuracy^[17].

In terms of force control methods, Liu Zhiheng of Harbin Institute of Technology realized the indirect control of contact force by using impedance algorithm, and designed an online correction and force real-time processing software system based on teaching playback method^[18]. Zhang Qingwei and Guo Xiaodong designed a force/movement hybrid control system^[19][20] which satisfies the requirements of simultaneous force control and position control. The deburring task of the robot is controlled in two mutually orthogonal directions, using the selection matrix in force, the conversion between the force control and the movement control is realized to improve the force control accuracy of the robot and make the contact more stable.