Research On Technology Of Precision Assembly And Flexible Packaging For Electro-acoustic Parts

Flexible production,a typical paradigm of Chinese smart manufacturing development strategy,has become a research hotspot in the electro-acoustic industry recently.As we all know,electro-acoustic products are characterized by small size,variety,irregular shape,fast replacement speed,and large labor cost.Therefore,the traditional "rigid" production in the electro-acoustic industry has certain shortcomings.Concretely,it can only achieve large-scale production of a single variety,and cannot adapt to the environment.The multi-sensor fusion robotic system provides an effective way for the flexible production in the electro-acoustic industry.In detail,it can improve the robot's adaptability and transferability,can handle a large number of similar but not identical tasks.Therefore,it can tackle the drawback of the traditional industrial robot's "poor flexibility",and become a research hotspot in the electro-acoustic industry.Aiming at this hot spot,this paper studies the key technologies for flexible production in electroacoustic industry,such as the construction of multi-sensor fusion robotic system,the precision assembly strategy and flexible packaging tasks.The project cooperated with Dongguan Jiahe Intelligent Technology Co.,Ltd.,combined with the production requirements of the automated assembly line,formulated the design requirements of the robot system,and constructed a collaborative robotic system based on multi-sensor information fusion.The system consists of a mancollaborative robot with six-degree-of-freedom,a 3D laser triangulation vision sensor,and a six-dimensional force / torque sensor,so as to achieve 3D precision visual recognition and positioning,compliance force control,and collision detection.In terms of the precise assembly task of electroacoustic parts with small size,irregular shape and different poses,a “vision-force” guided assembly strategy is proposed,and an admittance controller based on "learning from human demonstration" is established.This paper introduces the collection method of the "pose-force / torque" training set during human demonstration,integrates the weak Gaussian process regression machine(GPR)through the Ada Boost algorithm,and uses the training set to solve the admittance gain.Using support vector regression machine(SVR)and weak Gaussian process regression machine as the comparison method,the experimental comparison and verification were carried out in three different assembly poses.The results show that Ada Boost-based GPR shows better performance and more excellent generalization in different assembly poses.For shape recognition and prediction of flexible wires,in order to avoid the noise and outliers of the wire perceived by the robot,the Gaussian mixture model(GMM)is used to convert the non-rigid point set registration problem of the wire into a probability density estimation problem.In order to tackle the local occlusion problem,the coherent point drift regularization method is applied to the GMM model to establish a wire shape prediction model,which can predict the shape of point clouds in the blocked area while maintaining the existing point cloud sequence topology.A series of wire tracking experiment results show that the model can show good robustness in the presence of noise,outliers and local occlusion.Finally,the built multi-sensor fusion robotic system is applied to the actual assembly line in the electroacoustic industry,to verify its practicability.Firstly,the production line process was analyzed and formulated,according to the production line demand.Then the experiments verify the system's precision 3D visual recognition and positioning,intelligent end tool detection and collision detection functions.The results show that the system can meet the needs of the actual production line.