Dynamic Modeling Of 3-RRR PPM Based On Deep Learning Method

With the rapid development of science and technology now,intelligence is gradually becoming a research hotspot,which the application of neural networks in deep learning to solve traditional problems has become an important research field of artificial intelligence.Traditional planar 3-RRR parallel robot dynamic modeling requires a lot of formulas and the calculation process is so cumbersome that not easy to apply.Using RBF and BP neural networks of deep learning to replace traditional methods can establish a parallel robot dynamics model which can solve the above problems well.The main research contents are as follows:(1)Design of 3-RRR planar parallel robot mechanical system.According to the requirements of application scenarios,manufacturing and assembly processes,the three-dimensional structure and dimensions of each rod are determined.The innovative design of "counterhead pin+pin sleeve+axial positioning" solves the problems of portability and interference of the link hinge connection.(2)Design of 3-RRR plane parallel robot control system.The design of the control part includes two parts:software and hardware,the software part includes the design of the servo motor drive software and the supporting software design of the torque and speed sensor in the C++ environment;and the hardware part includes the selection of the servo drive motor,the driver of servo drive motor,torque and speed sensor and the integration of the entire experimental platform electromechanical system.(3)3-RRR planar parallel robot kinematics and dynamics modeling.Among them,the kinematics modeling uses the vector loop method that each member is regarded as a vector to establish the equation system to solve problems.However the dynamics modeling uses the inverse dynamics that the Lagrange equation is used to establish the equation system to solve.(4)Inverse dynamics modeling of 3-RRR planar parallel robot based on machine learning method.First design the BP and RBF neural network architecture,and then use the numerical solution data of the 3-RRR planar parallel robot inverse dynamics model and the data derived from the prototype dynamics test to train the BP and RBF neural networks.The comparison of the training results of the network through two different sources of data shows that the error of the inverse dynamic model based on BP and RBF neural network can meet the requirements of engineering applications and can replace complex theoretical models.