Research On Static And Dynamic Performance For Multi-dimensional Force/Torque Sensor

Multi-dimensional force/torque(F/T)sensors have been widely used in human-computer interaction,intelligent robot,biomedical research,medical devices,automotive,aerospace and other fields.A six-axis F/T sensor can measure the tangential force terms along X-,Y-,and normal force term along Z-axis(F_X,F_Y,and F_Z)as well as the moment terms about X-,Y-,and Z-axis(M_X,M_Y,and M_Z)simultaneously.In the field of human-computer interaction,haptic interaction is a new type of human-computer interaction technology with a bidirectional information transmission capability.The multi-dimensional F/T sensor transmits the perceived force information to the controller,and then transmits these information to the user through the actuator.In the field of intelligent robot,the accurate acquisition of force information is the basis and premise of the robot's accurate cognition of the environment and the successful completion of the follow-up force feedback and force control.In the field of aerospace,the multi-dimensional F/T sensor installed in the joints of space station manipulators can help the manipulators to realize force feedback function,which greatly expands the application range of manipulators.According to the requirement of force control of high-speed work robot,the static and dynamic performance of multi-dimension force sensor used in high-speed work robot is studied in this thesis,and the problem that multi-axis force sensor can not be used in high-speed robot is solved because of the lack of basic theories and methods such as structure design,dynamic performance measurement and analysis,dynamic performance compensation.Firstly,the six-dimensional F/T sensor system of the robot is designed,including the design of the mechanical structure of the sensor and the design of the subsequent signal conditioning circuit.The static and dynamic simulation analysis of the elastic beams of the force sensor is carried out by using the finite element analysis(FEA)software,the stress and strain of the elastic beams subjected to loads of various directions are obtained,and the vibration mode and resonance frequency of the elastic beams are analyzed and determined.In order to obtain the static and dynamic performance indexes of the six-dimensional F/T sensor,the static and dynamic calibration experiments must be carried out,so the calibration system and methods are introduced in the thesis.And then the possible error factors of the calibration system which affect the calibration accuracy are analyzed,and the error model of the calibration system is established.The measurement accuracy of the multi-dimensional F/T sensor is seriously restricted by coupling.In this thesis,the intelligent optimization algorithm is applied to adjust the initial weights of neural network,and a static decoupling algorithm based on genetic algorithm for optimizing BP neural network is proposed.The decoupling algorithm is compared with the most widely used method to solve the generalized inverse of the calibration matrix and the simple decoupling algorithm based on BP neural network.The experimental results show that the method can solve the problem of ill-conditioned matrix and local minimum easily,and the decoupling precision is higher than the traditional algorithm.The theoretical basis for mechanical analysis of multi-dimensional force sensors is generally to establish their mechanical models.However,at present static and dynamic models need to be established respectively for static analysis and dynamic analysis,and the multi-dimensional force sensor has different models in different directions,which increases the difficulty of modeling.Aiming at this problem,the static model of the elastic beam is established by the method of kinetostatic modeling based on the matrix displacement method.When the dynamic frequency is included into the stiffness matrix of the flexible beam in the model,the matrix becomes the dynamic stiffness matrix,and the frequency dependence between the nodal force and the displacement at any point on the beam is analyzed.Thus,a dynamic model of the elastic beam--a pseudo-static model can be obtained.The natural frequency of the elastic beam and the transfer function of the system can be obtained from the model,and the dynamic displacement characteristics of the system can be further analyzed.Simplified statics models of the elastic beam subjected to forces in different directions are established by using the Timoshenko Beam theory.The strain at any point on the elastic beam under various loads and the equivalent stiffness values in each direction are obtained.The natural frequency is related not only to the structure and the shape of the system but also to the load it carries,that is,the load characteristics.Generally speaking,the larger the load,the lower the natural frequency of the system.Therefore,when the multi-dimensional F/T sensor works with load,the related dynamic performance index of the sensor under no load condition is no longer applicable.At this point,it is necessary to know the quantitative relationship between the dynamic characteristics and the load.In this thesis,the load characteristics of the three-dimensional F/T sensor are analyzed,and the quantitative relationship between the natural frequency and the load is obtained.In order to solve the common problems of the multi-dimensional F/T sensor,such as the violent oscillation of step response and the long response time,the dynamic performance compensation method of multi-dimensional force sensor based on PID neural network controller is proposed by combining PID control technology with neural network.The simulation results show that the response time of the system is shortened from 250 ms to 120ms,and the working bandwidth is extended from 300 Hz to 610 Hz.At the same time,the dynamic decoupling between the channels is realized,and the method can be corrected in real time.And this method does not rely on the sensor's accurate mathematical model.To obtain a six-dimensional F/T sensor with higher sensitivity,a new resin material,Polyetheretherketone(PEEK),with good toughness,rigidity and good mechanical properties is used in this thesis for making elastic beam of the F/T sensor.Comparison with ordinary F/T sensor purpose was served by building simplified statics model that demonstrates the conclusion of highly sensitivity of the F/T sensor based on PEEK material.Moreover,the properties of this sensor were compared with a six-axis F/T sensor based on a metallic sensor with identical structure.Experimental results demonstrate that the PEEK sensor has a comparable static performance with the metallic sensor,and even higher sensitivity.However,its dynamic performance is poor,and the natural frequency is only about 1/4 of the aluminum alloy sensor under the same size.