Structural Optimization And Decoupling Of A Novel Three-dimensional Force Sensor Based On Flexible Structure

Multidimensional force sensor is widely used in the intelligent robot,mechanical assembly,automotive,medical and other fields.In order to achieve accurate measurement performance of three-dimensional force sensor,structural design,optimization,static and dynamic decoupling of three-dimensional force sensors have driven the development of three-dimensional force sensors.In this paper,a novel three-dimensional strain gauge force sensor based on flexible structure is designed by means of experimental design,and its performance optimization and output decoupling are carried out.The main contents are as follows:A novel three-dimensional strain gauge force sensor is designed.In order to improve the measurement accuracy of three dimensional force sensor,the three dimensional force sensor is designed and optimized based on the response surface method.The size parameters of the elastomers of three dimensional force sensor are taken as optimization variables,and the strain compliance matrix condition number of three dimensional force sensor is taken as the objective function of optimization.The condition number of the strain compliance matrix of the three dimensional force sensor is used as the objective function,and the size parameters of the elastomer of the sensor are optimized.The center combination test design method is used to get the test sample point.Through these test points,the correlation analysis is carried out with the software,and the relative large variables are taken as the optimal design variables.Then a response surface model is made for the overall performance index of the three dimensional force sensor by a number of sample points.Finally,the obtained response surface model is replaced by the optimization function to optimize the size parameters of the elastomer.Compared with the optimization before optimization,the optimized condition of the strain compliance matrix is reduced to 2.49.In order to verify the influence of the condition number of strain compliance matrix on the accuracy of force measurement,the error of force before and after optimization is compared,and the error of force measurement is reduced after optimization.The effectiveness of the optimization method is proved.In order to solve the coupling problem of linear output of three dimensional force sensors,a coupling error modeling method is used to decouple the output of three dimensional force sensors.Because of the coupling between the three dimensional force sensors,the force loading experiment of the three dimensional force sensor is carried out.The data obtained by the experiment are based on two methods,which are based on the generalized inverse matrix method and the coupling error modeling method,the decoupling model of three dimensional force sensor is obtained.The linear static decoupling of the three dimensional force sensor is carried out,and the accuracy of the decoupling of the two methods is compared.After solving the generalized inverse matrix method,the type I error of the force sensor is 1.218% and the maximum type II error is 0.298%.When the coupling error modeling method is used,the type I error of the decoupling force sensor is 0.273% and the maximum type II error is 0.239%.The results of the verification show that the method based on the coupling error modeling is effective for the static decoupling of the three dimensional force sensor.In order to solve the coupling problem of nonlinearity output of three dimensional force sensor,the nonlinearity static decoupling of three dimensional force sensor based on BP neural network is adopted.The reasons for the coupling of three dimensional force sensors are analyzed.The experimental data obtained from the loading experiment on the force of the three dimensional force sensor in the third chapter are used as the training samples of the BP neural network,the input data is normalized,the untrained data as the BP neural network test sample is also normalized.After continuous training,the threshold and weight of the BP neural network model are obtained.When the difference between the predicted and expected values reaches the presupposition requirement,the parameters of the network model are determined and the BP neural network model of the three dimensional force sensor is obtained.The error analysis of the BP neural network model is compared with the error obtained by the static linear decoupling method used in the third chapter.The results show that the nonlinearity decoupling method of BP neural network is superior to the linear static decoupling in reducing the coupling error.The maximum type I error of the BP neural network is 0.205% and the maximum type II error is 0.170%.The effectiveness of the BP neural network for the output decoupling of three dimensional force sensors is verified.In order to solve the error caused by the coupling of dynamic signals between three dimensional force sensors,a three-dimension force sensor is dynamically decoupled based on diagonal dominance dynamic compensation.The natural frequency measurement experiment is carried out by the force hammer knocking method,and the transfer function of the three dimensional force sensor is obtained.A diagonal dominance compensation matrix is obtained by using the optimization function in the MATLAB genetic algorithm.The transfer function of the three dimensional force transducer obtained by the experimental method is calculated by formula operation,and the transfer function after dynamic decoupling is obtained.The unit step of the output signal of the three dimensional force sensor before and after the dynamic decoupling of the diagonal dominance compensation can be seen.After decoupling,the output signal of the main direction is greatly improved and the coupling signal is obviously reduced.It shows the availability of this method.