Decoupling Of Triaxial Accelerometers Based On Surface Fitting And Time Frequency Analysis

In recent years,multidimensional accelerometers have been widely used in biomedical,inertial navigation,robotic dynamics control and seismic exploration due to their ability to measure acceleration in multiple axes.Practical applications such as geological exploration,robotic sensing,and mind/body source parameter measurement require high-precision triaxial accelerometers to accomplish.The inter-dimensional coupling between the axes of the triaxial accelerometers has become one of the main factors limiting the development of the triaxial accelerometers,which causes the sensor measurement accuracy and data authenticity to be affected.At present,the decoupling method of multidimensional accelerometer has low decoupling accuracy and efficiency,and most of them stay in the theoretical research stage,which is still far from the practical application.In order to suppress the inter-dimensional coupling phenomenon of triaxial accelerometers and apply the decoupling algorithm to the embedded decoupling module,this thesis carries out decoupling research based on the coupling data of triaxial accelerometers,and the specific research contents are as follows:Firstly,based on the coupling characteristics of the triaxial accelerometer,a coupling model based on vibration acceleration and vibration frequency are established based on higher order polynomials when the sensor structure and transfer function are unknown.Then,based on the coupling model,on the basis of direct surface fitting and point-line surface stepwise fitting method to solve the coupling function,the method of solving the coupling function based on line-surface stepwise fitting is proposed,and the method of determining the segmentation point by surface segmentation fitting is improved,and the segmentation method based on the extreme value point is proposed to improve the efficiency of surface segmentation.After that,decoupling experiments are conducted for the MEMS triaxial accelerometer.The experimental results show that the decoupling method based on segmented surface fitting can reduce the maximum inter-dimensional coupling error to less than 1%,and the average single-point decoupling time is 0.704ms,which has certain advantages in decoupling accuracy and efficiency compared with the classical BP neural network.Meanwhile,the decoupling method based on the line-surface stepwise fitting has better decoupling effect and can reduce the maximum inter-dimensional coupling error to less than 0.5%.The experiments verify the effectiveness and real-time performance of the decoupling method proposed in this thesis.Then,for practical applications,the acceleration signal generally obtained is a time domain two-dimensional signal with time corresponding to acceleration,and the signal frequency cannot be obtained directly,while the traditional Fast Fourier Transform cannot establish the relationship with time,therefore,this thesis further carries out the research on the time-frequency analysis of acceleration signal to obtain the signal frequency.For the acceleration signal characteristics,this thesis adopts the method of synchronous compression continuous wavelet transform to realize the time-frequency analysis of acceleration signal and obtain the vibration frequency of the sensor.In order to facilitate the subsequent transplantation of the algorithm,this thesis implements the time-frequency analysis method in C language.Compared with the actual frequency,the maximum relative error of the frequency obtained by using C language for time-frequency analysis does not exceed 1%.The experiment proves that the frequency of the signal can be obtained accurately by using this method.Finally,this thesis designs and implements an embedded decoupling module for triaxial accelerometers by analyzing the application requirements when decoupling is actually performed.By porting the decoupling algorithm and the time-frequency analysis algorithm proposed in this thesis,the F28335-based embedded decoupling module is finally implemented.In this thesis,single-point decoupling and real-time decoupling experiments are conducted for the existing MEMS triaxial accelerometers using the decoupling module.In the single-point decoupling experiments,the operational error of algorithm porting is in the order of 10^-6 and the decoupling error is in the order of 10^-3.The decoupling time of the segmented surface fitting-based decoupling method is about4.152ms,which can ensure the effectiveness and real-time performance of the decoupling module.In the real-time decoupling experiment,the decoupled signal can retain the original signal characteristics and the maximum inter-dimensional coupling error can be reduced to less than 3%,which can meet the design requirements of the decoupling module.Based on the characteristics of the coupling data of the actual triaxial accelerometer,this thesis establishes the coupling model of the triaxial accelerometer,uses the segmented surface fitting method to solve the coupling function,proposes the segmentation method based on the maximum value point,and provides a new decoupling method for the triaxial accelerometer.At the same time,combined with the synchronous compression continuous wavelet transform time-frequency analysis method to obtain the signal frequency,an embedded decoupling module for triaxial accelerometers is designed and implemented,which is of practical value for the development of multidimensional sensor decoupling methods and some multidimensional physical quantity measurement techniques with high accuracy and high real-time requirements.