Research On Temperature Compensation Technologies Of Silicon Micro-Accelerometer

Silicon micro-accelerometer is a kind of inertial sensor made of microelectromechan-ical technology,which can be used to measure the inertia force of an object.Silicon mi-cro-accelerometer has the advantages of miniaturization,intelligence,high integration,high accuracy and mass production.It is widely used in intelligent control,consumer elec-tronics,automotive electronics,high-speed railway,Aeronautics and Astronautics.Howev-er,the performance of silicon micro-accelerometer is affected by temperature,depending on the temperature sensitivity of silicon.It is crucial to study how to improve the tempera-ture stability of silicon micro-accelerometer with temperature compensation technologies.The current research status of silicon micro-accelerometer and the temperature com-pensation technologies of silicon micro-accelerometer at home and abroad are summarized in this dissertation.The working theory,signal detection theory and temperature character-istics of silicon micro-accelerometer are analyzed.Based on the characteristics of the ex-isting temperature compensation technologies,the software compensation methods are used to compensate the temperature of the silicon micro-accelerometer.Based on the sim-ulation and comparison of the advantages and disadvantages of several temperature com-pensation algorithms,such as the surface fitting,the BP neural network and the BP neural network optimized by genetic algorithm,an improved adaptive genetic algorithm is used to optimize the BP neural network to establish the temperature compensation model of accel-erometer in this dissertation.The hardware system and software system of temperature compensation based on STM32 and the test system software based on LABVIEW are de-signed.Finally,based on the test standards of silicon micro-accelerometer,the real-time temperature experiments of silicon micro-accelerometer are completed and the temperature compensation algorithms are validated.The experimental results show that,after the compensation of BP neural network op-timized by improved adaptive genetic algorithm,the temperature coefficient of scale factor,the full temperature range of zero bias,the stability of zero bias and the nonlinearity are reduced from 49.8 ppm/?,77.26 mg,23.81 mg and 4193 ppm to 7.5 ppm/?,2.86 mg,0.90 mg and 212 ppm respectively.The results testify the validity and correctness of tem-perature models established by the BP neural network optimized by IAGA.The BP neural network optimized by IAGA not only improves the prediction accuracy and global perfor-mance of BP neural network,but also improves the convergence speed of GA.These com-pensation technologies are suitable for other MEMS inertial sensors and have a wide ap-plication prospect.