Research Of Compensation Method For Pendulous Micro Machined Silicon Accelerometer

Micro mechanical silicon accelerometer is the key components of inertial measurement and navigation system, which is offered a broad application prospect in the aerospace, high-speed rail, consumer electronics and medical equipment,etc. However, the performance of accelerometer is affected by manufacturing process, internal structure and external environment, especially the temperature, so the research on how to improve the accelerometer output precision and stable performance has important practical value and engineering significance. A particular model of pendulous micro machined silicon accelerometer is carried out a series of error modeling and compensation method technology research in the paper. Main work is included:(1)According to the structure and working principle of pendulous micro machined silicon accelerometer, the temperature characteristic of accelerometer is analyzed theoretically. The performance parameters and test method of micro machined silicon accelerometer is introduced.(2)The temperature compensation model and nonlinear compensation model of pendulous micro machined silicon accelerometer are established by polynomial fitting, BP neural network and wavelet neural network fitting. The advantages and disadvantages of three methods of accelerometer compensation are summarized, which is provided guidance to how to choose the compensation method for the device.(3)Compensation circuit system for pendulous micro machined silicon accelerometer is designed based on STM32, including the hardware system and the software system.(4)Automation data acquisition system for SD6068 accelerometer is designed based on LABVIEW.(5)On the basis of reference related inertial device testing standard, the pendulous micro machined silicon accelerometer is carried on the real-time performance compensation test. After the compensation of polynomial fitting, the nonlinearity, the scale factor temperature coefficient and the temperature of zero bias are decreased from 17275 ppm, 298.3ppm/ ℃ and 98.13 mg to 662 ppm, 18.43ppm/ ℃ and 10.09 mg respectively. After the compensation of BP neural network fitting, three parameters were reduced to 88 ppm, 35.52ppm/℃ and 10.70mg/h. Test results show that: the correctness and validity of temperature model is set up by the polynomial fitting and BP neural network fitting. Compensation technology method is also suitable for other sensitive sensors, which has a broad application prospect.