Error Analysis And Uncertainty Research In Calibration Of Linear Accelerometers

Linear accelerometers are widely used in inertial systems. The precision of the inertial systems is highly affected by the precision of linear accelerometers. The main method to improve linear accelerometers’precision is to improve the precision of the calibration. Error analysis and uncertainty research are conducted to the error sources during the precision centrifuge testing, to help improve the scaling accuracy of calibration of linear accelerometers.First, the simplified model of acceleration load for the10"6level precision centrifuge test is derived, and the sensor misalignment about a vertical axis is proved to have little influence to the acceleration load. Based on the acceleration load model, the error of the acceleration load caused by the angular velocity, pitching alignment, static radius, and dynamic radius is analyzed separately. It’s proved that the influence of calibration caused by measurement method of dynamic radius could be ignored in the normal situation, and to satisfy the calibration of high precision, the static pitching alignment should be within1".Second, the linear coefficients in the accelerometer model should be calibrated in the tumbling testing in gravity, and the expressions of the second and third order coefficients calibrated in precision centrifuge testing are solved. Monte-Carlo simulation method is used to analyze the uncertainty of the second and third order coefficients, considered the relative uncertainty of acceleration load during scaling testing is about10-5to10-6. The results show that the relative uncertainty of second order coefficient is bigger the third order coefficient. The simulation results could be used to match the precision of equipment for the tumbling testing and precision centrifuge testing.Third, the coupling parts in the accelerometer model and the weighted least square method are analyzed. During the precision centrifuge testing, the output of linear accelerometer is combined with the acceleration loaded along the IA、PA、OA. The coefficients using two different accelerometer models, with or without the coupling items, are compared. As the condition of least square method could not be satisfied, the weighted least square method is introduced and the equations of accelerometer model coefficients are derived. Then the weighting factor is discussed. Further, the results using the least square method and the weighted least square method are compared. In the end, the calibration system of precision linear accelerometers is designed. The hardware in the system is designed to achieve the automation of calibration. Based on the demand analysis of the calibration system, hardware is chosen and the structure of the hardware is built. Then, to accomplish the automation, the flow chart of the software in the calibration system is built, and software frame is designed. Concretely, the function and algorithm of the main module are described.