A temperature compensation method for MEMS accelerometer based on LM_BP neural network

摘要

This paper reports an effective and practical temperature compensation method for improving the performance of MEMS torsional accelerometer. The reported method is developed based on LM_BP neural network, where the precise temperature dependency model of the MEMS accelerometer is established by utilizing the measured sensor response from -40°C to 60°C as learning data. Based on the sensor response measured under different temperatures, the temperature dependency model is optimized and the real-time temperature compensation is realized, which results in the improved performance of the accelerometer. A prototype temperature compensation system with digital output was designed. And the performance of the accelerometer with the temperature compensation system was tested. The measurement results indicate that the temperature coefficient of scale and the full-temperature zero bias stability are improved greatly, which decreased from 298.3ppm/°C and 16.62mg/h to 35.52 ppm/°C and 2.3mg/h respectively. Meanwhile, the maximum nonlinearity over the test temperature range decreased from 3329ppm to 603 ppm.