MEMS加速度计温漂预测补偿模型

Thermal drift prognosis and compensation model of MEMS accelerometer

温度漂移是影响MEMS加速度计实用效能的关键问题,为提高加速度计的精度和使用范围,需对温漂进行建模和补偿。首先通过温度循环试验对MEMS加速度计温漂特性进行了分析,然后分别采用曲面拟合法、SVM模型和RVM模型建立了温漂预测补偿模型,最后应用环境温度试验数据对模型进行检验和验证。结果表明:三种方法均能够有效的预测加速度计温漂,补偿后的温漂滞环开口由60 mg分别下降到5 mg、10 mg、15 mg。曲面拟合法简单、精度高,但对系统重复性要求高,且对噪声比较敏感;SVM模型法计算能力强,但计算量较大,模型参数较多;RVM模型法模型参数较少,计算量小,但训练时间较长,且预测精度不如其他两种方法高。

Thermal drift is a key factor that influences the practical effectiveness of MEMS accelerometer.The modeling and compensation of thermal drift are necessary for improving the precision and expanding the use scope.In this paper,the characteristics of thermal drift were analyzed according to the temperature cycling test.Thermal drift prognosis and compensation models were built by surface fitting,support vector machine and relevance vector machine,respectively.The testing results indicate that these three models are all effective in making prognosis on the thermal drift of accelerometer.After compensation,the hysteresis errors,with initial 60 mg,decreased to 5 mg,10 mg and 15 mg,respectively.The surface fitting method has the simplest structure and the highest precision,but its requirements on system repeatability are critical.The SVM model has greater calculation ability,but needs more parameters and computation.The RVM model needs less parameters and computation,but requires longer training time and has the lowest precision among the three models.