基于数据驱动的MEMS加速度计自检测自校正技术研究

Research on Data-driven MEMS Accelerometer Self-detection and Self-correction Technology

摘要:

MEMS加速度计是一种用于测量载体加速度的微型集成系统,已被广泛应用于生产生活中；然而MEMS器件在使用过程中易由内外因素影响出现故障,若不能及时检测故障并校正故障数据,将会使得系统无法准确感知外界环境进而导致控制出现偏差,因此及时检测MEMS加速度计的故障并校正其故障数据对于提高系统的鲁棒性、测量准确性以及控制稳定性等方面具有重要意义；现有检测校正方法大多依靠建立加速度计的物理模型或构建传感器冗余网络来实现加速度计的自检测与自校正,但这些方法存在建模复杂且引入额外误差或硬件资源需求高等问题；为了避免建模不准确引入的误差并减少算法对硬件资源的需求,基于近传感器计算的思想,设计了一种轻量化的、基于数据驱动的MEMS加速度计自检测自校正算法；测试结果表明,算法对冲击、偏差、信号丢失、恒定输出四种故障的检测率均达到90%,校正后数据与正常数据的平均绝对误差小于0.15 g,并且具有在2.55 ms内处理加速度计数据的能力。

Abstract:

MEMS accelerometers are miniature integrated systems widely utilized for measuring carrier acceleration in various industrial and domestic applications. However, these devices are susceptible to faults due to internal and external factors during operation. Failure to promptly detect and correct these faults may lead to inaccurate perception of the external environment, resulting in control deviations. Hence, timely detection and correction of MEMS accelerometer faults are crucial for enhancing system robustness, measurement accuracy, and control stability. Existing detection and calibration methods often rely on establishing the accelerometer"s physical model or constructing redundant sensor networks, which suffer from complexities in modeling and introduce additional errors or high hardware resource requirements. To mitigate inaccuracies introduced by modeling and reduce algorithmic hardware demands, a lightweight, data-driven self-testing and self-calibration algorithm for MEMS accelerometers is proposed based on the notion of proximal sensor computation. Test results demonstrate that the algorithm achieves a detection rate of 90% for four types of faults: shock, bias, signal loss, and constant output. The average absolute error between calibrated data and normal data is less than 0.15 g, with the ability to process accelerometer response data within 2.55 ms.