自确认多功能传感器的研究

The whole system will be affected once the sensors are faulty in the online detection. The self-diagnosis, self-repair and self-estimation functions are expected in application of sensors. The proposed self-validating multifunctional sensor can validate its own work condition. It not only can detect multiple parameters, but also can provide on-line measurement uncertainty, which benefits its reliability. Further, the fault diagnosis is done and the fault is substituted by an optimal estimated value to implement the fault recovery. This project centers on the online uncertainty evaluation, fault diagnosis and recovery by using multi-variable relevance vector machine. The correlation among multiple parameters is fully used under small sample. Traditional self-validating sensor only gives some measurement value status and it is absence of device health level. Combined with the historical health information, the health evaluation and health forecast of multifunctional sensor is extended, which is an important development in self-validating sensor. Both the gas quality detection system and water detection experimental system are designed to verify and improve above self-validating algorithms, and the corresponding strategy becomes more applicable. This project will play a significant role in the development of self-validating theory and self-validating sensors.

在线检测系统中的传感器若发生故障，将会影响整个系统运行。人们希望传感器具有自诊断、自恢复、自评估功能。我们提出的自确认多功能传感器可以确认自身工作状态，它不但能同时检测几个参数，而且可以输出测量值的在线不确定度，对测量值及传感器状态进行评估；对传感器进行故障诊断，并能输出最佳估计值代替错误数据，实现数据恢复。本项目研究多功能传感器的在线不确定度计算；研究利用多变量相关向量机等理论方法，充分利用多个被测参数间的相关信息，实现在线、小样本情况下的多功能传感器的故障诊断、数据恢复；并在传统自确认传感器对测量值评估的基础上，利用测量数据、历史数据及传感器模型，进一步实现对传感器健康状态的评估和预测，从而扩展传统的自确认传感器的功能。构建气体检测、水质检测的自确认多功能传感器实验系统，通过实验验证并完善自确认理论算法，使之更具有普适性。本项目对自确认理论及自确认多功能传感器的发展具有重要意义。

本项目研究的自确认多功能传感器是一种新型的智能传感器，它不但能测量多个参数，而且能对传感器自身工作状态进行自确认。本项目研究利用先进的智能信息处理的理论方法实现多功能传感器的自确认：利用主元分析、稀疏非负矩阵分解进行多故障检测；在传感器发生故障时利用小波相关向量机预测器实现数据恢复，以保证系统正常工作；利用确认的随机模糊变量和灰色自助法的动态不确定度方法对传感器进行在线不确定度评定；利用灰色理论和多源信息融合的健康可信度计算方法对自确认多功能传感器的健康状态进行评估和预测。本项目利用几种敏感元件和计算机组成检测氢气、甲烷等气体的自确认多功能传感器系统，并在组建的气体检测系统中进行大量实验，设置冲击故障、掉电故障、恒值输出、偏置故障、噪声干扰、漂移等故障，进行故障检测、数据恢复、健康评估等实验，完善并验证所研究的各种算法,实现自确认功能。大量实验证明了这些方法的可行性和正确性。智能化、集成化是传感器的发展方向，随着传感技术、计算技术、信息处理技术的发展，将这些算法嵌入到微型化的多功能传感器中，制成一体化的微型自确认多功能传感器。本项研究对自确认多功能传感器的发展具有重要意义。