一种自确认软测量方法的研究与应用

软测量技术在过程工业中得到了广泛应用于代替或部分代替仪表的功能进行监控。但是,长期困扰软测量应用的是其鲁棒性问题。因此,提出了新型的软测量模型即自确认软测量模型。利用主元分析对输入传感器数据进行了在线校验和故障重构,不仅能确保了输入传感器(辅助变量传感器)数据的可靠性。同时,利用一定的状态参数来指示当前输入传感器的状态。此外,通过输出方差和归纳区间估计两种方法对软测量输出的不确定性进行了描述。模型的输出将不是单个预测值的输出,而是同时输出五种信息:带概率区间的输出、模型的状态(故障状态,重构状态和迷失状态等等)、不确定性、故障信息和校验输出值,并对发生故障的输入传感器进行数据重构和修复以达到模型自校验和自诊断的目的。所提出方法在污水处理过程中得到了有效应用。

Research and Application on a Novel Self-validating Soft Sensor

A soft sensor is widely used as a alternative for process monitoring, but robust problems prevailing in soft sensor application hinder its further usage. A novel soft sensor such as self-validating soft sensor is proposed, which not only deals with input data online calibration and fault reconstruction problems, but also further ensures the reliability of input data by performing Principal Component Analysis (PCA) analysis. In addition, uncertainties are characterized by output variance and inductive confidence predictor (ICP). The output with uncertainties, model statuses uncertainties, faulty messages and validated output are implemented for ensuring proper fault reconstruction and model calibration. The effectiveness was proved in a wastewater treatment process.