Process Data Reconciliation And Its Application

Measurements are the base and starting point of the engineering technique work such as process design, simulation, optimization and control. However, process measurements from chemical plants are inherently containing random errors and possibly contaminated by gross errors, which will lead these measurements are not consistent with corresponding conservation laws and other process constraints. The goal of data reconciliation is to smooth the measured data then enhance their accuracy and reliability and to estimate unmeasured process variables. Our work is mainly concerned with steady state process and the model of data reconciliation and its solution and application are discussed. Firstly, a general weighted least squares(WLS) model of data reconciliation is introduced. The formulation and statistical properties of this model are extensively described. Meanwhile, a derivative model based on robust objective function is summarized as a contrast to WLS model. These provide a theoretical basis for solving data reconciliation problem and gross error detection. Secondly, a modified two-step matrix projection method for data classification has been proposed. A detailed deduction is given to ensure the modified data classification approach correct. Later a chemical flow sheet containing unmeasured variables in loop illustrates the limitation of existing two-step approach. The new modified approach can bring the same result as some prevalent data classification algorithms. Moreover, it appears much more straightforward than the latter in formulation and programming. Finally, an application example has been presented according to some chemical reaction processes. All the classical gross error detection methods can not identify measurement biases and process leaks. Generalized likelihood ratio(GLR) method is applied to these cases and it can identify single gross error or multiple gross errors efficiently. This approach provides a framework for identifying any type of gross error that can be mathematically modeled. So it can also be used in chemical reaction process including side reaction. For some complicated chemical reaction processes, a suggested scheme that utilizing the historical data from plants and simulation software to acquire some model parameters is given.