Research On Data Rectification Methods For Process Industry

The highly competitive nature of the world market, the increasing importance of product quality, and the growing number of environmental and safety regulations have increased the need to introduce fast and low-cost changes in chemical processes to improve their performance. The decision-making process about possible modifications in a system requires knowledge about its actual state. This information is obtained by collecting a data set. But measurements always contain some type of error which may make process fail, it is necessary to correct their values to know objectively the operating state of the process. Two types of errors can be identified in plant data:random and systematic errors. The target of data rectification is to make the measured data more effective using data redundancy. This dissertation studies the existed problem of data rectification, and makes processes in the following aspects:1. A GLR-NT combined method based on generalized likelihood ratio and nodal test is introduced for gross error detection and data reconciliation, making full use of the advantages of both generalized likelihood ratio and nodal test. The simulation results showed that the method could get better performance and was superior to both sole GLR method and NT-MT method for the system with more than one error, especially when the gross error's magnitude was small or several biased stream were counteracted at the same node. Finally an actual example is provided to indicate the usefulness of the proposed method.2. A new dynamic data reconciliation method based on constrained strong tracking filter was proposed. Strong tracking filter was used for dynamic data reconciliation, while algebraic constraints embedded in strong tracking filter. An adaptive strategy was introduced for dynamic data reconciliation of large-scale systems, which further reduces the computing time of reconciliation. Two simulations of examples prove the efficiency and superiority of the proposed algorithm. A new dynamic gross error's detection and processing method is applied in the introduced dynamic data reconciliation method, which combined the mahalanobis distance and history horizon method. A example indicates the algorithm's usefulness.3. A novel algorithm of unmeasured data classification based on pseudo-inverse matrix was presented. For the new algorithm, the unmeasured data can be classified properly only by calculating the multiplication result between the coefficient matrix's pseudo-inverse matrix of unmeasured variables and itself. The method is simple and prone to realize. The measured data is still classified through matrix projection algorithm. A comprehensive data reconciliation strategy is gotten by combining the data classification method with data reconciliation algorithm. Theory deduction demonstrated that the presented method can classify the unmeasured variables effectively and two examples prove the algorithm's validity.