Research On Data Reconciliation In Material Balance System

Data reconciliation to measure network redundancy analysis, significant error correction and data reconciliation, which applies the redundancy process measurement data, will reduce the impact of the error data for process measurement data and obtain more accurate measurements. However, tranditional data reconciliation algorithms have some flaws and limitation. Some problem about data correction in process industry has been discussed in this thesis.The contribution of the thesis is as follows.1) For the previous lack of process modeling methods, a hierarchical material balance modeling including the device layer, layer scheduling and statistical modeling of the system level has been proposed. All levels are given a direct relationship between the mathematical constraints. The first level results are constraints for the next level which make the use of measurement information maximize.2) For the problem of MT-NT algorithms such as matrix of reduced-rank and data pollution, a modified NT-MT combined algorithm, which makes full use of advantage of the NT, which will estimate data pollution, update the covariance and the solution measured values. The result of the algorithms does not cause matrix rank decreasing and reduces the effects of significantly error on other measured value and data pollution.3) Evidence theory has been studied. This thesis discusses about the reliability of measuring instruments and other information. By introducing the concept of the importance of evidence, we put forward a new significantly error detection methods based on the improved D-S evidence theory to avoid pollution data in the solution process and the reliability of measurement data can be obtained. There result is well applied in industrial practice.