Industrial Process Data Correction Technology Research And Analysis

In modern industry process, accurate and reliable data is the foundation of production decision-making, process monitoring and control, operations analysis and improvement, However, due to kinds of sources such as human errors, instrument degradation and malfunction, process-related errors, equipment leaks, and other unmeasured errors, measurements can be contaminated with random errors and/or gross errors which seriously affect the accuracy and reliability of the data. Data reconciliation (DR) is a procedure of optimally adjusting measured data so that the adjusted values obey the chemistry or physics conservation laws and other constraints. In this dissertation, robust DR and some methods of gross error detection are further researched and studied. And the methods with some defects are improved. The main contributions of this dissertation are described as follows:(1) Research of M estimate:By studying the characteristics of robust estimation and its application in the data correction, and anglicizing the structural conditions of the maximum likelihood estimate of commonly used robust estimation, a new robust estimation called LnRLS is presented, which not only meet the requirements of maximum likelihood estimation, but also has good robustness and stability.(2) Research of gross error distribution:By studying the robustness of Tjoa and Biegler's contaminated normal distribution and its application in the process of bilinear constraints, the model has some defects:estimate gross error variance,and priori probability of the occurrence of outers, limited magnitude of gross error and is less stable. According to robust adaptive distribution model presented by GAO Qian, a new robust adaptive distribution model is proposed, which can effectively avoid the deficiency of contaminated normal distribution model. Besides the new way has a good inhibitory effect for various ranges of gross errors.(3) Research of gross errors caused by leak:By studying the respective advantages of two gross errors detection methods:NT and GLR, the joint method NT-GLR detection method is raised, which not only effectively avoids the NT method does not determine the location of the gross errors, complex calculations and misjudgment of GLR method, but also can detect gross errors caused by leaks. The simulation results show that this method is effective.(4) Research of industrial process:To analyze and build a model of a methanol production process, by use of two-step metbod and LnRLS robust estimation it is successful to get a good coordination result of total flow rate and its composition.