Data Rectification Theoretical Research And Application

Due to the influence of many factors,the process data obtained through chemical manufacture devices or measuring meters may corrupted by kinds of noises,which makes the physical equations of chemical units cannot be met.Therefore, the collected data should be preprocessed before been applied in manufacture process analysis,as a sequence,data rectification technique receives wide spread applications.The rule of data rectification is to minimize square sum of errors between rectification value and its measurement value,under the conditions that materials balance,heat balance,chemical reaction measure balance and other materialization balances are satisfied.In general,data rectification can be divided into several subsections,namely the steady state testing, gross error detection,data reconciliation, etc.The corresponding research studies of theories and applications aim at the main sections of data rectification.In this dissertation,the reported research is directed at developing data reconciliation model,system error processing and associated optimization algorithms,based on the background of several chemical processes.Specifically,the contents of this dissertation comprise the following.(1) A new data reconciliation model is proposed based on the extensive analysis of the traditional model.The new model takes into account of the characteristics of measurement errors and the effects of corresponding balance equations on data reconciliation model.The validity and merits of the new model are verified theoretically through strict theoretical analysis, formula derivation,and contrastive study by using some mathematical methods such as theory of probability statistics,matrix theories,etc.The new model overcomes the disadvantages of traditional model and reduces effectively affection of big random error on reconciliation result.(2) The possibility of introducing Iterative Learning Control(ILC) into data rectification is investigated.The open-loop P-type ILC structure of iterable paramter setting of flowmeter or sensor is proposed based on convergence provability of open-loop P-type ILC method. Simulation results for efflux coefficient C indicate that the feasibility and effectiveness of using ILC in iterable paramter setting of flowmeter or sensor.It provides an effective solution for iterable paramter setting and avoids(reduces) system error caused by unreasonable flowmeter or sensor paramter setting.(3) A real-coded Quantum-inspired Genetic Algorithm (RICQGA) is presented.RICQGA adopts the interval division and real-number coding for chromosomes,which avoids fussy coding and decoding processes,low precisions,heavy computational cost,and difficult engineering applications.Simulation results of several benchmark functions validate the effectiveness of RICQGA.a linear data rectification result(secondary-air system of thermal power plant) and typical bilinear data rectification results(rectification process and extraction process of composing juice) indicate that RICQGA is feasible and effective for nonlinear data rectification.(4) An adaptive Differential Evolution (sFDE) algorithm is proposed.Different from other DE algorithms,sFDE avoids manual regulation of the scaling factor F and advances the optimization ability of DE by using nonlinear fitness function to regulate F.Besides,the multi-universe concept is added into sFDE fastening optimization search and improving algorithm efficiency.Simulation results of some benchmark functions validate the effectiveness of sFDE and dynamic data rectification results for continuously stirred tank reactor indicates the effectiveness of sFDE.In summary,the research reported in this dissertation provides a composite framework for chemical process data reconciliations.It covers advanced techniques of computational intelligence,also proposes effective solutions for both system models and optimization algorithms.