【作者】 张志军；

【导师】 顾幸生；

【作者基本信息】 华东理工大学 ， 控制科学与工程， 2012， 硕士

【摘要】 随着计算机技术的不断发展,现代连续工业生产过程越来越依赖计算机集成制造系统(CIMS)在过程控制、操作优化、决策分析和生产调度中的应用。作为对生产工况的直接反映,这些庞大的测量数据为控制优化、计划调度和决策分析提供依据。然而在实际的测量中,误差是难以避免的。由于测量误差的存在,导致测量数据不符合过程的变量平衡关系。一旦采用含有测量误差的数据源作为决策依据,可能会带来经济上的损失,严重的甚至可能会影响生产过程的稳定运行。本文首先针对某年产50万吨甲醇的甲醇合成装置工艺流程和控制要求进行了深入的分析,研究了该装置的数据校正问题。由于在负责的环境中运行,测量仪表难以避免在测量过程中产生偏差,从而导致测量数据可靠性下降。针对这种现象,本文对该甲醇合成装置的测量数据进行了数据校正,减少了测量误差的影响。本文引入了一种基于支持向量机的SV回归方法,该方法将显著误差视为回归模型的复杂度进行去除,同时还可以对测量数据进行数据协调。与此同时,针对粒子群算法的特点,本文提出了惯性权重动态更新策略与速度变异策略,得到了改进粒子群算法(IMPSO)、改进食物导向粒子群算法(IMFGPSO)和改进微分粒子群算法(IMDPSO)等,通过对不同测试函数的仿真,结果显示这些改进算法具有更好的全局搜索性能。本文将采用这些算法来处理甲醇合成装置的数据校正问题。本文采用了基于SV回归的数据校正方法,结合基于参数方程处理等式约束的粒子群算法及其改进算法,采用实际过程运行数据对某甲醇合成装置的测量数据进行了研究。结果显示,基于SV回归方法的数据校正算法可以有效地减少误差对测量数据的影响。本文研究结果对实际生产过程有指导意义。更多还原

【Abstract】 With the continuous development of computer technology, modern industrial processes are increasingly dependent on computer integrated manufacturing system (CIMS) for process control, operation optimization, decision analysis, and production scheduling. As a direct reflection of the production conditions, these measurement data provide a basis for the control optimization, scheduling and decision analysis. However, measurement errors are inevitable. Due to the presence of measurement errors, the process measurement data could’t meet the balance between the variables. Once the data source contains measurement errors is used as a basis for decision making, it may lead to economic losses, even worse may affect the stable operation of the production process.Firstly, the process and control requirements of some methanol synthesis unit, outputs 500,000 tons every year,are studied in this paper, followed by data reconciliation research of the unit. As a result of 6MPa working condition, the measuring instrument is difficult to avoid bias, which will lead to decreased reliability, in the measurement process. Thus, data rectification technical is applied for this unit to reduce the impact of measurement error. Secondly, this paper illustrates a SV regression method,which can remove gross error defined as regression model complexity,and implement data reconciliation at the same time. What’s more, according to the principle of particle swarm optimization (PSO), w-dynamic updated strategy and velocity variation strategy are proposed in this paper, which bring out some improved algorithms such as improved PSO (IMPSO), improved food guide PSO (IMFGPSO) and improved differential PSO (IMDPSO), followed by their testing in different testing functions, and the result shows that these improved algorithms have better global search performance. These algorithms are supposed to deal with the data reconciliation problem of methanol synthesis unit.In this paper, the data rectification technical based on SV regression is applied to the methanol synthesis unit, with differential particle swarm algorithm and it’s improved algorithms. The results show that the data rectification method based on SV regression can effectively decrease the error of process data.The study of this paper has a guiding signification on the actual production process.更多还原