数据分类的两步矩阵投影算法

化工过程的数据分类是进行数据校正和协调计算的基础。常用的两层次矩阵投影变换算法在对未测数据进行分类时,可能无法识别出所有的不可估计型数据。为了准确地将可估计型和不可估计型数据分开,采用Crowe等人提出的投影矩阵,引入矩阵的绝对线性无关列的概念,提出了新的数据分类方法。数学推导证明,此方法对化工过程未测数据分类彻底,并用一个示例将新旧算法对未测数据的分类结果进行了对比,验证了新算法作数据分类的正确性。

Two-step matrix projection algorithm for data classification

Process data classification is the foundation of data reconciliation. However, the existing two-level matrix projection transformation algorithm may not properly classify the unmeasured variables. An algorithm was developed to exactly divide unmeasured variables into determinable and indeterminable data, using matrix projection with a new concept called absolutely linearly independent columns of a matrix. Mathematical deduction demonstrated that this method can effectively classify unmeasured data. A mass reconciliation example was used to compare the previous unmeasured data classification algorithm with the new algorithm. The results show that the new algorithm accurately classifies all the variables.