Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
172827 | Computers & Chemical Engineering | 2012 | 13 Pages |
The methodology for computing correlations between continuous descriptors of catalytic materials and their performance in the catalysed process is addressed. Continuous descriptors are typically molar fractions of individual components of the catalyst, whereas the performance is represented most frequently by yield or selectivity of reaction products or conversion of key feed components. Measures of various kinds of correlation are recalled, and their descriptor-wise application to catalytic data for computing correlations between the composition and performance of catalysts is presented. The paper also compares the application of correlation measures to catalytic data on the one hand with the analysis of variance, on the other hand with the application of regression trees. As a case study, the presented approaches are applied to data from high-temperature synthesis of hydrocyanic acid.
Graphical abstractFigure optionsDownload full-size imageDownload as PowerPoint slideHighlights► Correlation measures applied descriptor-wise to data from catalyzed processes. ► Different kinds of correlation, e.g., general, intensity, concordance, linear. ► Comparison with the analysis of variance of catalytic data and with regression trees. ► Detailed case study using data from high-temperature synthesis of HCN. ► Case study results increase insight into the behaviour of HCN catalysts.