Output details
15 - General Engineering
University of Southampton
Detection of material property errors in handbooks and databases using artificial neural networks with hidden correlations
Significance of output:
During the search for property-property correlations in materials using artificial intelligence rather than traditional composition-structure-property mapping in materials science, we found artificial neural networks are also capable of checking scientific data in handbooks and databases, using hidden property-property correlations. Errors were found in widely used handbooks and databases even for properties of the elements. Manually checking of such big data sets is a very difficult job. This offers a possible automated data-checking approach that can be used in database cleansing.