Indexing technical data in a materials database

Much of the data used by design engineers lack precision, particularly materials property data which are measured in inexact physical experiments and engineering tests. The indexing of such data needs to suit the various types of proximity searches of interest to an engineer. It is shown that B/sup +/-tree performance is best when the data are stored as specific values. A new structure, the interval tree, is presented and shown to be useful when much of the stored data is expressed in the form of ranges.<>