In the present publication an automated computational method for cluster analysis of crystal structures (and/or their powder diagrams) is described. The clustering bases on the similarity index obtained by comparing of powder diagrams. This approach allows a quick detecting of manifold crystal structures in extensive data sets, an estimation of the number of minima for an energy landscape, a control over the progress of crystal structure generation during runtime, and a quick analysis of results of crystal structure predictions. (C) 2009 Elsevier B.V. All rights reserved.