Dutch Railways use a special train for the ultrasonic inspection ofrails. The output of the ultrasonic scanning system installed on thetrain consists of echo images -- so-called B-scans. The B-scans areclassified according to the images of rail constructions, noiseartefacts, and/or defects that they contain. Currently, at TNOInstitute of Applied Physics a hybrid rule-based / case-basedclassification system is being built for the train. This paperdescribes the way the cases are represented and matched in the system,the way the case-base is organised, and the way the case retrieval isdone. Interesting aspects of the system are: use of fuzzy membershipfunctions for case matching, decision tree (C4.5) based top-levelcase-base organisation, and clustered bottom-level case-baseorganisation.