Manufacturing processes produce a considerable amount of data as dimensions are measured, tests are performed and assembly checks are undertaken. Predominantly these data are used to inform and help improve the associated manufacturing processes and procedures. A variety of Knowledge Discovery techniques [1] have been introduced in the engineering field, most typically in areas with large quantities of data [2]. This paper describes research into the use of such techniques in the manufacture and assembly of large complex engineering products, an area which is characterised by low volume of data and dispersed databases. The developed methodology seeks to incorporate various approaches, with emphasis on using extracted knowledge to inform the implementation of subsequent techniques. This investigation centres on discovering and quantifying relationships between the various balance and vibration tests performed throughout assembly of gas turbine rotors, and to highlight critical parameters. Current assembly practices do not use forward prediction of test performance, and the first stages of this project aim to produce a model to enable this. The scope of this model will then be extended to feed this knowledge back to be used in the design and manufacture of future components.