Railway infrastructure managers have to achieve a safety performance that is economically sustainable for society. To realise this, it is necessary to work with continuous improvement, e.g., to bridge the gap between possibilities with new technologies and current maintenance practice. In addition, all safety measures have to be evaluated with regard to both availability and life cycle cost. Within this context, this paper describes a case study of continuous improvement of railway infrastructure maintenance. The purpose of the case study was to support a dynamic maintenance program by improving diagnostic support related to on-condition maintenance of railway sleepers. To fulfil the purpose, a four-step process was performed. First, data from the inspection and failure reporting systems was analysed, thereafter, failures were classified for improved fault diagnosis, then the fault diagnosis process was mapped, and finally a process risk analysis was performed. Some of the result was a failure classification structure, a failure diagnosis process and analysed risks of automated failure diagnosis. The results can be used to improve the current manual inspection of sleepers and also in the development of new automated inspection supported by Artificial Intelligence (AI).