Maintenance issues have in recent years grown in importance. Higher demands on availability and quality of products are some of the reasons. These demands have resulted in an increased use of condition monitoring systems, to support an effective maintenance concept. For a complex system, this technique generates an enormous amount of signals which is difficult to diagnose, and a great amount of information which is difficult to handle and interpret for the operator. Therefore, improved diagnosis and decision support tools are necessary, as a complement to the monitoring systems. This paper presents an ongoing project, trying to find a new approach to condition monitor and diagnose Kaplan hydro-power turbines, where the chain from transducer to decision making is under consideration. Risk analysis is adequate for decision making among alternative maintenance tasks and the ambition is a risk model that can be systematically updated with varying economic data and production requirements. Also, the availability is under consideration for the hydro power plant, as a total system, where the owners of the plant should be able to steer the availability according to demand. Another important aspect is the possibility of the condition monitoring system to test itself, which will prevent false alarms.