This paper describes the application of novelty detection for diagnosing damage in structures using the measured vibrational response signals. Novelty or anomaly detection is a technique for deciding whether measured data indicate departure from a pre-established normal operating condition. A scalar index is associated with a given measurement 'feature' of the system, and an alarm is raised if the index rises above a pre-determined threshold. As described in this article, the technique can be employed to detect the occurrence of faults in systems with time-varying parameters. In practice many systems exhibit sudden or continuous variation in their mass or stiffness while in operation. For example, the natural frequencies of offshore platforms undergo significant variations in time as a result of tides and changes of deck mass due to oil storage. As a consequence, in order to be detected using conventional techniques, damage must produce significantly greater alterations in the natural frequencies. However, the novelty detection approach can avoid this problem. Previous work has established that a novelty index based on a neural network is able to indicate damage when there are two disjoint normal operating conditions This paper extends this research with a more realistic case study, using a numerical model to simulate the dynamic behaviour of an offshore platform with a range of normal operating conditions The ultimate objective is to produce for structural damage detection a more sensitive diagnostic tool which can be employed generally in diverse applications.