According to IEC 61508, a safety-related system is regarded as type B if it presents a high complexity (i.e. the failure mode of at least one component is not well defined, or the behaviour under fault conditions cannot be completely determined), or if there is insufficient data to support claims for failure rates. This paper proposes a modelling method adapted to the evaluation of failure probabilities for systems with uncertain behaviour under fault conditions. To this aim, weighted “continuous gates” are introduced in a fault tree framework. By acting on weight values, it is then allowed to continuously graduate system part architectures between series and parallel structures. An intelligent transmitter is used as example. Probabilities of failure on demand are assessed, with both failure rates and behaviour uncertainty analyses. Results tend to show that the lack of knowledge in system behaviour can be partially handled by this kind of approach.