In business applications where wireless sensor networks (WSN) are applied, failures in essential parts of the system must be efficiently detected and automatically recovered to avoid major losses. In this thesis we investigate existing fault detection and isolation solutions and present a fault isolation method based on the identification of system patterns. This approach has the ability to learn the behaviour of the network when specific failures occur and to combine known patterns.