Due to the vast mileage of pipelines throughout the world, it is important that dependable leak detection systems (LDSs) are used to promptly identify when a leak has occurred so that appropriate response actions are initiated quickly. The swiftness of these actions can help reduce the consequences of accidents or incidents to the public, environment, and property. External leak detection systems  using dedicated measurement equipment such as probes and sensor cables are briefly presented, but the main focus is on internal leak detection systems which use existing field instrumentation and usually run continuously. Systems such as volume balance, mass balance or real-time transient model (RTTM) based methods are used successfully for leak detection. RTTM based methods offer excellent performance but more field sensors are needed than for simpler methods such as volume balancing, and therefore these methods are less robust because of their greater dependence on sensors which could fail. This paper therefore describes a new leak detection methodology which uses pattern recognition techniques to combine two or more internal methods seamlessly into one scheme hence improving performance, robustness and applicability. This new approach is a generalization of the extended RTTM (E-RTTM) technology presented in .