Combining neural networks, fuzzy logic, and Kalman filtering in an oil leak detector for underground electric power cables

This paper presents some of the issues that must be dealt with during the implementation of an oil leak detector in underground power cables. By using a very limited number of sensors, the detector must perform a considerable amount of signal processing in order to achieve reasonable security and dependability. Three original solutions making use of Neural Network, Fuzzy Logic, and Kalman Filtering are presented and compared.