A real-time shipboard fire-detection system based on grey-fuzzy algorithms

To improve on the performance of traditional ship fire alarm systems, this paper investigates a dual-sensor device employing a grey-fuzzy algorithm. The theoretical aspects of the device and its experimental evaluation are presented. In terms of the algorithm, first an adaptive fuzzy classification system with an automatically generated rule base is developed for accurate fire-detection response to the output of a sensor pair (one temperature K-type thermocouple and one analog photoelectric smoke detector). Second, two alternative grey GM(1,1) prediction models are developed for anticipating trends in real-time temperature and smoke data, thus allowing early fire warning. Finally, the fuzzy system is combined with the grey-prediction algorithms for final testing. In the engine room of a docked coastal fishing trawler, two experimentally controlled fires are created, one open flame and one smoldering, and results from the sensor pair are recorded. As-detected results for each fire are processed by computer which tests the response behaviour of the alternative fuzzy-grey options and selects the optimal options set, and also compares the dual-sensor pair as conventionally operated in a commercial detector. Results indicate grey-fuzzy algorithms combining fuzzy rule-based classification and grey GM(1,1) unified-dimensional new message modelling are feasible in real-time shipboard fire detection, allowing accurate fire alarm triggering from 30 to 60 s earlier than conventional methodology.