Switchboard Fire Detection System Using Expert Inference Method Based on Improved Fire Discrimination

Abstract

This paper proposes a switchboard fire detection system that uses the expert inference method based on improved fire probability for accurate fire identification. The conventional expert inference method has a disadvantage in that the range of the evaluation index for fire probability is limited because the defuzzification process uses a quantified set shape. To overcome this problem, we propose an improved expert inference method that uses a defuzzification correction operation after defuzzification. This technique applies a quadratic function to a range of reduced fire potentials, and then gradually expands the reduced range. The fire probability range was expanded using the improved expert inference method with the defuzzification correction operation, which also extended the fire detection range. As a result, it is expected that the fire awareness of this watchdog can be improved by the defuzzification correction operation.