FIRE DETECTOR AND FIRE DETECTION METHOD

• Main Authors: KIN TENKAI, OKAYASU KATSUYA, EHATA HIROMICHI, NODA YUSUKE, UTSUMI TERUHIRO
• Format: Patent
• Language: English; Japanese
• Published: 03.09.2020
• Subjects: ALARM SYSTEMS; CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ORDER TELEGRAPHS; PHYSICS; SIGNALLING; SIGNALLING OR CALLING SYSTEMS

To shorten a fire detection time and also to improve bearing force against a non-fire alarming factor by using a learning algorithm so as to determine a characteristic change in consideration of both of absolute temperature and relative temperature by a fire of a monitoring area.SOLUTION: A temperature input section 16 generates absolute temperature T and relative temperature ΔT based on detection temperature of a monitoring area detected by a temperature sensor 12, and, at the same time, inputs them to a hierarchical type mechanical learning section 20 so as to identify a fire or a non-fire. A learning control section 24 allows the hierarchical type mechanical learning section 20 to accept input of fire teacher data including absolute temperature and relative temperature to which a fire obtained in a fire experiment or the like is labelled and non-fire teacher data including absolute temperature and relative temperature to which a non-fire obtained in a non-fire experiment is labelled, and to learn both of data. The absolute temperature T to be input to the hierarchical type mechanical learning section 20 is detection temperature itself of the temperature sensor 12, and the relative temperature ΔT is difference temperature obtained by subtracting average temperature of a normal monitoring state as reference temperature from detection temperature of the temperature sensor 12.SELECTED DRAWING: Figure 1 【課題】監視領域の火災による絶対温度と相対温度の両方を加味した特徴的な変化を学習的アルゴリズムを利用して判断することにより、火災の検出時間を短縮すると共に非火災報要因に対する耐力を向上させる。【解決手段】温度入力部１６は、温度センサ１２で検出された監視領域の検出温度に基づいて絶対温度Ｔと相対温度ΔＴを生成し、同時に階層型機械学習部２０に入力して火災又は非火災を識別する。学習制御部２４は火災実験等で得られた火災がラベリングされた絶対温度と相対温度を含む火災教師データと、非火災実験で得られた非火災がラベリングされた絶対温度と相対温度を含む非火災教師データを階層型機械学習部２０に入力して学習させる。階層型機械学習部２０に入力する絶対温度Ｔは温度センサ１２の検出温度そのものであり、相対温度ΔＴは温度センサ１２の検出温度から通常監視状態の平均温度を基準温度として差し引いた差分温度である。【選択図】図１