Multi-parameter Joint Detection Technology For Aircraft Cargo Hold Combustible Fire

Photoelectric smoke fire detector is the mainstream of the civil aircraft cargo use fire detector,is the key to guarantee the safe operation of the aircraft,the probe is the use of smoke particles of light scattering intensity is more than set threshold for early warning,but as a result of suspended particles in the air(water vapor or dust particles)fire also with scattering effect,It also triggers a fire detector,a flaw in the principle that causes false positives to occur.According to the federal aviation administration(FAA)technology center data according to the survey,the aircraft fire alarm rate is only 0.5%,and the false alarm rate is 99.5%.the fire detector of high rate of false positives have the serious influence to the normal operation of the aircraft,brought inconvenience to the passengers travel,give the airlines are significant economic losses.Therefore,aiming at the problem of high false alarm rate of fire detector,this paper uses the principle that light of different wavelengths scatters different signals of particles to detect fire smoke,and introduces the average particle size of Sott to distinguish fire smoke particles from non-fire particles.Offered to blue light scattering signal(representation for the surface area of concentration),infrared light scattering signal(characterization for volume concentration),particulate sauter mean diameter and temperature joint detection technology is given priority to,video monitoring is complementary way of fire detection,on the basis of combining genetic algorithm to optimize the BP neural network,genetic BP neural network(GA-BP),The joint detection model of multi-fire characteristic parameters was established,and the fire identification accuracy of the model was trained by experimental data,so as to realize the identification of non-fire interference particles such as fire status of different kinds of combustible materials,water vapor and dust.Experimental results show that:(1)Through comparing the ratio of the received optical power to the transmitted optical power of blue and infrared light of corrugated paper,beech and cotton rope changes with time,and the smoldering decomposition ability is obtained: cotton rope;Scraps of paper;wood.The variation trend of blue light and infrared light PTR of combustible is related to the color of smoke generated during combustion.A large amount of white smoke will be generated when smouldering occurs,and a small amount of white smoke and a large amount of black smoke will be generated when open fire occurs.Compared with white smoke,the PTR of blue light and infrared light of black smoke is smaller.(2)Fire smoke particles and non-fire particles have different particle size ranges,and the particle size of non-fire particles is generally much larger than fire smoke particles.In this experiment,the average particle sizes of smoulter and flame cables of thesis,beech and cotton were similar,about 480±10nm,445±10nm and 530±10nm.The average particle sizes of steam and dust were 1025nm～1135nm and 2835± 290nm～3025 ±60nm,with significant differences.Therefore,salter mean particle size can be used as a characteristic parameter to distinguish fire smoke particles from non-fire smoke particles.(3)The temperature difference between smouldering and open fire is obvious for thesis,beech and cotton rope,so temperature can be used as a characteristic parameter to distinguish smouldering and open fire.As water vapor and dust disturb particles do not release heat,so temperature can also be used as an important parameter to distinguish true and false fire sources.(4)Based on the genetic BP neural network(GA-BP)algorithm,the joint detection model of multi-fire characteristic parameters was established.The experimental data verified that the predicted output values were basically consistent with the actual experimental values.The calculated average recognition rates of smouldering,open fire and interference sources were 98.52%,99.54% and 97.46%,respectively.In addition,the average absolute error rate of GA-BP neural network is 2.73%,and the recognition accuracy is 97.27%.Therefore,the multiparameter joint detection system based on GA-BP neural network model is feasible in fire detection.(5)The background elimination method is used to extract the suspected fire area,and the improved Surendra background update algorithm is used to effectively suppress the impact of aircraft cargo cabin vibration and improve the extraction accuracy of the suspected fire movement area.Finally,the non-maximum suppression algorithm is used to screen the identification box of smoke movement area and give the recognition result.