基于改进燃油流量神经网络的航空器巡航温室效应评估

Greenhouse effect assessment of aircraft cruise based on improved neural network of fuel flow-rate

为高效、深入地评估航空器巡航运行环境影响,该研究以全球绝对温变潜势为表征参数,建立航空器巡航阶段不同污染物排放的综合温室效应评估模型;构建用于评估过程中巡航燃油流量获取的三层神经网络,并引入遗传算法对网络初始权值和阈值进行优化。结果表明,经过遗传算法改进后的燃油流量神经网络相较于优化前,拟合优度由0.954 6提升至0.973 1,均方误差降低了48.15%,最大相对误差和平均相对误差也有效降低。提出了能够代替传统的BADA性能模型的污染物排放量计算方法,在保证可靠性的同时大大减少了对飞行性能参数的依赖,且能进一步给出污染物排放造成的综合温变影响。

In order to efficiently assess the environmental impact of the aircraft cruise, the global absolute temperature change potential is taken as a parameter to establish a comprehensive model for evaluating the greenhouse effect resulted from different pollutants during the aircraft cruise. A three-layer neural network is developed to obtain the aircraft fuel flow-rate during the cruise,and the genetic algorithm is introduced to optimize its initial weights and thresholds. Results show that R2 of the improved neural network is increased from 0.954 6 to 0.973 1 comparing with the primary network, the mean square error is decreased by 48.15%. In addition, the maximum and average value of relative percentage error are also effectively reduced. A pollutant emission calculation method which can replace the traditional one based on the BADA performance model is proposed. Based on the reliability of the model, the comprehensive temperature change caused by the pollutant discharge will be obtained with a greatly reduced dependence on flight performance parameters.