火灾烟颗粒粒径分布的反演是一个典型的病态问题,容易因为陷入局部最小值而丢失全局最优解。在详细分析了随机噪声对烟颗粒群光散射Mueller矩阵元随角度分布的影响之后,采用全局搜索能力很强的模拟退火算法,实现了对球形模型下单分散系和对数正态分布系的粒径反演,在有相对强度为信号最大值3%的随机噪声干扰下,反演结果的误差都小于0.3%。然后使用该反演程序对烟颗粒分形凝团的散射光数据进行反演,得出了不同分形维数的火灾烟颗粒分形凝团在球形模型下的光学等效半径,并且火灾烟颗粒分形凝团的回转半径与光学等效半径之间具有近似线性关系。 Inversion of particle size distribution of fire smoke particles is a typical pathological problem, and it is easy to lose the global optimal solution because of falling into local minimum. After analyzing the influence of random noise on angular distribution of light scattering Mueller matrix element of smoke particles, the simulated annealing algorithm with strong global search ability is used to realize the monodisperse and logarithmic normal distribution of spherical particles Particle size inversion, the error of the inversion results is less than 0.3% under the random noise interference whose relative intensity is 3% of the maximum value of the signal. Then the inversion procedure is used to inverse the data of the scattered light of the fractal agglomeration of the smoke particles, and the optical equivalent radius of the fractal agglomeration of the smoke particles with different fractal dimensions in the spherical model is obtained. There is an approximate linear relationship between the radius of gyration and the optical equivalent radius.