针对函数约束算法中传统的智能算法反演时存在鲁棒性差和易陷入局部最优的缺点, 提出了将正则化理论与细菌觅食优化算法相结合应用在颗粒粒度的测量中。引入Tikhonov平滑泛函来构建算法的目标函数, 采用L曲线法确定正则化参数; 再利用细菌觅食优化算法通过趋向、聚群、复制和迁徙等四种智能行为, 迭代计算来搜寻函数的最优解。实验仿真结果表明: 利用细菌觅食优化算法实现了在不同程度的随机噪声下的服从J-SB分布的单峰分布的均匀球形颗粒粒度分布反演, 其反演结果更稳定, 反演精度高, 对于实现稳定、快速、准确的颗粒粒度在线测量具有重要的意义。

A novel algorithm is presented which combines regularization theory and Bacterial Foraging Optimization Algorithm (BFOA), and is used to measure particle size distribution. BFOA can solve the traditional optimization algorithm’s weakness of falling into local optimum easily and having a poor robustness. Tikhonov smoothing algorithm is used to build functional objective function and L-curve method is used to determine the regularization parameter. BFOA contains four intelligent behaviors named chemotaxis, swarming, reproduction, elimination and dispersal. By using iterative calculation, the optimal solution of the function is searched. The simulation results show that BFOA achieve the inversion of the spheroidal particle of unimodal distribution that conforms to J-SB function distribution, which is retrieved at different levels of random noise. The inversion results are more stable and accurate, which have important significance for the realization of stable, fast and accurate particle size online measurement.