基于遗传算法和BP网络的精毛纺粗纱质量预报

Worsted roving forecast based on genetic algorithm and BP neural network

针对BP神经网络收敛速度慢,容易陷入局部最优解和遗传算法搜索范围广、效率高、鲁棒性强的特点,提出将二者结合用于精毛纺粗纱过程建模和质量预报。将BP网络初始权重和阈值按一定规律串接成字符串作为遗传算法的染色体,通过选择、交叉和变异操作对其优化,优化后的值作为该BP网络的初始权重和阈值进行二次训练。采用相同的数据训练表明,未优化的BP网络达不到预定精度或陷入局部最优解,经GA优化后收敛速度快且达到了所需精度。粗纱CV值和单重的20组数据预报表明:预报值与实测值间的相对平均误差率由之前的3.56%和3.48%分别降低到2.55%和2.23%;预报值和实测值间的相关系数较之前大为提高。

Due to slow rate of convergence and falling into part minimums easily in BP algorithm, and wide search space, high search efficiency and big robustness of genetic algorithm （GA）, a new improved genetic BP algorithm combining the two was put forward to model the fore-spinning process and forecast the quality. The weight and threshold matrix of the BP network were formed a string in an orderly way as the chromosome of GA. Through the operations of selection, crossover and mutation, they were optimized and used as the initial matrix of BP model to do second training. The verification for the same data indicates ： the pure BP models can not achieve the expected precision or fall into part minimums, the models after optimization of GA all have fast convergence and achieve the expected precision. The relative mean error percent （MEP） between the forecast and the measured value of the 20 groups of testing samples are reduced to 2.55 % and 2.23 % from 3.56 % and 3.48 % respectively; the correlation coefficient between them are also improved.