中文摘要：

将反馈型（BP）神经网络和遗传算法（GA）相结合用于激光多层熔覆厚纳米Al2O3-13%TiO2（质量分数）陶瓷涂层的工艺参数优化，根据3因素3水平正交试验结果对神经网络模型结构进行训练，建立了熔覆工艺参数（熔池闭环控制温度、超声振动频率及保温箱预热温度）与涂层性能（结合强度和显微硬度）之间的遗传神经网络预测模型。在此基础上，采用遗传算法对纳米陶瓷涂层结合强度和显微硬度进行了单目标和多目标参数优化。结果表明，遗传神经网络模型预测值与试验值误差较小，相对误差不超过2.5%。遗传算法优化的涂层最大结合强度和显微硬度分别为70.7 MPa和2025.5 HV；在结合强度和显微硬度两者权重相同的情况下，当熔池闭环控制温度为2472.0 ℃、超声振动频率为31.9 kHz和保温箱预热温度为400 ℃时涂层综合性能最优，对应的结合强度和显微硬度分别为69.1 MPa和1835.5 HV。

英文摘要：

Combination of back propagation （BP） neural networks and genetic algorithm （GA） is used to optimize process parameters of the thick nanostructured Al2O3-13%TiO2 （mass fraction） ceramic coating prepared by laser multi-layer cladding technique. The neural model is trained based on the experimental results of the orthogonal test including three factors and three levels. It is developed to express the relationship between coating properties （bonding strength and microhardness） and process parameters （closed-loop controlling temperature of molten pool, ultrasonic vibration frequency and preheating temperature of incubator）. Meanwhile, the bonding strength and microhardness of the nanostructured ceramic coating are optimized by single-objective and multi-objective optimization methods based on the genetic algorithms. The results show that the prediction data of genetic neural networks model agree well with the experimental values, and the relative error is less than 2.5%. The maximum bonding strength and microhardness of the coating are 70.7 MPa and 2025.5 HV, respectively. The process parameters of closed-loop controlling temperature in molten pool, ultrasonic vibration frequency and preheating temperature in incubator are set to 2472.0 ℃, 31.9 kHz and 400 ℃ when the bonding strength and microhardness have the same weight. At this state, the overall performance of the coating is the best and the bonding strength and microhardness of the coating are 69.1 MPa and 1835.5 HV, respectively.