基于ACO-BP算法的熔融沉积成型翘曲变形量的预测方法

Prediction Methods of the Warping Deformation Amount in the Fused DepositionModeling Based on ACO-BP Algorithm

针对熔融沉积成型翘曲变形量预测问题,提出了一种基于蚁群算法(ACO)-误差反向传播(BP)神经网络算法的预测方法。采用ACO算法优化BP神经网络的初始权值、阈值,防止其训练时收敛于局部极小。基于正交试验分别设计4因素4水平的训练样本集和4因素3水平的验证样本集。训练样本集用于预测模型的学习,验证样本集用于验证预测方法的精度。基于极差法分析了各工艺参数对翘曲变形量的影响程度。结果表明,工艺参数对翘曲变形量的影响程度从大到小分别为层高、填充率、喷头挤出温度和打印速度。采用训练样本集充分训练预测模型后,验证基于ACO-BP算法的翘曲变形量预测方法的效果。基于均方根误差(RMSE)、均方误差(MSE)、平均绝对误差(MAE)和平均绝对百分比误差(MAPE)评价模型预测精度。对于RMSE,BP算法的预测精度约为ACO-BP算法的1.7倍;对于MSE,BP算法的预测精度约为ACO-BP算法的2.9倍;对于MAE,BP算法的预测精度约为ACO-BP算法的1.6倍;对于MAPE,BP算法的预测精度约为ACO-BP算法的2.2倍。基于ACO算法优化的BP神经网络预测精度更高。

A method based on the ant colony algorithm(ACO)-error back propagation(BP)neural network algorithm was proposed for predicting the warping deformation amount in the fused deposition modeling.The initial weights and thresholds of the BP neural network were optimized by using the ACO algorithm to prevent it from converging to local minima during training.The 4-factor 4-level and 4-factor 3-level orthogonal experiments were designed as training and validation sample sets,respectively.The training sample set was used to learn the prediction model,and the validation sample set was used to verify the accuracy of the prediction method.The influences of various process parameters on warping deformation amount were analyzed using the range method.The results show that,the influences of process parameters on warping deformation amount ranged from large to small are layer height,filling rate,nozzle extrusion temperature,and printing speed.The prediction model was fully trained using a training sample set,and the prediction effect of warping deformation amount was validated using a validation sample set.The prediction accuracy was evaluated using root mean square error(RMSE),mean square error(MSE),mean absolute error(MAE),and mean absolute percentage error(MAPE).The results show that,for RMSE,the prediction accuracy of BP algorithm is approximately 1.7 times that of the ACO-BP algorithm.For MSE,the prediction accuracy of BP algorithm is approximately 2.9 times that of the ACO-BP algorithm.For MAE,the prediction accuracy of BP algorithm is approximately 1.6 times that of the ACO-BP algorithm.For MAPE, the prediction accuracy of BP algorithm is approximately 2.2 times that of the ACO-BP algorithm.The BP neural network optimized by the ACO algorithm has a higher prediction accuracy.