【摘要】 采用熔融沉积工艺(FDM)制造了碳纤维增强尼龙零件，基于响应面模型(RSM)研究了喷嘴温度(NT)、床层温度(BT)、打印速度(PS)和层高(LH)对零件拉伸强度的影响，基于RSM和混合机器学习模型(GA+ANN)预测了零件的拉伸强度。结合方差分析(ANOVA)、Pareto图和半正态分布曲线可知，BT和LH为影响零件拉伸强度的显著因素，床层温度的二阶项(BT~2)为影响显著的二阶效应。随着层高从0.2 mm增加到0.4 mm,零件拉伸强度从125 MPa线性降低至103 MPa;随着床层温度从80℃升高至120℃,零件拉伸强度从94 MPa线性增大至122 MPa。最优的ANN模型为4-7-5-1,在练习、验证、测试和全部数据集的回归系数分别达到了0.966 2、0.942 6、0.941 1和0.977 8,表现出较好的预测性能。通过对比ANN模型和响应面模型预测零件拉伸强度的准确性可知，ANN预测结果与实验值的均方误差(MSE)为1.30,而响应面模型预测结果与实验值的MSE达到了25.70。更多还原

【Abstract】 Carbon fiber reinforced nylon parts were manufactured using the fused deposition modeling technology(FDM). The effects of nozzle temperature(NT), bed temperature(BT), printing speed(PS), and layer height(LH) on the tensile strength of the parts were studied using response surface model(RSM). The tensile strength of the parts were predicted using RSM and hybrid machine learning model(GA+ANN). Based on the results of ANOVA, Pareto plot, and semi normal distribution curve, it was found that BT and LH were significant factors affecting the tensile strength of parts, while the second-order term(BT~2) of bed temperature was a significant second-order effect. As the layer height increased from 0.2 mm to 0.4 mm, the tensile strength of the parts decreased linearly from 125 MPa to 103 MPa. As the bed temperature increased from 80 ℃ to 120 ℃, the tensile strength of the parts linearly increased from 94 MPa to 122 MPa. The optimal ANN model was 4-7-5-1, with regression coefficients of 0.966 2, 0.942 6, 0.941 1 and 0.977 8 for training, validation, test, and all datasets, respectively, which demonstrated excellent predictive performances. Comparing the accuracy of ANN model and response surface model in predicting the tensile strength of parts, it could be found that the mean square error(MSE) between ANN prediction results and experimental values was 1.30, while the MSE between response surface model prediction results and experimental values was 25.70.更多还原