关键词: 熔丝制造, 性能预测, DDGA-BP模型, 数据切分策略

Abstract: The neural network model can effectively solve the problem of nonlinear coupling between the process parame鄄 ters of fused filament fabrication and the performance quality of molded parts through an adaptive learning mechanism, which provides a new way to achieve an accurate prediction for the performance of molded parts. In this study, taking nozzle temperature, printing speed, printing angle and printing layer thickness as the input parameters, on the basis of the data divide (DD) strategy and the genetic algorithm-back propagation (GA-BP) neural network model, we constructed a data divide-based genetic algorithm-back propagation (DDGA-BP) neural network model by synchronously locking the network structure and data division indices. This model can be used for predicting the performance of shape memory polymer (SMP) manufactured via fused filament fabri鄄 cation, including tensile strength, shape fixation rate and shape recovery rate. The model training set was con鄄 structed by experiment, and the training results showed that the mean absolute error (MAE) and mean squared error (MSE) between the predicted values of the improved DDGA-BP model and the experimental results were reduced within 0.2, and the correlation coefficients were increased to above 0.8. This research establishes a reli鄄 able data-driven framework for the prediction of mechanical and shape memory properties in SMP manufactured via fused filament fabrication.

Key words:  fused filament fabrication, performance prediction, DDGA-BP model, data divide strategy