A Quality Prediction Method for Dual-Nozzle FDM Molded Parts Based on CIWOA–BP

The fused deposition molding (FDM) process can manufacture complex molded parts based on molten filaments. The quality of parts molded by this process is measured primarily by tensile strength, flexural strength, and surface roughness. Due to the complexity and variety of FDM process parameters, there is a nonlinear variation relationship between the combination of process parameters and the quality of molded parts, which makes the accurate and stable prediction of quality particularly important. To this end, an improved whale optimization algorithm combined with a neural network is proposed to predict the quality of FDM molded parts and make up for the shortcomings of the traditional methods with low prediction accuracy and instability. Based on the FDM production process, four process parameters, namely, printing layer thickness, nozzle temperature, chamber temperature, and filling interval, were selected as input variables of the model. The prediction model was obtained by optimizing the hyperparameters of the neural network using the improved whale algorithm. A four-factor three-level experimental method was used to obtain experimental data on the tensile strength, bending strength, and surface roughness of the styles. The accuracy and stability of the model were verified according to Monte Carlo cross-validation based on the above data set. The prediction results of model CIWOA–BP are compared with those of model WOA–BP and BP in a comparison experiment. The results show that the FDM part quality prediction method based on the CIWOA–BP model has higher prediction accuracy and stability.