Artificial neural network optimized by Honey Badger Algorithm for predicting and explaining properties of 3D printed objects

The 3D printing technologies commonly applied in China’s manufacturing industries such as automotive and aerospace are hindered in the ability to forecast the mechanical properties (i.e., Tensile Strength (TS), Young’s modulus (YM)) of printed parts because the printing parameters such as the material type, layer thickness, and print speed interact in highly nonlinear ways. Conventional prediction models tend not to be able to capture these nonlinearities, and thus predictions are in accuracy rate while production processes become inefficient. The study utilizes a publicly available dataset of 3D-printed nylon specimens reinforced with short carbon fibres, obtained from tensile testing records published on Mendeley Data. The data include multiple process parameters and measured mechanical responses used for model development and validation. The proposed HBA–ANN model achieved an R2 of 0.9949 for tensile strength (TS) and 0.9928 for Young’s Modulus (YM), confirming its capability to accurately predict key mechanical properties of 3D-printed polymer components. The optimized model had an R2 = 0.9949, MSE = 0.00374149, and RMSE = 0.0161 N (normalized), which signifies the high predictive accuracy of the model. The sensitivity analysis showed that material type and layer height are the most significant factors influencing the predictions The HBA-optimized ANN outperformed conventional and advanced models, achieving a higher R2 value that demonstrates its superior predictive performance. This method significantly improves the accuracy of mechanical property predictions with increased process optimization and quality assurance for 3D printing that can be of great benefit in high-precision and material-reliability-required industries such as aerospace, automotive, and consumer goods production.