Bio-inspired geometries provide efficient and multifunctional design solutions; however, translating these complex natural forms into manufacturable digital models remains a significant challenge within conventional Computer-Aided Design workflows. This study proposes a parametric geometric modeling framework for generating bio-inspired geometries with integrated surface textures specifically intended for additive manufacturing. The proposed methodology consists of three sequential stages: (1) generation of biomimetic geometries from mathematical parametric surfaces, (2) creation of complex texture patterns through parametric formulations, and (3) integration of both components into a unified high-resolution geometric model suitable for fabrication. The framework is validated through representative case studies, including the development of a bio-inspired motorboat hull and multiple textured biomimetic surfaces. The resulting geometries were computationally generated, implemented in Python code, and successfully manufactured using fused deposition modeling, demonstrating the capability of the proposed approach to efficiently generate diverse design alternatives while preserving geometric fidelity and manufacturability. Unlike conventional sequential modeling approaches, the proposed method integrates global shape generation and local surface texture within a single mathematical formulation, enabling rapid morphological exploration during conceptual design and supporting Design for Additive Manufacturing. The proposed framework contributes to computational design by providing a unified mathematical strategy for texture-driven bio-inspired geometric modeling that expands the design space while reducing the complexity associated with traditional CAD modeling of organic geometries.
Parallel-shaft-planetary gear transmission systems are widely used in helicopter main reducers. When the load is not evenly shared among gear pairs, local overload, vibration, noise and reduced fatigue life may occur. To study this issue, a 36-degree-of-freedom bending-torsion coupled dynamic model of a parallel-shaft-planetary gear transmission system with spline-shaft connections is developed using the lumped-parameter method. The model includes time-varying mesh stiffness, mesh damping, backlash, equivalent mesh error, eccentricity error, support stiffness and spline-shaft torsional stiffness. To evaluate the load-sharing behaviour of different branches, load-distribution coefficients are defined for the input-stage gear pairs, planetary external gear pairs and planetary internal gear pairs. The effects of excitation parameters and operating conditions, such as rotational speed, acceleration and load torque, are then analysed through MATLAB-based dynamic calculation and ADAMS multibody simulation. The results show that the input-stage gear pairs are more prone to load imbalance than the planetary-stage gear pairs. This is mainly related to the dual-input structure, phase difference and high-speed excitation. Increasing speed and acceleration tends to worsen load-sharing performance, while increasing load torque improves relative load sharing within the studied range. The results also indicate that backlash and mesh stiffness need to be properly optimized rather than simply increased. Excessive backlash can strengthen impact excitation, while overly large mesh stiffness may make the system more sensitive to dynamic errors. This study demonstrates good numerical consistency between the analytical dynamic model and the multibody simulation. Experimental validation will be carried out in future work.
Fused Deposition Modeling (FDM) three-dimensional (3D) printing has made significant progress in the fabrication of complex structures and additive manufacturing processes. However, improper printing parameter settings can lead to defects, which are often complex in shape and prone to variation. Timely detection of these defects and the ability to halt the printing process are crucial for improving efficiency and reducing costs. Therefore, this study first analyzes six common surface defects in FDM 3D printing, which exhibit distinct multi-scale characteristics. A publicly accessible defect dataset is constructed by collecting and processing defect images. To address the challenges of multi-scale defect detection, an improved YOLOv8-based detection model is proposed. This model integrates Deformable Convolutional Networks (DCNv2), Multi-Scale Dynamic Attention (MSDA), and a lightweight Adaptive Downsampling (ADown) module, aiming to optimize both detection accuracy and computational efficiency to ensure high performance in real-time applications. Experimental results demonstrate that the proposed model achieves a mean Average Precision at an Intersection-over-Union threshold of 0.5 (mAP0.5) of 88.5
Wire Electrical Discharge Machining (WEDM) is a widely employed non-conventional machining technique capable of achieving complex shapes and high levels of accuracy. Accurate estimation and prediction of machining characteristics such as Surface Roughness (SR) and Material Removal Rate (MRR) are essential for ensuring the reliability and quality of the manufacturing process. In this research, the WEDM machining parameters, including pulse-on time (TON), pulse-off time (TOFF), peak current (IP), gap voltage (VG), and dielectric flow rate (DF), were varied while machining mild steel as the workpiece, and SR and MRR were obtained as responses. The target is to develop a predictive model utilizing Machine Learning (ML) algorithms to predict the responses with respect to WEDM process parameters. Appropriate ML algorithms, including Support Vector Machine (SVM), Random Forest (RF), Ridge Regression (RR), Artificial Neural Network (ANN), and K-Nearest Neighbors (KNN), were built after proper data preprocessing. The performance of the ML model was assessed through different metrics, such as the coefficient of determination (R2), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). The results prove the effectiveness of ML technology in forecasting SR and MRR in WEDM. The SVM model with the application of Grid Search hyperparameters tuning demonstrated the superiority to other algorithms in terms of R2 = 0.94712, MSE = 0.01064, RMSE = 0.10313, and MAE = 0.08902 for SR, whereas RF demonstrates higher results for the MRR prediction with R2 = 0.97236, MSE = 0.00658, RMSE = 0.08114, and MAE = 0.07006. Thus, the results show the perspective of applying ML-driven predictive analytics in WEDM, which will contribute to the improvement of the manufacturing process control, productivity, and the quality of products. The findings identify the prospect of applying ML-driven predictive metrology in the forecasting of surface integrity and machining efficiency in WEDM, which will contribute to the optimization of the manufacturing process and SDG 9: Industry, Innovation, and Infrastructure.
This study presents an integrated experimental computational and microstructural investigation of Gas Metal Arc Welding (GMAW) on AISI 446 high-chromium ferritic stainless steel, combining data-driven process optimisation with thermomechanical finite element analysis (FEA) and scanning electron microscopy energy dispersive X-ray spectroscopy (SEM–EDS) validation. A Central Composite Design Response Surface Methodology (RSM) framework was employed to characterise the effects of welding current (70–150 A), wire feed rate (3–11 mm/s), travel speed (2–6 mm/s), and shielding gas flow rate (10–18 L/min) on five weld quality responses as bead penetration, bead height, bead width, tensile strength and HAZ hardness measured as Vickers microhardness across 28 experimental observations using AISI 309 ferrite filler wire. Second-order RSM models achieved R2 values of 98.2–99.63
