Interpretable AI for predicting crack propagation analysis in 3D-printed gear systems

The significance of structural integrity tests is growing for 3D-printed gear systems utilised in automotive, aerospace, and industrial applications. Crack initiation and propagation in additive manufacturing must be predicted owing to the presence of anisotropic materials, layer-by-layer flaws, and residual loads. This work presents an explainable XAI framework for predicting crack propagation in dynamically stressed 3D-printed polymer and metal gear systems. Digital image correlation, vibration sensing, and thermography collect high-resolution microstructural, geometric, and operational data. A convolutional neural network informed recurrent unit hybrid deep learning model captures the spatial patterns of cracks and their temporal changes. Engineer friendly XAI approaches, including SHAP values, Grad-CAM visualisation, and feature attribution mapping, demonstrates the process of crack growth through microstructural flaws, print orientation, loading parameters, and environmental variables. The experimental evaluation of multimaterial 3D-printed gear prototypes provides accurate predictions and robust finite element fracture model connections. This predictive framework with XAI is trustworthy and simple to use for making next-generation 3D-printed mechanical components last longer, be of high quality, and regulate their lifespans.