Fusion of convolutional neural network and Bayesian model to predict stair life and usage

This study proposes an integrated framework that integrates convolutional neural network and Bayesian model to predict the life and usage of stairs, and achieves accurate inference by comprehensively analyzing wear characteristics and material degradation. The convolutional neural network extracts spatial features from the stair wear image to effectively distinguish the upward and downward usage patterns, while the Bayesian model combines observational data such as natural corrosion, artificial deformation and crack width to infer the life distribution and provide confidence intervals. The fusion framework integrates the outputs of both through a joint probability model, significantly improving the prediction performance. Experimental verification shows that the framework performs well in wear pattern classification and life prediction, provides an innovative tool for archaeological analysis and building maintenance, demonstrates the potential of combining deep learning with probability statistics, and has broad application prospects.