Industrial computed tomography (CT) technology is the core means of industrial non-destructive testing

and the accuracy and efficiency of its data reconstruction algorithm directly affect the credibility of the detection results. However

the presence of truncated projection data results in traditional reconstruction methods facing issues such as artifacts

low resolution

and high computational complexity. This article proposes an artificial intelligence based industrial CT truncated projection data reconstruction system

which combines deep learning technology with high response dataset construction methods to optimize data preprocessing

feature extraction

and model training processes. Experiments have shown that the system outperforms traditional iterative algorithms in both reconstruction accuracy and computational efficiency

and has the ability to dynamically adapt to complex industrial scenarios.