With the continuous development of urban rail transit in China, the demand for shield tunnel construction is increasing steadily. To ensure the safety of tunnel construction and the quality of backfill grouting meet standards, and to overcome the inefficiency of traditional manual GPR detection, the Loaded-to-Frame detection equipment is proposed to achieve automated rapid GPR inspection of backfill grouting quality. Detailed explanations are provided regarding its hardware composition, operational mode, and key recommended parameters. High-quality labeled GPR data is acquired based on model experiments, and machine learning model training tactics based on A-scan and B-scan are employed to achieve satisfactory performance of the artificial intelligence (AI) model. The "cloud-edge-end" architecture for intelligent backfill grouting detection is introduced; and the GPR-AI Master based on a partial "cloud-edge-end" architecture is proposed to enable cloud-based deployment and application of AI models. The dynamic feedback mechanism for intelligent applications based on the Loaded-to-Frame detection equipment is introduced, enabling overall tunnel excavation safety and application in numerous projects domestically and internationally.