Three-Dimensional Temperature Field Reconstruction and Dynamic Prediction of Grain Bulk Thermal Processes Based on Distributed Optical Fiber Sensing and STC-GRU

Temperature variations in grain bulk are characterized by spatial heterogeneity and temporal lag, making three-dimensional temperature information important for thermal process monitoring in grain storage. To address the limitations of conventional point-based monitoring methods, which rely on sparse measurement locations and cannot adequately characterize the full-field thermal state of grain bulk, wheat grain bulk in a flat warehouse was investigated by integrating distributed optical fiber temperature sensing, ANSYS 2024 R1 transient thermal simulation, and the STC-GRU model for three-dimensional temperature field reconstruction and short-term prediction. Internal grain bulk temperatures were first collected using a distributed optical fiber temperature sensing system and mapped to three-dimensional spatial coordinates. The temperature field was then reconstructed through ANSYS transient thermal analysis, and the reconstructed field was used as the learning target to train the STC-GRU model for next-step prediction. The ANSYS-reconstructed temperatures were generally consistent with the measured optical fiber temperatures, with MAE, RMSE, and R2 values of 0.143 °C, 0.178 °C, and 0.823, respectively. Time-step sensitivity analysis further confirmed the rationality of organizing ANSYS output at 1 h intervals. Compared with LSTM, GRU, BiLSTM, and BiGRU, STC-GRU reduced RMSE by 6.73%, 14.16%, 7.62%, and 10.19%, respectively, and MAE by 5.19%, 14.38%, 8.24%, and 6.23%, respectively, indicating its ability to learn the spatio-temporal evolution of the reconstructed temperature field. During the experimental period, the grain bulk remained under normal static storage conditions; therefore, the internal heat source term in the ANSYS model was set to zero, and the prediction reference for STC-GRU was the ANSYS-reconstructed temperature field rather than independently measured temperature data. These findings provide a reference for visual analysis of grain bulk temperature fields and short-term thermal trend assessment in flat warehouses.