This paper conducts a joint analysis of monitoring data in the hidden danger areas of railway subgrade deformation using a data-drivenmethod,therebyrealizing thesystematicrisk identification of regional hidden dangers.

The paper proposes a regional systematic risk identification method based on Bayesian and independent component analysis (ICA) theories. Firstly, the Gray Wolf Optimization (GWO) algorithm is used to partition each group of monitoring data in the hidden danger area, so that the data distribution characteristics within each sub-block are similar. Then, a distributed ICA early warning model is constructed to obtain prior knowledge such as control limits and statistics of the area under normal conditions. For the online evaluation process, the input data is partitioned following the above-mentioned procedure and the ICA statistics of each sub-block are calculated. The Bayesian method is applied to fuse online parameters with offline parameters, yielding statistics under a specific confidence interval. These statistics are then compared with the control limits – specifically, checking whether they exceed the pre-set confidence parameters – thus realizing the systematic risk identification of the hidden danger area.

Through simulation experiments, the proposed method can integrate prior knowledge such as control limits and statistics to effectively determine the overall stability status of the area, thereby realizing the systematic risk identification of the hidden danger area.

The proposed method leverages Bayesian theory to fuse online process parameters with offline parameters and further compares them with confidence parameters, thereby effectively enhancing the utilization efficiency of monitoring data and the robustness of the analytical model.