An intercorrelated multi-node skin temperature model using MODWT-GAT and a case study in plateau tunnels

Skin temperature is a critical parameter for human health and comfort. Existing data-driven models typically use either mean temperatures or local-node temperatures as whole-body or regional indicators, neglecting node intercorrelations. This study proposes a 10-node model that takes into account the intercorrelations among multiple nodes. By combining the time-frequency information of the maximal overlap discrete wavelet transform (MODWT) and the spatial information of the graph attention network (GAT), the model identifies the dynamic thermal responses among multiple nodes in different environments and characterizes the intercorrelations between nodes through feature weights. To evaluate the model's performance in complex scenarios, validation was conducted on six environmental cases (air temperatures: 31°C, 34°C, and 37°C; pressures: 65.7 and 96.5 kPa). The results show that the model achieves an average recognition accuracy of 93.18%. This study reveals the intercorrelations of multi-nodes, providing a valuable research direction for monitoring and evaluating skin temperature.