The efficient operation of urban freight transport systems is crucial to the economy. This study addresses its dynamic growth mechanism, which is challenging for traditional models due to node heterogeneity. Using the Xi'an City as a case study, we analyze 14 days of truck GPS data, POIs, and road networks. A freight network is constructed via OD flows derived from stops were identified by using the ST-DBSCAN algorithm. Analysis reveals non-equilibrium evolution: freight flow positively correlates with arterial road density but negatively with secondary roads. We propose a node heterogeneity-driven spatial-temporal growth model, incorporating attractiveness, geographic scale, and indirect interaction parameters, dynamically optimized via Bayesian optimization. Experimental results demonstrate that our model significantly outperforms traditional methods over 14 days, successfully revealing growth patterns and the dynamic relationship between hierarchical structure and functional coupling. These results provide a data-driven basis for urban freight planning, promoting innovative applications of complex network theory in logistics geography.