Abstract: Seismic exploration methods are widely used to investigate shallow crustal structures, and double-difference tomography is a commonly employed high-precision imaging technique. However, due to issues such as insufficient arrival-time picks and limited ray-density coverage in field data, the inversion results of double-difference tomography often exhibit uncertainties. To address this problem, this paper proposes a method based on the whale optimization algorithm (WOA)-optimized convolutional neural network (CNN) (WOA-CNN) for predicting seismic arrival times, thereby improving the reliability of double-difference tomography results. The feasibility of this method is evaluated through its application to velocity-structure imaging in the Jinchuan mining area. The Jinchuan large nickel-copper-platinum group element sulfide deposit is located in Gansu Province, China, and has been mined for over 50 years. As shallow ore bodies become depleted, the demand for deep mineral exploration has become increasingly urgent. To investigate the deep structure of the mining area, 69 short-period three-component seismometers were deployed in the Jinchuan mining area, continuously collecting seismic data for 27 days. Using the LOC-FLOW location method, 309 mining-induced seismic events were accurately located, including 4 316 P-wave arrival-time picks as input for the double-difference tomography. However, due to the limited number of P-wave arrivals, the imaging results were unsatisfactory. Subsequently, the WOA-CNN was used to predict the arrival times of the 309 events, obtaining a total of 11 124 P-wave arrival-time picks. By incorporating the predicted arrival times, the three-dimensional P-wave velocity structure of the study area was successfully constructed. Analysis combined with previous research results verifies the existence of two independent rock bodies in the deep part of the mining area, and this finding is highly consistent with existing geological data. Additionally, the typical “funnel-shaped” structure of the shallow ore body along the Ⅱ-36 survey line was verified. Based on the velocity variation patterns, it is inferred that there is still significant exploration potential below 2 km in the eastern section of the Jinchuan mining area.

Key words: WOA-CNN, arrival time prediction, 3D P-wave velocity structure, mine seismic location, metallic mineral exploration, Jinchuan mining area