This study aims to enhance communication signal stability in seismic surveys using dynamite in the Sichuan Basin’s mountainous regions. Traditional relay site selection, based on experience and field surveys, is inefficient and inconsistent. A new technology is proposed to optimize the classic Egli model with machine learning algorithms for model correction and parameter optimization, enhancing signal propagation prediction, coverage analysis, and relay station recommendations. The methodology integrates digital elevation model (DEM) data and satellite imagery with machine learning techniques, transforming site selection into a data-driven process. The enhanced Egli model, refined through linear regression and Bayesian optimization, corrects biases and fine-tunes parameters to align with actual data, improving predictive accuracy. Empirical trials conducted in the Sichuan Basin have illustrated that the novel technology is capable of pinpointing optimal radio relay locations in under 60 min. This represents a substantial leap in efficiency, outclassing traditional approaches by a factor of 5–8 times. The optimized model adapts to varying terrains and environmental changes, offering broader applicability and robustness. This technology significantly enhances seismic data acquisition efficiency and quality, providing reliable technical support. It transitions the process from experience-based to data-driven decision-making and pioneers quantitative analysis and modeling of radio signal propagation in seismic exploration.