This study explores the impact of biosignal-based attention monitoring on train driver performance in the context of smartphone usage, a critical factor influencing railroad safety. The persistent problem of smartphone distractions, which severely impair situational awareness and contribute to accidents, necessitates innovative solutions to enhance operational safety. To address this issue, this study develops an electroencephalogram (EEG)-based system for detecting smartphone usage in train drivers and analyzing its effects on cognitive performance. A full-type train simulator was used to replicate real-world train operations, where EEG data were collected from 25 participants under two experimental conditions: (1) train driving with smartphone usage, and (2) train driving without smartphone usage. A deep learning-based classification model, utilizing Long Short-Term Memory (LSTM) networks, was developed to analyze EEG signals and detect smartphone-related cognitive impairments. The model achieved an accuracy of 85.6% in distinguishing smartphone usage states, demonstrating its effectiveness in detecting cognitive changes associated with smartphone distractions. Furthermore, the findings indicate that smartphone usage leads to a 1.4x increase in response time to critical situations, significantly impacting reaction times and error rates. Unlike traditional behavior-based monitoring methods, this study pioneers an objective, real-time EEG-based smartphone usage detection system, offering a proactive strategy for accident prevention in railroad operations. By integrating deep learning with biosignal analysis, this research contributes to the advancement of real-time safety monitoring systems, providing new insights into human performance assessment in high-risk environments.