The impact of heat strain on pilots’ physiological characteristics and operational effectiveness is substantial, necessitating comprehensive assessments using multiple physiological indicators. Although Electroencephalography (EEG) is renowned for its real-time brain function monitoring, its efficacy as a sole indicator for accurately discerning heat strain requires further validation. This study employs a novel three-dimensional single-channel convolutional neural network (3DCNN) to analyze EEG data for detecting heat strain in pilots. Simulated flight experiments conducted under varied temperature and humidity conditions facilitated the extraction of both time-domain and entropy-domain EEG features. These features trained the 3DCNN, using the Composite Index of Heat Stress (CIHS) as the label. Results indicate that the 3DCNN effectively discriminates heat strain using both a single-feature approach based on differential entropy and a multi-feature fusion strategy incorporating time-domain features, differential entropy, and fuzzy entropy, with both methods achieving high accuracy.