This work addresses the significant challenge of aging in nuclear reactors by proposing a novel field reconstruction technique, the Noise and Vibration Tolerant ResNet (NVT-ResNet). This technique is designed to acquire detailed spatial data on reactor conditions, overcoming the limitations of sparse observer data and the presence of noise and sensor variability in operational environments. The NVT-ResNet is trained to incorporate these factors, resulting in a model that is robust against disturbances. An exploration of how varying sensor numbers affect performance reveals that the model maintains a high level of accuracy, with a relative L2 error consistently under 1% and a relative L∞ error averaging less than 5%, even under conditions of 5% noise and vibrations. The NVT-ResNet also demonstrates exceptional computational speed, completing field reconstruction in microseconds, which positions it as a strong candidate for online monitoring systems aimed at bolstering the safety and performance of nuclear reactors.