This paper proposes a novel method for bearing fault classification based on the FitzHugh-Nagumo (FHN) neuron model, known for its capability to spike in a chaotic manner. Using the high sensitivity of the FHN model in a chaotic mode to frequency components in external stimuli, we developed a fast and efficient method for classifying non-stationary signals which are fed into the neuron model. Processing of the neuron output is performed by analyzing the histogram of inter-spike intervals, which is common practice when studying signals of biological neurons. We demonstrate the effectiveness of our approach by classifying acceleration time-series of normal and damaged bearings from Paderborn University dataset. Our results show the classification accuracy achieving up to 99.1% with KNN classifier. These findings have important implications for the development of fast and reliable fault diagnosis systems in industrial applications.