Abstract: Rolling element bearings serve as critical supporting components in the drive motors of essential nuclear equipment, including reactor coolant pumps and motor-operated valves. Their operational condition directly determines the reliability of these rotating systems, thereby impacting the safety and stability of nuclear reactor operations. Incipient fault features of rolling element bearings are usually weak, and traditional methods are prone to poor diagnostic accuracy due to noise interference. To address this issue, an enhanced Kurtogram method for incipient fault diagnosis of rolling element bearings was proposed in this paper. First, the spectrum of fault signal is adaptively segmented on multiple scales using variational mode decomposition (VMD) to obtain a series of mode components. At each scale, a series of mode components located in different frequency bands are obtained, and these mode components are arranged in descending order of their center frequencies. Second, the correlated kurtosis value of each mode component is calculated, and the period is estimated using the autocorrelation function. When periodic impacts caused by bearing faults occur, the value of correlated kurtosis increases. The higher the correlated kurtosis value of a mode component, the more the fault information it contains. Therefore, this characteristic of correlated kurtosis is utilized to select the optimal mode component for envelope analysis. Third, the correlated kurtosis values of the mode components at each scale form the correlated kurtosis matrix. An enhanced Kurtogram is generated by ordering the scales from low to high. In this enhanced Kurtogram, the horizontal axis represents frequency, the vertical axis represents scale, and the color tone indicates the magnitude of the correlated kurtosis values. Finally, the node with the maximum correlated kurtosis is identified from the enhanced Kurtogram, and the corresponding mode component is extracted for envelope analysis. From the envelope spectrum, the fault characteristic frequencies of the bearing can be identified, enabling fault diagnosis. The proposed method was validated using bearing fault data from Case Western Reserve University (CWRU) and bearing life test data conducted in the lab. The vibration signal from the CWRU includes electrical noise. In the envelope spectrum of the mode components extracted by the proposed method, distinct bearing fault characteristic frequencies were identified. In contrast, the traditional Kurtogram method failed to identify characteristic frequencies in the filtered signal spectrum due to electrical noise interference. During the bearing life test, the damage occurred on the inner race of the bearings: One bearing exhibits a single localized defect, and the other bearing shows multiple distributed defects. The vibration signals of the incipient damage were processed using the proposed method, and prominent spectrum lines were observed at the theoretical bearing inner race fault characteristic frequencies in the envelope spectrum of the extracted mode components. These results indicate that the proposed method effectively enhances weak features in incipient fault signals, thereby facilitating accurate and reliable fault diagnosis.