To improve adaptability

feature resolution

and identification accuracy when diagnosing mechanical faults in an on-load tap changer（OLTC）of a transformer

an improved Hilbert-Huang transform containing the Ensemble Empirical Mode Decomposition（EEMD）and the Hilbert marginal spectrum analysis was proposed

combined with the Volterra model of chaotic time series to mechanical fault feature extraction OLTC. In detail

the EEMD is applied to decompose the multi-channel vibration signal in the OLTC switching process to obtain the intrinsic mode function（IMF）component

and then the Hilbert spectral spectrum of each IMF is obtained by Hilbert spectrum analysis. Furthermore

the Volterra feature matrix is constructed based on the Hilbert marginal spectrum by using the Volterra model

and the matrix singular values are used as the fault characteristic parameters. Finally

the test platform for simulating mechanical faults in the OLTC was built. On this basis

the vibration signals generated due to typical mechanical faults were acquired and analyzed. And multi-classification SVM was established to identify data sets

thus validating the effectiveness of the method. Compared with other methods

the new method achieved a higher fault recognition accuracy.