To improve adaptability, feature resolution and identification accuracy when diagnosing mechanical faults in an OLTC,a method combined EEMD,VolterraModelandDecision Acyclic Graph Support Vector Machine (DAG-SVM) was proposed. In this paper, the Volterra model for chaotic timeserieswasfirstlyappliedtoOLTCfaultdiagnosis,which could efficiently process non-stationary signals. Based on this, a new feature extraction method combining EEMD and Volterra model was proposed, which has high adaptability and feature resolution. Moreover, The DAG-SVM multiclassification model was applied to identify OLTC mechanical faults, which could realize the pattern recognition and automatic division of various mechanical faults of OLTC. Finally, an OLTC mechanical fault test platform was built, which could simulate some typical mechanical faults, such as loosening of moving contacts, lessening of transition contact and motor jam. Based on test platform, the new OLTC mechanical fault diagnosis method was verified by experiments.