Abstract:The integration and complexity of analog circuit are getting higher, the fault types are various and some faults are intermittent,unsteady and redundant. All these means that analog circuits is becoming more difficult to diagnosis.In this paper,a IHHO-SVM combining AVMD and PE and manifold learning is put forward. Adaptive variational modal decomposition is used to obtain IMF signals, which are computed with permutation entropy(PE) to construct fault features。Then, t-distributed stochastic neighbor embeddings(t-SNE) is combined to realize dimensionality reduction while remaining excellent discrimination power of fault features vectors.Finally, Harris Hawks algorithm are combined to optimize the support vector machine for fault classification. The simulation tests show that the alorithms proposed in this paper has an excellent effect of 100%.