In order to solve the problem of low accuracy in intelligent fault diagnosis for multi-input measuring circuit such as analog multiplier,this project studied the model building method for Multi-input Multi-output (MIMO) Circuit based on series of Volterra, which is used as the model for circuit fault diagnosis.Then, we proposed the method of feature extraction for whole annealing genetic features: by utilizing the global optimization property of Whole Annealing Genetic Algorithm (WAGA), we improved parameter extraction for fault diagnosis feature; then we select the feature with the largest feature difference between various fault status to improve the accuracy of fault diagnosis.Lastly, we conducted an experiment of intelligent optimization and extraction of model-building features and fault features, using analog multiplier as an example. The result of the experiment has proved that, the method described in this article is effective in model-building and improving the accuracy of intelligent fault diagnosis.