Abstract: In order to solve the limitations that exist in fault detection based on the negative selection algorithm,a fault detection model using the intelligence fusion approach based on negative selection principle is proposed.The principle and structure of the model are presented,and its training algorithm is derived.Taking the advantages of neural networks training,the information of anomalous patterns is stored in the distributed neural networks-based detectors.This model has the distinguished capability of adaptation,which is well suitable for dealing with practical problems under time-varying circumstances.A fault can be found out through the relevant activated detector.The simulations of anomaly detection in chaotic time series are carried out to investigate the effect of model parameters on the capability of fault detection.In the end the illustrative stall detection experiments of compressor in an aeroengine demonstrate that the proposed method can achieve precise discrimination in the pattern features of stall signals,which also testify that the neural networks-based detectors have better recognition ability than the binary encoding detectors.