Nowadays submersible deep well pumps are the most used irrigation systems in agriculture field. Efficient operation and economical life of pumps is an important issue. One of the most important parameters affecting pump efficiency and life is cavitation The cavitation is one of the problems frequently faced in the pump systems that widely used in the agriculture field. The cavitation could cause more undesired effects such as loss of hydraulic performance, erosion, vibration and noise. This paper presents a novel model for the detection of vortex cavitation in the deep well pump used in the agriculture system using adaptive neural fuzzy networks. The data submergence, flow rate, power consumption, pressure values, and noise values used for training the ANFIS (Adaptive-Network Based Fuzzy Inference Systems) network are acquired from an experimental pump. In this study, we use to the sixty-seven data for training process, while the fifteen data have used for testing of our model. The average percentage error (APE) has obtained as 0.08 % and as 0.34 % respectively for 67 training data and for 15 test data. The performance of the implemented model shows the advantages of ANFIS. The result of this study shows that ANFIS can be successfully used to detect vortex cavitation. This paper has two novel contributions which are the usage of noise value on cavitation detection and find out cavitation by using adaptive neural fuzzy networks. During the cavitation, the pump parameters must change by controller for prevent unwanted pump errors. The strategy proposed could be preliminary study of automatic pump control. Also proposed novel control strategy can be used for cavitation control in agriculture irrigation pumps, because of easy set up and no need extra cost. The ANFIS based model has real-time applicable thanks to rapid and easy control. It is possible to set safe boundaries in submergence in this model. Thus, users by adjusting controllable parameters can prevent cavitation and increase pump efficiency.