The development of engine health monitoring systems is one of the challenges of the aviation industry that can bring significant savings by improving engine availability and reliability. A health prediction framework reduces the unplanned repairs and maintenance through an accurate and timely estimate of the current engine health condition. In order to achieve higher efficiency in fault detection and also facilitate the rapid start of this process, in the aircraft, an algorithm is required that is less sensitive to the amount of training data in this problem. This paper proposes a fault detection algorithm for a turbojet engine based on a self-organizing neural network that uses the clustered data instead of the usual gathered data. This training method specifies the independent clusters of data through a support vector machine (SVM) and makes the training process more efficient. A two-layer feed-forward neural network identifies the fault patterns, and an SVM technique classifies them. The results on health, corrosion, erosion, fouling, and fuel nozzle faults from a validated thermodynamic model of the J85 engine show the capability of the proposed method in performance diagnostics. In addition, despite the significant reduction in the training data, the accuracy of fault detection has decreased from 99.8 to 97.7, which is satisfactory as a trade-off for less training data to be acquired for the faulty conditions.