Simultaneous fault diagnosis for aircraft engine using multi-label learning

Abstract

Fault detection and isolation system is crucial for the safety and reliability of aircraft engine. Traditional techniques of data-driven fault diagnosis for aircraft engine mainly focus on single fault diagnosis problems by means of the single-label learning strategy. However, the simultaneous fault diagnosis problems cannot be ignored in reality. In this research, two data-driven approaches based on multi-label learning and support vector machine are proposed to address the simultaneous fault diagnosis for an aircraft engine. Given that the simultaneous fault data are more difficult to obtain than single fault data, the proposed approaches have the ability to diagnose both single fault and simultaneous fault for aircraft engine when the fault diagnosis system is trained using single fault data only. The experimental results show that the proposed approaches can diagnose the simultaneous fault for an aircraft engine with high accuracy requiring low computation burden and a small number of single fault training data. In addition, the supplementary experiment confirms that the diagnosis accuracy of the proposed methods can be further improved by adding a small amount of the simultaneous fault data into the training dataset.