Multi-Objective Hybrid Artificial Intelligence Approach for Fault Diagnosis of Aerospace Systems

Abstract

Due to the space environment’s hazards and challenges, aerospace systems are continuously exposed to many failures, such as the degradation of the subsystem performance, sensor faults, connection loss, or equipment damage. Therefore, the effective fault diagnosis for detecting and identifying any failures or unusual behaviors can be recognized as the fundamental and critical role of aerospace systems’ predictive health management process. This paper proposes a novel fault diagnosis approach using Deep Learning (DL) technique. DL has recently become a popular approach in artificial intelligence (AI) due to its supremacy in accuracy and fast inference with the huge amount of data and highest order network structures. The proposed approach consists of two main phases; the feature selection phase by Binary Grasshopper Optimization Algorithm (BGOA), and the learning and prediction phase by Artificial Neural Networks (ANNs) with voting ensemble method. The proposed approach named BGOA-EANNs has been validated and evaluated its efficacy by comparing two existing diagnosis techniques using two types of aerospace health diagnosis datasets; satellite power system and aircraft engines. The experimental results demonstrated the effectiveness and superiority of the proposed approach.

Keywords

• Aerospace systems
• aircraft engines
• binary grasshopper optimizer
• deep learning
• fault diagnosis
• prognostic health management