Executive Summary : | Axial piston pumps (APP) are widely used in various applications, including aerospace and shipping, due to their high-pressure ratings, force-to-weight ratio, and ability to transmit power under pressure. They are central to indigenously developed light combat aircraft, Tejas, which provide hydraulic power to flight control surfaces and utility systems. The aircraft hydraulic system typically includes four axial piston pumps: two main high-power engine-driven pumps, one in operation and another on standby, and two as backup pumps. These pumps are manufactured to very fine tolerances, ensuring sufficient lubrication and tight seal even at high rotational speeds. However, they are highly prone to wear by contaminants, leading to leakage and a drop in system pressure. Wear can occur in multiple parts simultaneously, changing operating parameters and leading to rapid deterioration in pump performance. Additionally, wear may not be uniform from one cylinder to another, causing non-uniform delivery at the pump outlet and downstream fluid power system. This project aims to develop an experimentally validated real-time tool for leakage fault diagnosis based on time-domain/time-frequency domain and semi-supervised/supervised machine learning methods. This tool will help diagnose the exact type of leakage fault using flow, pressure, and temperature signals, and predict the behavior of a faulty pump under operating and boundary conditions encountered in an aircraft. |