Executive Summary : | In the aviation industry, flight risk management and flight safety are the top most priorities. Airlines need to constantly monitor/improve safety and fleet productivity. Accurate preventive maintenance is one of the key factors that impact safety & productivity. Advanced tools & technologies are required to process and analyze flight data to effectively identify safety risk, determine root causes and monitor corrective actions. Being a complex system, the aero engine is prone to have issues which need to be identified upfront to minimize disruptions in field operations.
Diagnostic and prognostic inspection of critical components is mandatory to monitor the health and ascertain safety & serviceability of aircraft. One of the inspection techniques widely used is “Boroscopic Inspection Technique” where a boroscopic inspection camera is inserted inside the engine through borescope ports and a thorough video of interiors of engine is recorded. The purpose is to detect & evaluate damage due to foreign/domestic object ingestion or severe aircraft operations without opening the engine. These videos are then inspected by engineering teams to identify and quantify any damage. Often processing of these videos takes high processing time and require expert intervention to take decisions related to continued flight operations. The process requires specialized skill and several labor hours to generate an inspection report. Being a manual process, it is prone to misses and human cognitive bias. With advances in Artificial Intelligence & Machine Learning techniques, intelligent & automated decision making is essential to ensure we keep the aircraft safe & in-operation.
The purpose of this project is to develop an automated cognitive technology using computer vision and deep learning techniques to generate automated inspection report from borescope inspection videos/images. It is very important to ascertain if the blade rubs against the casing & removes material from the casing and if any blade tip loss has happened due these rubs. Heavy casing rubs result in material property degradation and can lead to high cycle fatigue (HCF) crack initiation and these cracks can lead to critical hardware failures. Both the above conditions impose a risk for engine health & flight safety and are closely monitored.
A machine learning method resulting from a successful implementation of this project will use state of the art technology in computer vision and deep learning to automatically process the boroscopic videos and generate final inspection reports. This will play a very critical role in automating the process of monitoring & highlighting rub locations/magnitude and tip losses during engine operation faster & with higher accuracy. This will help in correlating performance over the engine lifecycle. A deep learning algorithm will be trained with existing data (inspection videos) and will then be used to detect events in new videos. |