International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 09 Issue: 05 | May 2022
www.irjet.net
p-ISSN: 2395-0072
Comparative Study on the Prediction of Remaining Useful Life of an Aircraft Engine 1Arpitha
V, 2A. Soumi Narayani, 3Abhilasha M S, 4Varshini S A, 5Sparsha Yadav M, 6Dr.Revathi V, 7Akshatha Venkatesh Prasad
1,2,3,4,5 Student,
Department of Computer Science, Dayananda Sagar University, Bangalore, India Professor, Department of Computer Science, Dayananda Sagar University, Bangalore, India 7Student, System Engineering and Engineering Management, Fachhochschule Südwestfalen, Soest, Germany ---------------------------------------------------------------------***------------------------------------------------------------------6Associate
Abstract - Aviation Safety is a major component of the Aviation Industry. The practice of preventive maintenance and Prognostic Health Management is a new area for research in the Aviation industry. The improvement of different safety systems and scheduled maintenance of airplane engines is beneficial in reducing operational and maintenance costs. The abundant information available from flight systems through different sensors recording data at different situations is helpful to obtain data patterns. The study of these data patterns obtained from the complex sensor data, post data-normalization and the removal of noise, helps to evaluate the ‘Remaining-Useful Life’ (RUL) of the airplane engine. ‘Supervised Machine Learning’ is a division of machine learning where the model is trained with labelled training data. To train the model we will be using the CMAPSS dataset for RUL prediction. This study aims to introduce a comparison between different supervised machine-learning models based on the RUL of the aircraft engines. The predictions made would be helpful to schedule maintenance of airplane engines only when required, preferably, once the threshold RUL value is reached. Key Words: Remaining Useful Life of an aircraft engine, PHM08 Prognostic Data Challenge, Supervised Machine Learning, CMAPSS.
1. INTRODUCTION Airplane systems are designed taking into accordance different failure scenarios that can occur during operation. Aviation safety is mainly dependent on the efficient and reliable operation of the engine as it is the most critical component. It provides thrust which is mainly required to keep the airplane air-borne. Hence, the prognostication of the RUL of the engine, ensures airplane’s safety. ‘Prognostics Health Management’ (PHM) is a novel technology aimed at predicting the occurrence of failures in components and consequently minimizing unexpected down-times of airplane engines, thereby, reducing the maintenance and the operational cost.
© 2022, IRJET
|
Impact Factor value: 7.529
|
The continuous operation of an aircraft engine makes it very vital for the different sensors to record data and report it to the main system. This ensures the engine is running both optimally and safely. The performance degradation process of the engine are due to different responses of various sensors due to noise/insensitivity, they show an uncertain tendency to degradation trend. The most sensitive sensors for this process are Temperature measurement sensors, Pressure measurement sensors, RPM measurement sensors, and Air Mass flow measurement sensors which are chosen as inputs to the RUL prediction. Currently, approaches for the RUL prediction of systems are categorized as: 1. ‘Physics based’ models Usually, Physics based model is used in the scenario with failed physical model of the components for predicting the RUL which is mostly dependent on the physics of failure of the components and they use failed historical samples or limited historical samples for the accurate prediction of RUL. However, for a complex system it is not economical to implement a physics-based approach. 2. ‘Data driven’ approach This is a robust approach requiring less prior knowledge, but quite a bit of Experimentation and Computation. The data-driven method uses a physics-based model. For RUL estimate, it makes use of the monitored operational data on system health. When the failure physics of a system is sophisticated or unavailable, the data-driven method relies on the system's degradation procedure and easily available data. In comparison to a physics-based model, the datadriven method for a system gives precise ‘RUL’ predictions which can be applied conveniently and affordably. Furthermore, data-driven method can be separated bifurcated into ‘statistical techniques’ utilizing regression methods and ‘AI techniques’ involving neural networks and SVM.
ISO 9001:2008 Certified Journal
|
Page 1448