Hybrid Physics-Infused Machine Learning Framework For Fault Diagnostics and Prognostics in Cyber-Physical System Of Diesel Engine

Author

Shubhendu Kumar SinghFollow

Date of Award

8-2024

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Automotive Engineering

Committee Chair/Advisor

Rahul Rai

Committee Member

Venkat Krovi

Committee Member

Srikanth Pilla

Committee Member

Zhen Li

Abstract

Fault diagnosis is required to ensure the safe operation of various equipment and enables real-time monitoring of associated components. As a result, the demand for new cognitive fault diagnosis algorithms is the need of the hour. Existing deep learning algorithms can detect, classify, and isolate faults. Still, most depend solely on data availability and do not incorporate the system's underlying physics into their prediction. Therefore, the results generated by these fault-detecting algorithms sometimes need to make more sense and deliver when tested in actual operating conditions.

Similar to diagnosis, the fault prognosis of diesel engines is paramount in numerous industries. Unexpected diesel engine failures can lead to significant operational disruptions and maintenance costs. Accurate Remaining Useful Life (RUL) estimation is crucial for proactive maintenance. Fault prognosis methods are essential for accurately estimating RUL. While current deep learning algorithms excel at identifying patterns in data, they often rely solely on data availability, neglecting to integrate the fundamental physics of the system into their predictions. Consequently, these algorithms fall short when subjected to real-world operational challenges.

The presented dissertation addresses the aforementioned issues. The specific contributions of this dissertation are: (1) Deploying a novel physics-infused one-dimensional Convolutional Neural Network (1D-CNN) based deep learning framework for diesel engine fault detection. (2) A hybrid ensemble learning framework that integrates physics principles into a 1D-CNN-based ensemble learning model to detect faults in diesel engines. (3) A physics-informed 1D CNN-based prognostics framework underpinned by a 1D CNN to estimate the RUL of the fuel injector.

Recommended Citation

Singh, Shubhendu Kumar, "Hybrid Physics-Infused Machine Learning Framework For Fault Diagnostics and Prognostics in Cyber-Physical System Of Diesel Engine" (2024). All Dissertations. 3732.
https://open.clemson.edu/all_dissertations/3732