Sina Piramoon

Date of Award

2024

Document Type

Dissertation

Publisher

Santa Clara : Santa Clara University, 2024

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical Engineering

First Advisor

Mohammad Ayoubi

Abstract

This Ph.D. thesis presents an integrated approach to mitigating lateral vibrations in vertical shaft rotary machines during transient responses by combining physics-based and data-driven methodologies. Two primary strategies are explored and validated through experimental studies on a laboratory test rig.

The research introduces a novel method that integrates Sparse Identification of Nonlinear Dynamical Systems (SINDy) with a physics-based approach to accurately reconstruct the governing nonlinear equations of vertical-shaft rotary machines. An extensive mathematical library supports this reconstruction, which underpins the design of a nonlinear controller using terminal sliding mode control (TSMC). This control technique effectively reduces lateral vibrations, ensuring the system’s stability, effectiveness, and robustness under transient conditions.

Additionally, the thesis explores the use of advanced time-series artificial neural networks, specifically Long Short-Term Memory (LSTM) and Time-Delay Neural Network (TDNN) models, to predict and control rotor vibrations in centrifuge systems with asymmetrical and variable-mass characteristics. By employing a data-driven reducedorder modeling (ROM) approach and integrating time-delay coordinates, the study achieves high-fidelity predictions of system dynamics over time. The trained neural networks facilitate the implementation of a nonsingular Terminal Sliding Mode Control (NTSMC) for real-time vibration mitigation under varying operational conditions.

To further enhance model accuracy and adaptability, the thesis also compares these neural network results with those from an Adaptive Neuro-Fuzzy Inference System (ANFIS), providing a comprehensive evaluation of different data-driven techniques.

Moreover, this thesis includes the design of a specialized test rig for the rotary system, along with mechatronic assembly diagrams. The Ph.D. test rig is specifically designed for the real-time implementation of the algorithms detailed in this thesis, providing tangible validation of their practical application in industry.

The integrated findings from these studies demonstrate the potential of combining physics-based models with machine learning and soft computing techniques to enhance the predictive accuracy and control capabilities of complex mechanical systems. This thesis contributes to the field by providing a robust framework adaptable to various types of rotary systems experiencing dynamic challenges in vibration control.

Download

Share

COinS