An experimental investigation of the interaction of water waves with flexible structures acting as breakwaters was carried out. Wave profiles, mapped out by water level measuring transducers, were studied to provide information on the performance of different breakwater models. A new signal analysis procedure for determining reflection coefficients based on wavelet theory was developed and compared to a conventional method. The reliability of using wavelet analysis to separate a partial standing wave into incident and reflected wave components was verified with a numerical example. It was also verified by the small variance in the estimates of the incident wave height from independent experimental measurements. Different geometries of rigid and flexible structures were constructed and examined. Reflection, transmission and energy loss coefficients were obtained over them. The influence of various properties of the models, such as the width and the internal pressure, on the effectiveness in reflecting or absorbing the incident wave energy was determined. Various factors which affect the performance of the breakwater, including the water depth, the wave length and the wave amplitude, were measured and documented. Suspended and bottom-mounted models were considered. The flow field over and near a hemi-cylindrical breakwater model was also examined using a flow visualization technique. An overall comparison among the models has also been provided. The results showed that the rectangular models, rigid and flexible, are the most effective structures to dissipate wave energy. The flow visualization technique indicated that the flow conforms with the circular geometry of a hemi-cylindrical breakwater model, yielding no flow separation. / Ph. D.

Time-Frequency decomposition is a signal processing method for analyzing and extracting information from aperiodic signals. Analysis of these signals are ineffective when done using the Fourier transform, instead these signals must be analyzed in the time and frequency domain simultaneously. The current tools for Time-Frequency analysis are either proprietary or computationally expensive making it prohibitive for researchers to use. This thesis investigates the computational aspects of signal processing with a focus on Time-Frequency analysis using wavelets. We develop algorithms that compute and plot the Time-Frequency decomposition automatically, and implement them in C++ as a framework. As a result our framework is significantly faster than MATLAB, and can be easily incorporated into applications that require Time-Frequency analysis. The framework is applied to identify the Event Related Spectral Perturbation of EEG signals; and to vibrational analysis by identifying the mechanical modal parameters of oscillating machines. / Thesis / Master of Applied Science (MASc)

In the industrial world, ensuring safety and operational efficiency, along with constant performance improvement, is of great importance. To achieve these goals, constant measurement of the parameters, such as vibration and sound, and monitoring of the system’s behavior are necessary. This master thesis will focus on the performance of the rotary parts of the machinery. Traditionally, human inspection and manual assessment are used to outline conclusions about the behavior and condition of the machine. Testing of the rotary parts involves analyzing audio signals by manual assessment. This thesis will focus on the vibrations produced by these parts and investigate ways to optimize the assessment of rotating systems. Utilizing the numerous advantages of embedded systems, in this case, STM32 microcontrollers, this master’s thesis explores signal processing methods such as the fast Fourier transform and Morlet wavelet transform. The detailed approach to applying both methods to analyze the data from the rotating system is described. It shows that both methods are good for detecting defects in rotating machinery, and the decision on which method to choose depends on the nature of the vibration signal as well as the nature of the faults that may occur. If machinery faults manifest in the form of periodic signals, the fast Fourier transform is a better option because it is more efficient and better for real-time systems, but for non-periodic faults, the Morlet wavelet transform is preferred. Additionally, through experimental analysis, this thesis gives new ideas on where to put sensors on rotating machines to get the best results. It shows that the sensors should be mounted close to the vibration source, on a flat surface and in the direction of the vibrations. This thesis lays a solid foundation for automating fault detection in rotating machinery, showing how to collect and analyze data that can be used for future implementation of machine learning models for predictive maintenance.