Since its inception in the nineteenth century, the Internal Combustion Engine (ICE) remains the most prevalent technology in transportation systems to date. In order to minimize emissions, it is important that ICE is operated according to its optimized design conditions. As such, condition monitoring and Fault Detection and Diagnosis (FDD) tools can play an important role in detecting conditions that would affect the operability of the engine. In this research, different signal-based Fault Detection and Diagnosis (FDD) techniques are researched and implemented for fault condition monitoring of ICE. The implementation of prognostics for the engine in an automated form has important consequences that include cost savings, increased reliability, reduction of GHG emissions, better safety, and extended life for the vehicle. In this research, in order to carry out FDD onboard, a low-cost and flexible internet-based data-acquisition system (DAQ) was designed and implemented. The main part of the system is an embedded hardware running a full desktop version of Linux. This sensory system leverages the positive aspects of both real-time and general-purpose architectures to ensure engine monitoring at high sampling rates. Unlike other commercial DAQ systems, the software of this device is open-source, free of charge, and highly expandable to suit other FDD applications. In addition to data collection at high sampling rates, the FDD system includes advanced FDD strategies. The Fault Detection and Diagnosis strategies considered use a combination of Fourier Transforms (FT), Wavelet Transforms (WT), and Principal Component Analysis (PCA). Meanwhile, Fault Classification was carried using Neural Networks consisting of the Multi-Layer Perceptron (MLP). Three strategies were comparatively considered for the training of the Neural Network (NN), namely the Levenberg-Marquardt (LM), the Extended Kalman Filter (EKF), and the Smooth Variable Structure Filter (SVSF) techniques. The proposed FDD system was able to achieve 100% accuracy in classifying a set of engine faults. / Thesis / Master of Applied Science (MASc)

Yes / Burns are one of the obnoxious injuries subjecting thousands to loss of life and physical defacement each year. Both high income and Third World countries face major evaluation challenges including but not limited to inadequate workforce, poor diagnostic facilities, inefficient diagnosis and high operational cost. As such, there is need to develop an automatic machine learning algorithm to noninvasively identify skin burns. This will operate with little or no human intervention, thereby acting as an affordable substitute to human expertise. We leverage the weights of pretrained deep neural networks for image description and, subsequently, the extracted image features are fed into the support vector machine for classification. To the best of our knowledge, this is the first study that investigates black African skins. Interestingly, the proposed algorithm achieves state-of-the-art classification accuracy on both Caucasian and African datasets.

As the planetary climate continues to evolve, it is important to build an accurate long-term climate record. State-of-the-art atmospheric science requires a variety of approaches to the measurement of the atmospheric structure and composition. This thesis supports the possibility of inferring cloud properties from sun photometer observations of the cloud solar aureole using an artificial neural network (ANN). Training of an ANN requires a large number of input and output parameter sets. A cloud radiance model is derived that takes into consideration the cloud depth, the mean size of the cloud water particles, and the cloud liquid water content. The cloud radiance model derived here is capable of considering the wavelength of the incident sunlight and the cloud lateral dimensions as parameters; however, here we consider only one wavelength—550 nm—and one lateral dimension—500 m—to demonstrate its performance. The cloud radiance model is then used to generate solar aureole profiles corresponding to the cloud parameters as they would be observed using a sun photometer. Coefficients representative of the solar aureole profiles may then be used as inputs to a trained ANN to infer the parameters used to generate the profile. This process is demonstrated through examples. A manuscript submitted for possible publication based on an early version of the cloud radiance model was deemed naïve by reviewers, ultimately leading to improvements documented here. / Master of Science / The Earth's climate is driven by heat from the sun and the exchange of heat between the Earth and space. The role of clouds is paramount in this process. One aspect of "cloud forcing" is cloud structure and composition. Required measures may be obtained by satellite or surface-based observations. Described here is the creation of a numerical model that calculates the disposition of individual bundles of light within water clouds. The clouds created in the model are all described by the mean size of the cloud water droplets, the amount of water in the cloud, and cloud depth. Changing these factors relative to each other changes the amount of light that traverses the cloud and the angle at which the individual bundles of light leave the cloud as measured using a device called a sun photometer. The measured amount and angle of bundles of light leaving the cloud are used to recover the parameters that characterize the cloud; i.e., the size of the cloud water droplets, the amount of water in the cloud, and the cloud depth. Two versions of the cloud radiance model are described.

Human interfaces to computers continue to be an active area of research. The keyboard is considered the basic interface for editing control as well as text input. Problems of correct typing and typing speed have urged research for alternative means for keyboard replacement, or at least "resizing" its monopoly. Pointing devices (e.g. a mouse) have been developed, and supporting software with icons is now widely used. Two other means are being developed and operationally tested, namely, the pen for handwriting text, commands and drawings, and spoken language interface, which is the subject of this thesis. Human/computer interface is an interactive man-machine communication facility that enjoys the following advantages. • High input speed: some experiments reveal that the rate of information input by speech is three times faster than keyboard input and eight times faster than inputting characters by hand. • No training needed: because the generation of speech is a very natural human action, it requires no special training. • Parallel processing with other information: production of speech works quite well in conjunction with gestures of hands and feet for visual perception of information. • Simple and economical input sensor: microphones are inexpensive and are readily available. • Coping with handicaps: these interfaces can be used in unusual circumstances of darkness, blindness, or other visual handicap. This dissertation presents a design of a Human Computer Interface (HCI) system that can be trained to work with an individual speaker. A new approach is introduced to extract key voice features, called Median Linear Predictive Coding (MLPC). MLPC reduces the HCI calculation time and gives an improved recognition rate. This design eliminated the typical Multi-layer Perceptron (MLP) problems of complexity growth with vocabulary size, the large training times required and the need for complete re-training whenever the vocabulary is extended. A novel modular neural network architecture, called a Pyramidal Modular Neural Network (PMNN), is introduced for recursive speech identification. In addition, many other system algorithms/components, such as speech endpoint detection, automatic noise thresholding, etc., must be tailored correctly in order to achieve high recognition accuracy. / Ph. D.

Yes / A 9 x 18 x 1 feed-forward neural network (NN) model trained using a resilient back-propagation algorithm and early stopping technique is constructed to predict the shear strength of deep reinforced concrete beams. The input layer covering geometrical and material properties of deep beams has nine neurons, and the corresponding output is the shear strength. Training, validation and testing of the developed neural network have been achieved using a comprehensive database compiled from 362 simple and 71 continuous deep beam specimens. The shear strength predictions of deep beams obtained from the developed NN are in better agreement with test results than those determined from strut-and-tie models. The mean and standard deviation of the ratio between predicted capacities using the NN and measured shear capacities are 1.028 and 0.154, respectively, for simple deep beams, and 1.0 and 0.122, respectively, for continuous deep beams. In addition, the trends ascertained from parametric study using the developed NN have a consistent agreement with those observed in other experimental and analytical investigations.

No / Optimum multi-layered feed-forward neural network (NN) models using a resilient back-propagation algorithm and early stopping technique are built to predict the shear capacity of reinforced concrete deep and slender beams. The input layer neurons represent geometrical and material properties of reinforced concrete beams and the output layer produces the beam shear capacity. Training, validation and testing of the developed neural network have been achieved using 50%, 25%, and 25%, respectively, of a comprehensive database compiled from 631 deep and 549 slender beam specimens. The predictions obtained from the developed neural network models are in much better agreement with test results than those determined from shear provisions of different codes, such as KBCS, ACI 318-05, and EC2. The mean and standard deviation of the ratio between predicted using the neural network models and measured shear capacities are 1.02 and 0.18, respectively, for deep beams, and 1.04 and 0.17, respectively, for slender beams. In addition, the influence of different parameters on the shear capacity of reinforced concrete beams predicted by the developed neural network shows consistent agreement with those experimentally observed.

We investigate the coherent gradient hypothesis and show that the coherence measurements are different on real and random data regardless of the network's initialization. We introduce "diffs," an attempt at an element-wise approximation at coherence, and investigate their properties. We study how coherence is affected by increasing the width of simple fully-connected networks. We then prune those fully-connected networks and find that sparse networks outperform dense networks with the same number of nonzero parameters. In addition, we show that it is possible to increase the performance of a sparse network by scaling the size of the dense parent network it is derived from. Finally we apply our pruning methods to ResNet50 and ViT and find that diff-based pruning can be competitive with other methods.

Tire-Pavement Interaction Noise (TPIN) becomes the main noise source contributor for passenger vehicles traveling at speeds above 40 kph. Therefore, it represents one of the main contributors to noise environmental pollution in residential areas nearby highways. TPIN has been subject of exhaustive studies since the 1970s. Still, almost 50 years later, there is still not an accurate way to model it. This is a consequence of a large number of noise generation mechanisms involved in this phenomenon, and their high complexity nature. It is acknowledged that the main noise mechanisms involve tire vibration, and air pumping within the tire tread and pavement surface. Moreover, TPIN represents the only vehicle noise source strongly affected by an external factor such as pavement roughness. For the last decade, new machine learning algorithms to model TPIN have been implemented. However, their development relay on experimental data, and do not provide strong physical insight into the problem. This research studied the correct configuration of such tools. More specifically, Artificial Neural Network (ANN) configurations were studied. Their implementation was based on the problem requirements (acoustic sound pressure prediction). Moreover, a customized neuron configuration showed improvements on the ANN TPIN prediction capabilities. During the second stage of this thesis, tire noise test was undertaken for different tires at different pavements surfaces on the Virginia Tech SMART road. The experimental data was used to develop an approach to account for the pavement profile when predicting TPIN. Finally, the new ANN configuration, along with the approach to account for pavement roughness were complemented using previous work to obtain what is the first reasonable accurate and complete tool to predict tire noise. This tool uses as inputs: 1) tire parameters, 2) pavement parameters, and 3) vehicle speed. Tire noise narrowband spectra for a frequency range of 400-1600 Hz is obtained as a result. / Master of Science / Tire-Pavement Interaction Noise (TPIN) becomes the main noise source contributor for passenger vehicles traveling at speeds above 40 kph. Therefore, it represents one of the main contributors to noise environmental pollution in residential areas nearby highways. TPIN has been subject of exhaustive studies since the 1970s. Still, almost 50 years later, there is still not an accurate way to model it. This is a consequence of a large number of noise generation mechanisms involved in this phenomenon, and their high complexity nature. It is acknowledged that the main noise mechanisms involve tire vibration, and air pumping within the tire tread and pavement surface. Moreover, TPIN represents the only vehicle noise source strongly affected by an external factor such as pavement roughness. For the last decade, machine learning algorithms, based on the human brain structure, have been implemented to model TPIN. However, their development relay on experimental data, and do not provide strong physical insight into the problem. This research focused on the study of the correct configuration of such machine learning algorithms applied to the very specific task of TPIN prediction. Moreover, a customized configuration showed improvements on the TPIN prediction capabilities of these algorithms. During the second stage of this thesis, tire noise test was undertaken for different tires at different pavements surfaces on the Virginia Tech SMART road. The experimental data was used to develop an approach to account for the pavement roughness when predicting TPIN. Finally, the new machine learning algorithm configuration, along with the approach to account for pavement roughness were complemented using previous work to obtain what is the first reasonable accurate and complete computational tool to predict tire noise. This tool uses as inputs: 1) tire parameters, 2) pavement parameters, and 3) vehicle speed.

This work explores physics-based and data-driven methods for material property prediction for metallic microstructures while indicating the context and benefit for microstructure- sensitive design. From this, the use of shape moment invariants is offered as solution to quantifying microstructure topology numerically using images. This offers a substantial benefit for computational time since image data is converted to numeric values. The goal of quantifying the image data is to help index grains based on their crystallographic orientation. Additionally, individual grains are isolated in order to investigate the effect of their shapes. After the microstructures are quantified, two methods for identifying the grain boundaries are proposed to make a more comprehensive approach to material property prediction. The grain boundaries as well as the grains of the quantified image are used to train artificial neural networks capable of predicting the material properties of the material. This prediction technique can be used as a tool for a microstructure-sensitive approach to design subtractively manufactured and Laser Engineered Net Shaping (LENS)-produced metallic materials. / Master of Science / Material properties are dependent on the underlying microstructural features. This work pro- poses numerical methods to quantify topology and grain boundaries of metallic microstruc- tures by developing physics-based and data-driven techniques for subtractively manufactured and Laser Engineered Net Shaping (LENS)-produced materials.

Machine learning is a powerful method to handle the self-driving problem. Researchers use machine learning to construct a neural network and train it to drive the car. A self-driving car is a safety-critical system. However, the neural network is not necessarily reliable. The output of a neural network can be easily influenced by many factors, such as the quality of training data and the runtime environment. Also, it takes time for the neural network to generate the output. That is, the self-driving car may not respond in time. Such weaknesses will increase the risk of accidents. In this thesis, considering the safety of self-driving cars, we apply a delay-aware shielding mechanism to the neural network to protect the self-driving car. Our approach is an improvement based on previous research on runtime safety enforcement for general cyber-physical systems that did not consider the delay to generate the output. Our approach contains two steps. The first is to use formal language to specify the safety properties of the system. The second step is to synthesize the specifications into a delay-aware enforcer called the shield, which enforces the violated output to satisfy the specifications during the whole delay. We use a lane keeping system as a small but representative case study to evaluate our approach. We utilize an end-to-end neural network as a typical implementation of such a lane keeping system. Our shield supervises those outputs of the neural network and verifies the safety properties during the whole delay period with a prediction. The shield can correct it if a violation exists. We use a 1/16 scale truck and construct a curvy lane to test our approach. We conduct the experiments both on a simulator and a real road to evaluate the performance of our proposed safety mechanism. The result shows the effectiveness of our approach. We improve the safety of a self-driving car and we will consider more comprehensive driving scenarios and safety features in the future. / Master of Science / Self-driving cars is a hot topic nowadays. Machine learning is a popular method to achieve self-driving cars. Machine learning constructs a neural network, which imitates a human driver's behavior to drive the car. However, a neural network is not necessarily reliable. Many things can mislead the neural network into making wrong decisions, such as insufficient training data or a complex driving environment. Thus, we need to guarantee the safety of self-driving cars. We are inspired to use formal language to specify the safety properties of the self-driving system. A system should always follow those specifications. Then the specifications are synthesized into an enforcer called the shield. When the system's output violates the specifications, the shield will modify the output to satisfy the specifications. Nevertheless, there is a problem with state-of-the-art research on specifications. When the specifications are synthesized into a shield, it does not consider the delay to compute the output. As a result, the specifications may not be always satisfied during the period of the delay. To solve such a problem, we propose a delay-aware shielding mechanism to continually protect the self-driving system. We use a lane keeping system as a small self-driving case study. We evaluate the effectiveness of our approach both on the simulation platform and the hardware platform. The experiments show that the safety of our self-driving car is enhanced. We intend to study more comprehensive driving scenarios and safety features in the future.