ASTU ETD

Adama Science and Technology University Library Electronic Theses & Dissertations (ASTU- ETD) is a new digital institutional repository system that collects, preserve and distribute the scholarly output of the university, mainly postgraduate Electronic Thesis and Dissertation (ETD), articles, proceedings. The system is dedicated to help users to find all the information they might require in order to format and successfully submit their graduate thesis, dissertation and publications electronically. The user friendly web interface enables to maximize & optimize the resource sharing among ASTU different Colleges.

Colleges and Departments in ASTU

Select a Colleges or Departments to browse its collections.

Now showing 1 - 6 of 6

Recent Submissions

Developing Robust Text Independent Speaker Recognition Using Deep Learning ModelsDissertation
(ASTU, 2024) Wondimu, Lambamo; Ramasamy Srinivasagan (Professor) ; Worku Jifara (Associate Professor)
Speaker recognition is the process of classifying/identifying a person from others based on speech characteristics. It has crucial applications in security, surveillance, forensics and financial transactions. The performance of the speaker recognition systems was good in the clean speech and without mismatch. However, the performance of the speaker recognition systems gets degraded under noisy and mismatched conditions. Several studies have been conducted in speaker recognition using machine learning methods to enhance performance in noisy environments. Recently, deep learning models outperformed machine learning methods in speaker recognition. Moreover, hybrid models of convolutional neural networks (CNN) and enhanced variants of recurrent neural networks (RNN) have shown better performance in image classification and natural language processing. However, only limited attempts have been conducted using hybrid CNN and RNN variants to enhance speaker recognition performance under noisy conditions. The features which have good performance in speaker recognition using machine learning methods were not as effective as spectrogram and cochleogram in deep learning-based speaker recognition. However, the noise robustness of the cochleogram and spectrogram was not analyzed in speaker recognition using deep learning models to employ the more robust feature in noisy conditions. In this study, a text-independent speaker recognition using deep learning models have been developed for noisy conditions. First, the noise robustness analysis of cochleogram and spectrogram in speaker recognition using deep learning were conducted to select the more robust feature. Then, the speaker recognition model using hybrid CNN and enhanced RNN variants have been developed to enhance the performance under noisy conditions. The enhanced RNN variants employed in this study include long short-term memory (LSTM), bidirectional LSTM (BiLSTM), gated recurrent unit (GRU) and bidirectional GRU (BiGRU). Cochleogram have shown better noise robustness t each signal-to-noise ratio (SNR) level and are used as an input in each of the speaker recognition models developed in this study. The experiments have been conducted on the VoxCeleb1 audio dataset with real-world and white Gaussian noises at the SNR level of -5dB to 20dB and without additive noises. The speaker recognition using hybrid CNN and BiGRU on the cochleogram input was proposed for noisy conditions in this study because of its higher performance. The proposed model has achieved speaker identification accuracy of 93.15% to 98.60% on the dataset with real-world noise at SNR of -5dB to 20dB, respectively and 98.85% on the dataset xvwithout additive noise. The equal error rate (EER) of the proposed model on the dataset with real-world noise at SNR of -5dB to 20dB ranges from 10.55% to 0.47%, respectively and 0.37% on the dataset without additive noise. The comparison with the existing works also confirmed that the proposed model has higher performance than existing works.
A Hybrid Genetic Algorithm Model for Early Detection and Classification of Breast Tumours.Dissertation
(ASTU, 2025-05) Abebe, Alemu; ANTENEH GIRMA, (PhD) Prof., Computer Science/CyberSecurity, University of the District of Columbia. MESFIN ABEBE, (PhD) Associate Prof., Computer Science and Engineering, Adama Science and Technology University.
The severity and fatality prevalence of breast cancer remain a global issue. The existing diagnosis and treatment are not efficient enough to successfully detect breast cancer tumours at an early stage. According to recent studies, breast cancer causes 25.84% of all cancer-related fatalities. New breast cancer cases account for 29.46% and 31.85% of all new cancer cases in Africa and Ethiopia, respectively. In Ethiopia, 72.56% of breast cancer diagnoses are in advanced stages (95% CI: 68.46-76.65%), emphasizing the critical need for early detection. Furthermore, a lack of radiologists causes delays in the annual reading and classification of Breast Imaging Reporting and Data Systems (BI-RADS). To address the problem areas of early detection, false positives and false negatives, research efforts continue to present different solutions using different advanced techniques. The techniques used are artificial intelligence (AI), machine learning (ML), and Genetic Algorithms (GA). Artificial intelligence and Machine learning are crucial in improving breast cancer diagnosis to reduce late discovery, which significantly affects survival chances. In this study, we collected 4092 mammographic image data from the diagnosis center with the radiologist’s recommendation for the diagnosis and future patient steps. In addition to this, we use the Wisconsin breast cancer public image data for training, validation, and testing of the model. We work on a hybrid model using genetic algorithms and machine learning for early detection and classification of the stage and type of breast cancer tumours to reduce the outliers for false-positive and false negative results. The implemented models are hybridizing Genetic Algorithms with K Nearest Neighbour (K-NN) and Support Vector Machine (SVM), (GA-KNN-SVM), a modified GA with Pontraygin Minimum Principles (PMP), and a hybrid GA model with K-Means++. The GAKNN-SVM model achieves superior performance in breast tumour detection, utilizing the Wisconsin Breast Cancer Diagnostic Dataset. The results show significant improvements, with accuracy, sensitivity, and specificity rates reaching 98.25%, 98.15%, and 98.41%, respectively, of the test sample data. At last, we have outperformed results in detecting breast tumours and classified mammography image data by BI-RADS scoring.
Interpretable Deep Learning Approaches For Identification And Classification Of Ethiopian Indigenous Medicinal Plant SpeciesDissertation
(ASTU, 2024-06) Mulugeta, Adibaru; Prof. DP. Sharma Dr. Mesfin Abebe
Ethiopia, known for its rich biodiversity, holds significant therapeutic potential in its diverse array of medicinal plants. Traditional medicines serve as cost-effective and culturally accepted healthcare solutions, used by the population in regions with limited healthcare infrastructure. However, identifying and classifying these Ethiopian indigenous medicinal plants species is a complex and time-intensive task requiring specialized scientific expertise. The main objectives of this research work is to identify andclassify Ethiopian indigenous medicinal plants using deep learning and interpretability. The study started using a systematic literature review aimed at investigating deep learning approaches to identifying and classifying medicinal plants. Subsequently, various deep learning approaches were employed to develop an efficient model through transfer learning and ensemble learning for identifying and classifying medicinal plants species. To tackle the interpretability issues of deep learning, interpretable deep learning models were designed using a multiple teacher-student approach with knowledge distillation concepts. It was done to present an integrated framework for identifying and classifying indigenous medicinal plants species using interpretable deep learning approaches. In the experimental phase, a dataset containing 12,438 labeled leaf images was prepared. Employing efficient pretrained models such as VGG19, VGG16, Xception, and InceptionNetV3, was adopted to enhance the model’s performance. In addition an ensemble EfficientNetB0, EfficientNetB2, and EfficientNetB4 are applied for the identification of parts and uses of Ethiopian indigenous medicinal plants. In the interpretable deep learning approach, a novel, distilled student model was designed using a collaborative teacher-student framework. The systematic review revealed disparities in global research due to resource and dataset variations, with most researchers uses private datasets and employing leaf shapes, transfer learning, and pre-trained models. The study effectively addressed the stated challenges and achieving a commendable accuracy of 95% through fine-tuning. The distilled student model attained an exceptional accuracy of 99.83%, facilitated by knowledge transfer metrics like cosine similarity and MSE. Integrating interpretability techniques such as LIME enhances model transparency and reliability, bridging traditional and modern medicine realms. Addressing the lack of globally accessible datasets for medicinal plants is essential to mitigating disparities in the field.
A Study on the Modelling and Controlling of a Smart Grid for Railway Power Supply A Case Study: Ethio-Djibouti Railway LineDissertation
(ASTU, 2024-10) Mebratu, Delelegn; Dr. CS.Reddy (Associate Professor) Dr. Endalew Ayenew (Assistant Professor)
Power quality analysis is a crucial aspect of managing railway power supply systems. These complex networks must maintain consistent, reliable electricity to power the various components essential for efficient and safe train operations. Analyzing power quality involves closely monitoring a range of electrical parameters, such as voltage, current, frequency, and waveform distortions, to identify any anomalies or deviations from optimal performance. The Ethio Djibouti railway line, a critical transportation link connecting the landlocked nation of Ethiopia to the port city of Djibouti, has been the subject of a comprehensive power quality analysis. This in-depth investigation sought to thoroughly examine the electrical systems and infrastructure supporting this vital rail network, ensuring reliable and efficient operation. Power quality is crucial for any railway, as consistent and stable electricity is required to power the locomotives, signaling systems, and other electrified components. The analysis delved into factors such as voltage regulation, harmonic distortion, power factor, and load balancing across the railway's distribution network. By closely monitoring these parameters, the study was able to identify any inconsistencies, anomalies, or areas for improvement, providing valuable insights to the railway operators. This research delves into investigating power quality phenomena on the Ethiopia-Djibouti railway line, specifically focusing on the Adegala and Aysha 230kV traction substations. Through the utilization of a power quality analyzer, measurements were taken at both 230kV and 25 kV to analyze the harmonic currents, power quantity, and overall distortion of voltage and current. The results of the measurements and simulations point towards exceeding IEEE standard 519-1992 limitations in terms of current and voltage harmonics. Significant voltage imbalances were detected within the train supply network's connecting spot, with the imbalance on the 230kV side surpassing 2%, failing to meet IEEE standard 1159-2009. Considering this, the research focuses on exploring the smart grid integration of wind energy with the railway's power supply. This study examines and maximizes the power generation from wind energy sources using artificial intelligence technique radial basis function network based maximum power point tracking controller to extract maximum power. The proposed MPPT controller design is applied to a 300kW wind energy structure, utilizing a conventional boost converter to maximize power output. The simulation is conducted in MATLAB and integrated with the railway power supply system to evaluate performance. The mean square error (MSE) obtained from the simulation results helps to validate the effectiveness of the control algorithm. Testing the MPPT approach under various wind speeds using MATLAB/Simulink shows promising results, with the system achieving an average maximum power of 289.3 kW at a wind speed of 20 m/sec. The low MSE value of 0.012 indicates the suitability of this MPPT controller for practical applications. The research delves into the intricate relationship between wind energy integration and railway power supply, exploring its viability in both grid-connected and island modes. By harnessing the power of the wind, this innovative approach aims to supplement the energy needs of railway infrastructure, reducing reliance on traditional fossil fuel-based sources and promoting sustainable transportation. In the grid-connected mode, the study analyzes how wind turbines can be seamlessly integrated into the existing power grid, allowing for the bidirectional flow of electricity and ensuring a reliable and uninterrupted supply of power to railway operations. All traction substations connected to the power grid in grid-connected mode are experiencing overload conditions. The recorded bus voltage fluctuates between 239kV and 249.9kV, failing to comply with the IEEE standard, which stipulates an acceptable voltage range of +5% and -5% of the nominal voltage and all traction substations connected to the power grid supply side operate underloaded when in island mode. The bus voltage at the traction load side ranges from 20.71 kV to 26.3 kV, adhering to European standards EN50163: 2004. According to these standards, the maximum non-permanent voltage permitted for short durations is 29 kV, while the minimum permanent voltage must not fall below 19.0 kV. This compliance ensures that the voltage levels remain within safe and efficient operational limits, thereby maintaining the reliability and stability of the traction power system. This research not only addresses the issue of unreliable electricity supply, but it also contributes to the development of railway power systems by introducing renewable energy sources for railway electrification. By incorporating smart grid technology, efficient energy management and utilization can be achieved, ensuring a more reliable and sustainable power supply for the railway line.
Design, Modeling and Performance Analysis of Smart Transformer for Distribution SystemDissertation
(ASTU, 2023-11) Yalisho, Girma; Dr. -Ing Getachew Biru Worku (Associate Professor) Dr. Chandra Sekhar Reddy (Associate Professor)
The changing architecture of electric power grid from centralized to decentralized form by integration of distributed energy resources (DERs) to ensure a reliable, efficient and environmentally friendly power supply to customers is becoming an opportunity and challenge for power companies. The decentralized electric power grid, poses many challenges to a hundreds of years old conventional transformers as it is a unidirectional power system infrastructure which doesn‟t compensate for harmonics and reactive power. Moreover, the conventional transformers in distribution system are bulky in size and volume; use a mechanical tap changer, which toggles between different tap positions many times per day, reducing life time of transformer by wear and tear. It doesn‟t allow integration of micro grids without investing on additional power converter equipment. The aforementioned short comings of conventional transformers are overcome by emerging technology known as a smart transformer. In this dissertation work, design, modeling and analysis of a three stage multiport smart transformer (ST) based on modular multilevel converter (MMC) is presented in Simulink/PLECS simulation platform using different controllers, modulation techniques and converter configurations. In addition, integration of smart transformer (ST) in to the existing feeder model of ArbaMinch distribution system containing 72 distribution transformers has been done and its impact on voltage regulation, loss minimization and power quality (PQ) improvement has been assessed. Observations of simulation result showed that Fuzzy Inference System (FIS) and Adaptive-Neuro-Fuzzy Inference System (ANFIS) controllers yield better output than Proportional-Integral (PI) controller for front stage of ST. Simulation result of Input-Series-Output-Parallel Dual Active Bridge (DAB-ISOP) configuration for DC-DC stage proved voltage sharing at input side and current sharing at output side reducing device voltage and current stress and increasing availability. The MMC topology for back-end converter yields a reduced voltage total harmonic reduction (THD)value (3.8%), higher efficiency (96.7%) and reduced current stress (61%) than its cascaded H bridge (CHB) counterpart having voltage THD of 13.7% and efficiency of 83% for the same sub-module (SM) used. It has been observed from the simulation result that the maximum voltage drop in the feeder model without ST integration is 781V with voltage regulation of 9.9%. However, the maximum voltage drop after integration of an ST model is reduced to 444V with voltage regulation of 5.4% (improvement by 4.5%). The line loss without ST integration is 1096.7kW, which is reduced to 823.2kW when ST is integrated in to the feeder model (power loss reduction by 273.5 kW).