Papers by Keyword: ANN

Abstract: Peak load shaving is a crucial strategy for enhancing grid reliability, efficiency, and sustainability by reducing maximum electricity demand. This study investigates the design and optimization of a 100-kW grid-connected Solar Photovoltaic (PV) system for peak load shaving and ancillary support at the Bahir Dar Distribution Substation (15 kV side). The system addresses the projected overloading of the substation within 22 months from 2023, as forecasted using an Artificial Neural Network (ANN) based on historical domestic customer data. Utilizing Bahir Dar’s abundant solar resources, the PV system replaces diesel gensets at the Ethiopian Electric Utility (EEU) datacenter, contributing to the national electrification goal by 2030. To enhance power quality and grid stability, the proposed system integrates an Adaptive Neuro-Fuzzy Inference System (ANFIS)-based boost converter for Maximum Power Point Tracking (MPPT), increasing the DC-link voltage from 437.6V to 730.04V. A five-level Modular Multilevel Converter (MMC) with Voltage Oriented Control (VOC) is implemented. Achieving a significant reduction in Total Harmonic Distortion (THD) from 16.27% to 1.12%, ensuring compliance with international standards. Additionally, the economic feasibility analysis indicates that the PV system, consisting of 41 panels in series and 8 in parallel, requires a total installation cost of $39,013 and generates approximately 170,209.18 kWh annually. Compared to the existing diesel-based power system, which incurs an annual operating cost of 571,663.35 ETB, the PV system offers substantial cost savings with an estimated payback period of 27 months. Despite these advantages, challenges such as weather variability, transient response analysis, and system scalability remain. Future work will focus on real-world validation through hardware-in-the-loop (HIL) testing, grid disturbance simulations including Low Voltage Ride Through (LVRT) and Low Frequency Ride Through (LFRT), and field trials to assess large-scale deployment feasibility. The findings highlight the potential of grid-connected PV systems to enhance energy reliability, reduce fossil fuel dependence, and support Ethiopia’s transition to a cleaner and more sustainable power infrastructure.

Abstract: Current devices using resin 3D printing technology are divided into three main lines, including: industrial, desktop and opensource. The use of desktop devices for producing end products is garnering significant attention thanks to several advantages. These include the capacity to utilize a variety of popular materials, affordable costs for both materials and components, and stable operation. However, some product quality categories such as surface properties, dimension and geometry accuracy, etc., of these devices depend greatly on the selection and adjustment of technological parameters. This article presents the process of building an Artificial Neural Network (ANN) model to control the dimension of test samples manufactured using Digital Light Processing (DLP) technology in three directions X, Y, Z. The testing process is designed according to the Face-centered central composite design (FCCCD) method with five input parameters. Test samples are chosen based on ASTM standards, produced using a desktop DLP resin 3D printer, and assessed with a Mitutoyo PJ-H3000F profile measuring instrument. Training results show that the ANN build model fits with the database and produces good predictions with the Mean Average Error (MAE) compared to actual data in the X direction being 0.038342, in the Y direction being 0.010258012, and in the Z direction being 0.012501036. The success of this research helps engineers and technicians reduce the time spent operating and calibrating related devices.

Abstract: The ability of hybrid renewable energy systems (HRES) to combine the advantages of several renewable energy sources has attracted a lot of attention. The intermittent nature of renewable energy sources, such wind and solar power, can make it difficult to keep the grid's power quality constant. Advanced intelligent control strategies are presented in this work with the goal of improving power quality in HRES. In order to reduce power quality difficulties, the research suggests a multidimensional approach that combines the capabilities of advanced control algorithms, intelligent decision-making, and predictive analytics. The main goal is to solve typical issues in HRES, such as harmonic distortions, voltage fluctuations and frequency variations. Proactive system management is made possible by the control approach, which forecasts renewable energy generation trends using machine learning techniques. In addition, real-time monitoring and control systems are included to enable prompt responses to modifications in the power generating mix. The HRES guarantees smooth integration and interaction by utilizing sophisticated hardware and software components to achieve these control mechanisms. This paper presents the results of an extensive simulation research that shows how well the suggested intelligent control solutions mitigate problems with power quality. The results show that the grid's frequency regulation, harmonic distortions, and voltage stability have all significantly improved.

Abstract: In the conventional turning process, cutting forces and cutting tool deflection plays an important role in tool wear mechanism. The cutting tool is subjected to uneven wear due to heavy cutting forces. This further leads to shear deformation of the tool which ultimately affects the surface texture of the workpiece. In this paper, an experimental setup is designed and fabricated which can hold rigidly the two cutting tools namely, the front and rear tool in opposite direction. Moreover, to check its robustness and feasibility, the numerical modelling of the experimental setup is performed using the experimental data. A total of nine simulation runs were performed according to the machining parameters used in the experimental study to check the accuracy of the Finite element model. Additionally, these experimental values of resultant cutting force were used to obtain the resultant deflection of both the cutting tools through simulation tests. Finally, the values of resultant cutting force and the resultant deflection of both the tools were predicted using Artificial Neural Networks (ANN) and compared to estimate the tool wear. It was found that the experimental results were in good agreement with the ANN predicted results.

Abstract: The surface roughness is a feature that is of tremendous relevance in the assessment of cutting performance, and it plays an essential part in the manufacturing process as well. In this research, an effort was made to construct a model based on artificial neural networks to replicate the hard turning of Monel K 500 in dry conditions. The results of this endeavor are presented. This model is anticipated to accurately estimate the surface roughness for various cutting settings. Networks that use Scaled Conjugate Gradient (SCG) were trained using a set of training data for several cycles. Then they were tested with a collection of input/output data that was specifically reserved for this purpose. For each of the designs that were considered, the root mean square error was calculated. As compared with other models, the RMSE that the SCG Produces better value-. Analysis was done on the ability of the ANN model to predict surface roughness (Ra). It was discovered that the predictions produced by the ANN model had a high degree of congruence with the experiment’s findings.

Abstract: This research aims to investigate the dry sliding wear behavior of Al-Cu-Zr (ACZ) metal matrix composite (MMC) at various Aluminium oxide (AOX) Nano particles compostion. ACZ alloy is widely used in on road and space mobility applications where the focus is on wear resistance. To enhance the resistance towards wear, Al alloy is reinforced with AOX nanoparticles at 3%, 6%, and 9% addition using stir casting process. The wear assessment is conducted at varying Composition (COMP), load (LD), speed (Ns), and sliding distance (SLDN). The wear rate (WRT) and Frictional force (FRFC) are analysed for different process parameters. To optimize the experiments, Taguchi signal-to-noise ratio (STNR) is used. Taguchi analysis show that the optimal conditions for minimum WRT and FRFC are at 6% AOX addition, 12.5 N load, 500 rpm speed, and 35 mm SLDN. Furthermore, an artificial neural network model (ANNM) is developed to forecast the WRT and FRFC. The neural network model is trained using the experimental data and the optimized process parameters. The neural network is a powerful tool that can learn the complex relationship between input and output variables. The model is validated using the experimental data, and the results show that the neural network model can predict the WRT and coefficient of friction with high accuracy. The Taguchi optimization and neural network model can provide a systematic approach to optimizing the process parameters and predicting the WRT and coefficient of friction. This approach can be applied to other materials and processes to improve their performance and reduce costs.

Abstract: In the electroless nickel-boron coating process, surfactant helps to minimize the surface tension between the substrate and the electrolyte in the bath. Despite, its high cost and the formation of micelles from monomeric surfactant molecules at its critical micelle concentration (CMC), it is essential to optimize the concentration while using in the bath. In this study, to solve this problem, mathematical models are developed using regression and artificial neural network (ANN) techniques to relate the concentration of amphoteric surfactant (0-0.162 g/L) as an independent variable and microhardness as a dependent variable. Then, the developed model was used to optimize microhardness at CMC using a genetic algorithm (GA). The goodness of fit of the models was evaluated using the coefficient of determination (R²). The ANN model was found to be the best fit with R² = 0.99. The maximum microhardness of 852 HV was achieved at the CMC of 0.064 g/L, from the GA using the validated model as a fitness function.

Abstract: The Incremental sheet metal forming (ISF) is come into the light due to its unique forming technique. In ISF, the metal sheets are transformed into the final product without using dedicated died. The plastic deformation of metal sheets is conducted through a simple forming tool. Its processing time resembles that the ISF is suitable for the formation of customized products, prototypes, and low volume production. Tool life in manufacturing processes is an important consideration for productivity. The present study is an approach to use image processing techniques to measure the exact location and amount of tool wear in the ISF tool which is made of 440C steel. Presently the complex histogram plot is under study. Therefore to predict the tool wear, feed-forward backpropagation (FFB) algorithm is utilized. It is reported that the maximum predicted tool wear of 0.0663mm is found in the trail run 05 with an error of 0.0104mm. The best-fitted value of the FFB model is observed at the epoch 05 with the value of 5.922e-005. The overall coefficient of performance i.e. R² of FFB modeling is reported as 90.52 % with the mean absolute error (MAE) of 0.0042 which shows a good agreement of the prediction model.

Abstract: In order to obtain flow curves from compression test results of a cold forging material and predict flow curves of the material at intermediate temperature and strain rate values, a model was developed using Python programming language in this study. The model consists of two parts: Flow curve determination and flow curve prediction. The compression test data including Force-Stroke values was processed to determine the flow curves in the first part, and the flow curve data constructed for certain temperature and strain rate values of the material was used as input for the machine learning algorithms to predict flow curve at desired intermediate temperature and strain rate values in the second part. Moreover, Ludwik material model parameters were estimated by using curve fitting methods in order to define the material model into the simulation software. Machine learning algorithms and various regression models in Python libraries were tested to predict the flow curves. The performances of different machine learning and regression models were compared with respect to the mean squared error and coefficient of determination performance measures. Support vector regression, k-Nearest Neighbour (kNN) and artificial neural network models were used to predict flow curves of cold forging materials and kNN regression model was able to found predictions with the lowest error rate. As a result, a model that can process the compression test data to predict flow curves at intermediate temperature or strain rate values was developed.

Abstract: Nanotechnology especially Zero Valent metals is a modern technology for the degradation of extensive ranges of biological wastewater contaminants. Due to their effectiveness, economically and safely properties, this study successfully prepared and characterized nanoZero Valent Iron (nZVI) to be encapsulated into natural alginate biopolymer. The effect of operating parameters was studied at different environmental conditions; pH, dose (g/L), contact time (min), stirring rate (rpm), and BOD concentrations. Adsorption isotherm, kinetic studies, and statistical analysis (Response Surface Methodology (RSM) and Artificial neural networks (ANNs)) were examined to describe the removal behavior. The obtained results indicated that the maximum removal efficiency was 81.2 % for initial BOD concentration 300 mg/L, at pH 7, using wet dose 3g/L, 25min, and stirring rate 200 rpm. Also, adsorption and kinetic data indicated that the adsorption mechanism runs toward the Sips model to approximate the Freundlich model at low concentration and to solve the Freundlich limitation at high concentration with a maximum adsorption capacity of 181mg/g. Kinetic results describe the solid transformation from one phase to another at a constant temperature by approving Avrami model. Finally, RSM results agree with ANNs results that the “Concentration effect” is the most significant variable that controls the removal efficiency.