International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 02 | Feb 2024
p-ISSN: 2395-0072
www.irjet.net
Electronics Component Classification Using Machine Learning Jishnu Nath Paul1, Mallika Roy2, Josita Sengupta3, Karan Kumar4, Swagata Bhattacharya5 Department of Electronics & Communication Engineering, Guru Nanak Institute of Technology, Kolkata, India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - This paper compares four convolutional neural
in various electronic systems. This innovative approach is poised to revolutionize electronic design, troubleshooting, and maintenance, offering engineers and technicians a powerful tool for rapid and precise component identification in an increasingly complex technological landscape.
network (CNN) models for classifying electrical components such as capacitors, diodes, resistors, ICs, inductors, and transformers. The CNN models are AlexNet, GoogleNet, ResNet50, and a proposed CNN model with a customized architecture. The paper describes the data collection, preprocessing, and augmentation steps, as well as the architecture and parameters of each CNN model. The paper evaluates the performance and accuracy of the CNN models using various metrics such as loss, accuracy, precision, recall, F1-score, and confusion matrix. The paper also analyses the factors affecting the results and provides insights and implications for the practical application of the CNN models. The paper concludes that ResNet50 is the best CNN model for the electrical component classification task, as it has the highest accuracy, precision, recall, and F1 score among the four models. It also has the lowest loss and the most accurate confusion matrix, indicating that it can learn and generalize the features of the electrical components very well.
2. MOTIVATION The motivation behind this project is rooted in the critical need to address the inefficiencies and challenges inherent in conventional electronic component identification processes. Traditional methods, often reliant on manual labor, are timeconsuming, error-prone, and struggle to keep pace with the expanding diversity of electronic components. This project is driven by a vision to revolutionize and modernize the identification process by harnessing the power of machine learning, specifically Convolutional Neural Networks (CNN). The overarching goal is to empower professionals and enthusiasts in the field by providing a robust and automated solution that significantly enhances accuracy and expedites the identification of electronic components. By combining machine learning with CNN, the project aspires to bridge the gap between the complexity of modern electronic systems and the limitations of traditional identification methods. This motivation is underpinned by a commitment to fostering efficiency, reducing errors, and advancing the capabilities of electronic systems in alignment with the ever-evolving landscape of technology. Ultimately, this project seeks to propel the electronics industry forward by offering a sophisticated yet accessible tool for streamlined component identification and classification.
Key Words: Convolution neural network (CNN), AlexNet, GoogleNet, ResNet50, F1-score, Confusion matrix, Accuracy, Precision.
1. INTRODUCTION The introduction of electronic component classifiers represents a significant leap forward in the field of electronics, particularly in the context of streamlining the identification process. In the face of a burgeoning array of electronic components, integrating machine learning, specifically Convolutional Neural Networks (CNN). Machine learning algorithms, powered by neural networks, are at the core of these classifiers [1]. This sophisticated approach allows for recognizing and categorizing electronic components based on intricate patterns, physical characteristics, and unique electrical properties. CNN, a specialized form of neural network tailored for image recognition [7], brings heightened accuracy to the identification process, making it particularly well-suited for categorizing electronic components with diverse visual features.
3. BACKGROUND OF THE PROJECT
This project's motivation is grounded in the imperative to overcome the limitations of traditional, time-consuming, and error-prone manual identification methods. Fusing machine learning, CNN, and electronic component classifiers not only enhances the accuracy and efficiency of the identification process but also paves the way for practical implementations
Machine learning, particularly Convolutional Neural Networks (CNN), emerges as a transformative solution in this context. The project's background is informed by the growing recognition of the potential of machine learning algorithms to decipher intricate patterns and features crucial for electronic component identification. CNN, which
© 2024, IRJET
|
Impact Factor value: 8.226
The background of this project is grounded in the persistent challenges faced by the electronics industry in the domain of electronic component identification. Traditional methods relying on manual identification are becoming increasingly inadequate due to the expanding complexity and diversity of electronic components. This necessitates a paradigm shift toward more efficient, accurate, and automated identification processes.
|
ISO 9001:2008 Certified Journal
|
Page 307