Title: Neural Network-Based Classification of Steel Annealing Data Using the UCI Annealing Dataset
Authors: Alaa Abusultan
Samy S. Abu-Naser
Volume: 10
Issue: 7
Pages: 49-65
Publication Date: 2026/07/28
Abstract:
Artificial Intelligence (AI) and Machine Learning (ML) technologies have become essential tools for improving efficiency, quality control, and decision-making in modern industrial manufacturing. Among various machine learning techniques, Artificial Neural Networks (ANNs) have demonstrated remarkable capability in solving complex nonlinear classification problems due to their ability to learn intricate relationships from large and heterogeneous datasets. In the steel manufacturing industry, accurate classification of annealed steel products is crucial for ensuring product quality, minimizing production defects, and optimizing manufacturing processes. However, industrial datasets often contain missing values, categorical attributes, and inconsistent feature scales, which require comprehensive preprocessing before effective neural network training can be achieved. This study presents a complete framework for classifying steel products using a feedforward artificial neural network and the UCI Annealing dataset, a widely recognized benchmark dataset for industrial machine learning research. The proposed methodology begins with a comprehensive data preprocessing stage that includes handling missing values, encoding categorical variables into numerical representations, applying Min-Max normalization to numerical attributes, and exporting the processed dataset in CSV format for compatibility with the JustNN neural network software. The processed dataset was divided into training and validation subsets using an 80:20 ratio, and a feedforward neural network with a single hidden layer was trained for 100 epochs using a learning rate of 0.01. Experimental evaluation demonstrated that the proposed model achieved a training accuracy of 95.1% and a validation accuracy of 92.7%, with training and validation loss values of 0.108 and 0.142, respectively. The confusion matrix and learning curves further indicated stable convergence, effective learning, and satisfactory generalization performance with limited evidence of overfitting. These findings demonstrate that careful data preprocessing significantly enhances neural network performance and enables reliable classification of industrial steel products. The proposed framework provides a practical and effective solution for industrial quality classification and may serve as a foundation for future research involving deep learning architectures, automated hyperparameter optimization, and comparative studies with other machine learning algorithms.