This study introduces a novel 1D Convolutional Neural Network (1DCNN) architecture designed for high-accuracy classification of glass varieties. The proposed model consists of three convolutional layers, followed by fully connected dense layers. To enhance training stability and generalization, we incorporate batch normalization and dropout techniques. Through meticulous hyperparameter tuning, the 1DCNN achieves an impressive classification accuracy of 88.84%, surpassing existing methods by approximately 10%. Furthermore, the model effectively addresses the challenge of imbalanced datasets, demonstrating robust performance across diverse glass varieties. This innovative approach enhances classification efficiency and consistency, contributing significantly to advancements in glass production and quality control.