Predictive maintenance in the manufacturing industry is crucial for minimizing downtime and optimizing machinery performance. This research focuses on using unsupervised learning algorithms to predict machinery failures by analyzing vibration data from a 0.75 kW, 230 V, S1 motor tested under normal and abnormal conditions. The motor's unique modification, allowing for imbalanced load testing, provided a diverse dataset for analysis. We implemented K-Means and Agglomerative Clustering on both raw and pre-processed data, using filters such as the Kalman Filter, Moving Average Filter, and Fast Fourier Transform (FFT) to enhance data quality. Our findings indicate that Agglomerative Clustering slightly outperforms K-Means on raw data, with accuracies of 68.50% and 67.63%, respectively. The Kalman Filter improved K-Means performance to 70.40%, while the Moving Average Filter yielded the highest accuracies for both algorithms, with Agglomerative Clustering at 71.51% and K-Means at 70.66%. Despite lower accuracies, the FFT method produced high Silhouette scores, indicating strong clustering quality. Overall, the Moving Average Filter combined with Agglomerative Clustering proved the most effective approach, emphasizing the importance of preprocessing in unsupervised learning for machinery failure prediction. These insights can enhance predictive maintenance strategies, improving reliability and efficiency in the manufacturing sector.