In industrial settings, unexpected equipment failures can have severe consequences, impacting both economic performance and human safety. Detecting or predicting potential failures at an early stage is crucial, as it lays the foundation for predictive maintenance strategies that optimize machine utilization and prevent costly downtimes. However, a major challenge in industrial applications is the lack of labeled data, which limits the effectiveness of supervised learning approach. Therefore, researchers have recently focused on unsupervised learning approaches to address anomaly detection. This paper investigates a case study of a pump operational dataset. Four anomaly detection algorithms (HBOS, KNN, LOF, and Isolation Forest) are selected and compared for their effectiveness in early failure detection. The results show that HBOS and Isolation Forest detect anomalies more effectively. Beside, HBOS is somewhat less effective in identifying early signs of impending failures. The goal is to apply machine learning algorithms to identify abnormal patterns in the data before actual failures occur, thereby enhancing predictive maintenance capabilities. This research contributes to improving industrial reliability by enabling proactive measures against equipment malfunctions.