This paper offers a method for identifying anomalies in rotary machinery based on the employment of multiple machine learning systems. Both normal and anomaly behaviors data that were preprocessed with feature scaling and rate of contamination, for which overhang, underhung, and speed of machine are taken as parameters of input. Isolation Forest (IF) and Local Outlier Factor (LOF) algorithms were chosen for this purpose and executed independently, after which they were combined to utilize the strengths of both models. Grid search was applied for hyperparameter tuning to provide the optimal performance. The models were developed with normal operating data and validated with hidden anomalous data, using recall, accuracy, F1-score, and precision as metrics for appraise. ROC and precision-recall curves were used to cross-validate the results. The results concluded that the addition of speed along with overhang and underhung enhanced anomaly detection, and the hybrid model performed better than single standalone methods, thus being a good solution for predictive maintenance in rotating equipment monitoring.