In the evolving landscape of Artificial Intelligence (AI) and Machine Learning (ML), society is increasingly leveraging these technologies to reduce effort and cost. Advanced ML models now offer near-certain predictions, making them indispensable in applications like condition monitoring, where their capabilities surpass human observation and calibration. This study focuses on a Sewage Pump within a Sewage Water Treatment Plant (SWTP), exploring its performance and potential failures through both practical testing and simulation. The pump having following key attributes, such as temperature, vibrations, current, voltage, and other general characteristics, are monitored and analysed. Failure scenarios, identified through literature reviews, surveys, and brainstorming sessions, and categorized into normal and broken conditions. A Simulink model simulates the pump operation, generating data for training and testing an ML model. The ML model, built using XGBoost, feature importance analysis, and hyperparameter tuning, is designed to predict failures before they occur, with severity levels indicated. This predictive model is then integrated with the Simulink model for real-time condition monitoring, signalling normal and abnormal pump conditions through a visual indicator. While the current model effectively predicts failures, it is acknowledged that further improvements are needed, particularly in reducing false positives. Future enhancements could involve replacing the Simulink model with a more sophisticated sensor fusion model, incorporating additional parameters such as fluid dynamics, environmental conditions, and specific location characteristics. Ultimately, this integration of ML and simulation offers a promising approach to proactive maintenance and failure prevention in SWTP operations.