This paper presents a new method to monitor fire pump failures using a long short-term memory (LSTM) network, which is suitable for analyzing time series data. Fire pumps are critical to safety, and high reliability and early detection of potential problems is necessary. By leveraging the LSTM network, we can accurately analyze historical sequential data to classify failure conditions for maintenance service. The LSTM-based model processes real-time sensor data from the firefighting pumps, identifying patterns and anomalies that indicate impending failure. This method enhances traditional maintenance strategies by providing early warnings and enabling proactive maintenance intervention. Our experimental results show that the LSTM method significantly improves the accuracy of failure predictions compared to conventional methods. Integrating the LSTM network into fault monitoring systems improves fire pumps’ operational reliability and contributes to sustainable maintenance practices by optimizing resource allocation and minimizing downtime.