The vibration of mobile cleaning robots can indicate performance degradation or operational safety issues. Therefore, it is crucial to identify the cause of vibrations at an early stage in order to prevent functional loss and hazardous working conditions. To overcome these drawbacks, a novel Maintenance using SCB-LSTM (MASS) Robot system has been proposed for enhanced maintenance planning and real-time fault detection in cleaning robots. Initially, vibration data is collected during the robot's operation. This data is processed through a Stacked Convolutional neural network Bi-directional Long Short Term Memory (SCB-LSTM) model to identify specific sources of vibration. The information is then sent wirelessly to a remote monitoring application, allowing users to track the robot's condition in real-time and diagnose issues efficiently. The suggested MASS technique has been assessed using a MATLAB simulator. The efficacy of the suggested MASS approach has been evaluated by utilizing parameters such as F1-score, recall, accuracy and precision respectively. The proposed MASS method achieves better accuracy of 79.8%, 85.4%, and 88.1% than GPM [20], DBF [23], and KPM [25] methods