With the emergence and development of Maritime Autonomous Surface Ships (MASS), reliable operation over extended periods for safety is crucial. Regular maintenance is vital for maintaining reliability, and predictive maintenance is particularly efficient in minimizing maintenance time and reducing resource waste. This paper proposes a condition prediction framework that considers component degradation to aid in developing effective predictive maintenance strategies for MASS. Five machine learning models were utilized to analyze a collected dataset. This framework evaluated the models using mean squared error (MSE), mean absolute error (MAE), and determination coefficient (R2) to identify the most effective one for condition prediction. To enhance the accuracy, the framework also considers component degradation. The framework’s applicability was demonstrated using a gas turbine system as a case study. The results show that the multilayer perceptron is the most fitting one for the condition prediction of the gas turbine system. The framework can support predictive maintenance in decision-making to improve MASS safety.