Marine engines are the core propulsion units in maritime transportation, where operational reliability directly impacts vessel safety and efficiency. With the advent of sensor-rich systems and advanced monitoring technologies, large-scale performance data can currently be collected and analyzed in real time. However, traditional diagnostic methods often struggle to handle high-dimensional datasets, leading to delayed or inaccurate fault detection. This research fills this gap by suggesting a proposal for a marine engine performance defect detection system based on machine learning (ML) and big data. The model utilizes a diverse dataset collected from onboard engine monitoring systems, incorporating operational variables like exhaust gas and cylinder pressure temperature, vibration levels, and lubrication oil properties. Data preprocessing is conducted using Robust Scaler Normalization to mitigate the influence of extreme values, followed by noise reduction through a Wavelet Denoising technique. Principal Component Analysis (PCA) is applied for feature extraction to reduce dimensionality while preserving key performance indicators. The proposed framework integrates these refined features into an Adaptive Sheep Flock Optimized Intelligent Random Forest Tree (ASF-IRFT) algorithm. The adaptive optimization mechanism tunes decision thresholds and tree parameters dynamically, enhancing fault classification accuracy and generalization. The system architecture enables continuous data acquisition, preprocessing, and classification to detect performance degradation and emerging faults at an early stage. Experimental evaluation demonstrates that the Python-implemented ASF-IRFT outperforms conventional models, achieving 98.3% accuracy with a training time of 20.23s. The proposed system offers a scalable, data-driven solution for real-time marine engine health monitoring, ultimately supporting safer and more efficient maritime operations.