Predictive maintenance has emerged as an effective strategy for improving equipment reliability and reducing operational downtime in industrial systems. Conventional maintenance practices in manufacturing plants are typically reactive, where maintenance actions are performed only after equipment failure or when abnormalities are manually detected by operators. Such approaches often fail to identify early failure mechanisms, resulting in prolonged downtime, increased maintenance costs, and decision-making bias. This study presents an Adaptive Neuro-Fuzzy Inference System (ANFIS)-based predictive maintenance framework for a 40-bar industrial air compressor. Historical and real-time sensor data including temperature, pressure, flow rate, and vibration were obtained from maintenance records and preprocessed for model development. The ANFIS model was trained to capture nonlinear relationships between operational parameters and component wear, enabling the identification of potential failure modes. Model performance was evaluated using Root Mean Square Error (RMSE), while a Gaussian Process Regression (GPR) model was employed to estimate the Remaining Useful Life (RUL) of compressor components. The models were developed using an incremental iterative approach, achieving prediction errors below 0.8 and an overall accuracy of approximately 92%. System evaluation was conducted under both healthy and faulty operating conditions. Under faulty conditions, the model identified wear as the dominant failure mode associated with elevated vibration levels, with a confidence level of 60.9%, and recommended targeted inspection of electric motors as a primary vibration source. Under healthy operating conditions, sensor data showed no significant anomalies, and RUL estimates remained above the predefined warning threshold. The proposed framework demonstrates the potential of data-driven predictive maintenance for improving reliability and maintenance planning in industrial compressed air systems.

Keywords

Remaining Useful Life, neural networks, neuro-fuzzy techniques, predictive maintenance, adaptive neuro-fuzzy inference system