Pump Sentinel AI: Hybrid Quantum Autoencoder for Intelligent Industrial Pump Fault Detection and Monitoring

Abstract

In the context of modern manufacturing and process industries, industrial pump systems are an important asset, and when anomalies in the system are not detected, they can lead to significant downtime, loss of efficiency, and unplanned maintenance. In this paper, we propose a hybrid quantum-classical autoencoder architecture for real-time anomaly detection of industrial pumps based on multivariate sensor data. A four-qubit variational quantum circuit with Angle Embedding and Strongly Entangling Layers is proposed to learn compact latent representations of normal operating conditions, while a deep classical encoder-decoder network is used to build the variational quantum circuit. The sensor measurements are preprocessed by means of standardization, and reconstruction error is used as the anomaly score to detect abnormal operating states through the hybrid AE. The entire framework is implemented as an intelligent monitoring platform with Flask for secure user authentication, real-time inference, log prediction results, and batch CSV analysis. Experimental tests on an industrial pump sensor dataset of more than 220,000 operational records show that the proposed model is effective for detecting normal and abnormal conditions. The developed system result is 95.33% accuracy, 58.46% precision, 99.99% recall, 73.78% F1 score, and an excellent ROC-AUC value of 0.9974, which shows that there's an extremely high anomaly detection sensitivity with an outstanding discrimination capability. The results obtained show that the integration of variational quantum learning and deep autoencoder reconstruction is an effective and efficient approach for predictive maintenance and intelligent health monitoring in industrial machinery.