Intelligent Model for the Reliability of the Non-Intrusive Continuous Sensors Used for the Detection of Fouling-Layer in Heat Exchanger System

Faults in this sensor must be detected on time to ensure the functionality of the entire system's architecture and to maintain system balance, which will keep false positive rates low during the system's operational period. False positives reduce diagnostic confidence and necessitate unnecessary and costly mitigation actions, lowering system productivity. It is on this basis that this study proposes a clustering model algorithm (K-mean clustering) to investigate and manage the reliability and performance of the sensors. The results from the implementation of the K-mean clustering method show that the running of the algorithm fits the model correctly, both for the training of the dataset and for the prediction of the cluster in each of the datasets considered. A reasonable grouping was found for the two and three clusters considered, which are represented by the colors (blue, orange, and green). These colors indicate the fault state, non-fault state, normal state, and abnormal state of the non-intrusive continuous sensor. The simulated results show the fault state in the blue region and the non-fault state in the orange region for the two clusters, while the normal state is in the blue region and the abnormal state is in the orange and green regions for the three clusters considered.