Sensors are placed at various locations in a production plant to monitor the state of its components and accordingly operate its control and protection. For the plant state monitoring to be effective, the sensors themselves must be monitored for detecting anomalies in their functioning and for reconstructing the correct values of the signals measured. In this work, the task of sensor monitoring and signal reconstruction is tackled with an ensemble of Principal Component Analysis (PCA) models. The novelty of the work consists in the investigation of local fusion (LF) strategies for the aggregation of the outcomes of the different models of the ensemble. In the reconstruction of a signal, each model of the ensemble is assigned a weight and a bias related to the error committed in the reconstruction of training patterns similar to the one under reconstruction. Iteration of the reconstruction procedure and use of past measurements of the signals are introduced for improved performance. The proposed methodology is applied to a case study concerning the reconstruction of seven signals in the pressurizer of a Pressurized Water Reactor (PWR) nuclear power plant.