This study proposes an analytical redundancy method that combines empirical models with a Kalman filter to ensure the reliability of measurements from axial deformation sensors in a turbine head-cover bolt-monitoring system. This integration enables the development of predictive models that optimally estimate the dynamic deformation of each bolt during turbine operation at full and partial load. The test results of the models under conditions of outliers, measurement noise, and changes in turbine operating mode, evaluated using accuracy and sensitivity metrics, confirmed their high accuracy ( Acc ≈ 0.146 µm) and robustness ( S A < 0.001). The evaluation of the models' responses to simulated sensor faults (offset, drift, precision degradation, stuck-at) revealed characteristic residual patterns for faults with magnitudes > 5 µm. These findings establish the foundation for developing a fault detection and isolation algorithm for continuous monitoring of these sensors' operational health. For practical implementation, the models require validation across all operational modes, and maximum admissible deformation thresholds must be defined.