Data-Driven Remaining Useful Life Prediction for Pt-Rh Thermocouples Using an Extended Kalman Filter.

Area of Science:

• Materials Science and Engineering
• Sensor Technology
• Artificial Intelligence

Background:

• Platinum-Rhodium (Pt-Rh) thermocouples are critical for high-temperature industrial monitoring.
• Harsh environments cause degradation, affecting measurement accuracy and leading to potential failure.
• Existing methods struggle with the slow, nonlinear drift characteristic of Pt-Rh thermocouples.

Purpose of the Study:

• To develop an accurate Remaining Useful Life (RUL) prediction model for Pt-Rh thermocouples under high-temperature conditions.
• To address the nonlinear errors and slow drift affecting thermocouple measurements.
• To enhance the reliability of real-time condition monitoring in industrial processes.

Main Methods:

• Developed a degradation model for Pt-Rh thermocouples based on the Seebeck effect and vapor-transport theory.
• Implemented a hybrid Extended Kalman Filter (EKF) and Back-Propagation Neural Network (BPNN) algorithm.
• Validated the degradation model with laboratory data and compared the EKF-BPNN algorithm with other prediction methods.

Main Results:

• The developed degradation model accurately reflects laboratory test data.
• The EKF-BPNN hybrid algorithm demonstrated superior prediction accuracy (MAE 0.0016%, RMSE 0.0019%, MAPE 0.039%, R2 0.9833).
• Algorithms with strong nonlinear estimation capabilities were found to be less suitable for this specific application.

Conclusions:

• The hybrid EKF-BPNN algorithm is optimal for RUL prediction of Pt-Rh thermocouples experiencing high-temperature degradation.
• This approach effectively compensates for nonlinear errors and slow drift.
• The findings support improved condition monitoring and predictive maintenance in industrial settings.