(146d) Combining Spectroscopic Measurements with Mass-Balance Predictions Via the Dual Kalman Filter for Real-Time Monitoring
AIChE Annual Meeting
2024
2024 AIChE Annual Meeting
Nuclear Engineering Division
Graduate Student and Early Career Investigations - Session I
Monday, October 28, 2024 - 2:05pm to 2:30pm
In the context of the Hanford Site, the three methods used to estimate compositions have distinct advantages and disadvantages.
- Grab-sampling allows accurate and complete characterization of waste samples, but also has risks associated with potential exposure of workers to radiation and it can be a rate-limiting-step in nuclear-waste processing.
- A mass-balance model instantaneously produces estimates of compositions that only require up-stream information. However, a mass-balance approach provides a static model and has limited adaptability to variable process conditions.
- Real-time monitoring provides instantaneous process feedback making measurements robust to changing process conditions. However, not every species can be detected with every spectroscopic technique, limiting the number of species measured directly. Additionally, there are practical challenges such as window probe fouling.
The goal of the present work is to combine grab-sampling, a mass-balance model, and real-time measurements to create a stream characterization framework that is robust, efficient, and accurate.
The methodology is developed using a simulation of a representative Hanford-like system. The simulation is comprised of a 3-tank batch system modeling the Concentrate Receipt Vessel, Melter Feed Preparation Vessel, and Melter Feed Vessel (shown in Figure 1). In addition, attention has been given to modeling aspects unique to the Hanford site such as mixing within the melter and sample acquisition time. For the simulation, time steps forward from an initial condition with uncertainty being introduced with every tank transfer and with the addition of glass-forming chemicals. At each time-step, the contents of one tank are emptied into the next tank, except for residuals (also known as heels) that are left behind [2]. Assumptions are made on which species can be readily measured with ATR-FTIR and Raman spectroscopies (and their respective measurement accuracies) based on prior experimental work.
A state- and parameter-estimation technique, the dual-Kalman Filter, is used in this study to combine grab-sampling and real-time measurements with a mass-balance model for chemical composition estimates and real-time parameter estimates of the mass-balance model [3]. Operationally, the filter can detect faults by comparing the measured mass-balance model to a nominal (expected) mass-balance model. When the heel-mass of the measured mass-balance model deviates from the heel-mass expected from the nominal mass balance model, a fault is suspected. In the context of Hanford and the current study of heel mass mixing, a fault may indicate that the stirring mechanism is broken, crystallization/precipitation of a species has occurred within a tank, there is buildup on the tank walls, or the size or morphology of insoluble solids in the waste stream have changed. Regardless of the cause of the fault, engineers should be alerted of the changed process conditions and assess the cause before a bad batch of glass is produced or equipment (such as the melter) sustains damage.
The dual-Kalman Filter, in general, is used to combine real-time measurements with a mass-balance model predictions and grab-sampling. The dual-Kalman filter and its varieties (Kalman filters) are well-established and have been applied to many monitoring tasks over the past 50 years. The dual-Kalman filter combines data sources by weighting them based on their uncertainty. Because of the incorporation of uncertainty into the model, error bounds can be reported in addition to the predictions themselves.
Vitrification of nuclear waste at the Hanford site is a large undertaking that will take decades and a hundreds of billions of dollars. Because of the substantial investment of capital and time, effective and fault-free waste processing is highly desirable to avoid prolonging remediation efforts. This study showcases how the dual-Kalman filter can integrate a mass balance model with grab-sample measurements and in-line measurements to detect faults in heel mass composition. Furthermore, the dual-Kalman filter in this work highlights how real-time measurements may be utilized with existing process models at Hanford to incorporate real-time information for process monitoring. Process faults are not planned, are difficult to predict, and cannot always be prevented. Therefore, fault detection at the Hanford site will have to rely on observation and measurement.
Figure 1. Different measurement scenarios of the three-tank system studied in this work (attached image).
References
[1] R. A. Peterson et al., âReview of the Scientiï¬c Understanding of Radioactive Waste at the U.S. DOE Hanford Site.pdf,â Environ. Sci. Technol., no. 52, pp. 381â396, 2018.
[2] G. F. Piepel, B. G. Amidan, A. Heredia-langner, D. R. Weier, and S. K. Cooley, âStatistical Methods and Results for WTP IHLW and ILAW Compliance,â 2005.
[3] T. A. Wenzel, K. J. Burnham, M. V. Blundell, and R. A. Williams, âDual extended Kalman filter for vehicle state and parameter estimation,â Veh. Syst. Dyn., vol. 44, no. 2, pp. 153â171, 2006, doi: 10.1080/00423110500385949.