[Paper Review] Scale-Bridging Model Development for Coal Particle Devolatilization

Benjamin Schroeder, Sean T. Smith+7 more|arXiv (Cornell University)|Sep 3, 2016

Probabilistic and Robust Engineering Design12 references3 citations

TL;DR

This paper develops a scale-bridging model (SBM) for coal particle devolatilization using a single-reaction kinetic model with distributed activation energy and a functional yield model, calibrated against a detailed Chemical Percolation Devolatilization (CPD) model. The SBM successfully captures thermodynamic trends and kinetic timescales at application-relevant scales while maintaining computational efficiency, with credibility enhanced through consistency testing and uncertainty quantification, achieving strong performance across diverse heating rates and conditions.

ABSTRACT

When performing large-scale, high-performance computations of multi-physics applications, it is common to limit the complexity of physics sub-models comprising the simulation. For a hierarchical system of coal boiler simulations a scale-bridging model is constructed to capture characteristics appropriate for the application-scale from a detailed coal devolatilization model. Such scale-bridging allows full descriptions of scale-applicable physics, while functioning at reasonable computational costs. This study presents a variation on multi-fidelity modeling with a detailed physics model, the chemical percolation devolatilization model, being used to calibrate a scale-briding model for the application of interest. The application space provides essential context for designing the scale-bridging model by defining scales, determining requirements and weighting desired characteristics. A single kinetic reaction equation with functional yield model and distributed activation energy is implemented to act as the scale-bridging model-form. Consistency constraints are used to locate regions of the scale-bridging model's parameter-space that are consistent with the uncertainty identified within the detailed model. Ultimately, the performance of the scale-bridging model with consistent parameter-sets was assessed against desired characteristics of the detailed model and found to perform satisfactorily in capturing thermodynamic trends and kinetic timescales for the desired application-scale. Framing the process of model-form selection within the context of calibration and uncertainty quantification allows the credibility of the model to be established.

Motivation & Objective

• To create a computationally efficient model that captures scale-appropriate physics from a detailed coal devolatilization model for use in large-scale simulations.
• To address the challenge of high computational cost in multi-physics simulations by bridging the gap between detailed physics models and application-scale requirements.
• To ensure model credibility by calibrating the SBM against the detailed model using consistency constraints and uncertainty quantification.
• To enable reliable prediction of devolatilization behavior across varying heating rates and temperatures with minimal parameterization.
• To provide a framework for model-form selection and validation that supports extrapolation confidence in engineering applications.

Proposed method

• A single-reaction kinetic model with distributed activation energy and a functional yield model is formulated as the scale-bridging model (SBM) form.
• The SBM is calibrated using consistency constraints derived from key quality-of-interest (QoI) metrics extracted from the detailed CPD model.
• Parameter-space exploration is performed to identify regions of the SBM’s parameter set that remain consistent with the uncertainty bounds of the detailed model’s predictions.
• Uncertainty quantification is applied to the detailed CPD model to define the range of acceptable behavior for calibration.
• The SBM’s performance is evaluated by comparing kinetic traces and cumulative yield distributions against the detailed model across multiple design of experiments (DOE) conditions.
• Model credibility is enhanced through iterative refinement of the yield model, informed by discrepancies between SBM and CPD predictions.

Experimental results

Research questions

• RQ1 How can a reduced-order scale-bridging model be constructed to accurately represent the thermodynamic and kinetic behavior of coal devolatilization at application scales while minimizing computational cost?
• RQ2 What constraints and quality-of-interest (QoI) metrics are necessary to ensure the SBM remains consistent with the uncertainty of the detailed CPD model?
• RQ3 To what extent can a single-reaction model with distributed activation energy and functional yield capture the complex behavior of coal devolatilization across varying heating rates and temperatures?
• RQ4 How does uncertainty quantification and consistency testing improve the credibility of the SBM for interpolation and limited extrapolation?
• RQ5 What improvements to the yield model can be made based on discrepancies between the SBM and detailed model predictions?

Key findings

• The SBM with four free parameters successfully captures the thermodynamic trends and kinetic timescales of the detailed CPD model across a range of heating rates, including 1×10⁶ K/s.
• Consistent parameter sets were identified through uncertainty-aware consistency testing, ensuring the SBM remains within the uncertainty bounds of the detailed model’s predictions.
• The model demonstrated strong performance across all DOE conditions, with visual and cumulative distribution comparisons confirming fidelity to the detailed model’s behavior.
• Refining the yield model based on observed discrepancies significantly improved the SBM’s accuracy, particularly in capturing high-temperature behavior.
• The updated yield model reduced model-form error and increased confidence in the SBM’s applicability for interpolative and limited extrapolative simulations.
• The methodology is transferable to other coal types by re-fitting the yield model to coal-specific CPD data and reapplying consistency testing with prior knowledge from the Utah Sufco bituminous coal results.