Forward Uncertainty Propagation Research Articles

Quantifying predictive uncertainty in damage classification for nondestructive evaluation using Bayesian approximation and deep learning

Magnetic flux leakage (MFL), a widely used nondestructive evaluation (NDE) method, for inspecting pipelines to prevent potential long-term failures. However, during field testing, uncertainties can affect the accuracy of the inspection and the decision-making process regarding damage conditions. Therefore, it is essential to identify and quantify these uncertainties to ensure the reliability of the inspection. This study focuses on the uncertainties that arise during the inverse NDE process due to the dynamic magnetization process, which is affected by the relative motion of the MFL sensor and the material being tested. Specifically, the study investigates the uncertainties caused by sensing liftoff, which can affect the output signal of the sensing system. Due to the complexity of describing the forward uncertainty propagation process, this study compared two typical machine learning (ML)-based approximate Bayesian inference methods, convolutional neural network and deep ensemble, to address the input uncertainty from the MFL response data. Besides, an autoencoder method is applied to tackle the lack of experimental data for the training model by augmenting the dataset, which is constructed with the pre-trained model based on transfer learning. Prior knowledge learned from large simulated MFL signals can fine-tune the autoencoder model which enhances the subsequent learning process on experimental MFL data with faster generalization. The augmented data from the fine-tuned autoencoder is further applied for ML-based defect size classification. This study conducted prediction accuracy and uncertainty analysis with calibration, which can evaluate the prediction performance and reveal the relation between the liftoff uncertainty and prediction accuracy. Further, to strengthen the trustworthiness of the prediction results, the decision-making process guided by uncertainty is applied to provide valuable insights into the reliability of the final prediction results. Overall, the proposed framework for uncertainty quantification offers valuable insights into the assessment of reliability in MFL-based decision-making and inverse problems.

Analytical Uncertainty Propagation in Neural Networks.

The usage of machine-learning techniques, such as neural networks, is common in a large variety of domains. Estimating the certainty of a predicted value is important when precise information is gained. Nevertheless, the forward propagation of uncertainty in machine-learning models is hardly understood. In general, providing error bars for measurements (measurement uncertainty) is crucial when high precision is needed for decision-making. The objective of this work is the development of an analytical method for aleatoric uncertainty forward propagation in neural networks, based on analytical uncertainty propagation well known from physics and engineering. With that, the method gives provable correct results. A benefit is that the method does not require a different training procedure, but only needs the weights and biases of the neural network and is computationally inexpensive. The analytical method is applied to real-world examples from the semiconductor industry (regression and image classification). Its usefulness is demonstrated by the provided examples, which show how meaningful error bars are when machine learning may be used for decision-making.

Probabilistic Printability Maps for Laser Powder Bed Fusion via Functional Calibration and Uncertainty Propagation

Abstract In this work, we develop an efficient computational framework for process space exploration in laser powder bed fusion (LPBF) based additive manufacturing technology. This framework aims to find suitable processing conditions by characterizing the probability of encountering common build defects. We employ a Bayesian approach towards inferring a functional relationship between LPBF processing conditions and the unobserved parameters of laser energy absorption and powder bed porosity. The relationship between processing conditions and inferred laser energy absorption is found to have good correspondence to literature measurements of powder bed energy absorption using calorimetric methods. The Bayesian approach naturally enables uncertainty quantification and we demonstrate its utility by performing efficient forward propagation of uncertainties through the modified Eagar-Tsai model to obtain estimates of melt pool geometries, which we validate using out-of-sample experimental data from the literature. These melt pool predictions are then used to compute the probability of occurrence of keyhole and lack-of-fusion based defects using geometry-based criteria. This information is summarized in a probabilistic printability map. We find that the probabilistic printability map can describe the keyhole and lack of fusion behavior in experimental data used for calibration, and is capable of generalizing to wider regions of processing space. This analysis is conducted for SS316L, IN718, IN625, and Ti6Al4V using melt pool measurement data retrieved from the literature.

Multi-target active subspaces generated using a neural network for computationally efficient turbulent combustion kinetic uncertainty quantification in the flamelet regime

Propagating uncertainties in kinetic models through combustion simulations can provide important metrics on the reliability and accuracy of a model, but remains a challenging and numerically expensive problem especially for large kinetic models and expensive turbulent combustion simulations. Various surrogate model and dimension reduction techniques have previously been applied in order to reduce the cost of forward uncertainty propagation in combustion simulations, but these are often limited to low-dimensional, simple combustion cases with scalar solution targets. In the current work, we developed a neural network-accelerated framework for identifying a low-dimensional active kinetic subspace that applies to the entire temperature solution space of a flamelet table and can capture the mixture fraction and strain rate dependent effects of the kinetic uncertainty. We then demonstrated the computational savings enabled by this framework through a proof-of-concept, flamelet-based application in a Reynolds-averaged Sandia Flame D simulation using a chemical model for methane combustion with 217 reactions. By leveraging the large dimensional compression and low-cost scaling of the active subspace method, offloading the initial dimension reduction gradient sampling onto the laminar flamelet simulations, and accelerating the gradient sampling process with a specifically designed neural network, we were able to estimate the temperature uncertainty profiles across the solution space of the turbulent flame with strong accuracy of 70−85% using just seven perturbed solutions. Additionally, as it occurs entirely within the flamelet table, the cost of identifying the reduced subspace does not scale with the cost of the turbulent combustion model, which is a promising feature of this framework for future application to larger-scale and more complex turbulent combustion applications.

An uncertainty quantification framework for logistic regression based geospatial natural hazard modeling

There is a class of data-driven global natural hazard predictive models that take advantage of broadly available geospatial proxies. These data-driven geospatial models have been commonly used for landslides and are becoming more available in recent years for liquefaction. Logistic regression is the most common method for predicting these ground failure occurrences. These models do not often include robust quantification of uncertainties although they are widely used in the pre-disaster planning and post-disaster response around the world. Taking the logistic regression based global geospatial liquefaction model (GGLM) (Zhu et al., 2017) as an example, we propose an uncertainty quantification (UQ) framework that consists of characterization of different sources of uncertainty, model sensitivity analysis, and forward uncertainty propagation. In this study, we have identified the main sources of uncertainty in such predictive models as parameter estimation uncertainty, modeling error, and geospatial input uncertainty. A Bayesian inference algorithm is used to quantify the posterior distribution of model parameters and quantify model parameter estimation uncertainties which are found to be negligible when a large amount of data is used in the parameter estimation process. Modeling errors are characterized based on the observed residuals between model predictions and measurements and by fitting a Gaussian distribution to the liquefaction probability residuals. The geospatial input uncertainties are characterized using the literature and expert judgment and propagated into model output. Second, we investigate the sensitivity of model output to different uncertain inputs and find that the variance of model output is largely controlled by the geospatial input uncertainties and model errors. Last, we propose an approximate forward uncertainty propagation method, which provides comparable results to a Monte Carlo simulation-based method with better computational efficiency. The proposed UQ framework provides a measure of uncertainty on model predictions and can be applied to any logistic-regression models and other geospatial modeling problems.

Robust Optimization of a Thermal Anti-Ice Protection System in Uncertain Cloud Conditions

This paper presents a framework for the robust optimization of the heat flux distribution for an anti-ice electrothermal ice protection system under uncertain conditions. The considered uncertainty regards a lack of knowledge concerning the characteristics of the cloud, i.e., the liquid water content and the median volume diameter of water droplets, and the accuracy of measuring devices, i.e., the static temperature probe. Uncertain parameters are modeled as random variables, and two sets of bounds are investigated. A forward uncertainty propagation analysis is carried out using a Monte Carlo approach exploiting a surrogate models. The optimization framework relies on a gradient-free algorithm (mesh adaptive direct search), and two different objective functions are considered, namely, the 95 quantile of the freezing mass rate and the statistical frequency of the fully evaporative operating regime. The framework is applied to a reference test case, revealing a potential to improve the heat flux distribution of the baseline design. A new heat flux distribution is proposed, and it presents a more efficient use of the thermal power, increasing flight safety even at nonnominal environmental conditions.

Probabilistic-based discrete model for the seismic fragility assessment of masonry structures

Classical Finite-Element and Discrete-Element strategies are expensive to carry when analysing masonry structures in the inelastic range, under a seismic excitation, and considering uncertainty. Their application to the seismic fragility assessment of masonry structures through non-linear time-history analysis becomes thus a challenge. The paper addresses such difficulty by presenting an alternative probabilistic-based numerical strategy. The strategy couples a discrete macro-element model at a structural-scale with a homogenization model at a meso-scale. A probabilistic nature is guaranteed through a forward propagation of uncertainty through loading, material, mechanical, and geometrical parameters. An incremental dynamic analysis is adopted, in which several assumptions decrease the required computational time-costs. A random mechanical response of masonry is provided by numerical homogenization, using Latin hypercube sampling with a non-identity correlation matrix, and only a reduced number of representative random samples are transferred to the macro-scale. The approach was applied to the seismic fragility assessment of an English-bond masonry mock-up. Its effectiveness was demonstrated, and its computational attractiveness highlighted. Results may foster its use within the seismic fragility assessment of larger structures, and the opportunity to better analyze the effect of material and geometric-based uncertainties in the stochastic dynamic response of masonry structures.

Harmonic transform-based non-parametric density estimation method for forward uncertainty propagation and reliability analysis

Reconstructing the probability density function (PDF) of the limit state function (LSF) is regarded as a non-intrusive method for forward uncertainty propagation and reliability analysis. In this paper, a novel non-parametric density estimation method is proposed to derive the unknown distribution of the LSF for this purpose, where the efficiency and accuracy are ensured. The main innovation of the paper is to derive the PDF of the LSF via a novel concept, i.e., the harmonic transform, regardless of the complexity of the problem. First, the non-parametric method for PDF estimation based on complex fractional moments/Mellin transform is briefly revisited, where the limitations are also pointed out. Inspired from Mellin transform, the concept of harmonic transform is put forward, which overcomes the disadvantages of Mellin transform. Then, the unknown PDF is recovered via the inverse harmonic transform, where the analytical formula is derived accordingly. Numerical estimation method is also given for practical implementations, where some critical issues are also pointed out. Detailed step-by-step procedures of the proposed method are also outlined. The proposed method is a non-parametric one since it does not require a distribution model in advance. Some commonly-used distributions are first applied to validate the efficacy of the proposed method. Then, four numerical examples involving different types of problems are extensively investigated to validate the proposed method for forward uncertainty propagation and reliability analysis. The results demonstrate that the proposed method is effective for deriving the PDFs for both static and dynamic reliability analyses with nonlinear explicit or implicit LSFs.

Uncertainty-quantified phenomenological optical potentials for single-nucleon scattering

Optical-model potentials (OMPs) continue to play a key role in nuclear reaction calculations. However, the uncertainty of phenomenological OMPs in widespread use---inherent to any parametric model trained on data---has not been fully characterized, and its impact on downstream users of OMPs remains unclear. Here we assign well-calibrated uncertainties for two representative global OMPs, those of Koning-Delaroche and Chapel Hill '89, using Markov-chain Monte Carlo for parameter inference. By comparing the canonical versions of these OMPs against the experimental data originally used to constrain them, we show how a lack of outlier rejection and a systematic underestimation of experimental uncertainties contributes to bias of, and overconfidence in, best-fit parameter values. Our updated, uncertainty-quantified versions of these OMPs address these issues and yield complete covariance information for potential parameters. Scattering predictions generated from our ensembles show improved performance both against the original training corpora of experimental data and against a new ``test'' corpus comprising many of the experimental single-nucleon scattering data collected over the last twenty years. Finally, we apply our uncertainty-quantified OMPs to two case studies of application-relevant cross sections. We conclude that, for many common applications of OMPs, including OMP uncertainty should become standard practice. To facilitate their immediate use, digital versions of our updated OMPs and related tools for forward uncertainty propagation are included as Supplemental Material.

Data-driven probabilistic quantification and assessment of the prediction error model in damage detection applications

Advances in computer hardware and sensor technologies have led to a surge in the use of data-driven modeling and machine learning for structural engineering applications, with Structural Health Monitoring (SHM) being one of them. Despite considerable interest, it remains a research topic due to the difficulty in accurately quantifying aleatoric and epistemic uncertainty in SHM systems. Sources of uncertainty are related to operational and environmental variability, as well as measurement noise and the model prediction error associated with the data used to train damage identification algorithms. In this work, the authors aim to explicitly quantify the statistical structure of model prediction error and assess its influence on the detection performance of strain-based SHM architectures under the existence of aleatoric variability. A structural beam, subjected to probabilistic static loading is used as the reference structure and strain measurements as the damage-sensitive features. Model prediction error is quantified explicitly using robust statistical tools through available laboratory observations and synthetic (Finite Element) data. Monte Carlo simulations enabled the forward propagation of uncertainty to the feature space to generate training data for three binary detectors (Likelihood Ratio Test, Quadratic Discriminant Analysis and Mahalanobis Distance), based on statistical pattern recognition. Detection performance was compared between the explicitly quantified prediction model error and the commonly assumed white Gaussian noise model, showcasing the influence of systematic error (bias) and correlation on the robustness of an SHM system using real-world data.

Design for ACP100 long term cooling flow resistance with random forests and inverse quantification

The ACP100 is an integrated small modular reactor with passive designs, and core cooling in the long-term cooling phase of the double-ended direct vessel injection (DEDVI) pipe break accident relies on the natural circulation in the sump recirculation pipeline. Therefore, the design of the recirculation pipeline flow resistance is very important. The ARSAC was used in the study to carry out the design for the recirculation pipeline flow resistance of the ACP100. The surrogate model of the accident simulation was constructed using the random forests algorithm, followed by using an inverse quantification method to determine the flow resistance distributions of the filter and related pipe, and finally the forward uncertainty propagation was performed for validation. The resistance design method utilized in the study has a high degree of freedom and is able to take the manufacturing uncertainty into consideration, which could facilitate the subsequent geometry design.

Uncertainty Quantification of Input Parameters in a 2-D Finite-Element Model for Broken Rotor Bar in an Induction Machine

In this article, a forward uncertainty propagation method is presented for a 2-D finite-element (FE) model in an induction machine. This method is applied to quantify the uncertainty of input parameters, for example, dimensions and material properties, and demonstrate their variability effect on harmonics related to the broken rotor bar (BRB) faults. To show the most influential input parameters in the case of BRB harmonics, a global sensitivity analysis is performed from the polynomial chaos expansion (PCE) approximation of the FE model. The results of this study indicate that BRB harmonics are highly sensitive to stator inner diameter, rotor outer diameter, rotor bar conductivity, and core materials. Moreover, the combined variability of these sensitive input parameters can attenuate the amplitude of the BRB harmonics 30%–90% compared to the simulation results at nominal values of input parameters and closely match with measurement results.

Quantifying the uncertainty of lake-groundwater interaction using the forward uncertainty propagation framework: The case of Lake Urmia

The interaction between a lake and groundwater has important implications to the quantity and quality of water in both environments. Quantification of lake-groundwater interaction (LGI) has been challenging in regions with limited in-situ data. LGI can be quantified by physically-based models, direct measurement of seepage, measurements of conservative chemical or isotopic tracers, and lake water balance. Despite the accuracy of the methods based on hydrochemical or isotopic measurements and analysis, they require extensive field data that are costly to collect in large lakes. Instead, the data required to quantify LGI by the lake water budget method can be obtained via typical ground measurements and satellite remote sensing. However, the influence on the estimated LGI imposed by the uncertainty associated with the other components of lake water budget (e.g., precipitation, evaporation, and inflow/outflow) is still poorly understood. In this study, we coupled a forward uncertainty propagation framework with the lake water budget to estimate the uncertainty of LGI using Latin-Hypercube sampling. We implemented the above framework in the hypersaline Lake Urmia (LU), located in the northwestern Iran, over five consecutive study periods between September 2017 and May 2020. The results revealed the exchange of flow from groundwater to the LU in all studied periods, ranging between 0.2 million cubic meters (MCM)/day and 7.6 MCM/day. The relative contribution of LGI in the water budget varied between 3% and 37%. The results of the uncertainty analysis further demonstrated that the cumulative probability of LGI < 0, which indicates flow direction from the LU to groundwater, was insignificant in all the study periods ranging between 0.001 and 0.35.

Combined effect of experimental and kinetic uncertainties on NO predictions in low-pressure premixed laminar H2/CH4/CO-air and H2/CH4/CO/C6H6-air flames

Accurate predictions of nitrogen oxides (NOx) are crucial for the design of low-emissions combustion technologies and require a truthful description of chemical kinetics and underlying thermo-chemical field. Nowadays, uncertainty remains in the reaction rates of different pathways of NOx formation. Experiments in simple systems, such as 1D flat premixed laminar flames, are necessary to shed light on the NOx formation process. However, experimental uncertainty cannot be avoided and influences the subsequent NOx formation modelling. In this context, forward propagation of experimental and kinetic uncertainties can be beneficial to assess their impact on the predictions of NOx emissions. In this work, forward Uncertainty Quantification (UQ) is performed via Polynomial Chaos Expansion (PCE), based on a sparse grid, to quantify the influence of the variability of temperature measurements and kinetic parameters on NO predictions in H2/CH4/CO-air 1D premixed laminar flames at low pressure, without and with 1.12% of benzene (C6H6) in the fuel blend. A significant uncertainty affects the prediction of NO for all the investigated flames. The measured temperature, with its uncertainty, is proven to be one of the most sensitive factor affecting the prediction of NO, particularly for the flames operated in lean and stoichiometric conditions. The results demonstrate that an accurate quantification of the uncertainty affecting the temperature measurements is critical for the development of predictive NOx formation models.

Toward reducing uncertainty quantification costs in DEM models of particulate flow: Testing simple, sensitivity-based, forward uncertainty propagation techniques

The performance of two conceptually-simple uncertainty quantification techniques are tested against the rigorous nested-loop sampling technique of Roy and Oberkampf (Comput Methods Appl Mech Eng, 200: 2131–2144, 2011) (herein called full-sampling) using two very small-scale DEM-based models of particulate flow (one gas-solid flow and one granular flow). The first simplified forward uncertainty propagation technique, reduced-sampling, uses a sensitivity analysis to eliminate uncertain inputs that have little impact on the model output prior to nested-loop sampling. The second technique, boundary-sampling, uses a sensitivity analysis to inform the selection of two bounding cases for each key model output. The uncertainties in the model outputs obtained via the reduced- and boundary-sampling methods agree well with those from full-sampling for both the gas-solid and granular flow models while yielding computational savings of 65–75% (reduced sampling) and 94–97% (boundary sampling).

MULTILEVEL QUASI-MONTE CARLO FOR INTERVAL ANALYSIS

This paper presents a multilevel quasi-Monte Carlo method for interval analysis, as a computationally efficient method for high-dimensional linear models. Interval analysis typically requires a global optimization procedure to calculate the interval bounds on the output side of a computational model. The main issue of such a procedure is that it requires numerous full-scale model evaluations. Even when simplified approaches such as the vertex method are applied, the required number of model evaluations scales combinatorially with the number of input intervals. This increase in required model evaluations is especially problematic for highly detailed numerical models containing thousands or even millions of degrees of freedom. In the context of probabilistic forward uncertainty propagation, multifidelity techniques such as multilevel quasi-Monte Carlo show great potential to reduce the computational cost. However, their translation to an interval context is not straightforward due to the fundamental differences between interval and probabilistic methods. In this work, we introduce a multilevel quasi-Monte Carlo framework. First, the input intervals are transformed to Cauchy random variables. Then, based on these Cauchy random variables, a multilevel sampling is designed. Finally, the corresponding model responses are post-processed to estimate the intervals on the output quantities with high accuracy. Two numerical examples show that the technique is very efficient for a medium to a high number of input intervals. This is in comparison with traditional propagation approaches for interval analysis and with results well within a predefined tolerance.

Quantifying the potential future contribution to global mean sea level from the Filchner–Ronne basin, Antarctica

Abstract. The future of the Antarctic Ice Sheet in response to climate warming is one of the largest sources of uncertainty in estimates of future changes in global mean sea level (ΔGMSL). Mass loss is currently concentrated in regions of warm circumpolar deep water, but it is unclear how ice shelves currently surrounded by relatively cold ocean waters will respond to climatic changes in the future. Studies suggest that warm water could flush the Filchner–Ronne (FR) ice shelf cavity during the 21st century, but the inland ice sheet response to a drastic increase in ice shelf melt rates is poorly known. Here, we use an ice flow model and uncertainty quantification approach to project the GMSL contribution of the FR basin under RCP emissions scenarios, and we assess the forward propagation and proportional contribution of uncertainties in model parameters (related to ice dynamics and atmospheric/oceanic forcing) on these projections. Our probabilistic projections, derived from an extensive sample of the parameter space using a surrogate model, reveal that the FR basin is unlikely to contribute positively to sea level rise by the 23rd century. This is primarily due to the mitigating effect of increased accumulation with warming, which is capable of suppressing ice loss associated with ocean-driven increases in sub-shelf melt. Mass gain (negative ΔGMSL) from the FR basin increases with warming, but uncertainties in these projections also become larger. In the highest emission scenario RCP8.5, ΔGMSL is likely to range from −103 to 26 mm, and this large spread can be apportioned predominantly to uncertainties in parameters driving increases in precipitation (30 %) and sub-shelf melting (44 %). There is potential, within the bounds of our input parameter space, for major collapse and retreat of ice streams feeding the FR ice shelf, and a substantial positive contribution to GMSL (up to approx. 300 mm), but we consider such a scenario to be very unlikely. Adopting uncertainty quantification techniques in future studies will help to provide robust estimates of potential sea level rise and further identify target areas for constraining projections.

A minimally invasive, efficient method for propagation of full‐field uncertainty in solid dynamics

AbstractWe present a minimally invasive method for forward propagation of material property uncertainty to full‐field quantities of interest in solid dynamics. Full‐field uncertainty quantification enables the design of complex systems where quantities of interest, such as failure points, are not known a priori . The method, motivated by the well‐known probability density function (PDF) propagation method of turbulence modeling, uses an ensemble of solutions to provide the joint PDF of desired quantities at every point in the domain. A small subset of the ensemble is computed exactly, and the remainder of the samples are computed with approximation of the evolution equations based on those exact solutions. Although the proposed method has commonalities with traditional interpolatory stochastic collocation methods applied directly to quantities of interest, it is distinct and exploits the parameter dependence and smoothness of the driving term of the evolution equations. The implementation is model independent, storage and communication efficient, and straightforward. We demonstrate its efficiency, accuracy, scaling with dimension of the parameter space, and convergence in distribution with two problems: a quasi‐one‐dimensional bar impact, and a two material notched plate impact. For the bar impact problem, we provide an analytical solution to PDF of the solution fields for method validation. With the notched plate problem, we also demonstrate good parallel efficiency and scaling of the method.

Methodology for holistic assessment of grey-green flood mitigation measures for climate change adaptation in urban basins

A novel methodological framework for urban drainage adaptation to climate change is introduced and tested in a small urban basin. The impact of climate change on Intensity-Duration-Frequency curves and on urban drainage is reported. The flood mitigation measures comprised both green and grey alternatives, i.e., permeable surfaces and sewer enlargement. The Storm Water Management Model (SWMM5) was calibrated and employed. Prior to calibration, sensitivity analysis of SWMM hydrologic-hydraulic and permeable surfaces input parameters was conducted using a global sensitivity analysis method. Two novel indices, the Index for Reliability of Urban Drainage Adaptation (IRUDA) and the Climate Variability Adaptation Index (CVAI), were also introduced. IRUDA assesses the total reliability of an urban drainage network in hydraulic, temporal and volumetric terms. Sizing of flood mitigation measures was undertaken employing construction cost along with IRUDA. CVAI was used for alternative selection among different flood mitigation scenarios. CVAI assesses the total adaptive capacity of the drainage network based on hydrological, hydraulic and economic aspects of the proposed adaptation plan. A forward uncertainty propagation scheme was employed on the selected alternative in order to examine input data and parametric uncertainty with respect to the peak discharge at the outlet of the basin. Finally, monetization of ecosystem services offered by green measures is presented. The simulation results for the study area illustrated that under the non-stationarity assumption future rainfall intensities and urban flooding risk will increase, and therefore, further flood mitigation measures will be required. Furthermore, the majority of the adaptation strategies examined were found to improve the capacity and reliability of the drainage network. Moreover, the results revealed that permeable pavements perform better for storm events with return period of up to five years. Overall, it can be concluded that design criteria need to be reviewed and also green measures, such as permeable surfaces, can be introduced in the design procedure.

Embedded-Error Bayesian Calibration of Thermal Decomposition of Organic Materials

Abstract Organic materials are an attractive choice for structural components due to their light weight and versatility. However, because they decompose at low temperatures relative to traditional materials, they pose a safety risk due to fire and loss of structural integrity. To quantify this risk, analysts use chemical kinetics models to describe the material pyrolysis and oxidation using thermogravimetric analysis (TGA). This process requires the calibration of many model parameters to closely match experimental data. Previous efforts in this field have largely been limited to finding a single best-fit set of parameters even though the experimental data may be very noisy. Furthermore, the chemical kinetics models are often simplified representations of the true decomposition process. The simplification induces model-form errors that the fitting process cannot capture. In this work, we propose a methodology for calibrating decomposition models to TGA data that accounts for uncertainty in the model-form and experimental data simultaneously. The methodology is applied to the decomposition of a carbon fiber epoxy composite with a three-stage reaction network and Arrhenius kinetics. The results show a good overlap between the model predictions and TGA data. Uncertainty bounds capture deviations of the model from the data. The calibrated parameter distributions are also presented. The distributions may be used in forward propagation of uncertainty in models that leverage this material.