Quantification Model Research Articles

Modelling parametric uncertainty in large-scale stratigraphic simulations

We combine forward stratigraphic models with a suite of uncertainty quantification and stochastic model calibration algorithms for the characterization of sedimentary successions in large scale systems. The analysis focuses on the information value provided by a probabilistic approach in the modelling of large-scale sedimentary basins. Stratigraphic forward models (SFMs) require a large number of input parameters usually affected by uncertainty. Thus, model calibration requires considerable time both in terms of human and computational resources, an issue currently limiting the applications of SFMs. Our work tackles this issue through the combination of sensitivity analysis, model reduction techniques and machine learning-based optimization algorithms. We first employ a two-step parameter screening procedure to identify relevant parameters and their assumed probability distributions. After selecting a restricted set of important parameters these are calibrated against available information, i.e., the depth of interpreted stratigraphic surfaces. Because of the large costs associated with SFM simulations, probability distributions of model parameters and outputs are obtained through a data driven reduced complexity model. Our study demonstrates the numerical approaches by considering a portion of the Porcupine Basin, Ireland. Results of the analysis are postprocessed to assess (i) the uncertainty and practical identifiability of model parameters given a set of observations, (ii) spatial distribution of lithologies. We analyse here the occurrences of sand bodies pinching against the continental slope, these systems likely resulting from gravity driven processes in deep sea environment.

Knowledge-informed generative adversarial network for functional calibration of computer models

Functional model calibration considers the input dependency of model parameters by explicitly considering them as functional calibration parameters. In this study, a novel knowledge-informed generative adversarial network (KI-GAN) is proposed for the functional calibration of computer models under uncertainty. The proposed KI-GAN leverages the superior distribution learning ability of generative models for uncertainty quantification of model parameters and uses prior knowledge to inform generative model training by formulating a knowledge-based loss function that can easily incorporate different types of physical and expert knowledge of the computer model and its parameters. Two experiments are presented to demonstrate the effectiveness of the proposed KI-GAN for calibrating the functional model parameters under various scenarios. The results indicate that the proposed KI-GAN can effectively encode various types of prior knowledge on the computer model and its parameters, such as partial differential equation constraints and derivative constraints, which can significantly improve the parameter estimation accuracy and reduce estimation uncertainty. Furthermore, with appropriate constraints applied, the proposed KI-GAN can estimate model parameters with reasonable accuracy even under sparse and noisy observations and quantify the underlying uncertainties. Finally, the trained generators can accurately infer the probabilistic distributions of unobserved model outputs without running the computer model.

Quantification of blood-brain barrier water exchange and permeability with multidelay diffusion-weighted pseudo-continuous arterial spin labeling.

To present a pulse sequence and mathematical models for quantification of blood-brain barrier water exchange and permeability. Motion-compensated diffusion-weighted (MCDW) gradient-and-spin echo (GRASE) pseudo-continuous arterial spin labeling (pCASL) sequence was proposed to acquire intravascular/extravascular perfusion signals from five postlabeling delays (PLDs, 1590-2790 ms). Experiments were performed on 11 healthy subjects at 3 T. A comprehensive set of perfusion and permeability parameters including cerebral blood flow (CBF), capillary transit time (τc ), and water exchange rate (kw ) were quantified, and permeability surface area product (PSw ), total extraction fraction (Ew ), and capillary volume (Vc ) were derived simultaneously by a three-compartment single-pass approximation (SPA) model on group-averaged data. With information (i.e., Vc and τc ) obtained from three-compartment SPA modeling, a simplified linear regression of logarithm (LRL) approach was proposed for individual kw quantification, and Ew and PSw can be estimated from long PLD (2490/2790 ms) signals. MCDW-pCASL was compared with a previously developed diffusion-prepared (DP) pCASL sequence, which calculates kw by a two-compartment SPA model from PLD = 1800 ms signals, to evaluate the improvements. Using three-compartment SPA modeling, group-averaged CBF = 51.5/36.8 ml/100 g/min, kw = 126.3/106.7 min-1 , PSw = 151.6/93.8 ml/100 g/min, Ew = 94.7/92.2%, τc = 1409.2/1431.8 ms, and Vc = 1.2/0.9 ml/100 g in gray/white matter, respectively. Temporal SNR of MCDW-pCASL perfusion signals increased 3-fold, and individual kw maps calculated by the LRL method achieved higher spatial resolution (3.5 mm3 isotropic) as compared with DP pCASL (3.5 × 3.5 × 8 mm3 ). MCDW-pCASL allows visualization of intravascular/extravascular ASL signals across multiple PLDs. The three-compartment SPA model provides a comprehensive measurement of blood-brain barrier water dynamics from group-averaged data, and a simplified LRL method was proposed for individual kw quantification.

Vision‐aided three‐dimensional damage quantification and finite element model geometric updating for reinforced concrete structures

AbstractThis article presents a vision‐aided framework to achieve three‐dimensional (3D) concrete damage quantification and finite element (FE) model geometric updating for reinforced concrete structures. The framework can process images and point clouds to extract damage information and update it in an FE model. First, a mask region convolutional neural network was used to realize highly precise damage detection and segmentation based on images. Second, a 3D point cloud was adopted in conjunction with the processed images for 3D damage qualification. The model‐updating method enables an FE model to delete concrete elements to update the variations in volume caused by structural damage. This framework supports interaction with mainstream FE software for further analysis. To demonstrate the efficiency of the proposed framework, it was used in an experiment on a reinforced‐concrete shear wall.

Droplet Digital PCR Assay with Linear Regression Models for Quantification of Buffalo-Derived Materials in Different Food Matrices

Droplet Digital PCR Assay with Linear Regression Models for Quantification of Buffalo-Derived Materials in Different Food Matrices

A Novel Cascaded Deep Learning Model for the Detection and Quantification of Defects in Pipelines via Magnetic Flux Leakage Signals

In this paper, we present a machine learning based quantitative method for the interpretation of signals gathered from non-destructive-testing (NDT) of steel pipelines via a semi-autonomous in-line-inspection (ILI) robot. The robot has a magnetic-flux-leakage (MFL) sensor that produces three axis data for each point of pipeline with specific intervals. Both the robot and the MFL sensor have been developed in-house. The signals collected via MFL sensor are converted into images to be used as input for the proposed defect detection model. We propose a combination of a defect detection model based on SwinYv5 object detection algorithm and a quantification model based on Cross-Residual Convolutional Neural Network (CR-CNN). The detected defect locations are used to extract the Region of Interest (ROI) images of defects that are used as input for the quantification model. In data collection phase, numerous tests have been conducted via a special test mechanism and a custom data augmentation technique has been deployed in order to increase the amount and variety of training data. According to test results, the proposed method is capable of detecting defects with a precision of 98.9% and quantifying them with maximum errors of 1.30, 1.65 and 0.47 mm for length, width and depth respectively.

Enhancing the accuracy of mid-infrared spectroscopy-based liver steatosis quantification using digital image analysis as a reference.

The assessment of liver steatosis is crucial in both hepatology and liver transplantation (LT) surgery. Steatosis can negatively impact the success of LT. Steatosis is a factor for excluding donated organs for LT, but the increasing demand for transplantable organs has led to the use of organs from marginal donors. The current standard for evaluating steatosis is a semi-quantitative grading based on the visual examination of a hematoxylin and eosin (H&E)-stained liver biopsy, but this method is time-consuming, subjective, and lacks reproducibility. Recent research has shown that infrared (IR) spectroscopy could be used as a real-time quantitative tool to assess steatosis during abdominal surgery. However, the development of IR-based methods has been hindered by the lack of appropriate quantitative reference values. In this study, we developed and validated digital image analysis methods for the quantitation of steatosis in H&E-stained liver sections using univariate and multivariate strategies including linear discriminant analysis (LDA), quadratic DA, logistic regression, partial least squares-DA (PLS-DA), and support vector machines. The analysis of 37 tissue samples with varying grades of steatosis demonstrates that digital image analysis provides accurate and reproducible reference values that improve the performance of IR spectroscopic models for steatosis quantification. A PLS model in the 1810-1052 cm-1 region using first derivative ATR-FTIR spectra provided RMSECV = 0.99%. The gained improvement in accuracy critically enhances the applicability of Attenuated Total Reflectance-Fourier Transform Infrared (ATR-FTIR) to support an objective graft evaluation at the operation room, which might be especially relevant in cases of marginal liver donors to avoid unnecessary graft explantation.

Building feature‐based machine learning regression to quantify urban material stocks: A Hong Kong study

AbstractUrban material stock (UMS) represents elegant thinking by perceiving cities as a repository of construction materials that can be reused in the future, rather than a burdensome generator of construction and demolition waste. Many studies have attempted to quantify UMS but they often fall short in accuracy, primarily owing to the lack of proper quantification methods or good data available at a micro level. This research aims to develop a simple but satisfactory model for UMS quantification by focusing on individual buildings. Generally, it is a “bottom‐up” approach that uses building features to proximate the material stocks of individual buildings. The research benefits from a set of valuable, “post‐mortem” ground truth data related to 71 buildings that have been demolished in Hong Kong. By comparing a series of machine learning‐based models, a multiple linear regression model with six building features, namely building type, building year, height, perimeter, total floor area, and total floor number, is found to yield a satisfactory estimate of building material stocks with a mean absolute percentage error of 9.1%, root‐mean‐square error of 474.13, and R‐square of 0.93. The major contribution of this research is to predict a building's material stock based on several easy‐to‐obtain building features. The methodology of machine learning regression is novel. The model provides a useful reference for quantifying UMS in other regions. Future explorations are recommended to calibrate the model when data in these regions is available.

Enhancing Natural-Hazard Exposure Modeling Using Natural Language Processing: a Case-Study for Maltese Planning Applications

The algorithmic processing of written language for tools such as predictive text, sentiment analysis, and translation services has become commonplace. The segment of computer science concerned with the interpretation of human language, NLP (Natural Language Processing), is a versatile and fast-developing field. In this paper, NLP is deployed unconventionally to gather insights into a building's multi-hazard exposure characteristics consistent with the GED4ALL attributes. NLP is used in this study to “read” the contents of digitally-submitted planning applications made on the Maltese archipelago. Maltese architects/engineers submit a concise but detailed description of the proposed works on any given site as part of a planning process. It is suggested that valuable insights exist within this description that can assist in classifying buildings within the bounds of the GED4ALL taxonomy. NLP can be used to layer additional, building-by-building information onto existing exposure models based on more conventional data. Although the results of this study are preliminary, NLP may prove a valuable tool for enhancing exposure modeling for multi-hazard risk quantification and management.

Nitroaromatic explosives' detection and quantification using an attention-based transformer on surface-enhanced Raman spectroscopy maps.

Rapidly and accurately detecting and quantifying the concentrations of nitroaromatic explosives is critical for public health and security. Among existing approaches, explosives' detection with Surface-Enhanced Raman Spectroscopy (SERS) has received considerable attention due to its high sensitivity. Typically, a preprocessed single spectrum that is the average of the entire or a selected subset of a SERS map is used to train various machine learning models for detection and quantification. Designing an appropriate averaging and preprocessing procedure for SERS maps across different concentrations is time-consuming and computationally costly, and the averaging of spectra may lead to the loss of crucial spectral information. We propose an attention-based vision transformer neural network for nitroaromatic explosives' detection and quantification that takes raw SERS maps as the input without any preprocessing. We produce two novel SERS datasets, 2,4-dinitrophenols (DNP) and picric acid (PA), and one benchmark SERS dataset, 4-nitrobenzenethiol (4-NBT), which have repeated measurements down to concentrations of 1 nM to illustrate the detection limit. We experimentally show that our approach outperforms or is on par with the existing methods in terms of detection and concentration prediction accuracy. With the produced attention maps, we can further identify the regions with a higher signal-to-noise ratio in the SERS maps. Based on our findings, the molecule of interest detection and concentration prediction using raw SERS maps is a promising alternative to existing approaches.

An on-line detection method for conveyor belt deviation faults

The conveyor belt deviation occurs frequently, and it will finally lead to an accident, so its detection has triggered skyrocketing attention from both industry and academia. In this paper, an adaptive segmentation model and a belt offset quantification model are established for continuous online detection of the conveyor belt deviation status. The results show that the degree of the conveyor belt deviation can be quantitatively calculated and its deviation status can be objectively evaluated. This technology has opened the path for a new way to on-line continuously detect the conveyor belt deviation.

Influence of Risk Management on Financial Performance of “Uzbekinvest” Insurance Company

Risk if not well managed could lead to collapse for most organizations especially those whose core business deals with day-to-day handling of risk. Risk management should, therefore, be at the core of an organization’s operations by integrating risk management practices into processes, systems, and culture of the entire organization. This involves identifying and analyzing risks, developing, and implementing risk handling techniques and monitoring the progress of these to avoid andor reduce the impact of risk on the financial performance of the company. Furthermore, insurance company under concern should follow current international leading practice by adopting Enterprise Risk Management (ERM) which incorporates other insurance risk quantification models. This will ensure that the companies remain afloat during such times of strict regulatory regimes such as solvency 11 and Basel.

Geomagnetically induced current uncertainty quantification based on the earth conductivity models constructed by the Kriging method

• Proposed and verified the superiority of Kriging method in constructing earth conductivity model. • Illustrate the necessity of uncertainty analysis for GIC. • Obtained the sensitivity of GIC input parameters(transformer parameters, transmission line parameters and ground resistance parameters) in 1000kV UHV Northern China power grid. With the expansion of power grid, the hazard of geomagnetically induced current (GIC) has attracted more attention. Due to the small number of Magnetotelluric detection points, the earth conductivity model based on limited data cannot be adapted to the GIC calculation of large-scale power grids. In addition, it is necessary to consider the effect of simultaneous changes in multiple input variables on GIC calculation results. Therefore, this paper presents a method for modeling the conductivity of the earth which is used in the case of less data, and uses the model to quantify the uncertainty. Based on the existing Magnetotelluric data in Northern China, the conductivity model within the power grid region is established using the Kriging method. The uncertainty quantification model of the GIC is constructed based on the Polynomial Chaos Expansions(PCE) method. The uncertainty quantification analysis is carried out for the 1000 kV ultra-high voltage Northern China Power Grid. The statistical information of the GIC and its sensitivity to each input parameter are obtained.

Advanced bridge visual inspection using real-time machine learning in edge devices

Conventional methods for bridge inspection are labor intensive and highly subjective. This study introduces an optimized approach using real-time learning-based computer vision algorithms on edge devices to assist inspectors in localizing and quantifying concrete surface defects. To facilitate a better AI-human interaction, localization and quantification are separated in this study. Two separate learning-based computer vision models are selected for this purpose. The models are chosen from several available deep learning models based on their accuracy, inference speed, and memory size. For defect localization, Yolov5s shows the most promising results when compared to several other Convolutional Neural Network architectures, including EfficientDet-d0. For the defect quantification model, 12 different architectures were trained and compared. UNet with EfficientNet-b0 backbone was found to be the best performing model in terms of inference speed and accuracy. The performance of the selected model is tested on multiple edge-computing devices to evaluate its performance in real-time. This showed how different model quantization methods are considered for different edge computing devices. The proposed approach eliminates the subjectivity of human inspection and reduces labor time. It also guarantees human-verified results, generates more annotated data for AI training, and eliminates the need for post-processing. In summary, this paper introduces a novel and efficient visual inspection methodology that uses a learning-based computer vision algorithm optimized for real-time operation in edge devices (i.e., wearable devices, smartphones etc.).

Modeling deadwood for rockfall mitigation assessments in windthrow areas

Abstract. Studying how deadwood mitigates the rockfall hazard in mountain forests is key to understanding the influence of climate-induced disturbances on the protective capacity of mountain forests. Both experimental quantification and numerical process modeling are needed to address this question. Modeling provides detailed insights into the rock–deadwood interaction and can therefore be used to develop effective forest management strategies. Here, we introduce an automatic deadwood generator (ADG) for assessing the impact of fresh woody storm debris on the protective capacity of a forest stand against rockfall. The creation of various deadwood scenarios allows us to directly quantify the mitigation potential of deadwood. To demonstrate the functionality of the proposed ADG method, we compare deadwood log patterns, deadwood effective height, and mesoscale surface ruggedness observed in field surveys in a natural windthrow area with their simulated counterparts. Specifically, we consider two sites near Lake Klöntal, Switzerland, where a major windthrow event occurred in 2019. We perform rockfall simulations for the time (a) before, (b) directly after, and (c) 10 years after the windthrow event. We further compare the results with (d) a simulation with complete clearing of the thrown wood: in other words, a scenario with no standing forest remaining. We showcase an integration of deadwood into rockfall simulations with realistic deadwood configurations alongside a diameter at breast height (DBH)- and rot-fungi-dependent maximum deadwood breaking energy. Our results confirm the mitigation effect of deadwood, which significantly reduces the jump heights and velocities of 400 kg rocks. Our modeling results suggest that, even a decade after the windthrow event, deadwood has a stronger protective effect against rockfall than that provided by standing trees. We conclude that an ADG can contribute to the decision-making involved in forest and deadwood management after disturbances.

Estimating and visualizing the exposure to tower crane operation hazards on construction sites

As an increasing number of building components are modularized and prefabricated, tower cranes are required to lift larger and heavier loads. This situation increased the risk of crane failure and the severity of crane accidents, challenging the effectiveness of conventional reactive safety management approaches. Thus, early hazard recognition and proactive control (Prevention through Design, PtD) are necessitated to manage Tower Crane Operation (TCO) hazards. This study proposes an automatic PtD method to manage the Hazard Exposure (HE) brought by one or multiple tower crane operations on construction sites in the construction planning stage by leveraging a path-finding algorithm and Building Information Modelling (BIM)-enabled spatial–temporal analyses. The dynamic distribution of TCO hazards is estimated based on safe and practical lift paths generated by a path-finding algorithm, and visualized as HE heatmaps in BIM models. Three application scenarios of the HE estimation and visualization are explored, including managing critical lifts, improving daily hazard communication on site, and contextualizing scheduling/planning tasks with hazard information. Estimating and visualizing HE in these scenarios are expected to allow planners and safety managers arrange activities and workspaces to minimize the risks of TCO hazards. The contributions of the paper include (1) proposing a generic energy-based quantification model for different types of TCO hazards, (2) increasing the spatial–temporal granularity of the HE estimation and visualization, and (3) promoting proactive controlling measures over the TCO hazards through planning, scheduling, and the coordination of crane operations with other on-site activities.

Comparison of Models for Quantification of Tomato Brown Rugose Fruit Virus Based on a Bioassay Using a Local Lesion Host.

Considering the availability of serological and molecular biological methods, the bioassay has been paled into insignificance, although it is the only experimental method that can be used to demonstrate the infectivity of a virus. We compared goodness-of-fit and predictability power of five models for the quantification of tomato brown rugose fruit virus (ToBRFV) based on local lesion assays: the Kleczkowski model, Furumoto and Mickey models I and II, the Gokhale and Bald model (growth curve model), and the modified Poisson model. For this purpose, mechanical inoculations onto Nicotiana tabacum L. cv. Xanthi nc and N. glutionosa L. with defined virus concentrations were first performed with half-leaf randomization in a Latin square design. Subsequently, models were implemented using Python software and fitted to the number of local lesions. All models could fit to the data for quantifying ToBRFV based on local lesions, among which the modified Poisson model had the best prediction of virus concentration in spike samples based on local lesions, although data of individual indicator plants showed variations. More accurate modeling was obtained from the test plant N. glutinosa than from N. tabacum cv. Xanthi nc. The position of the half-leaves on the test plants had no significant effect on the number of local lesions.

Lipid Profile Quantification and Species Discrimination of Pine Seeds through NIR Spectroscopy: A Feasibility Study

Pine seeds are known for their richness in lipid compounds and other healthy substances. However, the reference procedures that are commonly applied for their analysis are quite laborious, time-consuming, and expensive. Therefore, it is important to develop rapid, accurate, multi-parametric, cost-effective and, essentially, environmentally friendly analytical techniques that are easily implemented at an industrial scale. The viability of using near-infrared (NIR) spectroscopy to analyse the seed lipid content and profile of three different pine species (Pinus halepensis, Pinus brutia and Pinus pinaster) was investigated. Moreover, species discrimination using NIR was also attempted. Different chemometric models, namely partial least squares (PLS) regression, for lipid analysis, and partial least square-discriminant analysis (PLS-DA), for pine species discrimination, were applied. In relation to the discrimination of pine seed species, a total of 90.5% of correct classification rates were obtained. Regarding the quantification models, most of the compounds assessed yielded determination coefficients (R2P) higher than 0.80. The best PLS models were obtained for total fat, vitamin E, saturated and monounsaturated fatty acids, C20:2, C20:1n9, C20, C18:2n6c, C18:1n9c, C18 and C16:1. Globally, the obtained results demonstrated that NIR spectroscopy is a suitable analytical technique for lipid analysis and species discrimination of pine seeds.

Importance of size representation and morphology in modelling optical properties of black carbon: comparison between laboratory measurements and model simulations

Abstract. Black carbon (BC) from incomplete combustion of biomass or fossil fuels is the strongest absorbing aerosol component in the atmosphere. Optical properties of BC are essential in climate models for quantification of their impact on radiative forcing. The global climate models, however, consider BC to be spherical particles, which causes uncertainties in their optical properties. Based on this, an increasing number of model-based studies provide databases and parameterization schemes for the optical properties of BC, using more realistic fractal aggregate morphologies. In this study, the reliability of the different modelling techniques of BC was investigated by comparing them to laboratory measurements. The modelling techniques were examined for bare BC particles in the first step and for BC particles with organic material in the second step. A total of six morphological representations of BC particles were compared, three each for spherical and fractal aggregate morphologies. In general, the aggregate representation performed well for modelling the particle light absorption coefficient σabs, single-scattering albedo SSA, and mass absorption cross-section MACBC for laboratory-generated BC particles with volume mean mobility diameters dp,V larger than 100 nm. However, for modelling Ångström absorption exponent AAE, it was difficult to suggest a method due to size dependence, although the spherical assumption was in better agreement in some cases. The BC fractal aggregates are usually modelled using monodispersed particles, since their optical simulations are computationally expensive. In such studies, the modelled optical properties showed a 25 % uncertainty in using the monodisperse size method. It is shown that using the polydisperse size distribution in combination with fractal aggregate morphology reduces the uncertainty in measured σabs to 10 % for particles with dp,V between 60–160 nm. Furthermore, the sensitivities of the BC optical properties to the various model input parameters such as the real and imaginary parts of the refractive index (mre and mim), the fractal dimension (Df), and the primary particle radius (app) of an aggregate were investigated. When the BC particle is small and rather fresh, the change in the Df had relatively little effect on the optical properties. There was, however, a significant relationship between app and the particle light scattering, which increased by a factor of up to 6 with increasing total particle size. The modelled optical properties of BC are well aligned with laboratory-measured values when the following assumptions are used in the fractal aggregate representation: mre between 1.6 and 2, mim between 0.50 and 1, Df from 1.7 to 1.9, and app between 10 and 14 nm. Overall, this study provides experimental support for emphasizing the importance of an appropriate size representation (polydisperse size method) and an appropriate morphological representation for optical modelling and parameterization scheme development of BC.

Amyloid PET: A Questionable Single Primary Surrogate Efficacy Measure on Alzheimer Immunotherapy Trials.

Using amyloid PET imaging as a single primary surrogate efficacy measure in Alzheimer's disease immunotherapy trials, as happened when the FDA granted accelerated approval of aducanumab, is unjustified. In vivo evidence indicates that PET quantification of amyloid deposition is distorted and misrepresents effects of anti-amyloid treatments due to lack of specificity of the PET imaging probe, effects of amyloid-related imaging abnormalities, spill-over from high white matter signals, and questionable quantification models. Before granting approval to other immunotherapy candidates, the FDA should require rigorous evidence of all imaging claims and irrefutable documentation that proposed treatments are clinically effective and harmless to patients.