Multiscale Bayesian Rectification of Data from Linear Steady-State and Dynamic Systems without Accurate Models

The development and scale up of cost effective carbon capture processes is of paramount importance to enable the widespread deployment of these technologies to significantly reduce greenhouse gas emissions. The U.S. Department of Energy initiated the Carbon Capture Simulation Initiative (CCSI) in 2011 with the goal of developing a computational toolset that would enable industry to more effectively identify, design, scale up, operate, and optimize promising concepts. The first half of the presentation will introduce the CCSI Toolset consisting of basic data submodels, steady-state and dynamic process models, process optimization and uncertainty quantification tools, an advanced dynamic process control framework, and high-resolution filtered computationalfluid- dynamics (CFD) submodels. The second half of the presentation will describe a high-fidelity model of a mesoporous silica supported, polyethylenimine (PEI)-impregnated solid sorbent for CO2 capture. The sorbent model includes a detailed treatment of transport and amine-CO2- H2O interactions based on quantum chemistry calculations. Using a Bayesian approach for uncertainty quantification, we calibrate the sorbent model to Thermogravimetric (TGA) data.

Steady-state process models are common in virtual flow meter applications due to low computational complexity, and low model development and maintenance cost. Nevertheless, the prediction performance of steady-state models typically degrades with time due to the inherent nonstationarity of the underlying process being modeled. Few studies have investigated how learning methods can be applied to sustain the prediction accuracy of steady-state virtual flow meters. This paper explores passive learning, where the model is frequently calibrated to new data, as a way to address nonstationarity and improve long-term performance. An advantage with passive learning is that it is compatible with models used in the industry. Two passive learning methods, periodic batch learning and online learning, are applied with varying calibration frequency to train virtual flow meters. Six different model types, ranging from data-driven to first-principles, are trained on historical production data from 10 petroleum wells. The results are two-fold: first, in the presence of frequently arriving measurements, frequent model updating sustains an excellent prediction performance over time; second, in the presence of intermittent and infrequently arriving measurements, frequent updating in addition to the utilization of expert knowledge is essential to increase the performance accuracy. The investigation may be of interest to experts developing soft-sensors for nonstationary processes, such as virtual flow meters.

Bone suppression of chest radiographs (CXRs) is potentially useful for diagnosing lung diseases for radiologists and computer–aided diagnosis. This paper presents a cascaded convolutional network model in wavelet domain (Wavelet-CCN) for bone suppression in single conventional CXR. Wavelet coefficients are sparse and suitable as the output of convolutional network. The convolutional networks are trained to predict the wavelet coefficients of bone images from the wavelet coefficients of CXRs, using real two-exposure dual energy subtraction (DES) CXRs as training data. By combining the multi-level wavelet decomposition and a cascaded refinement framework, the Wavelet-CCN model can work automatically with a multi-scale approach and progressively refine the prediction in terms of accuracy and spatial resolution. Compared with previous work of CamsNet model which preforms bone prediction in gradient domain, the Wavelet-CCN model predicts the wavelet coefficients to reconstruct bone images and can avoid the inconsistent background intensity caused by 2D integration of gradients. The predicted bone image is subtracted from the original CXR to produce a soft-tissue image. The Wavelet-CCN model and its variants with different wavelet basis are evaluated on a dataset that consists of 504 cases of real two-exposure DES CXRs (404 cases for training and 100 cases for test). Experimental results show that among all the variants and different wavelet bases, the Wavelet-CCN model with Haar wavelet performs best. The average peak signal-to-noise ratio and structural similarity index of the soft-tissue images produced by the proposed Wavelet-CCN model are both higher than those of the previous CamsNet model in gradient domain, reaching values of 39.4 (±0.94) dB and 0.977 (±0.004), respectively. The results also demonstrate that the Wavelet-CCN model can process the CXRs acquired by four types of X-ray machines.

For grid-scale intermittent electricity storage, liquid air energy storage (LAES) is considered to be one of the most promising technologies for storing renewable energy. In this study, a steady-state process model was developed for an LAES, by combining a Linde liquefaction process and an open Rankine power cycle. To investigate the system performance and achieve global optimization, a single-factor analysis approach and multifactor genetic algorithm (GA) optimization model were built using MATLAB software. The effects of the charging pressure, storage pressure, discharging pressure, and isentropic efficiency of the compressor/turbine on the LAES performance parameters, such as the inlet temperature of the Joule–Thomson valve, round-trip efficiency (RTE), liquefaction ratio (LR), and power consumptions in the compressor, cryo-pump, and turbine, were investigated. Subsequently, the optimal conditions were obtained using the GA optimization method to achieve the optimal performance of the LAES. The results showed that the charging pressure, discharging pressure, and isentropic efficiencies of the compressor and turbine had significant effects on the RTE; an increase in the discharging pressure resulted in an improved expansion power output; the GA optimization could achieve RTE, LR, the system energy storage and recovery exergy efficiencies of 53.33%, 86.96%, 81.00%, and 78.16%; the power consumption in the compressor of GA optimization was 10.02% maximum saving. The proposed optimization method can be used to further explore the global optimization of cryogenic energy storage systems, such as different-layout LAES systems and different cryogenic liquefaction media energy storage systems.

Mathematical models of the gas metal arc (GMA) welding process may be used to study the influence of various welding parameters on weld dimensions, to assist in the development of welding procedures, and to aid in the generation of process control algorithms for automated applications. In this work, a three-dimensional (3-D), steady-state thermal model of the GMA welding process has been formulated for a moving coordinate framework and solved using the finite-element method. The model includes temperature-dependent material properties, a new finite-element formulation for the inclusion of latent heat of fusion, a Gaussian distribution of heat flux from the arc, plus the effects of mass convection into the weld pool from the melted filler wire. The influence of weld pool convection on the pool shape was approximated using anisotropically enhanced thermal conductivity for the liquid phase. Weld bead width and reinforcement height were predicted using a unique iterative technique developed for this purpose. In this paper, the numerical model is shown to be capable of predicting GMA weld dimensions for individual welds, including those with finger penetration. Also, good agreement is demonstrated between predicted weld dimensions and experimentally derived relations that describe the effects of process variables and their influence on average weld dimensions for bead-onplate GMA welds on steel plate.

In view of the limitation in term of direction that the existing interpolation reconstruction algorithm of single image and super resolution (SR) image cannot fully utilize the specific geometric features of the image itself, this paper proposed a novel SR algorithm of single image based on the integration of Bandelet geometric-flow-oriented interpolation and total variation in wavelet domain (IBITV). This method seamlessly took advantages of both wavelet and super-wavelet transformations. Firstly, the framework of image SR reconstruction and optimization was presented based on Bandelet and wavelet transformation. The framework made full use of the acquisition ability of the accurately geometric flow characteristics of the image based on Bandelet transformation. Furthermore, it introduced the idea that Bandelet transformed two-dimensional image and one-dimensional signal were projection mapping interactively, and the structures of image are similar at the same high frequency components of different scales after wavelet transformation to apply the Bandelet precise direction discrimination ability rather than sparse representation ability. Finally, in order to avoid the possibility of block effect, the influence of difference between low frequency component of the high resolution image and low resolution image on the final reconstructed image was considered based on the initially estimated image with high resolution. The total variation method of wavelet domain was utilized to carry out iterative solution of SR so as to obtain the final high resolution image. Experimental results show that the proposed method achieved favorable results in both iterative convergence characteristic of the algorithm and the image SR.

Process modelling and analysis of intensified CO2 capture using monoethanolamine (MEA) in rotating packed bed absorber

BioTFuel may set up facility at Total site in Dunkerque

This paper presents a double wavelet denoising (DWAD) method, which can preserve more details of an original signal. Although the noise removal method based on wavelet transform has been widely used, it still performs poorly for the signals with a low signal-to-noise ratio (SNR) or frequency overlap. Different from the wavelet denoising methods based on a single basis function, the DWAD considers filtering the wavelet coefficients of the noisy signal by threshold functions under two different wavelet domains, simultaneously. It considers using the difference of wavelet coefficient distribution and forcing the denoised signals under two wavelet domains to be the same to achieve more retention of details. In addition, the arctangent function is employed as a penalty function for wavelet coefficients to induce strong sparse wavelet coefficients. The DWAD is applied to one-dimensional signals and it is found that some wavelet coefficients which are smaller than the threshold could be retained during noise removal. The experiment results show that the average SNR of different noise levels is improved by at least 4.2 and 2.1 dB compared with the classical soft threshold method for the one-dimensional and image signals, respectively. Besides, the DWAD tends to obtain better performance on the details of original signals.

Compact steam bottoming cycles: Model validation with plant data and evaluation of control strategies for fast load changes

Incremental identification of hybrid process models

We propose and investigate dynamic spatial-temporal point process models for independent and interacting events. The models for independent events are dynamic spatial-temporal Poisson point process (DSTPPP) model that account for temporal and spatial clustering. The models proposed for events with interaction are Markov (Gibbs) space-time point process models. We model the intensity function of a DSTPPP via conditioning arguments that allow for additional interpretations and inclusion of well-known point process models as special cases. Depending on the nature of the questions to be answered, the in- tensity function of a DSTPPP can be modeled in several ways. First, we develop models via conditioning on the time component of the events and then consider conditioning on the location component of the events. Modeling, simulation and computation are accomplished in a fully hierarchical Bayesian framework. Finally, we focus on dynamic marked Markov space-time point processes, where the events are allowed to interact with each other across time and space. Once again the hierarchical Bayesian framework is invaluable in this case, since it allows us to introduce dynamic process models via conditioning. The methodolo- gies are illustrated using simulated data, and several applications, including modeling and inference for earthquake events in California, and tornado events in Missouri.

Most wavelet-based reconstruction methods of compressive sensing (CS) are developed under the independence assumption of the wavelet coefficients. However, the wavelet coefficients of images have significant statistical dependencies. Lots of multivariate prior models for the wavelet coefficients of images have been proposed and successfully applied to the image estimation problems. In this paper, the statistical structures of the wavelet coefficients are considered for CS reconstruction of images that are sparse or compressive in wavelet domain. A multivariate pursuit algorithm (MPA) based on the multivariate models is developed. Several multivariate scale mixture models are used as the prior distributions of MPA. Our method reconstructs the images by means of modeling the statistical dependencies of the wavelet coefficients in a neighborhood. The proposed algorithm based on these scale mixture models provides superior performance compared with many state-of-the-art compressive sensing reconstruction algorithms.

This work presents pattern recognition-based methods for controller adaptation and performance evaluation. These methods comprise a passive model-based adaptive control algorithm that is simple to use, easy to understand, stable, and fairly robust in a wide variety of applications. Controller adaptation in this work uses excitation diagnostics to initiate batch-wise regression of a process model to dynamic closed-loop process data. The process model is then employed in model-based controller tuning relations to update the controller's character. Controller performance evaluation is used to determine appropriate adjustments to the tuning relations such that an accurate process model will produce desired controller performance. These adaptive techniques are implemented using vector quantizing neural networks as efficient pattern recognition tools. The adaptive algorithm is presented in a structure that allows for the implementation of these advanced techniques without requiring the replacement of an existing feedback controller. This is demonstrated using a simulated nonlinear third order process and an IMC tuned PI controller with Smith Predictor.

In the last decade, technical advancements have greatly improved the design and execution efficiency of well completions, leading to improved recovery from unconventional reservoirs. However, analyzing fracture diagnostic tests in unconventional plays are still challenging due to high uncertainty in predictive capabilities in the context of fracture dynamics during treatment. The main objective of this study is to identify fracture behavior during injection and pressure fall-off periods in hydraulic fracturing treatments and diagnostic fracture injection tests (DFIT), respectively. In this study, discrete wavelet transformation (DWT) was used to analyze real field injection and fall-off data in the wavelet domain. The analyzed data are from multi-stage hydraulic fracturing operations and DFIT in unconventional horizontal wells. DWT coefficients reveal very crucial information related to the nature of the events within recorded signals; they also reveal various patterns that are hard to recognize otherwise. The high-frequency components of the pressure and rate signals (detail coefficients) that are calculated by the wavelet transformation determine localization and separation of various events. We compared the identified events for injection and fall-off periods with moving reference point (MRP) and G-function analysis, respectively. The main advantage of our proposed approach is that it is based on real-time data and does not require any assumptions related to existing or created fractures. Also, it is very sensitive to physical changes in the system; thus, it reveals hidden information related to those changes. Consequently, the energy of detail coefficients represents several events at different frequencies. We used pseudo-frequency of wavelet coefficients as a diagnostic tool for an accurate comparison of fracture propagation and fracture closure events to determine similarities and differences between them. For example, the signal energy of detail coefficients from the wavelet transformation of hydraulic fracturing data demonstrates abrupt frequency changes during dilation or fracture height growth during fracture propagation. Therefore, we were able to identify those events by energy density analysis in both time and pseudo-frequency domains in an objective manner, which otherwise was not possible with conventional methodologies such as G- function derivative analysis. This paper details the successful methodology for effective implementation of a new fracture diagnostic technique for fracturing operations or DFITs in unconventional horizontal wells. This new fracture diagnostic method does not require any reservoir or fracture pre-assumptions; it mainly relies on the pressure behavior, which is a result of various events at different frequencies. Pressure fall-off behavior of a DFIT gives essential information related to closure event of the created mini-fracture. Identification of these events at different pseudo-frequency ranges improves the understanding of the dynamic fracture behavior also the characteristics of the reservoir. Unlike many other diagnostic techniques, this data-driven approach requires minimum input/data for analysis. This approach also lends itself to real-time application quite easily.