Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “computational efficiency”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 397 records · Page 22

Optimization of Water-Alternating-CO2 Injection Field Operations Using a Machine-Learning-Assisted Workflow

Summary This paper will present a robust workflow to address multiobjective optimization (MOO) of carbon dioxide (CO2)-enhanced oil recovery (EOR)-sequestration projects with a large number of operational control parameters. Farnsworth unit (FWU) field, a mature oil reservoir undergoing CO2 alternating water injection (CO2-WAG) EOR, will be used as a field case to validate the proposed optimization protocol. The expected outcome of this work would be a repository of Pareto-optimal solutions of multiple objective functions, including oil recovery, carbon storage volume, and project economics. FWU’s numerical model is used to demonstrate the proposed optimization workflow. Because using MOO requires computationally intensive procedures, machine-learning-based proxies are introduced to substitute for the high-fidelity model, thus reducing the total computation overhead. The vector machine regression combined with the Gaussian kernel (Gaussian-SVR) is used to construct proxies. An iterative self-adjusting process prepares the training knowledge base to develop robust proxies and minimizes computational time. The proxies’ hyperparameters will be optimally designed using Bayesian optimization to achieve better generalization performance. Trained proxies will be coupled with multiobjective particle swarm Optimization (MOPSO) protocol to construct the Pareto-front solution repository. The outcomes of this workflow will be a repository containing Pareto-optimal solutions of multiple objectives considered in the CO2-WAG project. The proposed optimization workflow will be compared with another established methodology using a multilayer neural network (MLNN) to validate its feasibility in handling MOO with a large number of parameters to control. Optimization parameters used include operational variables that might be used to control the CO2-WAG process, such as the duration of the water/gas injection period, producer bottomhole pressure (BHP) control, and water injection rate of each well included in the numerical model. It is proved that the workflow coupling Gaussian-SVR proxies and the iterative self-adjusting protocol is more computationally efficient. The MOO process is made more rapid by squeezing the size of the required training knowledge base while maintaining the high accuracy of the optimized results. The outcomes of the optimization study show promising results in successfully establishing the solution repository considering multiple objective functions. Results are also verified by validating the Pareto fronts with simulation results using obtained optimized control parameters. The outcome from this work could provide field operators an opportunity to design a CO2-WAG project using as many inputs as possible from the reservoir models. The proposed work introduces a novel concept that couples Gaussian-SVR proxies with a self-adjusting protocol to increase the computational efficiency of the proposed workflow and to guarantee the high accuracy of the obtained optimized results. More importantly, the workflow can optimize a large number of control parameters used in a complex CO2-WAG process, which greatly extends its utility in solving large-scale MOO problems in various projects with similar desired outcomes.

Energy & Fuels↗

Enhancing high-fidelity nonlinear solver with reduced order model

Abstract We propose the use of reduced order modeling (ROM) to reduce the computational cost and improve the convergence rate of nonlinear solvers of full order models (FOM) for solving partial differential equations. In this study, a novel ROM-assisted approach is developed to improve the computational efficiency of FOM nonlinear solvers by using ROM’s prediction as an initial guess. We hypothesize that the nonlinear solver will take fewer steps to the converged solutions with an initial guess that is closer to the real solutions. To evaluate our approach, four physical problems with varying degrees of nonlinearity in flow and mechanics have been tested: Richards’ equation of water flow in heterogeneous porous media, a contact problem in a hyperelastic material, two-phase flow in layered porous media, and fracture propagation in a homogeneous material. Overall, our approach maintains the FOM’s accuracy while speeding up nonlinear solver by 18–73% (through suitable ROM-assisted FOMs). More importantly, the proximity of ROM’s prediction to the solution space leads to the improved convergence of FOMs that would have otherwise diverged with default initial guesses. We demonstrate that the ROM’s accuracy can impact the computational efficiency with more accurate ROM solutions, resulting in a better cost reduction. We also illustrate that this approach could be used in many FOM discretizations (e.g., finite volume, finite element, or a combination of those). Since our ROMs are data-driven and non-intrusive, the proposed procedure can easily lend itself to any nonlinear physics-based problem.

97 MATHEMATICS AND COMPUTING↗

Coupling fast fluid dynamics and multizone airflow models in Modelica Buildings library to simulate the dynamics of HVAC systems

Multizone models are widely used in building airflow and energy performance simulations due to their fast computing speed. However, multizone models assume that the air in a room is well mixed, consequently limiting their application. In specific rooms where this assumption fails, the use of computational fluid dynamics (CFD) models may be an alternative option. Previous research has mainly focused on coupling CFD models and multizone models to study airflow in large spaces. While significant, most of these analyses did not consider the coupled simulation of the building airflow with the building’s Heating, Ventilation, and Air-Conditioning (HVAC) systems. This paper tries to fill the gap by integrating the models for HVAC systems with coupled multizone and CFD simulations for airflows, using the Modelica simulation platform. To improve the computational efficiency, we incorporated a simplified CFD model named fast fluid dynamics (FFD). We first introduce the data synchronization strategy and implementation in Modelica. Then, we verify the implementation using two case studies involving an isothermal and a non-isothermal flow by comparing model simulations to experiment data. Afterward, we study another three cases that are deemed more realistic. Furthermore, this is done by attaching a variable air volume (VAV) terminal box and a VAV system to previous flows to assess the capability of the models in studying the dynamic control of HVAC systems. Finally, we discuss further research needs on the coupled simulation using the models.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

An airfoil-based synthetic actuator disk model for wind turbine aerodynamic and structural analysis

Here, this study introduces an airfoil-based refinement technique to enhance the Actuator Disk Model (ADM) for improved wind turbine aerodynamic load prediction and structural simulation in conjunction with Large Eddy Simulations of the wind flow. While ADM offers higher computational efficiency than the more detailed but resource-intensive Actuator Line Model (ALM), it traditionally lacks the resolution needed to capture the localized blade forces accurately. To address this limitation, we introduce a refinement technique that uses airfoil-specific data and employs interpolation-based grid point refinement, achieving ALM-comparable accuracy while preserving ADM's efficiency. Unlike conventional ADM that provides only rotor-disk averaged forces, our synthetic method tracks transient aerodynamic load variations over multiple blade revolutions, allowing us to calculate the distributions of maximum and minimum loads during typical cycles. Applied to the NREL 5 MW reference turbine, our enhanced ADM accurately predicts key aerodynamic parameters (angle of attack, axial velocity, lift, drag, axial and tangential forces along the blades) as well as structural responses (blade tip deflection, maximum stress, and stress concentration). Our results show that the tip deflection ranges from 2.33m (3.69 % of blade length) to 4.28m (6.79 %), with maximum stress concentration occurring near the blade root. This research demonstrates that a refined synthetic ADM approach can serve as a computationally efficient alternative for both aerodynamic analysis and structural simulation of wind turbine blades subjected to realistic wind fields.

17 WIND ENERGY↗

AI Improves the Accuracy, Reliability, and Economic Value of Continental‐Scale Flood Predictions

Accurate flood early warnings are critical to minimize damage and loss of life. Current large‐scale operational forecasting systems, however, have limited accuracy, description of uncertainty, and computational efficiency. While Artificial intelligence (AI) can address these limitations in principle, the accuracy and reliability of AI forecasts have thus far proven insufficient. Here we present a novel hybrid framework that integrates AI‐based machinery termed Errorcastnet (ECN) with the National Water Model (NWM) to showcase the potential of ensemble AI flood forecasts over the contiguous U.S. ECN boosts prediction accuracy four‐ to six‐fold across lead times of 1–10 days, while providing uncertainty quantification. It also outperforms Google's state‐of‐the‐art global AI model. ECN‐based forecasts offer superior economic value (up to four‐fold) for decision‐making as compared to those from NWM alone. ECN performs well in varied ecoregions, physiography, and land management conditions. The framework is computationally efficient, enabling national‐scale ensemble forecasts in minutes.

artificial intelligence↗

Machine learning the Hubbard U parameter in DFT+U using Bayesian optimization

Abstract Within density functional theory (DFT), adding a Hubbard U correction can mitigate some of the deficiencies of local and semi-local exchange-correlation functionals, while maintaining computational efficiency. However, the accuracy of DFT+U largely depends on the chosen Hubbard U values. We propose an approach to determining the optimal U parameters for a given material by machine learning. The Bayesian optimization (BO) algorithm is used with an objective function formulated to reproduce the band structures produced by more accurate hybrid functionals. This approach is demonstrated for transition metal oxides, europium chalcogenides, and narrow-gap semiconductors. The band structures obtained using the BO U values are in agreement with hybrid functional results. Additionally, comparison to the linear response (LR) approach to determining U demonstrates that the BO method is superior.

Chemistry↗

Comparative Study of Nonlinear Black-Box Modeling for Power Electronics Converters

With the increasing penetration level of renewable sources and power electronics loads in modern power systems, accurate and computationally efficient models are needed. Black-box model (BBM) could be a useful method in such systems. However, not very extensive research efforts have been made for power electronics BBM so far, and existing works mostly focus on steady-state operation, neglecting the important transient behaviors such as load transients, voltage transients, and faults. This paper presents a comparative study of three commonly used nonlinear BBM approaches for transient behaviors of power electronics converters. Comparison methods are proposed, and the evaluations are conducted under different transients using a grid-connected single-phase photovoltaic inverter. The findings of this study provide valuable references for further feasibility investigations on implementing BBMs in large-scale power electronics-rich power systems.

Qiao, Liang↗

A voxel-based finite element model for the prediction of bladder deformation

Purpose: A finite element (FE) bladder model was previously developed to predict bladder deformation caused by bladder filling change. However, two factors prevent a wide application of FE models: (1) the labor required to construct a FE model with high quality mesh and (2) long computation time needed to construct the FE model and solve the FE equations. In this work, we address these issues by constructing a low-resolution voxel-based FE bladder model directly from the binary segmentation images and compare the accuracy and computational efficiency of the voxel-based model used to simulate bladder deformation with those of a classical FE model with a tetrahedral mesh. Methods: For ten healthy volunteers, a series of MRI scans of the pelvic region was recorded at regular intervals of 10 min over 1 h. For this series of scans, the bladder volume gradually increased while rectal volume remained constant. All pelvic structures were defined from a reference image for each volunteer, including bladder wall, small bowel, prostate (male), uterus (female), rectum, pelvic bone, spine, and the rest of the body. Four separate FE models were constructed from these structures: one with a tetrahedral mesh (used in previous study), one with a uniform hexahedral mesh, one with a nonuniform hexahedral mesh, and one with a low-resolution nonuniform hexahedral mesh. Appropriate material properties were assigned to all structures and uniform pressure was applied to the inner bladder wall to simulate bladder deformation from urine inflow. Performance of the hexahedral meshes was evaluated against the performance of the standard tetrahedral mesh by comparing the accuracy of bladder shape prediction and computational efficiency. Results: FE model with a hexahedral mesh can be quickly and automatically constructed. No substantial differences were observed between the simulation results of the tetrahedral mesh and hexahedral meshes (<1% difference in mean dice similarity coefficient to manual contours and <0.02 cm difference in mean standard deviation of residual errors). The average equation solving time (without manual intervention) for the first two types of hexahedral meshes increased to 2.3 h and 2.6 h compared to the 1.1 h needed for the tetrahedral mesh, however, the low-resolution nonuniform hexahedral mesh dramatically decreased the equation solving time to 3 min without reducing accuracy. Conclusions: Voxel-based mesh generation allows fast, automatic, and robust creation of finite element bladder models directly from binary segmentation images without user intervention. Even the low-resolution voxel-based hexahedral mesh yields comparable accuracy in bladder shape prediction and more than 20 times faster in computational speed compared to the tetrahedral mesh. This approach makes it more feasible and accessible to apply FE method to model bladder deformation in adaptive radiotherapy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Practical Operations of Energy Storage Providing Ancillary Services: From Day-Ahead to Real-Time

As renewable resources are increasingly penetrating power systems, energy storage systems (ESSs) become essential in providing both energy arbitrage and ancillary services. Because of its high flexibility, ESSs have been taken into account in the daily operation of the bulk power system in many places. This paper proposes a general framework in the current electricity market environment to help system operators model the participation of multi-type ESSs and evaluate the performance of both energy arbitrage and reserve provision. We discuss different levels of ESSs' flexibility in providing ancillary services and also consider a deliberate and practical implementation of the modeling. This framework can be seamlessly integrated into the daily market operation without sacrificing computational efficiency. Numerical experiments validate the efficacy of the proposed framework and show that ESSs possess excellent potential in providing ancillary services for the bulk power system.

41 EE - Solar Energy Technologies Office (EE-4S)↗

On the application of the MARTINI coarse-grained model to immersion of a protein in a phospholipid bilayer

An important step in the simulation of a membrane protein in a phospholipid bilayer is the correct immersion of the protein in the bilayer. Crystal structures are determined without the bilayer. Particularly for proteins with monotopic domains, it can be unclear how deeply and in which orientation the protein is being inserted in the membrane. We have previously developed a procedure combining coarse-grain (CG) with all-atom (AA) molecular dynamics (MD) simulations to insert and simulate a cytochrome P450 (CYP) possessing an N-terminal transmembrane helix connected by a flexible linker region to a globular domain that dips into the membrane. The CG simulations provide a computationally efficient means to explore different orientations and conformations of the CYP in the membrane. Converged configurations obtained in the CG simulations are then refined in AA simulations. Here, we tested different variants of the MARTINI CG model, differing in the water model, the treatment of long-range non-bonded interactions, and the implementation (GROMACS 4.5.5 vs 5.0.4), for this purpose. We examined the behavior of the models for simulating a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayer in water and for the immersion of CYP3A4 in a POPC bilayer, and compared the CG-MD results with the previously reported experimental and simulation results. We also tested the methodology on a set of four other CYPs. Finally, we propose an optimized protocol for modeling such protein-membrane systems that provides the most plausible configurations and is computationally efficient; this incorporates the standard non-polar water model and the GROMACS 5.0.4 implementation with a reaction field treatment of long-range interactions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Response of hypoxia to future climate change is sensitive to methodological assumptions

Climate-induced changes in hypoxia are among the most serious threats facing estuaries, which are among the most productive ecosystems on Earth. Future projections of estuarine hypoxia typically involve long-term multi-decadal continuous simulations or more computationally efficient time slice and delta methods that are restricted to short historical and future periods. We make a first comparison of these three methods by applying a linked terrestrial–estuarine model to the Chesapeake Bay, a large coastal-plain estuary in the eastern United States. Results show that the time slice approach accurately captures the behavior of the continuous approach, indicating a minimal impact of model memory. However, increases in mean annual hypoxic volume by the mid-twenty-first century simulated by the delta approach (+ 19%) are approximately twice as large as the time slice and continuous experiments (+ 9% and + 11%, respectively), indicating an important impact of changes in climate variability. Our findings suggest that system memory and projected changes in climate variability, as well as simulation length and natural variability of system hypoxia, should be considered when deciding to apply the more computationally efficient delta and time slice methods.

54 ENVIRONMENTAL SCIENCES↗

Chiral Effective Field Theory and the High-Density Nuclear Equation of State

Born in the aftermath of core-collapse supernovae, neutron stars contain matter under extraordinary conditions of density and temperature that are difficult to reproduce in the laboratory. In recent years, neutron star observations have begun to yield novel insights into the nature of strongly interacting matter in the high-density regime where current theoretical models are challenged. At the same time, chiral effective field theory has developed into a powerful framework to study nuclear matter properties with quantified uncertainties in the moderate-density regime for modeling neutron stars. In this article, we review recent developments in chiral effective field theory and focus on many-body perturbation theory as a computationally efficient tool for calculating the properties of hot and dense nuclear matter. We also demonstrate how effective field theory enables statistically meaningful comparisons among nuclear theory predictions, nuclear experiments, and observational constraints on the nuclear equation of state.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Convolutional Neural Networks for Hydrometeor Classification using Dual Polarization Doppler Radars

Traditional fuzzy logic hydrometeor classification algorithm is a common way to classify precipitation type from dual polarization doppler radar. We propose a deep learning-based method to estimate hydrometeors efficiently using observed radar variables such as horizontal reflectivity (Z H ), differential reflectivity (Z DR ), correlation coefficient (ρ HV ) and specific differential phase (K DP ) from National Weather Service NEXRAD collected at Vance AFB facility at the first elevation angle from January 1st, 2015 to July 31th, 2019. We stack matrices of values from these four polarimetric variables as one 3D array. Samples are preprocessed and divided into training, validation and test set with four target hydrometeor categories (Ice Crystals (IC), Dry Snow (DS), Light and/or Moderate Rain (RA) and Big Drops (rain) (BD)). We developed and optimized five Convolutional Neural Networks (CNNs) architectures and achieved an accuracy of 87.23% and 93.736% respectively using modified ResNet with two different input data selection strategies for hydrometeor classification. Training data selection strategies were important to ensure use of available samples in training for robust performance evaluated by applying the models to novel time period beyond what was use to train the model. Seasonal variation in atmospheric conditions lead to seasonal patterns of liquid vs solid forms of precipitation, that poses challenge for classifier and offer insights into domain specific approaches required for problem of hydrometeor identification. Computationally efficient and scalable approach for classification of hydrometeors offer opportunties to effectively use the large volumes of rich time series of radar observations that are becoming increasingly available.

Lu, Yuping↗

Simulation of multi-photon emission isotopes using time-resolved SimSET multiple photon history generator

Abstract-Multiple-photon emitters, such as In-111 or Se-75, have enormous potential in the field of nuclear medicine imaging. For example, Se-75 can be used to investigate the bile acid malabsorption and measure the bile acid pool loss. The simulation system for emission tomography (SimSET) is a well-known Monte Carlo simulation (MCS) code in nuclear medicine for its high computational efficiency. However, current SimSET cannot simulate these isotopes due to the lack of modeling of complex decay scheme and the time-dependent decay process. To extend the versatility of SimSET for simulation of those multi-photon emission isotopes, a time-resolved multiple photon history generator based on SimSET codes is developed in present study. For developing the time-resolved SimSET (trSimSET) with radionuclide decay process, the new MCS model introduce new features, including decay time information and photon time-of-flight information, into this new code. The half-life of energy states were tabulated from the Evaluated Nuclear Structure Data File (ENSDF) database. The MCS results indicate that the overall percent difference is less than 8.5% for all simulation trials as compared to GATE. To sum up, we demonstrated that time-resolved SimSET multiple photon history generator can have comparable accuracy with GATE and keeping better computational efficiency. The new MCS code is very useful to study the multi-photon imaging of novel isotopes that needs the simulation of lifetime and the time-of-fight measurements. (authors)

07 ISOTOPE AND RADIATION SOURCES↗

Geometry of matrix product states: Metric, parallel transport, and curvature

We study the geometric properties of the manifold of states described as (uniform) matrix product states. Due to the parameter redundancy in the matrix product state representation, matrix product states have the mathematical structure of a (principal) fiber bundle. The total space or bundle space corresponds to the parameter space, i.e., the space of tensors associated to every physical site. The base manifold is embedded in Hilbert space and can be given the structure of a Kähler manifold by inducing the Hilbert space metric. Our main interest is in the states living in the tangent space to the base manifold, which have recently been shown to be interesting in relation to time dependence and elementary excitations. By lifting these tangent vectors to the (tangent space) of the bundle space using a well-chosen prescription (a principal bundle connection), we can define and efficiently compute an inverse metric, and introduce differential geometric concepts such as parallel transport (related to the Levi-Civita connection) and the Riemann curvature tensor.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

An indirect approach to optimize the reaction rates of thermal NO formation for diesel engines

With stringent emission regulations, it has become more important for modern diesel engine manufacturers to accurately predict engine-out nitrogen oxide (NO x ) emissions across a wide range of operating conditions. Thermal NO is the major source of engine-out NO x in modern diesel engines. For thermal NO formation, several earlier studies have recommended the forward and reverse reaction rate coefficients of the rate-limiting reaction (O + N 2 ⇌ NO + N). However, due to deficiencies in sub-models and inadequacies of reduced chemical mechanisms to represent diesel combustion, these recommended values more often than not need to be adjusted in reduced order combustion models to accurately predict engine-out NO x . Hence, in this work a systematic and computationally efficient approach has been proposed to streamline the process of determining the optimum reaction rate coefficients. Here, to develop the optimization approach, four different production diesel engines with different operating conditions in terms of speed, load, and exhaust-gas recirculation have been considered. Numerical simulations have been performed using a detailed zero-dimensional velocity-composition-frequency transported probability density function (0D-VCF-tPDF) model that uses hundreds of notional particles to capture in-cylinder stratification. Four different combinations of hydrocarbon and NO x chemical mechanisms were used to represent chemistry. It was found that for the rate-limiting reaction, the pre-exponent factors (A f1 , A r1 ) and activation energies (E A,f1 , E A,r1 ) of the forward and reverse reaction rates follow a linear band in A f1 - E A,f1 and A r1 - E A,r1 space where predicted engine-out NO x match the measured values closely. By encompassing such bands from different engines and considering constraints on activation energies, a reduced search domain of pre-exponent factors and activation energies was constructed that is expected to be applicable to any diesel engine. Eventually, computationally efficient three-line and one-line search approaches were proposed to determine the optimum values of the pre-exponent factors and activation energies that led to a minimum error between measured and predicted engine-out NO x . Finally, these three-line and one-line NO x optimization approaches were applied to a fifth production diesel engine for which the 0D-VCF-tPDF model showed a very good predictive performance in terms of predicting peak pressure, 50% burn rate, and engine-out NO x when compared to measured and 3D-CFD values.

33 ADVANCED PROPULSION SYSTEMS↗

Use of radial symmetry for the calculation of cylindrical absorption coefficients and optimal capillary loadings

The problem of numerically evaluating absorption correction factors for cylindrical samples has been revisited using a treatment that fully takes advantage of the sample symmetry. It is shown that the path lengths for all points within the sample at all possible diffraction angles can be trivially determined once the angle-dependent distance distribution for a single line of points is calculated. This provides advantages in both computational efficiency and in gaining an intuitive understanding of the effects of absorption on diffraction data. A matrix of absorption coefficients calculated for µR products between 0 and 20 for diffraction angles θ D of 0° to 90° were used to examine the influence of (1) capillary diameter and of (2) sample density on the overall scattered intensity as a function of diffraction angle, where µ is the linear absorption coefficient for the sample and R is the capillary radius. Based on this analysis, the optimal sample loading for a capillary experiment to maximize diffraction at angles of 0 – 50° is in general expected to be achieved when the maximum radius capillary compatible with the beam is used, and when the sample density is adjusted to be 3/(4µR) of its original density.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Automatic extraction of cancer registry reportable information from free-text pathology reports using multitask convolutional neural networks

Objective: In this work we implement 2 different multitask learning (MTL) techniques, hard parameter sharing and cross-stitch, to train a word-level convolutional neural network (CNN) specifically designed for automatic extraction of cancer data from unstructured text in pathology reports. We show the importance of learning related information extraction (IE) tasks leveraging shared representations across the tasks to achieve state-of-the-art performance in classification accuracy and computational efficiency. Materials and Methods: Multitask CNN (MTCNN) attempts to tackle document information extraction by learning to extract multiple key cancer characteristics simultaneously. We trained our MTCNN to perform 5 information extraction tasks: (1) primary cancer site (65 classes), (2) laterality (4 classes), (3) behavior (3 classes), (4) histological type (63 classes), and (5) histological grade (5 classes). We evaluated the performance on a corpus of 95 231 pathology documents (71 223 unique tumors) obtained from the Louisiana Tumor Registry. We compared the performance of the MTCNN models against single-task CNN models and 2 traditional machine learning approaches, namely support vector machine (SVM) and random forest classifier (RFC). Results: MTCNNs offered superior performance across all 5 tasks in terms of classification accuracy as compared with the other machine learning models. Based on retrospective evaluation, the hard parameter sharing and cross-stitch MTCNN models correctly classified 59.04% and 57.93% of the pathology reports respectively across all 5 tasks. The baseline models achieved 53.68% (CNN), 46.37% (RFC), and 36.75% (SVM). Based on prospective evaluation, the percentages of correctly classified cases across the 5 tasks were 60.11% (hard parameter sharing), 58.13% (cross-stitch), 51.30% (single-task CNN), 42.07% (RFC), and 35.16% (SVM). Moreover, hard parameter sharing MTCNNs outperformed the other models in computational efficiency by using about the same number of trainable parameters as a single-task CNN. Conclusions: The hard parameter sharing MTCNN offers superior classification accuracy for automated coding support of pathology documents across a wide range of cancers and multiple information extraction tasks while maintaining similar training and inference time as those of a single task–specific model.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗