Reduction In Computation Time Research Articles

Recommender System Metaheuristic for Optimizing Decision-Making Computation

We implement a novel recommender system (RS) metaheuristic framework within a nonlinear NP-hard decision-making problem, for reducing the solution search space before high-burden computational steps are performed. Our RS-based metaheuristic supports consideration of comprehensive evaluation criteria, including estimations of the potential solution set’s optimality, diversity, and feedback/preference of the end-user, while also being fully compatible with additional established RS evaluation metrics. Compared to prior Operations Research metaheuristics, our RS-based metaheuristic allows for (1) achieving near-optimal solution scores through comprehensive deep learning training, (2) fast metaheuristic parameter inference during solution instantiation trials, and (3) the ability to reuse this trained RS module for traditional RS ranking of final solution options for the end-user. When implementing this RS metaheuristic within an experimental high-dimensionality simulation environment, we see an average 91.7% reduction in computation time against a baseline approach, and solution scores within 9.1% of theoretical optimal scores. A simplified RS metaheuristic technique was also developed in a more realistic decision-making environment dealing with multidomain command and control scenarios, where a significant computation time reduction of 87.5% is also achieved compared with a baseline approach, while maintaining solution scores within 9.5% of theoretical optimal scores.

Reduced-order modeling of solid-liquid mixing in a stirred tank using data-driven singular value decomposition

Stirred tanks are widely used across the (bio)chemical and process industries for solid-liquid mixing. Predicting solid suspension behavior under varying agitation speeds is critical for process control and optimization. However, inherent turbulence and multiphase interactions challenges the simulation in terms of accuracy and speed. In response, increasing attention has been paid to machine learning algorithms to enhance fluid dynamics simulations. In this work, a reduced-order model (ROM) to simulate solid-liquid flows in a stirred tank was developed, which uses singular value decomposition (SVD) to learn the flow patterns from computational fluid dynamics (CFD) simulations. The impact of mode numbers and design points were further investigated. The results show that the use of the ROM can result in a reduction of computation time of up to three orders of magnitude with reasonable accuracy. This study contributes by offering an exploration into extending ROM to multiphase flows, with a particular focus on solid-liquid mixing processes.

A novel low‐complexity model predictive control for Vienna rectifier

SummaryA Vienna rectifier is a kind of three‐phase converter with complex operation constraints. Traditional control methods suffer from poor dynamic responses and total harmonic distortion (THD), particularly when operating with adjustable wide‐range power. A novel low‐complexity model predictive control (LC‐MPC) algorithm is proposed based on the optimal switching vector sequence in this paper. First, a model predictive optimization control (MPOC) method is designed to search for the voltage vector sequence and its acting time. Second, the equivalent transformation and coordinate mapping of MPOC are efficiently achieved through the derived correlation factors and lookup table. Supported by the correlation factors, the redundant objective function calculation and repetitive online optimization are eliminated. Meanwhile, the simplified optimal over‐modulation strategy is implemented. Finally, the effectiveness and superiority of the algorithm are verified by comparative experiments. The results show that the proposed LC‐MPC is beneficial in terms of the computation time, dynamic response, over‐modulation, and harmonic content reduction.

LASSO–penalized clusterwise linear regression modelling: a two–step approach

In clusterwise regression analysis, the goal is to predict a response variable based on a set of explanatory variables, each with cluster-specific effects. In many real–life problems, the number of candidate predictors is typically large, with perhaps only a few of them meaningfully contributing to the prediction. A well–known method to perform variable selection is the LASSO, with calibration done by minimizing the Bayesian Information Criterion (BIC). However, existing LASSO-penalized estimators are problematic for several reasons. First, only certain types of penalties are considered. Second, the computations may sometimes involve approximate schemes. Third, variable selection is usually time consuming, due to a complex calibration of the penalty term, possibly requiring several multiple evaluations of an estimator for each plausible value of the tuning parameter(s). We introduce a two–step approach to fill these gaps. In step 1, we fit LASSO clusterwise linear regressions with some pre–specified level of penalization (Fit step). In step 2 (Selection step), we perform covariate selection locally, i.e. on the weighted data, with weights corresponding to the posterior probabilities from the previous step. This is done by using a generalization of the Least Angle Regression (LARS) algorithm, which permits covariate selection with a single evaluation of the estimator. In addition, both Fit and Selection steps leverage on an Expectation Maximization (EM) algorithm, fully in closed forms, designed with a very general version of the LASSO penalty. The advantages of our proposal, in terms of computation time reduction, and accuracy of model estimation and selection, are shown by means of a simulation study, and illustrated with a real data application.

Dynamic Simulation of Three Phase Induction Machines Based on Reduced Order Model for Power Systems Analysis

Induction machines are a crucial type of electrical equipment utilized as motors in various industries and single-phase forms in household applications. They make up more than 80% of the industrial motors used today. This paper presents a dynamic simulation of three-phase induction machines based on the (d-q) model, providing a clear and easy-to-understand explanation of the behavior of the induction motor in the synchronous reference frame. The simulation is implemented using SIMULINK/MATLAB software, and full-order model analysis is conducted for transient analysis. However, modeling induction machines as part of power system analysis can be done with a less detailed model than the full-order one. In this paper, a reduced-order model is employed to simulate the induction motor using MATLAB/SIMULINK. The results of the reduced-order system and the full-order framework are compared to investigate the model's limits, and the computational benefits of saving time during large power system simulations justify the relative decrease in accuracy. The use of a reduced-order model results in a significant reduction in computation time when simulating large and very large-scale power systems, with a much higher accuracy in transient analysis than the conventional steady-state model.

On Dynamic Ray Tracing and Anticipative Channel Prediction for Dynamic Environments

Ray tracing algorithms, that can simulate multipath radio propagation in presence of geometric obstacles such as buildings, objects or vehicles, are becoming quite popular, due to the increasing availability of digital environment databases and high-performance computation platforms, such as multi-core computers and cloud computing services. When objects or vehicles are moving, which is the case of industrial or vehicular environments, multiple successive representations of the environment ("snapshots") and multiple ray tracing runs are often necessary, which require a great human effort and a great deal of computation resources. Recently, the Dynamic Ray Tracing (DRT) approach has been proposed to predict the multipath evolution within a given time lapse on the base of the current multipath geometry, assuming constant speeds and/or accelerations for moving objects, using analytical extrapolation formulas. This is done without re-running a full ray tracing for every "snapshot" of the environment, therefore with a great reduction in both human labour and computation time. When DRT is embedded in a mobile radio system and used in real-time, ahead-of-time (or anticipative) field prediction is possible that opens the way to interesting applications. In the present work, a full-3D DRT algorithm is presented that allows to account for multiple reflections, edge diffraction and diffuse scattering for the general case where moving objects can translate and rotate. For the purpose of validation, the model is first applied to some ideal cases and then to realistic cases where results are compared with conventional ray tracing simulation and measurements available in the literature.

Numerical and artificial neural network analysis of an axisymmetric co-flow-focusing microfluidic droplet generator using active and passive control

Droplet generation in microscale has gained enormous attention in recent years especially in the pharmaceutical industry due to their application in targeted drug delivery into droplets. In most of these applications, monodispersity and uniformity of droplets are essential. Microfluidic devices can generate droplets at high throughput, enabling thousands of droplet compound encapsulation per second. The monodispersity of the droplets is ensured hydrodynamically through the dripping regime and their uniformity is controlled by active and passive microflow control methods. Here, we study numerically a microfluidic chip that uses a non-embedded co-flow-focusing geometry, so that the droplet generation throughput can take advantage of the flow-focusing devices while the non-embedded co-flow geometry forecloses the surfactant addition necessity. The continuous and dispersed phases were light mineral oil and water, respectively. We investigated the formation of droplets and studied how changing the external diameter of the chip affects the transition between the dripping regime (which corresponds to monodispersity) and the jetting regime. The number of parameters to be taken into account for the optimization of the device is enormous; therefore, in order to account for the effect of many geometrical and hydrodynamical parameters, we trained an artificial neural network based on our simulation data. Using this neural network, we evaluated droplet formation in 3240 different cases. This approach resulted in a remarkable reduction of computation time, from months to seconds. Examining numerous cases in such a short period lets us choose the optimum geometry and flow rate based on the application. The optimization was able to find the best geometry to extend the region of dripping regime in the flow rate map. Finally, to harness the droplet generation frequency, we also simulated a periodically switched laser and we were able to predict the generation of droplets with the same frequency as the switching frequency. Therefore, altering and controlling the frequency and dimensions of the droplets for a given flow rate ratio could be achieved with this technique, even without satellite droplets.

Mapping confinement potentials and charge densities of interacting quantum systems using conditional generative adversarial networks

Accurate and efficient tools for calculating the ground state properties of interacting quantum systems are essential in the design of nanoelectronic devices. The exact diagonalization method fully accounts for the Coulomb interaction beyond mean field approximations and it is regarded as the gold-standard for few electron systems. However, by increasing the number of instances to be solved, the computational costs become prohibitive and new approaches based on machine learning techniques can provide a significant reduction in computational time and resources, maintaining a reasonable accuracy. Here, we employ pix2pix, a general-purpose image-to-image translation method based on conditional generative adversarial network (cGAN), for predicting ground state densities from randomly generated confinement potentials. Other mappings were also investigated, like potentials to non-interacting densities and the translation from non-interacting to interacting densities. The architecture of the cGAN was optimized with respect to the internal parameters of the generator and discriminator. Moreover, the inverse problem of finding the confinement potential given the interacting density can also be approached by the pix2pix mapping, which is an important step in finding near-optimal solutions for confinement potentials.

Superpixel-based linear reconstruction method for dual-camera compressed hyperspectral imaging system

Compressive spectral imaging systems have received significant research attention in recent years. Among these systems, the dual-camera compressive hyperspectral imaging (DCCHI) design, has the potential to improve imaging resolution while maintaining the snapshot advantage. In this paper, we propose superpixel-based linear reconstruction methods for DCCHI system. Specifically, we assume that the spectral intensity of the superpixels generated from the RGB image varies linearly in the spectral dimension. Instead of addressing an ill-posed problem, we reconstruct hyperspectral images via solving overdetermined equations using the least squares method, resulting in a significant reduction in both computation time and memory usage. To further enhance the reconstruction accuracy, we extend the spectra of superpixels from a 1D linear space into a 3D space with an RGB spectral response matrix. Despite the simplicity of our method, it offers improved reconstruction accuracy of hyperspectral images compared to other state-of-the-art methods. Furthermore, it is more efficient in terms of memory usage and computational time.

A Coherent Wideband Acoustic Source Localization Using a Uniform Circular Array.

In modern applications such as robotics, autonomous vehicles, and speaker localization, the computational power for sound source localization applications can be limited when other functionalities get more complex. In such application fields, there is a need to maintain high localization accuracy for several sound sources while reducing computational complexity. The array manifold interpolation (AMI) method applied with the Multiple Signal Classification (MUSIC) algorithm enables sound source localization of multiple sources with high accuracy. However, the computational complexity has so far been relatively high. This paper presents a modified AMI for uniform circular array (UCA) that offers reduced computational complexity compared to the original AMI. The complexity reduction is based on the proposed UCA-specific focusing matrix which eliminates the calculation of the Bessel function. The simulation comparison is done with the existing methods of iMUSIC, the Weighted Squared Test of Orthogonality of Projected Subspaces (WS-TOPS), and the original AMI. The experiment result under different scenarios shows that the proposed algorithm outperforms the original AMI method in terms of estimation accuracy and up to a 30% reduction in computation time. An advantage offered by this proposed method is the ability to implement wideband array processing on low-end microprocessors.

PyTorch-FEA: Autograd-enabled finite element analysis methods with applications for biomechanical analysis of human aorta

Background and objectivesFinite-element analysis (FEA) is widely used as a standard tool for stress and deformation analysis of solid structures, including human tissues and organs. For instance, FEA can be applied at a patient-specific level to assist in medical diagnosis and treatment planning, such as risk assessment of thoracic aortic aneurysm rupture/dissection. These FEA-based biomechanical assessments often involve both forward and inverse mechanics problems. Current commercial FEA software packages (e.g., Abaqus) and inverse methods exhibit performance issues in either accuracy or speed. MethodsIn this study, we propose and develop a new library of FEA code and methods, named PyTorch-FEA, by taking advantage of autograd, an automatic differentiation mechanism in PyTorch. We develop a class of PyTorch-FEA functionalities to solve forward and inverse problems with improved loss functions, and we demonstrate the capability of PyTorch-FEA in a series of applications related to human aorta biomechanics. In one of the inverse methods, we combine PyTorch-FEA with deep neural networks (DNNs) to further improve performance. ResultsWe applied PyTorch-FEA in four fundamental applications for biomechanical analysis of human aorta. In the forward analysis, PyTorch-FEA achieved a significant reduction in computational time without compromising accuracy compared with Abaqus, a commercial FEA package. Compared to other inverse methods, inverse analysis with PyTorch-FEA achieves better performance in either accuracy or speed, or both if combined with DNNs. ConclusionsWe have presented PyTorch-FEA, a new library of FEA code and methods, representing a new approach to develop FEA methods to forward and inverse problems in solid mechanics. PyTorch-FEA eases the development of new inverse methods and enables a natural integration of FEA and DNNs, which will have numerous potential applications.

Numerical Modeling of Thermal Behavior during Lunar Soil Drilling

This paper presents a detailed thermal simulation analysis of the drilling process for icy soil in the lunar polar region. The aim is to investigate the temperature changes that occur in the debris removal area during the drilling process. We developed a multi-level particle size simulation model that includes a thermal sieve based on geometric constraints to evaluate the influence of specific heat capacity and thermal conductivity on particle temperature. Using the central composite design method, we carried out the simulation test design and analyzed the average temperature difference of particles within and outside the range of the thermal sieve. The parameters of the discrete element model were determined by comparing the temperature of the debris removal zone in the lunar environment with the temperature simulated by the discrete element method. The results show that the thermal conductivity of the sieve ranges from 100 to 400 W/m, and the average temperature inside the thermal sieve is negatively related to the specific heat capacity. The temperature deviation of the chip removal area is ±10 °C, which is consistent with the temperature deviation observed in the lunar environment and the lunar icy regolith drilling test. Furthermore, the addition of the thermal sieve to the multi-stage particle size simulation modeling significantly reduces the calculation time by 86%. This reduction in computational time may potentially increase the efficiency of drilling operations in the future. Our study provides insights into the thermal behavior of lunar icy regolith during drilling, and proposes a numerical model of heat transfer with a thermal sieve that can effectively reduce computational time while ensuring accurate temperature calculations.

Betweenness Centrality-Based seismic risk management for bridge transportation networks

Although bridges are integral parts of transportation networks, they are vulnerable to natural hazards. The damage incurred during these natural hazards events leads to network disruptions and hinders post-event recovery. Therefore, improving the seismic performance of transportation networks is critical. This paper proposes an origin–destination pair betweenness centrality (BC)-based methodology to rank bridges according to their importance in a complex bridge network. The proposed methodology is an easy-to-use approach that considers user equilibrium and topology and significantly shorten the computational time. In addition, this study proposes two strategies including the retrofit of bridges and construction of new bridges based on the BC importance of bridges to mitigate the seismic risk of bride networks. A well-studied bridge network in California was used to demonstrate the application and the effect of different strategies was compared in this study. Bridges located at the shortest path significantly affect the seismic performance of bridge networks, indicating there is no need to rehabilitate or replace all bridges under budget constraints. Also, constructing five new bridges based on BC importance improved seismic performance of the bridge network much more than retrofitting 21 bridges. Moreover, the proposed methodology presents a significant reduction in computational time (only 1/30 compared with an existing study) in resolving the importance of bridges in a network and can be applied to provide useful information for disaster management teams to mitigate the seismic risk of bridge networks.

A Spatiotemporal Deep Learning Approach for Urban Pluvial Flood Forecasting with Multi-Source Data

This study presents a deep-learning-based forecast model for spatial and temporal prediction of pluvial flooding. The developed model can produce the flooding situation for the upcoming time steps as a sequence of flooding maps. Thus, a dynamic overview of the forthcoming flooding situation is generated to support the decision of crisis management actors. The influence of different input data, data formats, and model setups on the prediction results was investigated. Data from multiple sources were considered as follows: precipitation information, spatial information, and an overflow forecast. In addition, models with different layers and network architectures such as convolutional layers, graph convolutional layers, or generative adversarial networks (GANs) were considered and evaluated. The data required to train and test the models were generated using a coupled hydrodynamic 1D/2D model. The model setup with the inclusion of all available input variables and an architecture with graph convolutional layers presented, in general, the best results in terms of root mean square error (RMSE) and critical success index (CSI). The prediction results of the final model showed a high agreement with the simulation results of the hydrodynamic model, with drastic reductions in computation time, making this model suitable for integration into an early warning system for pluvial flooding.

IMPROVED DESIGN OF SINGLE-LAYERED WIRE STRAND FOR COMBINED TENSILE AND CRIMPING APPLICATION WITH MESHING OPTIMIZATION

Faced with increasing competition in the global automotive market, reducing costs and design time has become a major focus of development among manufacturers. Faced with these needs, the numerical simulation of shaping processes has become an essential tool. It saves considerable time in the upstream phase of a project by reducing the number of experimental campaigns on prototypes. The studied mechanical system consists mainly in a part crimping at one end of a multi-stranded wire rope, which is stressed in tension. In order to reduce calculation time, it is important to get the optimal meshing parameters which provide a high accuracy. In this paper, a numerical model of a single layered wire rope is studied. The objective is to determine the optimal size and shape of meshing elements to get both accurate and fast results. First of all, the wire strand is investigated in tension and compared with the analytical formulation by G.Costello. Then, the organized wire strand is investigated in crimping process. The process reveals an optimal solution for wire strand structure with a 95% reduction of computation time without losing accuracy. This model is then used to simulate the behavior of a strand in one of the first-ever combined crimping-tensile test in order to obtain a minimal crimping length. This work can be very important for designers to reduce considerably the calculation time of their numerical models while guarantying a great accuracy in the mechanical behavior of a wire rope. Keywords: Wire rope, Finite element simulation, meshing process, Optimization, Mesh elements parameters optimization, Single-layered wire strand numerical model, Wire strand optimal mesh parameters, Wire rope crimping simulation, Tensile stressed wire strand

Assessment of most critical project delays on a multi-stage transmission expansion plan using particle swarm optimization

Transmission reinforcement projects of a Transmission Expansion Plan can suffer implementation delays that can impact the short-term power system operation in unpredicted ways. These delays may lead project managers to paying particular care to the development of those plans proved to be most critical from this point of view. This paper proposes a multi-objective dynamic-programming based method to evaluate the impacts of such delays in a short-term transmission plan with respect to load shedding, reliability and losses. It uses the Binary Pareto-Optimal Particle Swarm Optimization as optimization tool. For the trade-off analysis and the final decision to be made, the paper proposes indices to rank the projects in the plan from most to least critical, should they be delayed, based on the severity of the impacts and the probability of occurrence, both on an annual basis and for the whole planning horizon. The proposed method has been tested on an actual Brazilian 40-bus subtransmission system for a 3-year and a 6-year planning horizon, and with the 24-bus IEEE RTS reliability test system for a 3-year planning horizon. To assess its accuracy and efficiency, the results are compared to those using the Non-dominated Sorting Genetic Algorithm II method (NSGA-II) and to the exhaustive search that considers every delay configuration, showing the same accuracy as the exhaustive search and better convergence properties than NSGA-II. The most critical projects are accurately ranked helping the project manager to decide which projects should receive higher attention to guarantee its construction on the planned date. The method obtained a reduction in computational time of 35% for the Brazilian system and of 42% for the IEEE RTS in relation to the exhaustive approach, demonstrating its efficiency.

Learning assisted column generation for model predictive control based energy management in microgrids

Model predictive control is an effective approach for microgrid energy management. However, the main downside of such method is its expensive online computational cost, which is not amenable to most practical microgrid implementations. To address this issue, we propose a deep neural network assisted column generation approach that can accelerate the solution procedure of model predictive control. In each iteration, our approach leverages different deep neural networks to predict the optimal solutions of all the subproblems in column generation, which can accelerate the computation of all the subproblems and the entire process of column generation. The pre-existing knowledge of the microgrid is also exploited to guarantee the feasibility of the neural network outputs using multi-parametric programming theory. The numerical results show that our approach leads to a reduction in computational time of approximately 50% in a medium-sized microgrid compared with the full mixed integer solution based on traditional branch and bound method, while the optimality loss is only 0.02% in terms of operating costs.

An ultra-fast optimization algorithm for unit commitment based on neural branching

The computational efficiency of unit commitment (UC) is important for power system operations. Traditionally the unit commitment problem is solved per hour in a day, but with the scale of the power system and the electricity market continuing to expand, the large-scale UC problem will be hard to be solved within 1 h which will affect the power system operation and market clearing. To reduce the solving time of the large-scale UC problem, an ultra-fast optimization algorithm of neural branching for unit commitment (NBUC) is proposed. NBUC learns the branch and bound (B&B) decision made by full strong branching (FSB), which can generate the perfect B&B order to minimize the iterative process but take a lot of time to decide the perfect order by using graph convolutional neural network according to the historical data, and then makes the perfect order prediction with a certain precision without spending a lot of time to make B&B decision which minimizes the solving time including the iteration time and decision time in order to solve the large-scale UC problem quickly. A modified RTS-96 bus system is used to validate the effectiveness of the proposed NBUC. The results show 9.3% and 23.2% reductions in computational time compared with commercial software CPLEX and SCIP.

Generating functions for intersection products of divisors in resolved F-theory models

Building on the approach of 1703.00905, we present an efficient algorithm for computing topological intersection numbers of divisors in a broad class of elliptic fibrations with the aid of a symbolic computing tool. A key part of our strategy is organizing the intersection products of divisors into a succinct analytic generating function, namely the exponential of the Kähler class. We use the methods of 1703.00905 to compute the pushforward of this function to the base of the elliptic fibration. We implement our algorithm in an accompanying Mathematica package IntersectionNumbers.m that computes generating functions of intersection products for resolutions of F-theory Tate models defined over smooth base of arbitrary complex dimension. Our algorithm appears to offer a significant reduction in computation time needed to compute intersection numbers as compared to previously explored implementations of the methods in 1703.00905; as an illustration, we explicitly compute the generating functions for all F-theory Tate models with simple classical groups of rank up to twenty and highlight the growth of the computation time with the rank of the group.

Wave Propagation in Laminated Cylinders with Internal Fluid and Residual Stress

Numerical computation of wave propagation in laminated cylinders with internal fluid and residual stress is obtained using a Wave Finite Element formulation for 2D waveguides. Only a very small segment of the system is modelled, resulting in a very low-order finite element (FE) model to which the theory of wave propagation in 2D periodic structures is applied. The method uses standard FE formulations and exploits the capability of commercial FE software to model both fluid and structure and their interaction, resulting in a very large reduction in computational time. The presented approach is general, and can be applied without the need to make assumptions related to shell theory or low-frequency analysis. In particular, the laminated structure is discretised using 3D solid elements, thus representing the through-thickness dynamics with high accuracy. Residual radial and hoop stresses are included in the model by adding the FE pre-stress stiffness matrix to the original stiffness matrix of the system. The method provides simultaneously a very substantial reduction of computational cost, accurate solutions up to very high frequency and prediction of the dispersion curves for selected circumferential orders without the need for any further analysis. Here, the formulation of the method is introduced and its application to laminated cylinders filled with an acoustic fluid is presented. A composite, reinforced rubber cylinder, pre-stressed by a circumferential tension, is also shown as an example of a laminated pipe for high-pressure applications.