Optimal Network Research Articles

Investigation of the ANNs' potential for reliable assessment of r/c frame's seismic damage using different performance evaluation metrics

The development of a reliable method for the rapid assessment of the expected level of seismic damage of buildings constructed in countries with high seismicity areas is one of the crucial issues of current research, so that the authorities can take the necessary decisions for their rehabilitation or retrofit. A new approach to the problem is the application of methods that fall within the field of Artificial Neural Networks (ANNs). In this paper, an application of ANNs is attempted to predict the level of seismic damage in reinforced concrete frames. For this purpose, 27 frames with different structural characteristics were selected, designed and analyzed by nonlinear dynamic analysis. Then, ANNs were used to test their ability to reliably predict the level of seismic damage. The parameters that configure the networks were also investigated and their performance was evaluated using a number of metrics. The results showed that the optimal network can estimate the seismic damage level with significant reliability, provided that the training sample and the network modeling parameters are properly selected through a testing procedure.

Optimal quantum key distribution networks: capacitance versus security

The rate and security of quantum communications between users placed at arbitrary points of a quantum communication network depend on the structure of the network, on its extension and on the nature of the communication channels. In this work we propose a strategy for the optimization of trusted-relays based networks that intertwines classical network approaches and quantum information theory. Specifically, by suitably defining a quantum communication efficiency functional, we identify the optimal quantum communication connections through the network by balancing security and the quantum communication rate. The optimized network is then constructed as the network of the maximal quantum communication efficiency connections and its performance is evaluated by studying the scaling of average properties as functions of the number of nodes and of the network spatial extension.

"They Just Love Me"-An Examination of Social Support Experiences and Values Among People With TBI.

Social support is important for health and functional outcomes after traumatic brain injury (TBI), but many adults with TBI report inadequate social support. Little research has examined the social support priorities of adults with TBI or what an optimal social support network should include. The objective of this study was to describe the social support structures and experiences of adults with TBI. Community-based rehabilitation program. Community-dwelling adults with TBI, predominantly from racial and ethnic minority groups. Cross-sectional collection of quantitative and qualitative data. Number of close social contacts and total social contacts, along with a report of the frequency of social support using the Medical Outcomes Study Social Support Survey. Qualitative interviews were used to examine the quality of social support and participant values and priorities. Participants had an average of 4.3 close contacts (median 3.5, range 1-10), an average of 8.5 total social contacts (median 7, range 2-28), and a low frequency of social support (mean 28.4, SD=24.7). Interview themes fell under 2 grand themes: structure of social networks (social networks are made up of families; proximity influences the type of support) and quality of social support (commitment vs indifference; doing things with and for others adds meaning; and "they just love me"). Participants in this study had small social networks and a low frequency of social support. Findings suggest that people with TBI value committed support partners, social inclusion, bidirectional relationships, and emotional connection, yet experience lower support in these areas compared to tangible support. Future research should examine how to adapt existing models of social support to more closely align with participant values. Rehabilitation services may need to focus more heavily on individual and family training, addressing strategies to develop and maintain relationships, emotional connection, and communication to enhance relationship and social support quality.

An Efficient Flood Forecasting Model Using an Optimal Deep Belief Network

<p>Floods inflict significant damage globally each year, underscoring the importance of accurate and timely flood prediction to mitigate property loss and life. Precise flood prediction provides governments with crucial preemptive alerts regarding potential flood disasters, enabling timely evacuations and life-saving measures. Although various ML (machine learning) models have demonstrated improved performance compared to traditional statistical models in flood prediction, they often overlook the spatial features crucial to understanding the generation of floods. DL (deep learning) is used in flood prediction to enhance the promptness and efficiency of flood levels predictions. This work presents an optimized DL model to forecast floods using time series data. Initially, the data set was cleaned and normalized by linear interpolation. Then, the DL model ODBN (optimal deep belief network) is utilized for flood forecasting time series prediction. ODBN is the integration of DBN and SCA (Sinh-Cosh algorithm). The experimental analysis is carried out on the real-time dataset and achieved better MSE and RMSE values of 0.75 and 0.94 respectively. The findings suggest that the use of an ODBN is an effective method of accurately forecasting floods.</p>

Prediction of pressure evolution in non-venting self-pressurized liquid hydrogen tanks using artificial neural network approach

Liquid hydrogen (LH2) is a future energy carrier compared to most conventional fuels. One of the practical bottlenecks in non-venting LH2 tanks is to accurately predict the pressure evolution. Accordingly, many thermal models and correlations were developed; however significant deviations arose due to inherent complex thermal phenomena which have not been addressed yet. A general artificial neural network (ANN) model has been first developed based on 448 datasets amassed from different literature sources to predict the pressure evolution under different heat fluxes, initial filling ratios, as well as tank volumes and shapes. The model has been configured with a typical multi-layer feed-forward structure using a back-propagation learning algorithm and Levenberg-Marquardt training algorithm. For the ANN model, a preliminary optimization process has been conducted for reaching the optimal values for training-to-testing datasets ratio, learning parameters and optimum network structure. Besides, some criteria were set to evaluate the ANN model performance. The ANN model with the configuration of 5-11-11-1 was found to be adopted for such a prediction task. Compared to experimental results and literature models and correlations, the ANN model has been proven to predict the pressure evolution in non-venting self-pressurized LH2 tanks with a mean square error of 1.1779 × 10−5 s2, root mean square error of 0.0034 s, mean absolute error of 6%, mean relative error of 0.2% as well as a satisfactory correlation coefficient of 0.99934 between the network outputs and corresponding targets. This research work provides an effective modeling approach based on artificial neural network to solve one complex problem related with LH2 tanks with fast, accurate and consistent results. Also, the results of the present work can offer some practical guidance for the design of LH2 tanks. Further, the present work gives some insights into new application areas of ANN which will be the emphasis of future research work.

Simultaneous Synthesis of Single- and Multiple-Contaminant Water Networks Using LINGO and Excel Software

Controlling the distribution of water and wastewater between industrial processes is vital to rationalize water usage and preserve the environment. In this paper, a mathematical technique is proposed to optimize water–wastewater networks, and a nonlinear program is introduced to minimize the consumption of freshwater and, consequently, the flowrate of wastewater discharge. A general mathematical model, able to handle industrial plants containing up to eight sources and eight sinks, is developed using LINGO optimization software to facilitate dealing with complex case studies. The introduced model can handle single-contaminant networks as well as multiple-contaminant ones. The optimal water network is synthesized through two steps; the first step involves the introduction of the case study data into the developed mathematical model. The second step considers using the optimal solution produced after running the developed LINGO model as feed data for a pre-designed Excel sheet able to deal with these results and simultaneously draw the optimal water–wastewater network. The proposed mathematical model is applied to two case studies. The first case study includes actual data from four fertilizer plants located in Egypt; the water resources and requirements are simultaneously integrated to obtain a sensible cutting in both freshwater consumption (lowered by 52.2%) and wastewater discharge (zero wastewater discharge). The second case study regards a Brazilian petrochemical plant; the obtained results show noticeable reductions in freshwater consumption by 12.3%, while the reduction percentage of wastewater discharge is 4.5%.

Static Equivalence Method of Power Grid Based on Genetic Algorithm

When the electromagnetic transient simulation of a large-scale power grid is carried out, because the simulation is limited by the scale of simulation software, it is necessary to divide the grid into internal and external grids, and the external grid is modeled with equivalent simplification. Aiming at resolving the difficulties of the traditional grid equivalence method, such as a cumbersome calculation process and harsh calculation conditions, a static grid equivalence method based on the genetic algorithm is proposed in this paper. The method first constructs an external grid equivalence network, which includes coupling branches between boundary nodes, and each boundary node is connected to the external grid equivalence power supply through the branches. Then, the external grid equivalence model is used to establish an optimization model for solving the equivalence network parameters based on the information of the internal network, the difference of the power input from the external grid to the boundary nodes before and after equivalence is used as the objective function, and the genetic algorithm is used to realize the iterative optimization of the objective function to obtain a set of optimal post-equivalence network model parameters so that the state variables of the internal power grid before and after the equivalence are matched. The accuracy and effectiveness of the proposed method are verified through simulation with the CERPI36V7 node system.

Biphase routing scheme for optimal throughput in large-scale optical satellite networks

In the large-scale optical satellite network (LS-OSN), hundreds to thousands of low Earth orbit (LEO) satellites will be interconnected via laser links, offering global coverage characterized by high throughput and low latency. LS-OSNs present an attractive strategy to cultivate a comprehensively connected, intelligent world. However, the dynamic nature of the satellites, as they orbit the Earth, results in frequent changes in the LS-OSN topology. Thus, there is a pressing need for efficient routing algorithms that not only cater to massive traffic demands but also swiftly adapt to these constant topological changes. Traditional routing algorithms for services with specific bandwidth requirements often compromise on either computational speed or throughput efficiency. In response, this study introduces a routing scheme based on flow optimization and decomposition (FOND). This seeks to shorten the computation time while preserving optimal network throughput. Expanding upon the FOND scheme, we further devised two heuristic algorithms: the flow-based greedy path (FGP) and the flow-based greedy width (FGW). Simulation results from a 288-satellite constellation network indicate that both the FGP and FGW outpace contemporary methods in terms of the routing computation time while maintaining a consistent throughput equal to 100% of the network capacity. Notably, the FGP has exhibited an impressive capability, reducing the routing computation time to 0.23% compared to the baseline incremental-widest-path (IWP) algorithm, which operates on Dijkstra’s algorithm principles.

Fault Diagnosis of Hydraulic Components Based on Multi-Sensor Information Fusion Using Improved TSO-CNN-BiLSTM.

In order to realize the accurate and reliable fault diagnosis of hydraulic systems, a diagnostic model based on improved tuna swarm optimization (ITSO), optimized convolutional neural networks (CNNs), and bi-directional long short-term memory (BiLSTM) networks is proposed. Firstly, sensor selection is implemented using the random forest algorithm to select useful signals from six kinds of physical or virtual sensors including pressure, temperature, flow rate, vibration, motor power, and motor efficiency coefficient. After that, fused features are extracted by CNN, and then, BiLSTM is applied to learn the forward and backward information contained in the data. The ITSO algorithm is adopted to adaptively optimize the learning rate, regularization coefficient, and node number to obtain the optimal CNN-BiLSTM network. Improved Chebyshev chaotic mapping and the nonlinear reduction strategy are adopted to improve population initialization and individual position updating, further promoting the optimization effect of TSO. The experimental results show that the proposed method can automatically extract fusion features and effectively utilize multi-sensor information. The diagnostic accuracies of the plunger pump, cooler, throttle valve, and accumulator are 99.07%, 99.4%, 98.81%, and 98.51%, respectively. The diagnostic results of noisy data with 0 dB, 5 dB, and 10 dB signal-to-noise ratios (SNRs) show that the ITSO-CNN-BiLSTM model has good robustness to noise interference.

Utilizing optimal physics-informed neural networks for dynamical analysis of nanocomposite one-variable edge plates

This study presents an innovative Physics-Informed Neural Network (PINN) approach designed to predict the dynamic responses of the One-Variable Edge Plate (OVEP), a unique plate structure characterized by an edge defined by arbitrary mathematical functions. The OVEP is constructed from a nanocomposite material reinforced with graphene nanoplatelets. Utilizing an optimized PINN pipeline, this research successfully predicts the vibration characteristics of the OVEP, encompassing linear, nonlinear vibrations, and beating phenomena. The study also demonstrates that the optimal PINN outperforms conventional neural network in terms of stability, accuracy (achieving above 99% accuracy), and efficiency in predicting long-duration vibrations. Additionally, the computational time required for generating testing results is notably diminished compared to traditional partial differential equation (PDE) solvers (reduced by about 3 to 12 times). To demonstrate model robustness, synthetic noise is intentionally introduced into the training data. The results not only enhance our understanding of the complex dynamics of the OVEP but also highlight the effectiveness of the proposed PINN framework in capturing and forecasting the dynamic behaviors of advanced plate structures. This research presents a promising potential for addressing dynamic problems in the fields of aerospace, civil, and mechanical engineering.

V ALL-UKRAINIAN SCIENTIFIC AND PRACTICAL CONFERENCE WITH INTERNATIONAL PARTICIPATION «PERSONALITY IN THE EDUCATIONAL ENVIRONMENT OF MODERN SOCIETY»

In recent years, we have openly declared our move away from the Soviet legacy in the educational sphere. And today it is important to offer interesting practice-oriented classes to the participants of the educational process that would not push a person away from the learning process, but rather encourage them to actively immerse themselves in it. After all, the high-tech world has long recognized the special importance of science and knowledge in the civilization processes of scientific and technological progress, sustainable development of national economies, and improvement of citizens' lives. The modern interactive and meaningful educational environment has a positive impact on all participants in the educational process, from children who become more confident, curious and open-minded to adults who do not stop but continue to improve and develop professionally. Despite the fact that today Ukraine is experiencing military actions of the occupier, more than a hundred destroyed and damaged schools and higher education institutions, we must lay a stable foundation for democracy, civic engagement and professional environment of people who will continue to build the Ukrainian state. Education and science predict the prospects for human development and prepare people for tomorrow, so we must start today to form conscious citizens who will rebuild the country after our victory. We are active participants in reforms in secondary education under the New Ukrainian School Concept and transformations in higher education, which are focused on the future of Ukraine and a democratic society. Therefore, the entire education system of the region and the country is balanced and focused on the optimal network of educational institutions, on improving the content and quality of educational services, which will outline new dimensions of educational activities and performance, and provide space for the use of new digital resources for the educational development of the individual. In these areas, we find the experience of European national models of education, practices of educational activities and teaching life competencies quite interesting. 16 Ukrainian science and education is gaining unique European integration experience that promotes the use of innovative pedagogical teaching methods, the development of algorithms for using digital technologies and supporting individual entrepreneurial thinking, and the development of personal language skills as a means of active communication and professional development. All of this is done to expand educational opportunities for students' learning and development. In the above contexts, the question arises: is a new strategy in the field of education and science needed? We can confidently answer that we need a strategic plan and, in general, we need to fully adopt European principles of strategic planning - to actively fill them with new principles, plans and results of the education reform. Everything is aimed at developing and building the state and a person.

Deep Reinforcement Learning Lane-Changing Decision Algorithm for Intelligent Vehicles Combining LSTM Trajectory Prediction

Intelligent decisions for autonomous lane-changing in vehicles have consistently been a focal point of research in the industry. Traditional lane-changing algorithms, which rely on predefined rules, are ill-suited for the complexities and variabilities of real-world road conditions. In this study, we propose an algorithm that leverages the deep deterministic policy gradient (DDPG) reinforcement learning, integrated with a long short-term memory (LSTM) trajectory prediction model, termed as LSTM-DDPG. In the proposed LSTM-DDPG model, the LSTM state module transforms the observed values from the observation module into a state representation, which then serves as a direct input to the DDPG actor network. Meanwhile, the LSTM prediction module translates the historical trajectory coordinates of nearby vehicles into a word-embedding vector via a fully connected layer, thus providing predicted trajectory information for surrounding vehicles. This integrated LSTM approach considers the potential influence of nearby vehicles on the lane-changing decisions of the subject vehicle. Furthermore, our study emphasizes the safety, efficiency, and comfort of the lane-changing process. Accordingly, we designed a reward and penalty function for the LSTM-DDPG algorithm and determined the optimal network structure parameters. The algorithm was then tested on a simulation platform built with MATLAB/Simulink. Our findings indicate that the LSTM-DDPG model offers a more realistic representation of traffic scenarios involving vehicle interactions. When compared to the traditional DDPG algorithm, the LSTM-DDPG achieved a 7.4% increase in average single-step rewards after normalization, underscoring its superior performance in enhancing lane-changing safety and efficiency. This research provides new ideas for advanced lane-changing decisions in autonomous vehicles.

Degradation prediction of proton exchange membrane fuel cell based on mixed gated units under multiple operating conditions

The aging process of proton exchange membrane fuel cell (PEMFC) is affected by different operating factors, and its practical application scenarios involve multiple operation conditions. To improve the accuracy and controllability of aging prediction, a mixed recurrent neural network (RNN) model based on Gated Recurrent Unit (GRU) and Minimal Gated Unit (MGU) is proposed. The model dynamically adjusts the mixed weights of GRU and MGU throughout the entire training process to obtain the optimal prediction network structure throughout the entire lifecycle. Furthermore, an attention mechanism is incorporated into the mixed gated unit (MIXGU) model. The effectiveness of the MIXGU model is evaluated by utilizing experimental data from static condition, quasi-dynamic condition, and dynamic load cycling conditions. The predictive performance of MGU, GRU, MIXGU, and MIXGU model with attention mechanism (AT-MIXGU) are compared under different operating conditions. The validation results indicate that MIXGU model exhibits superior predictive performance compared to single gated unit, the attention mechanism enhances prediction accuracy. And the AT-MIXGU model demonstrates strong generalization capabilities, and well-suited for accurately predicting PEMFC aging under diverse operating conditions.

A mathematical model for the integrated problem of production network design and inventory positioning

In this paper, an integrated approach is addressed to tackle the production network design and inventory positioning problem in a single-sourcing, multi-product, and multi-period context. The objective of our model is to simultaneously determine the optimal network structure, safety stock amounts, and their respective locations, considering normal demand conditions. The main goal is to minimize the overall production network cost. The proposed model merges the traditional network design formulation with that for inventory positioning using the concept of guaranteed service. The developed model is implemented and tested on a case study of office furniture manufacturing, where the bill of materials is considered. The effectiveness of the proposed model is demonstrated by comparing it to a sequential approach where the network structure and safety stock decisions are made sequentially. The results reveal that the integrated approach surpasses the sequential approach, with cost savings ranging from 1.7% to 3.7%. In addition, a higher optimal cycle service level is obtained by the integrated approach compared to that of the sequential approach. To further evaluate the model's performance, a sensitivity analysis is conducted to examine the influence of the model's parameters, i.e., committed service time and coefficient of variation of demand, on its solutions. The analysis shows that when the coefficient of the demand variation remains unchanged, longer committed service times lead to lower safety stock costs and higher optimal service levels. Moreover, it is observed that more safety stocks of components are kept in the upstream layer of the network than those of finished goods.

Changes in the land-use landscape pattern and ecological network of Xuzhou planning area

Ongoing rapid urbanization has triggered significant changes in land use, rendering landscape patterns adversely impacted and certain habitat patches degraded. Ecological networks have consequently contracted overall. As such, an investigation into how land-use landscape patterns and ecological networks change over time and space is of major significance for ecological restoration and regional sustainability. Taking Xuzhou Planning Area as a case study, we examined spatiotemporal changes and features of the landscape pattern by employing the land-use change degree, the land-use transition matrix, and quantified landscape pattern indices. An ecological network analysis, which studies the changes in network connectivity and robustness, as well as their causes and contributors, was undertaken to probe into the features and trends of spatiotemporal changes in the land-use landscape pattern and ecological network amid expeditious urbanization. Analysis results unveiled the following: (1) From 1985 to 2020, there was a decline in the area of farmland, forest, and grassland, accompanied by an increase in land for construction, water bodies, and unused land. The southwestern research area witnessed farmland substantially give way to land for construction for this period, and the most dramatic change in land use occurred between 2000 and 2010. (2) The area of dominant patches in the research area shrank, along with more fragmented, complex landscapes. The land for construction was emerging as the dominant landscape by area, whereas patches of farmland, forest, grassland, and water bodies became less connected. (3) The ecological network was densely linked in the northeast, with sparser connections in the southwest. Spatial shrinkage was observed in the research area’s southwestern and central ecological corridors. Overall, the number of ecological sources and corridors rose and subsequently dropped before a rebound. (4) The ecological network grew more connected and robust from 1985 through 1990, as portions of farmland were converted into water bodies, which led to an increase in ecological sources. Given a reduction in ecological sources and corridors in the southwestern and central regions between 1990 and 2010, network connectivity and robustness declined, which was reversed from 2010 onward with the addition of two ecological sources—Pan’an Lake and Dugong Lake. With an optimal ecological network in 1990, however, it deteriorated significantly by 2010. The research area saw the minimum value of its network connectivity indices of network stability index (α), evenness index (β), and connectivity index (γ), in 2010, when its ecological network was highly fragmented and vulnerable, attributing to a strong contrast between the maximal connected subgraph’s relative size and connectivity robustness. The research findings can lay scientific groundwork for addressing ecological issues, restoring landscape patterns, and developing ecological networks amid urbanization.

Empirical Insights into 5G Deployments in Highway Operational Environments and Comparative Performance with 4G

Despite the promising benefits, the integration of 5G, particularly through Non-Standalone (NSA) architectures that rely on existing 4G infrastructures, introduces challenges in maintaining optimal radio network performance and service quality. This study evaluates 4G and 5G radio network performance through empirical field trials across highway operational scenarios, a domain less scrutinized compared with urban environments. By conducting simultaneous measurements and comparing all available networks, this research focuses on dissecting the performance of critical quality indicators to gauge the intricacies of radio network behavior in high-speed travel conditions. Our findings illuminate distinct behavioral patterns, highlighting the unique challenges and optimization opportunities in these scenarios. Variability in signal strength across the highway routes underscores the influence of geography and infrastructure on coverage, while the relative stability in signal quality suggests the networks’ capability to maintain signal quality amid fluctuating strength. Interference results indicate effective management of signal interference, crucial for high-quality links, whereas latency and throughput metrics highlight a lag behind anticipated goals for reduced latency but promising data rates. This study not only showcases the variances in network quality and performance, thus pinpointing areas for operator-specific enhancements, but also emphasizes the comparison between the robustness of 4G infrastructures and the challenges in optimizing 5G networks.

A method for quantifying the generalization capabilities of generative models for solving Ising models

For Ising models with complex energy landscapes, whether the ground state can be found by neural networks depends heavily on the Hamming distance between the training datasets and the ground state. Despite the fact that various recently proposed generative models have shown good performance in solving Ising models, there is no adequate discussion on how to quantify their generalization capabilities. Here we design a Hamming distance regularizer in the framework of a class of generative models, variational autoregressive networks (VANs), to quantify the generalization capabilities of various network architectures combined with VAN. The regularizer can control the size of the overlaps between the ground state and the training datasets generated by networks, which, together with the success rates of finding the ground state, form a quantitative metric to quantify their generalization capabilities. We conduct numerical experiments on several prototypical network architectures combined with VAN, including feed-forward neural networks, recurrent neural networks, and graph neural networks, to quantify their generalization capabilities when solving Ising models. Moreover, considering the fact that the quantification of the generalization capabilities of networks on small-scale problems can be used to predict their relative performance on large-scale problems, our method is of great significance for assisting in the Neural Architecture Search field of searching for the optimal network architectures when solving large-scale Ising models.

Design of an Iterative Method for Dynamic Resource Management in 5G Networks with IoT Integration Operations

In the realm of fifth-generation (5G) wireless networks, the escalating demands for high-speed, reliable, and efficient communication are paramount, especially with the widespread deployment of Internet of Things (IoT) devices & scenarios. Despite the advancements in 5G technologies, existing network management strategies often fall short in addressing the dynamic nature of network conditions, the heterogeneity of IoT device requirements, and the need for stringent privacy measures. These limitations underscore the necessity for innovative approaches that can adapt in real-time to varying demands while ensuring optimal network performance and user privacy. This paper introduces a suite of machine learning models designed to enhance the efficiency and reliability of 5G networks, catering specifically to the diverse needs of IoT applications. At the forefront, DynamicSlicerNet, a deep reinforcement learning-based model, dynamically slices 5G network resources tailored to IoT devices' requirements, addressing device mobility, application demands, and network congestion. This model demonstrates a substantial reduction in latency by up to 30% and improvement in reliability by up to 20%, outperforming static resource allocation methods. Further enhancing edge computing capabilities, FedEdgeAI leverages federated learning to train models directly on edge devices, a move that not only slashes latency by minimizing data transmission to centralized servers but also fortifies data privacy. Experimental evaluations highlight FedEdgeAI's efficacy in maintaining model accuracy while halving communication overhead. PredictiveNetCare, employing time-series analysis and anomaly detection, anticipates network failures, facilitating preemptive maintenance strategies. This predictive approach has shown a marked precision over 90% in identifying potential disruptions, significantly reducing maintenance-related downtime by 30% and bolstering network reliability by 15%. OptiAllocRL and AdaptiveQoSDL, both harnessing reinforcement and deep learning techniques, respectively, optimize resource allocation and manage Quality of Service (QoS) parameters adaptively. OptiAllocRL's strategy results in a 40% latency reduction and a 25% throughput increase, while AdaptiveQoSDL minimizes packet loss by up to 50% and enhances end-to-end delay by up to 35%, ensuring high QoS levels under fluctuating network conditions. This comprehensive approach sets a new benchmark for future 5G network management, paving the way for a more connected, efficient, and secure digital world.

Optimization Method for Fracture-Network Design under Transient and Pseudosteady Conditions Using Unified-Fracture-Design and Deep-Learning Approaches

Summary In unconventional shale and tight reservoirs, the concept of stimulated reservoir volume (SRV) is used to correlate the volume of total injected proppant with well performance. The SRV configuration consists of primary fractures connected to the wellbore and secondary fractures intersecting primary fractures. SRV productivity is determined by fracture conductivity, fracture dimensions, and network complexity, which also vary with time. This work presents an extension of the unified-fracture-design (UFD) approach to account for not only the pseudosteady state (PSS) but also transient flow regimes and ultimately optimize SRV for maximizing well performance. A generalized productivity index (PI) for both the transient and PSS regimes is presented to improve well performance by searching for the maximum PI over time. In addition, a surrogate model is developed to accelerate the optimization. This study demonstrates that the UFD enables the determination of the optimal fracture network conductivity and complexity that contribute to the maximum PI with a given proppant volume. The optimal SRV design is time-dependent until the PSS is reached. The surrogate model not only improves the computational efficiency but also delivers high precision, which means far less computational burden than the traditional parametric-sensitivity analysis.

Deep-learning-assisted design of multi-degree-of-freedom metamaterial absorber

In recent years, the development of metamaterial absorbers (MAs) based on deep learning methods has become a popular research topic. Nevertheless, conventional analytical approaches have limitations, leading to MAs characterized by low-degree-of-freedom cell structures, thereby restricting their absorption characteristics. In this study, we propose a multi-degree-of-freedom MA with multiple variable parameters in terms of structures, materials, and number of layers, which are distinguished by coding. Additionally, deep learning models, including a fully connected neural network, recurrent neural network, AlexNet, and residual neural network, are designed to predict structures and absorption spectra. We compare the prediction accuracy of these four neural network models and identify the optimal network model for spectral prediction and on-demand design. Using trained neural network models, we successfully designed broadband, dual-band, and single-band MAs in on-demand design. Compared with other design methods, this method provides a greater degree of freedom in device design, which allows the design of MAs to have more absorption characteristics