Neural Network Modeling Method Research Articles

Research on Chebyshev Graph Convolutional Neural Network Modeling Method for Rotating Equipment Fault Diagnosis under Variable Working Conditions

Research on Chebyshev Graph Convolutional Neural Network Modeling Method for Rotating Equipment Fault Diagnosis under Variable Working Conditions

Искусственный интеллект для планирования тактики ведения пациентов после перкутанной нефролитотрипсии.

Introduction. Urolithiasis remains one of the most common urological pathologies that deserves due attention in the healthcare system. A current direction in optimizing percutaneous nephrolithotripsy (PСNL) is a personalized prognosis for patient management based on artificial intelligence (AI) decisionmaking algorithms. In our work, we applied an algorithm to create the most optimal tactics for patient management after PCNL. Purpose. Optimize the management of patients after PCNL based on AI. Materials and methods. The total number of patients treated at N. Lopatkin Scientific Research Institute of Urology and Interventional Radiology andthose who took part in the study amounted to 1000 people. The number of men was 419 (41.9%), women 581 (58.1%). The age of patients included in the studies ranged from 18 to 88 years (52.3±13.47). The division into Training and Holdout partitions varied depending on the target variable to obtain the most accurate result. The maximum difference in the samples was 600n and 400n (60% and 40%), the minimum 800n and 200n (80% and 20%). Data processing was carried out using IBM SPSS Statistics and Modeler programs using the neural network modeling method. Results. The accuracy of prediction for the choice of nephroscope size was 82.2%, requirement for intraoperative stent placement 93.9%, requirement for ureteroscopy or contact ureterolithotripsy 98.5%. The accuracy of the forecast for the number of puncture accesses was 92.6%, for access through the upper group 95%, for access through the middle group 91.2%, for access through the lower group 91.2%. The algorithm allows us to predict the presence of residual stones with an accuracy of 84.1%, the duration of surgical intervention is 87.3% Conclusion. Further development of AI technology will not only provide a «second opinion» and reduce the time for processing large volumes of information, but will also standardize the approach to the treatment of complex clinical cases and help optimize the work of the urology department.

Development of an Economic Cost Model for Gold and Associated Minerals Using Economic Analysis and Artificial Intelligence Approach

Limited accuracy due to complex geological processes, insufficient data granularity, and challenges in predicting market dynamics impact mineral forecast models. In order to develop a cost-based model for gold and related minerals in Birnin Gwari (Kaduna State) and Kagara (Niger State), this study is being carried out in Nigeria. The created concept aims to increase the profitability and economic viability of open-pit mines operated by craftsmen and mining investors. By maximizing mine life cycles and preventing the early closure of mining sites, this was intended to encourage the best recovery rates. This study used economic and artificial neural network (ANN) modeling methods to create a cost-based model for ten gold-associated minerals. The coefficient of correlation (R2), root mean square error (RSME), and variance accounting factor (VAF) were used to assess the created model's correctness. The model evaluator demonstrates that the proposed cost-based models are economically viable and the quantitative results are as presented in Equations (14-23). The cost-benefit connection between gold and the ten related minerals was also determined using a least-squares regression analysis. According to geochemical results, the average values of gold and related minerals in the study region range from 4.87 g/t of gold (Au), 1.501 g/t of copper (Cu), and 187.13 g/t of iron (Fe) in Birni Gwari to 2.29 g/t of gold (Au), 1.358 g/t of copper (Cu), and 173.75 g/t of iron (Fe) in Kagara. In Birnin Gwari, the equivalent financial estimations are N876,600 for Au, N60,040,000 for Cu, and N1,129,375 for Fe, and N5,780,000 for Au, N49,500 for Cu, and N363,060,000 for Fe in Kagara. The model's scope demonstrates that running a single mineral can result in low recovery rates and a loss of profit margin; thus, the model's application to all open pit mines is relevant as a reference for profit margin cash flow.

Presentation of artificial neural network models based on optimum theories for predicting accident severity on rural roads in Iran

This study focuses on predicting the number and severity of rural accidents using nine accident-related variables by the multi-layer perceptron (MLP) approach as a type of artificial neural network modeling method. The models were developed using the input parameters of shoulder width, roadway width, roadside hazard, access density, passing zone ratio, speed limit, pavement condition index, shoulder rumble strips, and centerline rumble strips. MLP models were created based on accident data that occurred from 2019 to 2020 on the Tehran–Qom and Tehran-Saveh rural roads in Iran. In order to achieve the highest accuracy, twenty MLP models have been built with various structures. The MLP model, consisting of a multilayer feedforward network with hidden sigmoid and softmax output, has been used in this study. In this regard, an attempt was made to present an optimum theory to select the best model. The most accurate model was chosen based on the R-value and root mean square error (RMSE), mean absolute error (MAE), f, SD, and R2. Results indicated that the R-value obtained from the optimum model was 0.912, representing the accurate performance of the selected model. In addition, access density, roadside hazard, and roadway width were identified as the most significant variables.

Merging Counter-Propagation and Back-Propagation Algorithms: Overcoming the Limitations of Counter-Propagation Neural Network Models.

Artificial neural networks (ANNs) are nowadays applied as the most efficient methods in the majority of machine learning approaches, including data-driven modeling for assessment of the toxicity of chemicals. We developed a combined neural network methodology that can be used in the scope of new approach methodologies (NAMs) assessing chemical or drug toxicity. Here, we present QSAR models for predicting the physical and biochemical properties of molecules of three different datasets: aqueous solubility, acute fish toxicity toward fat head minnow, and bio-concentration factors. A novel neural network modeling method is developed by combining two neural network algorithms, namely, the counter-propagation modeling strategy (CP-ANN) with the back-propagation-of-errors algorithm (BPE-ANN). The advantage is a short training time, robustness, and good interpretability through the initial CP-ANN part, while the extension with BPE-ANN improves the precision of predictions in the range between minimal and maximal property values of the training data, regardless of the number of neurons in both neural networks, either CP-ANN or BPE-ANN.

Искусственный интеллект для персонализированного подхода к перкутанной нефролитотрипсии.

Introduction. In recent years, there has been more and more research on the creation of artificial intelligence (AI) to help diagnose and treat patients in various fields of medicine. The use of algorithms allows us to reduce the time for processing large volumes of information and give a «second opinion» to a specialist in complex and non-standard clinical cases. In order to increase the effectiveness of percutaneous nephrolithotripsy, we have created an AI algorithm to personalize approach in the surgical treatment of nephrolithiasis. Purpose. To develop a universal tool for predicting the most optimal tactics of percutaneous nephrolithotripsy depending on the clinical case. Materials and methods. The total number of patients treated at N. Lopatkin Scientific Research Institute of Urology and Interventional Radiology and those who took part in the study amounted to 1000 people. The number of men was 419 (41.9%), women 581 (58.1%). The age of patients included in the studies ranged from 18 to 88 years (52.3±13.47). The division into Training and Holdout partitions varied depending on the target variable to obtain the most accurate result. The maximum difference in the samples was 600n and 400n (60% and 40%), the minimum 800n and 200n (80% and 20%). Data processing was carried out using IBM SPSS Statistics and Modeler programs using the neural network modeling method. Results. The accuracy of prediction for the choice of nephroscope size was 82.2%, requirement for intraoperative stent placement 93.9%, requirement for ureteroscopy or contact ureterolithotripsy 98.5%. The accuracy of the forecast for the number of puncture accesses was 92.6%, for access through the upper group 95%, for access through the middle group 91.2%, for access through the lower group 91.2%. The algorithm allows us to predict the presence of residual stones with an accuracy of 84.1%, the duration of surgical intervention is 87.3%.The number of test observations was 30. The algorithm demonstrated high accuracy of prediction (from 73.8% to 94.5%) of surgical tactics in complex clinical cases. Conclusion. The results of using the algorithm for personalized prognosis in the surgical treatment of nephrolithiasis showed high accuracy and efficiency. Further development of technology to create algorithms for each nosology and surgical interventions will reduce the time of data analysis by the attending physician, provide a «second opinion» to young specialists and standardize the approach to treatment in rare and complex clinical cases.

A calibration and prediction method of the camera imaging parameters in variable temperature environment

In the on-orbit application of photogrammetry, constraints on thermal control resources or aging of thermal control equipment will trigger increased fluctuations in the camera operating environment temperature. Consequently, the parameters of the camera imaging model vary, such that the measurement accuracy is reduced. Accordingly, the accurate prediction of the camera imaging parameters (e.g., Camera Interior Parameters (CIP) and Lens Distortion Parameters (LDP)) at variable temperatures takes on critical significance.First, a novel method, termed as Virtual Multi-station Camera Calibration (VMCC), is proposed to calibrate the CIP and LDP with each frame at variable temperatures. Second, the Variational Mode Decomposition method combined with the Northern Goshawk Optimizer (NGO-VMD) is employed to decompose the camera imaging parameter signals obtained from VMCC, yielding the n Intrinsic Mode Functions (IMFs) contained in the respective parameter signal. Third, the Generalized Regression Neural Network combined with the Dung Beetle Optimizer (DBO-GRNN) is employed, such that the smoothness parameter q of the GRNN is iteratively optimized based on the camera imaging parameter prediction error. Lastly, the CIP and the LDP are accurately predicted at different temperatures in accordance with the temperature data from multiple parts of the camera.Based on this study, a reliable dataset of camera imaging parameter responses at variable temperatures can be generated for the on-orbit photogrammetry system. Furthermore, the prediction accuracy and immense application potential of neural network modeling methods are demonstrated in camera parameter prediction.

Automatic sign language translation system using neural network technologies and 3D animation

Implementation of automatic sign language translation software in the process of social inclusion of people with hearing impairment is an important task. Social inclusion for people with hearing disabilities is an acute problem that must be solved in the context of the development of IT technologies and legislative initiatives that ensure the rights of people with disabilities and their equal opportunities. This substantiates the relevance of the research of assistive technologies, in the context of software tools, such as the process of social inclusion of people with severe hearing impairment in society. The subject of research is methods of automated sign language translation using intelligent technologies. The purpose of the work is the development and research of sign language automation methods to improve the quality of life of people with hearing impairments in accordance with the "Goals of Sustainable Development of Ukraine" (in the "Reduction of Inequality" part). The main tasks of the research are the development and testing of methods of converting sign language into text, converting text into sign language, as well as automating translation from one sign language to another sign language using modern intelligent technologies. Neural network modeling and 3D animation methods were used to solve these problems. The following results were obtained in the work: the main problems and tasks of social inclusion for people with hearing impairments were identified; a comparative analysis of modern methods and software platforms of automatic sign language translation was carried out; a system combining the SL-to-Text method is proposed and investigated; the Text-to-SL method using 3D animation to generate sign language concepts; the method of generating a 3D-animated gesture from video recordings; method of implementing the Sign Language1 to Sign Language2 technology. For gesture recognition, a convolutional neural network model is used, which is trained using imported and system-generated datasets of video gestures. The trained model has a high recognition accuracy (98.52%). The creation of a 3D model for displaying the gesture on the screen and its processing took place in the Unity 3D environment. The structure of the project, executive and auxiliary files used to build 3D animation for the generation of sign language concepts includes: event handler files; display results according to which they carry information about the position of the tracked points of the body; files that store the characteristics of materials that have been added to certain body mapping points. Conclusions: the proposed methods of automated translation have practical significance, which is confirmed by the demo versions of the software applications "Sign Language to Text" and "Text to Sign Language". A promising direction for continuing research on the topic of the work is the improvement of SL1-to-SL2 methods, the creation of open datasets of video gestures, the joining of scientists and developers to fill dictionaries with concepts of various sign languages.

The path of management of dispute cases of legal issues of webcasting bandwagon industry in the information age

Abstract In this paper, for the common live streaming with cargo dispute cases in the context of the information technology era, the live streaming with cargo industry dispute case retrieval algorithm based on case-based reasoning calculates the attribute similarity measure and weights of the dispute cases in order to realize the live streaming with cargo industry dispute case retrieval. Three thousand judicial documents with live streaming with cargo disputes as the cause of the case were crawled from the China Judicial Documents Network as the initial data for the study and the method of BIOES was used to annotate the initial data and formulate eight basic legal elements of live streaming with cargo dispute cases. The form of extracting legal elements of live streaming with goods industry is transformed into a multi-label text classification task, and the neural network modeling method combined with the attention mechanism is used to design the extraction model of legal elements of the live streaming with goods dispute cases, and then the live streaming with goods dispute cases on the Internet are analyzed by examples. The results show that the Acc value of this paper’s model is 0.973, the Loss value is 0.051, and the F1 value is 0.981, and on the whole, this paper’s model performs better compared with the other five models, i.e., it shows that this paper’s model can be more intelligent and accurate to collect the case element information under the field of live streaming with goods disputes. This study aims to contribute to the development of a new sales model of live streaming with goods benignly.

Modeling of nonlinear characteristics of switched reluctance motor based on optimized RBF neuron network

The strong nonlinearity of switched reluctance motor (SRM) makes it difficult to establish an accurate mathematical model of its electromagnetic characteristics. In this paper, an artificial neural network modeling method is adopted to estimate the nonlinear flux linkage characteristics of a four-phase SRM. Based on the samples obtained by experimental measurement, a radial basis function neural network is trained and further optimized using the K-Means algorithm. The modeling accuracy of the optimized flux linkage model is analyzed in detail and also verified by dynamic simulation of a direct instantaneous torque control system of the motor.

Application of neural network modeling methods in solving initial boundary value problems for partial differential equations

Machine learning allows you to solve a variety of data analysis problems, but its use for solving differential equations has appeared relatively recently. The approximation of the solution of the boundary value problem for differential equations (ordinary and partial derivatives) is constructed using neural network functions. The selection of weighting coefficients is carried out during the training of the neural network. The criteria for the quality of training in this case are inconsistencies in the equation and boundary-initial conditions. This approach makes it possible, instead of grid solutions, to find solutions defined on the entire feasible region of the boundary value problem. Specific examples show the features of the application of physics-informed neural networks to the solution of boundary value problems for differential equations of various types. Physics-informed neural networks training methods can be used in the tasks of retraining intelligent control systems on incomplete sets of input data.

Research on the construction method of neural network model based on memristors

Abstract With the continuous development of neural networks, the hardware requirements are getting higher and higher, and the emergence of memristors is expected to optimize this challenge. In this paper, by studying the inter-mapping relationship between memristor arrays and neural networks, a memristor-based convolutional neural network circuit module is designed, a binarization method is used to quantize the neural network, an acceleration module is constructed based on multiple computational arrays, and an additive tree is used to accumulate the intermediate results of the output values of multiple computational units. Simulation experiments were conducted with the help of simulation software to compare and analyze the memristor-based CNN with other neural networks. Compared with LSTM, RNN, and ELM, the accuracy of the memristor-based CNN is 4.17%, 7.48%, and 2.01% higher compared to other neural networks, respectively. In the performance analysis, the recognition rate of the memristor CNN is almost unaffected by the programming error and still achieves a recognition rate close to 98% in the best case. This study provides a new idea for implementing and applying convolutional neural networks in memristor arrays, which is expected to promote the further development of memristor neuromorphic computing.

Hydra: An ensemble deep learning recognition model for plant diseases

Indian agriculture contributes about 17% to the total GDP and employs over 60% of the population. Due to the subjectiveness and time-consuming nature of manual approaches to disease detection, large or small-scale farming faces many challenges. Cotton diseases such as Areolate, Mildew, Sore Shin, Fusarium, Wilt, and Myrothecium hugely affect crop yield. The Hydra framework-based ensemble deep learning model proposes to conduct symptom-wise recognition for cotton disease. Hydra Framework is an ensemble method of Convolution Neural Network (CNN) and VGG16 model with SoftMax function and ReLU (Rectified Linear Unit) that improves the performance of CNN by removing the negative values obtained from each layer of feature extraction. A real-time field study from Thadikombu village, Dindigul District, Tamil Nadu, is considered. The data augmentation technique is used to overcome the overfitting issue, thereby enlarging the available dataset. With 15,600 images containing healthy and diseased leaves, the ensemble Hydra model achieved an accuracy of 95% for recognizing diseases and was validated with pictures collected from Thadikombu village, Dindigul District, Tamilnadu. A comparison is made with the results obtained from the proposed Ensemble Hydra model with CNN and fine-tuned VGG model. Results showed Ensemble Hydra model was a helpful tool for recognizing cotton diseases.

Hybrid mechanistic and neural network modeling of nuclear reactors

An accurate and efficient model is the foundation for dynamic characteristics analysis and control design of nuclear reactors. Nevertheless, the widely-used mechanistic models of nuclear reactors inevitably have some deviations from the actual reactors due to inaccurate model parameters and model assumptions. This paper proposes a hybrid mechanistic and neural network modeling method for nuclear reactors to eliminate or minimizing these deviations. The high-efficiency point reactor kinetics model was adopted as the mechanistic reactor core model, and the simulation results obtained using the commercial program RELAP5 were taken as the reference operational data, based on which the reactivity feedback in the mechanistic model that is difficult to be accurately determined during reactor operations was calibrated using a genetic algorithm optimized BP neural network. Dynamic simulation results of a Generation III large pressured water reactor under three typical reactivity change transients show that the core power and outlet coolant temperature obtained with the hybrid model match much better with the reference results than those of the mechanistic model, demonstrating that the hybrid modeling method can improve the accuracy of the mechanistic reactor model effectively. This study can provide a valuable reference for accurate and efficient modeling of complex systems.

The Use of Neural Network Modeling Methods to Determine Regional Threshold Values of Hydrochemical Indicators in the Environmental Monitoring System of Waterbodies.

The regulation of the anthropogenic load on waterbodies is carried out based on water quality standards that are determined using the threshold values of hydrochemical indicators. These applied standards should be defined both geographically and differentially, taking into account the regional specifics of the formation of surface water compositions. However, there is currently no unified approach to defining these regional standards. It is, therefore. appropriate to develop regional water quality standards utilizing modern technologies for the mathematical purpose of methods analysis using both experimental data sources and information system technologies. As suggested by the use of sets of chemical analysis and neural network cluster analysis, both methods of analysis and an expert assessment could identify surface water types as well as define the official regional threshold values of hydrochemical system indicators, to improve the adequacy of assessments and ensure the mathematical justification of developed standards. The process for testing the proposed approach was carried out, using the surface water resource objects in the territory of the Republic of Tatarstan as our example, in addition to using the results of long-term systematic measurements of informative hydrochemical indicators. In the first stage, typing was performed on surface waters using the neural network clustering method. Clustering was performed based on sets of determined hydrochemical parameters in Kohonen's self-organizing neural network. To assess the uniformity of data, groups in each of the selected clusters were represented by specialists in this subject area's region. To determine the regional threshold values of hydrochemical indicators, statistical data for the corresponding clusters were calculated, and the ranges of these values were used. The results of testing this proposed approach allowed us to recommend it for identifying surface water types, as well as to define the threshold values of hydrochemical indicators in the territory of any region with different surface water compositions.

A data-driven multidimensional assessment model for English listening and speaking courses in higher education

Based on multiple assessment approach, this study used factor analysis and neural network modeling methods to build a data-driven multidimensional assessment model for English listening and speaking courses in higher education. We found that: (1) Peer assessment, student self-assessment, previous academic records, and teacher assessment were the four effective assessors of the multi-dimensional assessment of English listening and speaking courses; (2) The multidimensional assessment model based on the four effective assessors can predict the final academic performance of students in English listening and speaking courses, with previous academic records contributing the most, followed by peer assessment, teacher assessment, and student self-assessment. Therefore, a multidimensional assessment model for English listening and speaking courses in higher education was proposed: the academic performance of students (on a percentage basis) should be composed of 29% previous academic records, 28% peer assessment, 26% teacher assessment, and 17% student self-assessment. This model can guide teachers to intervene with students who need help in a timely manner, based on various assessors, thereby effectively improving their academic performance.

Research on fire distribution algorithm based on Hopfield neural network

the problem of air defense fire distribution in ECM command and control system is studied. Based on the mathematical model of fire distribution, the classical solution algorithm Hungarian method is introduced, and the shortcomings of Hungarian method in solving fire distribution are pointed out. According to the basic principle of Hopfield neural network model, the algorithm process of solving specific optimization problems is deduced, and the fire distribution problem is simulated. According to the simulation results, the application conclusion and improved method of Hopfield neural network model in fire distribution problem are given, which proves that the model can solve the fire distribution problem more effectively.

Systematic Evaluation of Deep Neural Network Based Dynamic Modeling Method for AC Power Electronic System

Since the high penetration of renewable energy complicates the dynamic characteristics of the AC power electronic system (ACPES), it is essential to establish an accurate dynamic model to obtain its dynamic behavior for ensure the safe and stable operation of the system. However, due to the no or limited internal control details, the state-space modeling method cannot be realized. It leads to the ACPES system becoming a black-box dynamic system. The dynamic modeling method based on deep neural network can simulate the dynamic behavior using port data without obtaining internal control details. However, deep neural network modeling methods are rarely systematically evaluated. In practice, the construction of neural network faces the selection of massive data and various network structure parameters. However, different sample distributions make the trained network performance quite different. Different network structure hyperparameters also mean different convergence time. Due to the lack of systematic evaluation and targeted suggestions, neural network modeling with high precision and high training speed cannot be realized quickly and conveniently in practical engineering applications. To fill this gap, this paper systematically evaluates the deep neural network from sample distribution and structural hyperparameter selection. The influence on modeling accuracy is analyzed in detail, then some modeling suggestions are presented. Simulation results under multiple operating points verify the effectiveness of the proposed method.

ВИКОРИСТАННЯ ПРОГНОЗУВАННЯ ЗА ДОПОМОГОЮ ШТУЧНИХ НЕЙРОННИХ МЕРЕЖ ПРИ ВИВЧЕННІ КОНКУРЕНТОСПРОМОЖНОСТІ ПРИМОРСЬКИХ РЕГІОНІВ

The use of the mechanism of regional competitiveness enhancement is to maintain and strengthen the region's resource, intellectual, social and human capital. The article is devoted to the study of the possibilities of using artificial neural networks for forecasting indicators of competitiveness, which have a direct impact on subsystems and elements of competitiveness of regions. The practical application of the method of learning artificial neural networks allows you to predict the scenarios of the development of the indicator, confidence intervals, and the amount of the forecasting error. The analysis data can be used in planning the strategic development of the region, studying trends in indicators. The use of software greatly facilitates and simplifies the process of calculations and allows you to obtain data in graphical and calculated form, which is a necessary condition for the interpretation and analysis of data by economists. Forecasting regional competitiveness indicators with the help of neural networks allows testing, obtaining several forecast results and evaluating the forecast results by calculating the average error. Comparison of the obtained values of the calculated average error allows you to choose the forecasting result that is the most acceptable and possible for further use. Taking into account the values of forecast indicators of the region's competitiveness is necessary when planning the region's development strategy, when choosing strategic priorities and initiatives, emphasis is placed on achieving the forecast values of the «markers» of the strategy. Using the example of the indicator of the gross regional product of the region, calculations of the forecast of changes in the indicator were carried out using the Statistica software by the method of artificial neural network modeling. The gross regional product is an indicator chosen for analysis due to certain of its strong positive aspects: universality of the comparative base, availability of statistical data, economic sense. The study of competitiveness requires an understanding of the prospects for the development of its indicators, since their management affects the overall indicator. Competitiveness forecasting makes it possible to create a prognostic scenario of the development of both competitiveness factors, its components, and the object of research in general. However, the use of forecasting tools requires knowledge of certain nuances that accompany forecasting: peculiarities of using software tools, checking the forecast by calculating the standard error, etc.

Modeling and implementation of a novel active voltage balancing circuit using deep recurrent neural network with dropout regularization

SummaryRecurrent neural networks (RNN) emerged as powerful tools to model and analyze the nonlinear behavior of electronic circuits accurately and quickly. Efforts to improve the accuracy of RNN will lead to the design of better‐quality products, which is essential in various fields such as the design of energy harvesting (EH) systems. EH techniques can provide the electrical energy needed for low‐power electronics without the need for a battery or with minimal dependency. Due to the importance of the active voltage balancing circuit in EH systems, we have proposed a new macromodeling method called dropout local‐feedback deep recurrent neural network (Dropout‐LFDRNN) to model and analyze this circuit along with two other nonlinear circuits as examples. This technique is an advance over the LFDRNN macromodeling method, and based on the obtained results from the measurements, we were able to build a fast macromodel for the active balancing circuit, which outperforms conventional deep recurrent neural network modeling method in terms of accuracy without sacrificing speed. It is worth mentioning that this proposed technique in this paper can be considered a viable approach for modeling and analysis of nonlinear electronic components and circuits. In addition to the advantage of generating a more accurate model, the model based on the Dropout‐LFDRNN approach is much faster than the existing transistor‐level models.