BazTech - Yadda

1

Artificial intelligence-powered pulse sequences in nuclear magnetic resonance and magnetic resonance imaging: historical trends, current innovations and perspectives

Tokarz Paweł

Scientiae Radices

|

2024

|

Vol. 3, Iss. 1

30--52

EN

This review article explores the historical background and recent advances in the application of artificial intelligence (AI) in the development of radiofrequency pulses and pulse sequences in nuclear magnetic resonance spectroscopy (NMR) and imaging (MRI). The introduction of AI into this field, which traces back to the late 1970s, has recently witnessed remarkable progress, leading to the design of specialized frameworks and software solutions such as DeepRF, MRzero, and GENETICS-AI. Through an analysis of literature and case studies, this review tracks the transformation of AI-driven pulse design from initial proof-of-concept studies to comprehensive scientific programs, shedding light on the potential implications for the broader NMR and MRI communities. The fusion of artificial intelligence and magnetic resonance pulse design stands as a promising frontier in spectroscopy and imaging, offering innovative enhancements in data acquisition, analysis, and interpretation across diverse scientific domains.

2

Hybrid wavelet transform - MLR and ANN models for river flow prediction: Case study of Brahmaputra river (Pancharatna station)

Khandekar Sachin Dadu, Aswar Dinesh Shrikrishna, Sabale Pandurang Digamber, Khandekar Varsha Sachin, Bajad Mohankumar Namdeorao, Khaple Shivakumar

Przegląd Naukowy Inżynieria i Kształtowanie Środowiska

|

2024

|

Vol. 33, No. 1

69--94

EN

In this research, discrete wavelet transform (DWT) is combined with MLR and ANN to develop WMLR and WANN hybrid models, respectively, for the Brahmaputra river (Pancharatna station) flow forecasting. Daily flow data for the period of 10 year were decomposed (up to fifth level) into detailed and approximation coefficients (using Daubechies wavelets db1, db2, db3, db8 and db10) which were fed as input to MLR and ANN to get the predicted discharge values two days, four days, seven days and 14 days ahead. For all lead times, the WMLR-db10 model was found to be superior as compared to WANN-db1, WANN-db2, WANN-db3, WANN-db8, WMLR-db1, WMLR-db2, WMLR-db3, WMLR-db8 and single MLR and ANN models. During testing period, the values of determination coefficient (R2) and RMSE for WMLR-db10 model for two-, four-, seven- and 14-day lead time were found to be, respectively, 0.996 (751.87 m3·s–1), 0.991 (1,174.80 m3·s–1), 0.984 (1,585.02 m3·s–1), and 0.968 (2,196.46 m3·s–1). Also, it was observed that for lower order wavelets (db1, db2, db3) WANN’s performance was better, and for higher order wavelets (db8, db10) WMLR’s performance was better. Correspondingly, it was observed that all hybrid models’ efficiency increased with increase in the decomposition level.

3

Predicting open-pit mine production using machine learning techniques

Kumah Faustin Nartey, Alex Kwasi Saim, Oppong Millicent Nkrumah, Arthur Clement Kweku

Journal of Sustainable Mining

|

2024

|

Vol. 23, iss. 2

118--131

EN

In mining, where production is affected by several factors, including equipment availability, it is necessary to develop reliable models to accurately predict mine production to improve operational efficiency. Hence, in this study, four (4) machine learning algorithms - namely: artificial neural network (ANN), random forest (RF), gradient boosting regression (GBR) and decision tree (DT)) - were implemented to predict mine production. Multiple Linear Regression (MLR) analysis was used as a baseline study for comparison purposes. In that regard, one hundred and twenty-six (126) datasets from an open-pit gold mine were used. The developed models were evaluated and compared using the correlation coefficient (R2), mean absolute percentage error (MAPE) and variance accounted for (VAF). It has been shown in this study that the ANN model can best estimate open-pit mine production by comparing its performance to that of the other machine learning models. The R2, MAPE, RMSE and VAF of the models were 0.8003, 0.7486, 0.7519, 0.6538, 0.6044, 4.23%, 5.07%, 5.44%, 6.31%, 6.15% and 79.66%, 74.69%, 74.10%, 65.16% and 60.11% for ANN, RF, GBR, DT and MLR, respectively. Overall, this study has shown that machine learning algorithms predict mine production with higher accuracy.

4

Modeling of Dispersed Red 17 Dye Removal from an Aqueous Solution Using Artificial Neural Network

Ibrahim Abdullah I., Asmel Nabel K., Alabdraba Waleed M. S., Al-Nima Raid R. O.

Journal of Ecological Engineering

|

2024

|

Vol. 25, nr 2

10--19

EN

A significant amount of hazardous compounds has leaked into the environment due to the widespread usage of organic dyes, and it is essential that these dangerous contaminants be removed in a sustainable way. This study used varying amounts of H2O2 (0, 0.5, 1.5, 3, and 5) mM/L to extract the dye from the aqueous solution. Furthermore, concentrations of 0.4, 1, 1.7, and 2.3 mM/L of Fe+2 as FeSO4•7H2O were also utilized. Batch Advanced Oxidation Process (AOP) was carried out under various working conditions, including: contact time (5–60 min), mixing speed (100–300 rpm), and UV light intensity (0–40 W). Utilizing experimental data, the AOP efficiency of Dispersed Red 17 Dye was calculated. Genetic Cascade-forward Neural Network (GCNN) was employed as a machine-learning tool to forecast the oxidation efficiency and the amount of dye that would be removed from the aqueous solution, specifically Dispersed Red 17. When compared to experimental data, the best model had an R2 correlation value of 0.955. The findings of the importance analysis showed that the studied parameters affected the discoloration efficiency with order of: H2O2, UV, Fe+2, mixing speed, and contact time. The obtained results demonstrated the effectiveness of GCNN as a novel approach in forecasting the AOP efficiency of Dispersed Red 17 Dye.

5

Estimation of Photovoltaic Module Performance with L-Shaped Aluminum Fins Using Weather Data

Hamdan Mohammad, Abdelhafez Eman, Musa Akram, Ajib Salman

Journal of Ecological Engineering

|

2024

|

Vol. 25, nr 1

336--344

EN

Photovoltaic (PV) power prediction is vital for efficient and effective solar energy utilization within the energy ecosystem. It enables grid stability, cost savings, and the seamless integration of solar power into the broader energy infrastructure. In this work, previously obtained data on the estimation of the power produced by a PV, which is cooled by L-shaped aluminum fins attached to the backside of the PV at different spacings, is used to predict the power produced by the PV. This is achieved by employing both neural network models and multiple linear regression (MLR) techniques to assess the correlation between power generated by PV with L-shaped aluminum fins and its input variables. Two distinct approaches were employed for this purpose. The first approach involved the conventional MLR model, while the second utilized a neural network, specifically the multilayer perceptron (MLP) model. The estimated outcomes were subsequently compared against the previously measured data. The MLP model showed a great ability to identify the relationship between input and output variables, it was noted. The statistical error study provided evidence of data mining’s acceptable accuracy when using the MLP model. Conversely, the results indicated that the MLR technique exhibited the least ability to estimate the power generated by PV with L-shaped aluminum fins.

6

Detection and identification of fault in transmission lines based on ANN

Abdullah Ahmed G., Ibrahim Mohammad A., Ibrahim Waseem Kh.

Diagnostyka

|

2024

|

Vol. 25, No. 2

art. no. 2024206

EN

Along with power transmission lines' efficiency, another crucial factor in electrical power transmission networks is reliability, which guarantees power transmission stability. One of the crucial and essential tasks for maintaining the continuity and stability of power transmission in transmission networks Capacity without any significant failures is identifying errors and malfunctions in power transmission lines as soon as possible. The goal of this article is to develop and apply ANN technology to overcome the obstacles faced by the electrical power transmission network. In order for the ANN to learn useful patterns and features from raw current measurements, pre-processing and feature extraction techniques are used during the training process. Real-time applications can benefit from the ANN's architecture, which is optimized for high accuracy, quick response times, and scalability. To validate the performance of the ANN-based fault detection system, extensive simulations are conducted using data from different transmission line scenarios, including various fault types that short-circuit. The results demonstrate the capability of the ANN model to accurately detect and classify faults, as well as disconnect the power grid after detect any fault. The results showed the accuracy and high speed of the proposed method using a neural network compared to traditional methods.

7

Application of artificial neural networks for the prediction of the service conditions of an elastohydrodynamic EHL contact in the presence of solid pollutant

Mattallah Sabrina, Kelaiaia Ridha, Louahem M’Sabah Hanane, Kerboua Adlen

Diagnostyka

|

2024

|

Vol. 25, No. 1

art. no. 2024107

EN

Lubricated mechanical mechanisms operate under service conditions influenced by several environmental parameters, and their life times may be threatened due to inappropriate use or by the presence of solid contaminants. The objective of this work is to study the effect of three operating parameters, namely: rotational speed 𝑉, load 𝑄 and kinematic viscosity 𝜈 in the presence of three sizes of solid contaminants 𝑇, on the degradation of an EHL contact, to predict the ranges of effects that may lead to the damage of the contacting surfaces. In our investigation, anexperimental design of nine trials is used to combine four factors with three levels each to accomplish the experimental investigation. Artificial neural network regression and the desirability function were used for the interpretation and modelling of the responses, whichare: wear 𝑊, arithmetic mean height 𝑅𝑎, total profile height 𝑅𝑡 and maximum profile height 𝑅𝑧. From these methods we observed that the sand grain sizes have a significant impact on the wear 𝑊 and the roughness 𝑅𝑎, but that viscosity has the primary influence on the variation of the roughnesses 𝑅𝑡 and 𝑅𝑧. We also found that the quality of the predicted models is very good, with overall determination coefficients of 𝑅2 learning = 0.9985 and 𝑅2 validation = 0.9996. Several levels of degradation depending on the operating conditions are predicted using the desirability function.

8

A novel artificial intelligence-based hybrid system to improve breast cancer detection using DCE-MRI

Akgül İsmail, Kaya Volkan, Karavaş Erdal, Aydin Sonay, Baran Ahmet

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2024

|

Vol. 72, nr 3

art. no. e149172

EN

The interpretation of breast magnetic resonance imaging (MRI) in the healthcare field depends on the good knowledge and experience of radiologists. Recent developments in artificial intelligence (AI) have shown advances in the field of radiology. However, the desired levels have not been reached in the field of radiology yet. In this study, a novel model structure is proposed to characterize the diagnostic performance of AI technology for individual breast dynamic contrast material–enhanced (DCE) MRI sequences. In the proposed model structure, Inception-v3, EfficientNet-B3, and DenseNet-201 models were used as hybrids together with the Yolo-v3 algorithm to detect breast and cancer regions. In the proposed model, DCE-MRI sequences (T2, ADC, Diffusion, Non-Contrast Fat Non-Suppressed T1, Non-Contrast Fat Suppressed T1, Contrast Fat Suppressed T1, and Subtraction T1) were evaluated separately and validation was made, thus providing a unique perspective. According to the validation results, the model structure with the best performance was determined as Yolo-v3 + DenseNet-201. With this model structure, 92.41% accuracy, 0.5936 loss, 92.44% sensitivity, and 92.44% specificity rates were obtained. In addition, it was determined that the results obtained without using contrast material in the best model were 91.53% accuracy, 0.9646 loss, 92.19% sensitivity, and 92.19% specificity. Therefore, it is predicted that the need for contrast material use can be reduced with the help of this model structure.

9

Prediction of Buckling Behaviour of Composite Plate Element Using Artificial Neural Networks

Falkowicz Katarzyna, Kulisz Monika

Advances in Science and Technology. Research Journal

|

2024

|

Vol. 18, no 1

231--243

EN

This article presents the use of Artificial Neural Networks (ANNs) to analysis of the composite plate elements with cut-outs which can work as a spring element. The analysis were based on results from numerical approach. ANNs models have been developed utilizing the obtained numerical data to predict the composite plate’s flexural-torsional form of buckling as natural form for different cut-outs and angels configurations. The ANNs models were trained and tested using a large dataset, and their accuracy is evaluated using various statistical measures. The developed ANNs models demonstrated high accuracy in predicting the critical force and buckling form of thin-walled plates with different cut-out and fiber angels configurations under compression. The combination of numerical analyses with ANNs models provides a practical and efficient solution for evaluating the stability behaviour of composite plates with cut-outs, which can be useful for design optimization and structural monitoring in engineering applications.

10

A comparative study of artificial neural networks and multivariate regression for predicting groundwater depths in the Arak aquifer

Hanifian Soadat, Khaleghi Mohammad Reza, Najarchi Mohsen, Jafarnia Reza, Varvani Javad

Acta Geophysica

|

2024

|

Vol. 72, no. 1

419--432

EN

In recent years, the groundwater resources of Arak plain have been under severe stress, so in some areas, due to the drying up of wells, the depth of wells has increased to access water. In some areas, the groundwater depth is high, which will lead to the salinization of those lands in the future. Regional modeling was used to organize and measure the response of the groundwater resources of Arak plain against the implementation of different management and implementation scenarios. This study aims to investigate the effective factors in the groundwater depth to provide a regional model with multiple linear regression (MLR) methods for Arak plain aquifer. For this purpose, the average groundwater potential maps (GPMs) in the Arak plain, as a dependent variable, and the transmissivity of the aquifer formations, groundwater exploitation values, altitude, average precipitation of the region, the amount of evaporation, and the distance from water resources are considered independent variables and regression analysis is done in SPSS software media. It was done to present a linear model. In the next stage, the presented model was evaluated by applying it to places where its statistics and information were not used to present the model, and finally, by applying this model in the GIS environment, the GPMs for the region were created. The study was prepared. Also, an artificial neural network (ANN) was used to simulate the depth of underground water. The performance of the ANN was measured through parameters such as root-mean-square error (RMSE) and correlation coefficient between real and desired outputs (R). The results of both methods indicate that factors such as the transmissivity of aquifer formations, GPMs drawdown, topography (the height of the well site on the level of the watershed), the groundwater exploitation values at the maximum operating radius of the well, and the distance from water resources are the main factors of GPMs drawdown. But the effectiveness of ANN in estimating GPMs drawdown is higher than the MLR method. The implemented methodology could be generalized to other watersheds with water scarcity problems for groundwater management.

11

Research and applications of artificial neural networks in spatial analysis: Review

Garczyńska Ilona

Zeszyty Naukowe Akademii Morskiej w Szczecinie

|

2023

|

nr 74 (146)

35--45

EN

The conducted review presents the possibility of using artificial neural networks in sectors related to environmental protection, agriculture, forestry, land uses, groundwater and bathymetric. Today there is a lot of research in these areas with different research methodologies. The result is the improvement of decision-making processes, design, and prediction of certain events that, with appropriate intervention, can prevent severe consequences for society. The review shows the capabilities to optimize and automate the processes of modeling urban and land dynamics. It examines the forecasts of assessment of the damage caused by natural phenomena. Detection of environmental changes via the analysis of certain time intervals and classification of objects on the basis of different images is presented. The practical aspects of this work include the ability to choose the correct artificial neural network model depending on the complexity of the problem. This factor is a novel element since previously reviewed articles did not encounter a study of the correlation between the chosen model or algorithm, depending on the use case or area of the problem. This article seeks to outline the reason for the interest in artificial intelligence. Its purpose is to find answers to the following questions: How can artificial neural networks be used for spatial analysis? What does the implementation of detailed algorithms depend on? It is proved that an artificial intelligence approach can be an effective and powerful tool in various domains where spatial aspects are important.

12

Modern method based on artificial intelligence for safe control in the marine environment

Mohamed-Seghir Gdynia Maritime University, Gdynia, Poland

TransNav : International Journal on Marine Navigation and Safety of Sea Transportation

|

2023

|

Vol. 17 No. 2

283--288

EN

The article presents an approach to formulating a ship control process model in order to solve the problem of determining a safe ship trajectory in collision situations. Fuzzy process properties are included in the model to bring it closer to reality, as in many situations the navigator makes a subjective decision. A special neural network was used to solve the presented problem. This artificial neural network is characterized by minimum and maximum operations when set. In order to confirm the correctness of the operation of the proposed algorithm, the results of the simulations obtained were presented and an discussion was conducted.

13

The modelling of layered rocks using a numerical homogenisation technique and an artificial neural network

Urbański Aleksander, Ligęza Szymon, Przecherski Piotr

Technical Transactions

|

2023

|

Vol. 120, iss. 1

art. no. e2023007

EN

A method of creating a constitutive model of layered rocks based on an artificial neural network (ANN) is reported in this work. The ANN gives an implicit constitutive function Ʃⁿ⁺¹= F( Ʃⁿ , ΔE), relating the new state of homogenized stresses Ʃⁿ⁺¹ with the old state Ʃⁿ and with the increment of homogenized strains ΔƩ. The first step is to repeatedly run a strain- controlled homogenisation on an uni-dimensional finite element model of a periodic cell with elastic-plastic models (Drucker-Prager) of the components. Paths are created in (Ʃ, E) space, from which, a set of patterns is formed to train the ANN. A description of how to prepare this data and a discussion on ANN training issues are presented. Finally, the procedure based on trained ANN is put into a finite-element code (ZSoil.PC) as a user-delivered constitutive function. The approach is verified by comparing the results of the developed model basing on ANN with a direct (single-scale) analysis, which showed acceptable accuracy.

14

Reduction of Numerical Model in Some Geotechnical Problems

Góral Artur, Lefik Marek, Wojciechowski Marek

Studia Geotechnica et Mechanica

|

2023

|

Vol. 45, nr s1

333--349

EN

The concept of equivalence of the realistic, initial reference model and the simplified, reduced model is proposed. In reduced models, the action of the soil on the structure is replaced by the action of a layer with prescribed properties, defined by a set of parameters. The main difficulty here is to find the parameter values required by the simplified theory. The subject of this work is to find the dependence of the parameters of the reduced model on the parameters of the full model, including realistic soil behavior, in order to ensure the equivalence of both models. We show the potential of the method by presenting two examples: Winkler and Pasternak's model of a plate on the ground. We assume that both models are equivalent if they give identical results (displacements) at a finite number of observation points. An artificial neural network (ANN) is built in order to approximate and record the dependence of the parameters of the reduced model (at the network output) from the parameters of the full model (given at the network input). The complex network acts as a formula that assigns the parameters of the reduced model to a realistic description of the soil structure that is used for finite element method (FEM) modeling. The formalism we propose is quite general and can be applied to many engineering problems. The presented procedure is entirely numerical; it allows to calculate the parameters of the reduced model without resorting to symbolic calculations or additional theoretical considerations.

15

A Novel Method for Optimizing Parameters influencing the Bearing Capacity of Geosynthetic Reinforced Sand Using RSM, ANN, and Multiobjective Genetic Algorithm

Lafifi Brahim, Rouaiguia Ammar, Soltani El Alia

Studia Geotechnica et Mechanica

|

2023

|

Vol. 45, nr 2

174--196

EN

In this study, a novel method is proposed to optimize the reinforced parameters influencing the bearing capacity of a shallow square foundation resting on sandy soil reinforced with geosynthetic. The parameters to be optimized are reinforcement length (L), the number of reinforcement layers (N), the depth of the topmost layer of geosynthetic (U), and the vertical distance between two reinforcement layers (X). To achieve this objective, 25 laboratory small-scale model tests were conducted on reinforced sand. This laboratory-scale model has used two geosynthetics as reinforcement materials and one sandy soil. Firstly, the effect of reinforcement parameters on the bearing load was investigated using the analysis of variance (ANOVA). Both response surface methodology (RSM) and artificial neural networks (ANN) tools were applied and compared to model bearing capacity. Finally, the multiobjective genetic algorithm (MOGA) coupled with RSM and ANN models was used to solve multi objective optimization problems. The design of bearing capacity is considered a multi-objective optimization problem. In this regard, the two conflicting objectives are the need to maximize bearing capacity and minimize the cost. According to the obtained results, an informed decision regarding the design of the bearing capacity of reinforced sand is reached.

16

Performance enhancement of pv system using artificial neural network (ANN) as a maximum power point tracking

Bensiali Nadia, Benalia Nadia, Zerzouri Nora

Przegląd Elektrotechniczny

|

2023

|

R. 99, nr 4

262--265

EN

Since the generalization of the use of electricity, energy consumption, conversion and energy storage has led to research and development of new sources of supply. This interest has increased due to the depletion of fossil fuels, their impact on the environment and the waste they generate. Photovoltaic (PV) technology is an attractive solution as a replacement or complementary to conventional sources of electricity supply due to its many advantages. This work focuses on studying, modeling and designing of an advanced controller for a step up converter that uses an artificial neural network (ANN) as a maximum power point tracking (MPPT) algorithm to generate maximum power. The effectiveness was proved for climatic conditions variation. A comparative study between the proposed method and the conventional ones was carried out using Matlab/Simulink software.

PL

Od czasu uogólnienia wykorzystania energii elektrycznej jej zużycie, konwersja i magazynowanie energii doprowadziły do badań i rozwoju nowych źródeł zaopatrzenia. Zainteresowanie to wzrosło ze względu na wyczerpywanie się paliw kopalnych, ich wpływ na środowisko i generowane przez nie odpady. Technologia fotowoltaiczna (PV) jest atrakcyjnym rozwiązaniem jako zamiennik lub uzupełnienie konwencjonalnych źródeł zaopatrzenia w energię elektryczną ze względu na swoje liczne zalety. Praca ta koncentruje się na badaniu, modelowaniu i projektowaniu zaawansowanego sterownika do przetwornicy podwyższającej, która wykorzystuje sztuczną sieć neuronową (ANN) jako algorytm śledzenia punktu mocy maksymalnej (MPPT) do generowania maksymalnej mocy. Skuteczność została wykazana przy zmienności warunków klimatycznych. Badanie porównawcze proponowanej metody z metodami konwencjonalnymi przeprowadzono z wykorzystaniem oprogramowania Matlab/Simulink.

17

Investigation of the possibility of using artificial neural networks in the process of assessing the threat of lightning

Sobieska Emilia, Sobolewski Konrad

Przegląd Elektrotechniczny

|

2023

|

R. 99, nr 1

211--214

EN

The article presents issues related to the possibility of using an artificial neural network to classify factors describing the lightning hazard. It also analyzed the selection of the most optimal type of neural network for this type of task, as well asthe issue of input data, which both teach the neural network and form the basis for the final classification. Finally, the conclusions and assumptions necessary for the practical implementation of the task were drawn. The practical part of the task is planned as the following consideration stage.

PL

W artykule została przedstawiona problematyka związana z możliwością wykorzystania sztucznej sieci neuronowej do przeprowadzenia klasyfikacji współczynników opisujących zagrożenie piorunowe. Przeanalizowano wybór najbardziej optymalnego rodzaju sieci neuronowej dla tego rodzaju zadania, a także kwestię danych wejściowych, które posłużą zarówno do uczenia sieci neuronowej, jak również stanowią podstawę do końcowej klasyfikacji. Na koniec wyciągnięte zostały wnioski oraz założenia niezbędne do praktycznej realizacji zadania. Praktyczna część zadania będzie stanowić kolejny etap rozważań.

18

Comparative analysis of optimization value between artificial neural network and long short-term memory for prediction particulate matter (PM2.5) in Bangkok Thailand

Atiwanwong Thanpitcha, Jantakun Adirek, Sangsongfa Adisak

Przegląd Elektrotechniczny

|

2023

|

R. 99, nr 7

89--95

EN

The crisis that the general public is worried about is particulate matter as small as 2.5 microns, which is invisible to the naked eye, causing a great lack of awareness of health hazards. One of the key goals and visions of government leaders around the world is to tackle PM2.5 particulate matter, but without measurements, reports and predictions, how will it lead to emission reduction and remedial steps? Therefore, the prediction of PM2.5 is considered as the main factor that will help to reduce the pollution of PM2.5. So, Neural networks have been widely used in predictive research, but the problem is What type of neural network would be most suitable for predicting the value of PM2.5? In this research, the predictions were compared between Artificial Neural Network (ANN) and Long Short -Term Memory (LSTM) using values measured from the performance test results with accuracy. The results showed that when the values of similar hyperparameters were given different results, the average ANN accuracy is 91.1460%. The average accuracy LSTM is 96.8496%. The values obtained from the comparison clearly show that for the prediction of PM2.5, the LSTM neural network was significantly more suitable than the ANN neural network.

PL

Kryzys, który niepokoi opinię publiczną, to pył zawieszony o wielkości zaledwie 2,5 mikrona, który jest niewidoczny gołym okiem, powodując ogromny brak świadomości zagrożeń dla zdrowia. Jednym z kluczowych celów i wizji przywódców rządów na całym świecie jest rozwiązanie problemu pyłu zawieszonego PM2,5, ale bez pomiarów, raportów i prognoz, w jaki sposób doprowadzi to do redukcji emisji i działań zaradczych? Dlatego prognoza PM2,5 jest uważana za główny czynnik, który pomoże zmniejszyć zanieczyszczenie PM2,5. Tak więc sieci neuronowe były szeroko stosowane w badaniach predykcyjnych, ale problem polega na tym, jaki typ sieci neuronowej byłby najbardziej odpowiedni do przewidywania wartości PM2,5? W tym badaniu porównano przewidywania między sztuczną siecią neuronową (ANN) a pamięcią długokrótkoterminową (LSTM) przy użyciu wartości zmierzonych z wynikami testu wydajności z dużą dokładnością. Wyniki pokazały, że przy różnych wartościach podobnych hiperparametrów średnia dokładność ANN wynosi 91,1460%. Średnia dokładność LSTM wynosi 96,8496%. Uzyskane z porównania wartości jednoznacznie wskazują, że do predykcji PM2,5 sieć neuronowa LSTM okazała się znacznie bardziej odpowiednia niż sieć neuronowa ANN.

19

Autonomiczna kamera do wykrywania wolnych miejsc parkingowych

Lamża Aleksander

Przegląd Elektrotechniczny

|

2023

|

R. 99, nr 4

192--195

PL

W artykule została opisana architektura systemu autonomicznej kamery do wykrywania wolnych miejsc parkingowych. System został zrealizowany i przetestowany zarówno w warunkach laboratoryjnych, jak i rzeczywistych. Omówione zostały kluczowe elementy systemu, ich współdziałanie, a także możliwości modyfikacji i rozwoju.

EN

The article describes the architecture of an autonomous camera system to detect free parking spaces. The system was implemented and tested both in laboratory and real conditions. The key elements of the system, their interaction, as well as the possibilities of modification and development were discussed.

20

Neural Network-Based Optimisation of Sinusoidal PWM Controller for VSI-Driven BLDC Motor

Owusu George, Annan John Kojo, Nunoo Solomon

Power Electronics and Drives

|

2023

|

Vol. 8 (43)

275-298

EN

Although increasing the number of switches increases the switch losses, most designed controllers focus on controlling an inverter circuit with more than six switches. The paper aims to address this issue that arises in implementation of the voltage source inverter (VSI) for brushless DC (BLDC) motors. It optimises the sinusoidal pulse width modulation (PWM) controller, minimising total harmonic distortion (THD) while keeping the VSI’s circuit at six switches to avoid increased switching losses. This was achieved by applying an artificial neural network (ANN) to generate a signal, which combines with the already existing reference and carrier signals. The addition of the new signal to the existing signals contributed to generating more pulses compared with the conventional sinusoidal PWM. Simulink was used to design the system and analyse its performance with the conventional and neutral point clamped (NPC) VSI systems. Results indicated that the proposed system performs better when controlled with an LCC filter. Compared with the control experiments, its output waveform has the lowest THD value, which is 6.04%. The switching losses of all the systems were also computed. Results from the computation indicated that the proposed system is capable of reducing the switching losses by 0.6 kW compared with the NPC VSI brushless DC motor (BLDCM) system. BLDCM speed was tested across various conditions; the results are reported in Section 5.