Browse

Journals By Subject

Life Sciences, Agriculture & Food
Chemistry
Medicine & Health
Materials Science
Mathematics & Physics
Electrical & Computer Science
Earth, Energy & Environment
Architecture & Civil Engineering
Education
Economics & Management
Humanities & Social Sciences
Multidisciplinary

Books

Publish Conference Abstract Books
Upcoming Conference Abstract Books
Published Conference Abstract Books
Publish Your Books
Upcoming Books
Published Books

Proceedings

Events

Events
Five Types of Conference Publications

Join

Join as Editorial Board Member
Become a Reviewer

Information

For Authors
For Reviewers
For Editors
For Conference Organizers
For Librarians
Article Processing Charges
Special Issue Guidelines
Editorial Process
Manuscript Transfers
Peer Review at SciencePG
Open Access
Copyright and License
Ethical Guidelines
Submit an Article
Research Article | | Peer-Reviewed

Classification Panel Defects from Infrared Thermal Images Using Deep Neural Networks

Mabikana Voula Boniface Herve*,  Mabiala Louboto Antoine Victorien,  Nkombo Mazouka Michel,  M’Passi Mabiala Bernard
Published in American Journal of Modern Energy (Volume 12, Issue 4)
Received: 23 June 2026     Accepted: 11 July 2026     Published: 11 August 2026
Views:       Downloads:
Download PDF
Abstract

Development research has today, through different technologies, boosted the solar panel manufacturing industry. These technologies allow these solar panels to make the most of solar radiation to produce electricity, and as a result, improve efficiency. These photovoltaic solar panels, whenever they are installed, often have their performance affected by faults and anomalies that lead to a drop in energy efficiency. Many studies have looked into diagnosing, detecting, and locating these faults using algorithms. Building on this research and the experience gained from these systems in operation, several faults have been identified in photovoltaic cells. The goal of this study was to design, based on these different defects, artificial intelligence algorithms to detect, classify, and predict defects on solar panels over time. The model was trained using a neural network with the public database available on GitHub (intended for researchers, public services, solar project developers, and decision-makers). It contains infrared imaging data highlighting anomalies in a photovoltaic solar field. This dataset, made up of 20,000 images, allowed us to create an approximator (neural network) linking the pixels of a thermal image of a photovoltaic module to the presence of defects. During model training, the decrease in training and validation losses shows that this model has learned and can improve its performance. Beyond 20 epochs, the model converges, which means that no additional learning can improve the model's performance. This developed, trained, and validated model was exported to Simulink and allowed us to classify and predict solar panel faults, with an average accuracy of over 95% for some faults and 100% for others.

Published in American Journal of Modern Energy (Volume 12, Issue 4)
DOI 10.11648/j.ajme.20261204.11
Page(s) 59-69
Creative Commons

This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited.

Copyright

Copyright © The Author(s), 2026. Published by Science Publishing Group
Previous article
Keywords

Diagnosis, Panel Defects, Neural Network, Photovoltaic Solar Panels

1. Introduction
The performance of solar panels is strongly influenced by their surface condition; dirt buildup or the appearance of defects leads to a noticeable drop in energy efficiency
[1]
. This performance is crucial for maximizing electricity production. Several environmental and technical factors can affect this performance (changes in sunlight, temperature effects, etc.)
[1]
.
Artificial neural networks (ANNs) are a key part of artificial intelligence and machine learning, offering advanced capabilities to model complex relationships and handle large amounts of data. These systems are especially relevant in the field of renewable energy, where they can be used to optimize photovoltaic system performance and to efficiently detect and classify solar panel defects
[2]
.
The main causes of PV module degradation include gradual power loss, optical losses in the encapsulant, dirt on the front surface of the module, glass breakage, hot spots, degradation of bypass diodes, encapsulant delamination, water accumulation inside the module, etc.
[3]
.
Accelerated degradation tests (for safety, mechanical integrity, and thermal cycling) have been developed by ASTM
[4]
and IEC
[5]
to simulate the various modes of degradation in solar modules.
Module efficiency may differ from STC efficiency, primarily due to four meteorological factors (temperature, humidity, irradiance, and wind)
[6-9]
.
In this article, we aim to develop a neural network-based model for the detection, classification, and prediction of defects observed in solar panels during operation. Most manufacturers guarantee a lifespan of approximately 25 to 30 years for solar cells; however, as part of this work, it was not straightforward for us to track the degradation of a solar panel in order to obtain data on its degradation. Therefore, to compensate for the lack of data on solar panel defects, we used publicly available databases on GitHub (accessible to researchers, public agencies, solar developers, and decision-makers). This data consists of infrared imagery.
2. Methodology
A neural network is a collection of interconnected neurons; it is a directed graph. A neural network can be used to solve many artificial intelligence problems.
Formalization of the problem: Consider n input neurons and m output neurons.
Let X=x1, x2, …,xp be a set of p input vectors,
Sd=Sd1, Sd2, …,Sdp, a set of p desired outputs, and
S=S1, S2,…,Sp, the set of outputs actually obtained at the end of the forward propagation of the vectors in the set X.
The pair (X, Sd) forms the basis of the network’s learning.
Consider the kth element xk of the set X; we denote Xk=xk1, xk2…xkx as the components of this vector.
Similarly, we denote:
Sdk=Sdk1,Sdk2,Sdkm, the components of the kth element of the set Sd, and
Sk=Sk1,Sk2,…Skm, the components of the kth element of the set S
We also denote as the activation function of a neuron (and ’ as its derivative) and yj as the output of neuron j in a hidden layer, such that
yj= ∑wijei-Ѳj(1)
where wij are the weights of the connection between neuron $j$ and neuron $i$ in the previous layer.
Where ei is the ni input to neuron J; θi is the bias, if any
After presenting the kth example, we minimize the observed quadratic error defined as the difference between the actual output and the desired output:
E=12∑Sdki-Ski2(2)
This error propagates through the network, resulting in a change in the synaptic weights. If the error is less than a chosen threshold, the network is said to converge
[10, 11]
. The training phase then minimizes the squared error of the entire model by adjusting the weights to reduce the gradient.
[12].
This error is then minimized using the gradient method, by updating the weights in the direction indicated by the gradient of E.
[12].
The rule for modifying the weights of the connection linking neuron j to neuron i for iteration k is given by:
Wijk=Wijk-1-αδiyj(3)
where
α: the learning rate
δi: the error signal of unit i, that is, the contribution of unit i’s input to the squared error observed at the output. This error signal is calculated using the backpropagation principle.
δi=Si-Sdi'ei(4)
The calculation of the weights in the last layer depends on the elements in the previous layers.
[13].
If i is now considered the index of a hidden neuron and the layer of neurons following unit i is denoted by j, the quadratic error E associated with the presentation of an example depends functionally on the set of all outputs yj that depend on yi. In this case, the error signal δi is given by:
δi='ei∑δc Wij(5)
2.1. Activation Function
The activation function is a mathematical formula that is applied under certain circumstances. When neurons calculate the weighted sum of the input values, these values are passed to the activation function, which checks whether the calculated value exceeds the required threshold.
If the calculated value exceeds the required threshold, the activation function is triggered to calculate the output value. This output value is then passed to subsequent or preceding layers (depending on the network’s complexity), which can help the networks adjust the weights of their neurons.
It can take various forms (linear, nonlinear, sigmoid, etc.), and it allows for a comparison between the sum and the threshold to validate the output value. The output of the formal neuron can be binary or real.
This output corresponds to a weighted sum of the weights and inputs plus the neuron’s bias b. The result of the weighted sum is called the neuron’s activation level.
The bias b is also called the neuron’s activation threshold
[14, 15]
.
When the activation level reaches or exceeds the threshold b, the argument of f becomes positive (or zero). Otherwise, it is negative
[16, 17]
, and based on this value, an activation function calculates the output value of neuron j.
The activation function defines the state of the neuron based on its total input.
It describes the relationship between the neuron’s activation and decision potentials.
The three main stages in the development of a neural network:
1) Creation of a database of infrared images showing the main anomalies in solar modules;
2) Selection of the neural network architecture;
3) Training the network using the database and validating the resulting neural network model.
2.2. Fault Detection Methodology
Observing the various types of degradation in solar panels requires a fairly long period of time
[18]
. For a solar farm, data related to this degradation can be detected using satellites, drones, and infrared methods
[19-21]
.
This dataset is a collection of labeled data designed to facilitate research and solve problems well-suited to machine learning that may have an impact on the environment.
This dataset consists of 20000 infrared images, each 24 x 40 pixels in size, of solar modules exhibiting various anomalies found in solar farms. This dataset comprises 12 groups (classes) of solar panel images, including 11 classes of different anomalies and one class consisting of solar panels with no anomalies. These 12 image classes are listed in Table 1.
Table 1. Classification of images of different defects.

N°

Class Name

Number of images

Description

0

Single-cell

1877

Hot spot appearing on a square module

1

Multi-cell

1288

Hot spot appearing on a square module within a cell

2

Cracking

941

Cracking on the module's surface.

3

Hot Spot

251

Hot spot on a module

4

Hot-Spot-Multi

247

Multiple hot spots on a module

5

Shading

1056

Sunlight blocked by vegetation, man-made structures, or adjacent rows.

6

Diode

1499

One shunt diode is activated, typically affecting 1/3 of the module.

7

Multi-Diode

175

Several shunt diodes are activated, generally affecting 2/3 of the module.

8

Vegetation

1639

Panels blocked by vegetation.

9

Fouling

205

Dirt, dust, or other debris on the module's surface.

10

Module not connected to the grid

828

The entire module is overheating.

11

No anomalies

10000

Nominal solar module.
This table shows the classification and description of defects in the 20000 infrared images, which are divided into 11 defect classes—ten of which represent specific defect types, while the eleventh class comprises all 10000 defect-free images.
This data was used to develop a model for predicting defects based on the 20000 infrared images of solar panels, categorized into 11 classes as shown in Table 1.
This selection aims to capture a wide range of common conditions that affect the performance of solar panels.
Figure 1. Classification of panel defects using Deep Network Designer (DND).
In MATLAB, a pre-trained TensorFlow model allows for transfer learning, a process that involves using the pre-trained model to refine and adapt it to the problem at hand.
As a result, using Deep Network Designer (DND), we can perform transfer learning either programmatically or iteratively.
The network model is designed for training data using a few lines of MATLAB code and built-in functions that replace, remove, or add layers to any part of the network architecture. Simulink is a block diagram environment used for designing systems with multi-domain models.
Artificial intelligence models are integrated into Simulink to simulate entire systems. Using the TensorFlow model, an AI system reads and classifies the imported images.
The convolutional neural network was created and trained for classification, as this type of network is well-suited for image recognition.
To make predictions, we used an LSTM network that takes sequence data as input.
Therefore, a large number of annotated images is required to use a CNN for image classification problems. Although there is no formal rule specifying the exact number of images needed to guarantee a minimum level of performance for a CNN, several empirical observations have highlighted the relationship between the complexity of the classification problem to be solved and the size of the training image dataset. Thus, as the number of classes and their variability increase, the number of annotated images required will also increase.
[22].
3. Results and Discussion
In this section, we present the simulation results of our model.
Step 1: Designing the network model.
Figure 2. Network model for simulation.
Step 2: Import the data (images)
Figure 3. Importing Images.
Figure 4. Data Validation.
Step 3: Training
Figure 5. Launching the model for training.
Figure 6. Training Curves.
Figure 7. Evolution of the Training Process.
Figure 8. Evolution of the Training Process (continued).
Figure 9. Evolution of the Training Process (continued).
Figure 10. The End of the Training Process.
Defect Identification Tests
Figure 11. Simulation of the classification system.
Hot spot defect classification test
Figure 12. Identification of the Hot Spot Defect.
Diode Fault Classification Test
Figure 13. Diode Fault Identification.
Multi-diode Fault Classification Test
Figure 14. Diode Fault Diagnosis.
The shape of the curves in Figures 5, 6, 7, 8, 9, and 10 reflects the model’s learning process, characterized by an increasing trend in accuracy over the course of iterations, with a frequency of 50 iterations out of 360.
This trend indicates that the model is learning information from the images and improving its accuracy, thereby justifying the model’s performance. This trend over the course of iterations supports the hypothesis that the number of convolutional layers and the number of iterations are key factors in improving the performance of a convolutional neural network in terms of accuracy and error, on the one hand. On the other hand, as the number of iterations increases, we tend to minimize the classification error.
Based on the verification of this hypothesis, we can confirm that we have a convolutional neural network model that performs well in terms of accuracy, minimizes error, and enables us to classify images.
In terms of accuracy, the results obtained were validated with an accuracy of over 90% and an error rate of less than 5%.
The second part, devoted to testing, focused on the detection and classification of images based on defect type using the convolutional neural network model that was implemented.
The results obtained demonstrate impressive performance in terms of accuracy and robustness, which reinforces the potential for practical applications in predictive maintenance and yield optimization for photovoltaic systems.
Meanwhile, for the classification of solar module defects, a neural network model was designed using a dataset consisting of 20,000 infrared images, each measuring 24x40 pixels, of each module. This model, which was trained, validated, and exported to Simulink, enabled us to classify and predict defects in solar panels with an accuracy of over 95%. Specifically, 96% for hot spot defects, 95.5% for diode defects, and 100% for multi-diode defects.
4. Conclusion
Among the various deep learning algorithms, we used a convolutional neural network for image classification. In this article, we developed a convolutional neural network model for defect prediction by training it on infrared image data representing different classes of defects recorded during the operation of solar panels. The results of this model are consistent with those reported in the literature. The effectiveness of the designed model allowed us to classify images and predict defects with acceptable accuracy and minimized error.
During training, certain layers slow down information processing, leading to overfitting. To avoid this phenomenon, the dropout method can be used to randomly eliminate neurons. Using this method would help limit marginal effects and enable the network to better generate concepts during training. For the next phase of this work, it would be advisable to conduct an experimental validation of the proposed algorithm using practical tools specific to these techniques.
Abbreviations

STC

Standard Test Conditions

ASTM

American Society for Testing and Materials

ANN

Artificial Neural Networks

DND

Deep Netwok Designer

LSTM

Long Short-Term Memory

CNN

Convolutional Neural Network
Author Contributions
Mabikana Voula Boniface Herve: Conceptualization, Data curation, Formal Analysis, Funding acquisition, Investigation
Mabiala Louboto Antoine Victorien: Investigation, Methodology, Project administration, Resources
Nkombo Mazouka Michel: Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing
Conflicts of Interest
The authors declare no conflicts of interest.
References
[1] Oussema Aloulou ‘Use of Artificial Neural Networks for MPPT Control Optimization and Solar Panel Fault Detection’, University of Quebec in Abitibi-Témiscamingue, 2024.
[2] Anouar boucheham, ‘Identification of Biomarkers and Non-Coding RNA Using Computational Intelligence-Based Approaches, Constantine 2 University, Abdelhamid Mehri, Faculty of New Information and Communication Technologies, Department of Fundamental Computer Science and Its Applications’, 2016.
[3] R. Ross, Proc. 17th IEEE Photovoltaic Specialist Conf., 464–472, 1984.
[4] C. Osterwald, ASTM Standards Development Status, Proc. 18th IEEE Photovoltaic Specialist Conf., 749–753, 1985.
[5] Qualification Test Procedures for Photovoltaic Modules, Commission of the European communities, Joint Research Center, ISPRA Establishment, Specification 502, 1984.
[6] K. Bücher, Site dependence of the energy collection of PV modules, Solar Energy Materials and Solar Cells, 47 85-94, 1997.
[7] K. Bücher, G. Kleiss and D. Bätzner, Photovoltaic Modules in Buildings: Performance and Safety, Renewable Energy 15 545-551, 1998.
[8] A. Parretta, A. Sarno and L. R. M. Vicari, Effects of irradiation conditions on the outdoor performance of photovoltaic modules, Optics Communications 153 153.
[9] Mohamed Bentoum Tools for Detection and Classification: Application to the Diagnosis of Rail Surface Defects, Henri Poincaré University - Nancy I, 2006.
[10] Veïs Oudjail, Spike Neural Networks Applied to Computer Vision, Ph. D. dissertation, University of Lille, 2022.
[11] Erwann Martin’ Neural Networks at the Nanoscale: Which Learning Models? Dissertation, Paris-Saclay University’, 2022.
[12] Guillaume Lacharme’ Optimization of Deep Learning Model Hyperparameters’ University of Tours Graduate School: MIPSIS University of Tours Graduate School, 2024.
[13] youssef barkaoui ‘Deep Classification and Neural Networks: Applications in Data Science at the University of Quebec at Trois-Rivières ‘2022 Godi Tchere Moustapha.’ Detection and Inspection Using Neural Networks in Ellipsometric Scatterometry. Optics / Photonics. Jean Monnet University - Saint-Étienne’, 2023. Français.
[14] Gabrielle Gregoire ‘On Nonlinear Autoregressive Models with Smooth Transitions and the Calculation of Their Forecasts University of Montreal Faculty of Graduate and Postdoctoral Studies toward the degree of Master of Science (M.Sc.) in Statistics’ 2019.
[15] NOUAR Ahcene’ Numerical Solution of Nonlinear Equations Arising from Fluid Flow Using Stochastic AlgorithmsUuniversity BADJIMOKHTAR- ANNABA 2022.
[16] Realini A. Mean Time before Failure of Photovoltaic Modules. Final Report (MTBF Project), Federal Office for Education and Science Tech. Rep., BBW 99.0579, 2003.
[17] Carlson D. E., Romerol R., Willing F., Meakin D., Gonzalez L., Murphy R., Moutinho H. R., Al-Jassim M. “Corrosion Effects in Thin-Film Photovoltaic Modules”. Progress Photovoltaics: Research and Applications, 11: 377–386, 2003.

https://doi.org/10.1002/pip.500

[18] Lucie Pirot–Berson ‘Study of the degradation of photovoltaic modules based on heterojunction and TOPcon silicon technologies in a humid environment. Physics. Université Grenoble Alpes’, 2025. Français.
[19] Mohamed Chakchouk. ‘Design of a smart mobile mechatronic system for observing molecules in the gas phase using IR spectroscopy. Mechanical Engineering [physics. class-ph]. Normandy University; University of Sfax (Tunisia)), 2024. Français.
[20] Emilien ALVAREZ-VANHARD’ Synergies Between Drone- and Satellite-Based Optical Remote Sensing: Scale Change and Application to Wetland Grassland Conservation Thesis presented and defended in Rennes on October 25, 2021 LETG Research Unit (Coastal Zone - Environment - Remote Sensing - Geomatics) UMR 6554.
[21] Hassina AIT ISSADD ‘Smart Drone Deployment for the Agriculture of the Future: Mouloud Mammeri University of Tizi-Ouzou ‘ 2020.
[22] Antoine Acremont. ‘Deep Neural Networks for Object Classification in Infrared Imaging: Contributions of Learning from Synthetic Data and Anomaly Detection. Neural Network [cs. NE]. ENSTA Bretagne - National School of Advanced Technology of Brittany,’2020. Français.
Cite This Article
Plain Text BibTeX RIS

  • APA Style

    Herve, M. V. B., Victorien, M. L. A., Michel, N. M., Bernard, M. M. (2026). Classification Panel Defects from Infrared Thermal Images Using Deep Neural Networks. American Journal of Modern Energy, 12(4), 59-69. https://doi.org/10.11648/j.ajme.20261204.11

    Copy | Download

    ACS Style

    Herve, M. V. B.; Victorien, M. L. A.; Michel, N. M.; Bernard, M. M. Classification Panel Defects from Infrared Thermal Images Using Deep Neural Networks. Am. J. Mod. Energy 2026, 12(4), 59-69. doi: 10.11648/j.ajme.20261204.11

    Copy | Download

    AMA Style

    Herve MVB, Victorien MLA, Michel NM, Bernard MM. Classification Panel Defects from Infrared Thermal Images Using Deep Neural Networks. Am J Mod Energy. 2026;12(4):59-69. doi: 10.11648/j.ajme.20261204.11

    Copy | Download 

  • @article{10.11648/j.ajme.20261204.11,
  author = {Mabikana Voula Boniface Herve and Mabiala Louboto Antoine Victorien and Nkombo Mazouka Michel and M’Passi Mabiala Bernard},
  title = {Classification Panel Defects from Infrared Thermal Images Using Deep Neural Networks},
  journal = {American Journal of Modern Energy},
  volume = {12},
  number = {4},
  pages = {59-69},
  doi = {10.11648/j.ajme.20261204.11},
  url = {https://doi.org/10.11648/j.ajme.20261204.11},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ajme.20261204.11},
  abstract = {Development research has today, through different technologies, boosted the solar panel manufacturing industry. These technologies allow these solar panels to make the most of solar radiation to produce electricity, and as a result, improve efficiency. These photovoltaic solar panels, whenever they are installed, often have their performance affected by faults and anomalies that lead to a drop in energy efficiency. Many studies have looked into diagnosing, detecting, and locating these faults using algorithms. Building on this research and the experience gained from these systems in operation, several faults have been identified in photovoltaic cells. The goal of this study was to design, based on these different defects, artificial intelligence algorithms to detect, classify, and predict defects on solar panels over time. The model was trained using a neural network with the public database available on GitHub (intended for researchers, public services, solar project developers, and decision-makers). It contains infrared imaging data highlighting anomalies in a photovoltaic solar field. This dataset, made up of 20,000 images, allowed us to create an approximator (neural network) linking the pixels of a thermal image of a photovoltaic module to the presence of defects. During model training, the decrease in training and validation losses shows that this model has learned and can improve its performance. Beyond 20 epochs, the model converges, which means that no additional learning can improve the model's performance. This developed, trained, and validated model was exported to Simulink and allowed us to classify and predict solar panel faults, with an average accuracy of over 95% for some faults and 100% for others.},
 year = {2026}
}

    Copy | Download 

  • TY  - JOUR
T1  - Classification Panel Defects from Infrared Thermal Images Using Deep Neural Networks
AU  - Mabikana Voula Boniface Herve
AU  - Mabiala Louboto Antoine Victorien
AU  - Nkombo Mazouka Michel
AU  - M’Passi Mabiala Bernard
Y1  - 2026/08/11
PY  - 2026
N1  - https://doi.org/10.11648/j.ajme.20261204.11
DO  - 10.11648/j.ajme.20261204.11
T2  - American Journal of Modern Energy
JF  - American Journal of Modern Energy
JO  - American Journal of Modern Energy
SP  - 59
EP  - 69
PB  - Science Publishing Group
SN  - 2575-3797
UR  - https://doi.org/10.11648/j.ajme.20261204.11
AB  - Development research has today, through different technologies, boosted the solar panel manufacturing industry. These technologies allow these solar panels to make the most of solar radiation to produce electricity, and as a result, improve efficiency. These photovoltaic solar panels, whenever they are installed, often have their performance affected by faults and anomalies that lead to a drop in energy efficiency. Many studies have looked into diagnosing, detecting, and locating these faults using algorithms. Building on this research and the experience gained from these systems in operation, several faults have been identified in photovoltaic cells. The goal of this study was to design, based on these different defects, artificial intelligence algorithms to detect, classify, and predict defects on solar panels over time. The model was trained using a neural network with the public database available on GitHub (intended for researchers, public services, solar project developers, and decision-makers). It contains infrared imaging data highlighting anomalies in a photovoltaic solar field. This dataset, made up of 20,000 images, allowed us to create an approximator (neural network) linking the pixels of a thermal image of a photovoltaic module to the presence of defects. During model training, the decrease in training and validation losses shows that this model has learned and can improve its performance. Beyond 20 epochs, the model converges, which means that no additional learning can improve the model's performance. This developed, trained, and validated model was exported to Simulink and allowed us to classify and predict solar panel faults, with an average accuracy of over 95% for some faults and 100% for others.
VL  - 12
IS  - 4
ER  -

    Copy | Download

Author Information

  • Mabikana Voula Boniface Herve

    Laboratory of Electrical and Electronic Engineering, National Higher Polytechnic School, Marien NGOUABI University, Brazzaville, Republic of the Congo

    Contact Email

    http://orcid.org/0009-0003-1677-940X

  • Mabiala Louboto Antoine Victorien

    Laboratory of Electrical and Electronic Engineering, National Higher Polytechnic School, Marien NGOUABI University, Brazzaville, Republic of the Congo

  • Nkombo Mazouka Michel

    Laboratory of Electrical and Electronic Engineering, National Higher Polytechnic School, Marien NGOUABI University, Brazzaville, Republic of the Congo

  • M’Passi Mabiala Bernard

    Faculty of Science and Technology, Department of Physics, Marien Ngouabi University, Brazzaville, Republic of the Congo

    http://orcid.org/0000-0002-7340-7165

Download PDF
Submit an Article

About Us

Science Publishing Group (SciencePG) is an Open Access publisher, with more than 300 online, peer-reviewed journals covering a wide range of academic disciplines.

Learn More About SciencePG

Products

Information

Important Link

Copyright © 2012 -- 2026 Science Publishing Group – All rights reserved.