Wyniki wyszukiwania - BazTech

1

Identifying region specifc seasonal crop for leaf borne diseases by utilizing deep learning techniques

Jayagopal Prabhu, Rajendran Sukumar, Mathivanan Sandeep Kumar, Sathish Kumar Sree Dharinya, Raja Kiruba Thangam, Paneerselvam Surekha

Acta Geophysica

|

2022

|

Vol. 70, no 6

2841--2854

EN

India economy depends on agriculture with severe climatic changes and a heavy infestation of diseases depleting food crop yield substantially. Rapid identification and real-time infestation feedback that affects plants are accomplished through computer vision and IoT, thereby providing a reliable system for farmers to increase the season’s growth yield. With LSTM, CNN provides an efficient way of identifying diseases specific leaf in plants through image recognition techniques. An extensive collection of plant leaf images is trained to recognize season-specific diseases like early blight and late blight, leaf mold, and yellow leaf curl. The proposed CNN model identifies the infestation with high accuracy and precision with significantly fewer training epochs. The proposed model provides an efficient way of identifying leaf borne infestation pertained to a particular agricultural region. Furthermore, there is a need to increase and improve different region-specific infestations that arise due to climatic and seasonal changes.

2

Deep back propagation–long short-term memory network based upper-limb sEMG signal classification for automated rehabilitation

Wang Ying, Wu Qun, Dey Nilanjan, Fong Simon, Ashour Amira S.

Biocybernetics and Biomedical Engineering

|

2020

|

Vol. 40, no. 3

987--1001

EN

Autonomous rehabilitation training for assisted patients with injured upper-limbs promotes the regenerative communication between muscle signals and brain consciousness. Surface electromyographic (sEMG) is a type of electrical signals of neuromuscular activity recorded by electrodes on the surface of the human body, which is widely applied for detecting gestures and stimuli reactions. Experimental results proved the importance of the sEMG signals for extracting such reactions, in which, the segmentation and classification of the sEMG are vital tasks. The objective of the present work is to segment and classify the sEMG signals of patients to assist the design of clinical rehabilitation devices based on the classification of sEMG signals. In the pre-processing stage, a dual-tone multi-frequency signaling is designed for signal coding; subsequently, the pre-processed sEMG signal is transformed by the Fast Fourier Transfer. Afterward, a time-series frequency analysisis performed by applyingHiddenMarkov Models.A basic traditional longshort- term memory (LSTM) model is addressed for waveform-based classification to be compared to the proposed improved deep BP (back-propagation)–LSTM for sEMG signal classification. Seventeen performance features are selected for evaluating the proposed multi-classification, deep learning model for classifying six actions, namely moving gesture of grip, slowly moving, flexor, straight lift, stretch, and up-high lift; which were proposed by rehabilitation physician. The experiment results indicated the superiority of the proposed method compared to other well-known classifiers, such as the neural network, support vector machine, decision trees, Bayes inference, and recurrent neural network. The proposed deep BP–LSTM network achieved 92% accuracy, 89% specificity, 91% precision, and 96% F1-score, in the multi-classification of the sEMG signals.

3

Complex lithology prediction using mean impact value, particle swarm optimization, and probabilistic neural network techniques

Gu Yufeng, Zhang Zhongmin, Zhang Demin, Zhu Yixuan, Bao Zhidong, Zhang Daoyong

Acta Geophysica

|

2020

|

Vol. 68, no. 6

1727--1752

EN

Lithology prediction is a fundamental problem because the outcome of lithology prediction is the critical underlying data for some basic geological work, e.g., establishing stratigraphic framework or analyzing distribution of sedimentary facies. As the geological formation generally consists of many diferent lithologies, the lithology prediction is always viewed as a tough work by geologists. Probabilistic neural network (PNN) shows high efciency when solving pattern recognition problem since learning data do not need to do any pre-training of learning data and calculation results are universally reliable, and then, this model could be considered as an efective solution. However, there are two factors that seriously limit the PNN’s performance: One is existence of the interference variables of learning samples, and the other is selection of the window length of probability density distribution. In view of adverse impact of those two factors, two techniques, mean impact value (MIV) and particle swarm optimization (PSO), are introduced to improve the PNN’s calculation capability. Thus, a new prediction method referred as MIV–PSO–PNN is proposed in this paper. The proposed method is validated by three well-designed experiments, and the corresponding experiment data are recorded by two cored wells of the LULA oilfeld. For the three experiments, prediction accuracies of the results provided by the proposed method are 81.67%, 73.34% and 88.34%, respectively, all of which are higher than those provided by other comparative approaches including backpropagation (BP), PNN, and MIV-PNN. The experiment results strongly demonstrate that the proposed method is capable to predict complex lithology.

4

Control of the weld quality using welding parameters in a robotic welding process

Park M.-H., Jin B.-J., Yun T.-J., Son J.-S., Kim C.-G., Kim I. S.

Journal of Achievements in Materials and Manufacturing Engineering

|

2018

|

Vol. 87, nr 1

32--40

EN

Purpose: Since the welding automations have widely been required for industries and engineering, the development of the predicted model has become more important for the increased demands for the automatic welding systems where a poor welding quality becomes apparent if the welding parameters are not controlled. The automated welding system must be modelling and controlling the changes in weld characteristics and produced the output that is in some way related to the change being detected as welding quality. To be acceptable a weld quality must be positioned accurately with respect to the joints, have good appearance with sufficient penetration and reduce low porosity and inclusion content. Design/methodology/approach: To achieve the objectives, two intelligent models involving the use of a neural network algorithm in arc welding process with the help of a numerical analysis program MATLAB have been developed. Findings: The results represented that welding quality was fully capable of quantifying and qualifying the welding faults. Research limitations/implications: Welding parameters in the arc welding process should be well established and categorized for development of the automatic welding system. Furthermore, typical characteristics of welding quality are the bead geometry, composition, microstructure and appearance. However, an intelligent algorithm that predicts the optimal bead geometry and accomplishes the desired mechanical properties of the weldment in the robotic GMA (Gas Metal Arc) welding should be required. The developed algorithm should expand a wide range of material thicknesses and be applicable in all welding position for arc welding process. Furthermore, the model must be available in the form of mathematical equations for the automatic welding system. Practical implications: The neural network models which called BP (Back Propagation) and LM (Levenberg-Marquardt) neural networks to predict optimal welding parameters on the required bead reinforcement area in lab joint in the robotic GMA welding process have been developed. Experimental results have been employed to find the optimal algorithm to predict bead reinforcement area by BP and LM neural networks in lab joint in the robotic GMA welding. The developed intelligent models can be estimated the optimal welding parameters on the desired bead reinforcement area and weld criteria, establish guidelines and criteria for the most effective joint design for the robotic arc welding process. Originality/value: In this study, intelligent models, which employed the neural network algorithms, one of AI (Artificial Intelligence) technologies have been developed to study the effects of welding parameters on bead reinforcement area and to predict the optimal bead reinforcement area for lab joint in the robotic GMA welding process. BP (Back Propagation) and LM (Levenberg-Marquardt) neural network algorithm have been used to develop the intelligent model.

5

Decoupling control for permanent magnet in-wheel motor using internal model control based on back-propagation neural network inverse system

Li Y., Zhang B., Xu X.

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2018

|

Vol. 66, nr 6

961--972

EN

The permanent magnet in-wheel motor (PMIWM) is a nonlinear, multivariable, strongly coupled and highly complex system. The key to the development and application of the PMIWM consists in the improvement of its control accuracy and dynamic performance. In order to effectively decouple the PMIWM, this paper presents a novel internal model control (IMC) approach based on the back-propagation neural network inverse (BPNNI) control method. First, theoretical analysis is conducted to show the existence of the PMIWM inverse system, to be modeled mathematically. The inverse system approximated and identified by the back-propagation neural network (BPNN) constitutes the back-propagation neural network inverse (BPNNI) system. Then, by cascading the BPNNI system on the left side of the original PMIWM system, a new decoupling, pseudo-linear system is established. Moreover, the 2-DOF internal model control (IMC) method is employed to design the extra closed-loop controller that further improves disturbance rejection and robustness of the whole system. Consequently, the proposed decoupling control approach incorporates the advantages of both the BPNNI and the IMC. Effectiveness of thus proposed control approach is verified by means of simulation and real-time hardware-in-the-loop (HIL) experiments.

6

Optimization of Neural Network Model Design: An Electoral Cooperative Particle Swarm Optimization Approach

Li D., Xu H.

Przegląd Elektrotechniczny

|

2013

|

R. 89, nr 1b

165--167

EN

This paper proposes an electoral cooperative particle swarm optimization approach to optimize the model of neural network from both structure and linked weights. Different with other related research work, a new encoding method is adopted to divide the neural network into several modules, each of them corresponding to a sub-swarm. Based on the experiments on typical problems and classic dataset, the results show that the proposed algorithm outperforms all the compared ones in perspective of test error, correctness, connection number, and the CPU time of the training phase.

PL

W przedstawionym artykule opisano zastosowanie metod optymalizacji roju cząstek do optymalizacji struktury i współczynników wagowych sieci neuronowej. Zaimplementowano nową metodę analizy, do dzielenia podzielenia sieci na moduły, reprezentujące mniejsze roje. Weryfikacja eksperymentalna i porównanie z metodami klasycznymi wykazały wysoką sprawność i skuteczność analizy.

7

Sterowanie nieholonomicznym robotem mobilnym przy wykorzystaniu oscylatora kinematycznego.

Kozłowski K., Majchrzak J.

Prace Naukowe Instytutu Cybernetyki Technicznej Politechniki Wrocławskiej. Konferencje

|

2001

|

Vol. 103, nr 47

31-44

PL

W pracy przedstawiono problem sterowania nieholnomicznym robotem mobilnym z uwzględnieniem jego kinematycznego i dynamicznego modelu dla przypadku odtwarzania trajektorii odniesienia. Wykorzystywany algorytm oscylatora kinematycznego umożliwia ekspotencjalną zbieżność błędu śledzenia modelu kinematycznego. Sterownik zawierający oscylator zmienny w czasie, oparty na przekształceniu kinematycznym, umożliwia uzyskanie sterowania bezpośrednio dla układu nieholonomicznego. Do budowy złożonego sterownika ruchu robota wykorzystano metodę propagacji wstecznej, która umożliwia wielopoziomowe sterowanie obiektem. Przedstawiono też zjawiska związane z ograniczeniem sygnału sterowania.

EN

In this paper a problem of trajectory control of nonholonomic mobile robot with its kinematic and dynamic model has been presented. It has been shown that by making use of the kinematic oscillator it is possible to obtain the exponential convergence of the kinematic error with an accuracy e. Next, a backstepping scheme which includes both dynamic model of a mobile robot and servo current-voltage loop is outlined. It is shown ; that the whole system is globally asymptotically stable. Simulation results illustrate theoretical considerations. Experimental work, though not reported, is under way.

8

Zastosowanie sieci neuronowej do określania własności połączeń spawanych

Pamuła T.

Zeszyty Naukowe. Transport / Politechnika Śląska

|

2000

|

z. 40

13-21

PL

W artykule przedstawiono próbę zastosowania sieci neuronowej do oceny struktur spoin niskowęglowych. Zaproponowano użycie trójwarstwowej jednokierunkowej nieliniowej sieci do przetwarzania obrazów mikroskopowych spoin i powiązania ich z bazą danych własności. Zastosowano sigmoidalną funkcję przejścia. Naukę sieci przeprowadzono wikorzystując metodę propagacji wstecznej z różnymi współczynnikami uczenia dla każdej z warstw. Mikroskopowe obrazy spoin potraktowano jak tekstury i przygotowano ciągi uczące dla sieci wykorzystując metody klasyfikacji tekstur.

EN

The paper presents an attempt to apply a neural network for assesing the properties of low carbon welds. A three layer unidirectional non linear net was designed to interpret the contents of microscopic images of the welds and combine it with the contents of a database of weld properties. For this purpose a sigmoidal transfer functions were used. The course of the net training was carried out using backpropagation idea with different learning rates and momentum coefficients for the net layers. The weld images were treated as textures. A method for preparing training sequences was devised, based on texture analysis methods.

9

Hierarchical intelligent control of a learning robot using a camera and system of artificial neural networks

Miljković Z.

Journal of Applied Computer Science

|

2000

|

Vol. 8, nr 2

79-97

EN

This paper presents interaction of mechatronic subsystems in order to achieve an adaptive behavior of learning robots. A learning robot is able to deal flexibly with changes in its environment and to execute intelligent tasks. The control strategy for the learning robot is established by using a recognition system and machine learning. The recognition system that utilizes artificial intelligence techniques is used in order to test and verify the hypothesis that learning robots can achieve sensor-actuator co-ordination and team successfully. The hypothesis is tested and verified on the basis of visual information obtained from the camera and an artificial neural network system. For this purpose the experimental set of software packages Make it, ART-1 Simulator and BPNET, as well as the physical model of anthropomorphic mobile robot Don Kihot with four degrees of freedom, are realized.

10

On-line adaptive learning rate BP algorithm for MLP and application to an identification problem

Sha D. H., Bajić V. B.

Journal of Applied Computer Science

|

1999

|

Vol. 7, nr 2

67-82

EN

An on-line algorithm that uses an adaptive learning rate is proposed. Its development is based on the analysis of the convergence of the conventional gradient descent method for three-layer BP neural networks. The effectiveness of the proposed algorithm applied to the identification and prediction of behavior of non-linear dynamic systems is demonstrated by simulation experiments.