neural network

The behavior of engineering structures affected by dynamic and static loads are different depending on the type of design, the type of materials used in them and the operating time of the structures. These behaviors can be observed and evaluated with the time series obtained from GPS observations in the time and the frequency domains. The current study is conducted with the aim of investigating the behavior of Quds cable bridge of Ardabil, using the geodetic monitoring of the GPS system at a rate of 30 seconds. Microgeodesy network is used for static data collection and RTKLIB software with Kalman filter capability used for initial processing and time series formation. The processed results have an error of 3 mm. In order to evaluate the movement of the bridge, moving average, time average, and median filters have been used to estimate the semi-static, static, and dynamic components of the bridge movement, respectively. The results of applying the moving average filter indicate that the semi-static component is extracted with 4 mm accuracy in the East, 1 mm in the North and 2 mm in the Height directions. The RMSE between the obtained dynamic components and the short period components are 0.8, 0.1 and 4 mm in three directions E, N and H respectively. In order to extract the dominant frequencies of the bridge, the Fast Fourier Transform (FFT) method was used. Then, based on FFT, the periodogram diagram was obtained and the dominant frequencies were identified. In practice, numerical methods such as the periodogram diagram and its visual inspection were used to extract the existing dominant frequencies. A periodogram diagram presents the relation between the power spectrum and the frequency. In the next step, this diagram is used to compare the power spectrum of each frequency with its neighboring frequencies. Finally, the frequencies with a higher power spectrum were considered as the dominant frequencies. Finally, the Artificial Neural Network model (ANN) with 10 hidden layers and 10 delays was used to predict bridge deformation and the accurate evaluation of the components. The results obtained from the processing of GPS observations in the time and frequency domains indicated the least changes compared to the safety design range of the bridge behavior. Besides, the frequency analysis of the bridge movement time series and the Neural Network model, can be used to detect significant frequency changes and study the bridge performance rigidity, respectively.Given that, since bridges are designed with a high reliability factor for static and dynamic loads more than the loads received during bridge monitoring, the natural frequencies of the bridge are extracted with the existing process with high accuracy, which can be the biases of early warnings. On the other hand, the bridge health monitoring system should be based on modern measurement methods. In the case of Quds Bridge, its complete behavioral investigation shows that the bridge is safe in its current condition and within the design limits.In order to evaluate the frequency content other than the fast Fourier transform, it is also suggested to use short time Fourier transform and wavelet transforms.

The plasma focus device can be utilized as an X-ray generator in radiographic applications. A few tens of nanoseconds after the pinch moment, the hard X-ray of this device is released in a time interval of 100-150 ns. The pulsed form of X-ray emitted from this device makes the common method of detection and spectrometry useless for characterizing them. In this article, passive spectrometry, using radiographic film with Al attenuation filters, was used to determine the pulsed X-ray spectrum emitted from the UIPF-1. Neural network technology was used to determine the spectrum according to recorded doses. The neural network was trained using MCNPX simulation results. The results showed that the X-ray spectrum extends from 6 keV to 50 keV with a maximum value of 8 keV when the UIPF-1 device is operated with the copper anode, the copper insert, the working voltage of 21 kV, and 0.9 mbar air gas injection. The average spectrum energy was also obtained at 17.5 keV.

Calculation of stopping power and range of different ions is extremely important due to their special importance at controversy of shielding and health physics. Although the so-called Bethe-Bloch relationship describes well how to reduce the energy of particles, the difference in the interaction mechanisms of ions and light and heavy particles in a wide range of energy values causes that the accurate estimation of this relationship requires various corrections. In this article, in order to simulate stopping power changes in terms of proton inclining beam energy with different ions, artificial neural network and multilayer perceptron techniques will be used. And high efficiency of suggested way for stopping power and range and valid artificial model are achieved by comparing results of the neural network with results of calculating SRIM code and also available experimental data.

Sonar systems are of special importance in many ways, including military applications, shipping, fishing, etc. Therefore, the classification of sonar data is always of interest to experts in this field. In this article, two data preparation methods were used. In the first method, all the features extracted from the data and in the proposed method were averaged out of the time period used to extract the feature in the form of ten period. Different structures of artificial neural network and hybrid neural network were compared with particle swarm algorithm (PSO) to achieve the highest performance in classifying sounds emitted by floats based on float length. The results showed that in the case of using raw extracted features in the use of artificial neural network, the 2-2-2 structure in the hidden layer had the highest performance for training and testing conditions equal to 90.61 and 90% respectively. By using the hybrid neural network, the classification accuracy increased and reached 94.44% in the test conditions. In using the proposed method to prepare the extracted data, the simple structure of one layer with 6 neurons in the hidden layer provided the highest performance in the classification of the extracted features by 100% for training and testing.

The proton induced prompt gamma spectrum is applied for elemental analysis of irradiation tissues. The main purpose of the analysis in proton therapy is to track oxygen concentration in abnormal tissues. Online monitoring of oxygen concentration over a full course of treatment could provide a direct method for evaluating the response of these tissues to proton therapy and this information on the response of the tumor and healthy tissues to irradiation could then be used by the oncologist to adjust the patient’s treatment plan to ensure proper dose delivery to the tumor. In this study, the prompt gamma spectrum of a human eye phantom is simulated in an HPGe detector using the Geant4 toolkit, and the elemental analysis is accomplished using an artificial neural network (ANN). In the analysis, 33 eye phantoms are considered with different tumors, including different densities and elemental compositions. 21 samples were used as train data in ANN and 6 samples for testing and 6 samples for validation. The results show that there is a good correlation between outputs and targets. They show that the error percent of oxygen, carbon, and nitrogen in test samples are less than 5.8, 12.7, and 25%, respectively. Finally, the potential of quantitative elemental analysis of inhomogeneous targets is confirmed for providing a method to track change in the oxygen level of tumors.

The ionosphere is a layer of the Earth's atmosphere that extends from an altitude of 60 km to an altitude of 1,500 km. Knowledge of electron density distribution in the ionosphere is very important and necessary for scientific studies and practical applications. Observations of global navigation satellite system (GNSS) such as the global positioning system (GPS) are recognized as an effective and valuable tool for studying the properties of the ionosphere. Studies on ionosphere modeling in the Iranian region have shown that the global ionosphere maps (GIM) model as well as empirical models such as IRI2016 and NeQuick have low accuracy in this region. The main reason for the low accuracy of these models is the lack of sufficient observations in the Iranian region. For this reason, this paper presents the idea of using learning-based methods to generate a local ionosphere model using observations of GNSS stations. Therefore, the main purpose of this paper is to use three models of artificial neural networks (ANNs), adaptive neuro-fuzzy inference system (ANFIS) and support vector machine (SVM) to model and predict the time series of ionospheric TEC variations in Tehran GNSS station.An adaptive neuro-fuzzy inference system (ANFIS) is a kind of ANN that is based on Takagi–Sugeno fuzzy inference system. The technique was developed in the early 1990s (Jang, 1993). Since it integrates both neural networks and fuzzy logic principles, it has potential to capture the benefits of both in a single framework. Its inference system corresponds to a set of fuzzy IF–THEN rules that have learning capability to approximate nonlinear functions. Hence, ANFIS is considered to be a universal estimator. ANFIS architecture consists of five layers: fuzzy layer, product layer, normalized layer, defuzzy layer, and total output layer.In machine learning, support-vector machines (SVM) are supervised learning models with associated learning algorithms that analyze data used for classification and regression analysis. More formally, a SVM constructs a hyperplane or set of hyperplanes in a high- or infinite-dimensional space, which can be used for classification, regression, or other tasks like outliers detection (Vapnik, 1995). In SVM method, using nonlinear functions φ(x), the input vector (x) is depicted from N-dimensional space to M-dimensional space (M>N). The number of hidden units (M) is equal to the number of support vectors that are the learning data points, closest to the separating hyperplane.The results of this paper show that the SVM has a very high accuracy and capability in modeling and predicting the ionosphere TEC time series. This model has a higher accuracy in the period of severe solar activity than GIM and IRI2016 models, which are the traditional ionospheric models in the world. Due to the fact that global models in the region of Iran do not have acceptable accuracy due to lack of sufficient observations, therefore, the SVM can be used as a local ionosphere model with high accuracy. Using this model, the TEC value can be predicted with high accuracy for different times and during periods of severe solar activity. This model can be used in studies related to the physics of the ionosphere as well as its temporal variations.

Sea current velocity measurement plays an important role in engineering design and measurements. Studies in the Persian Gulf and the Strait of Hormuz have conducted field studies or numerical modeling of flows in this region. In the present research, the surface currents of the Strait of Hormuz are predicted using artificial neural network approaches. In order to determine the model’s inputs, the eastern and western time series of surface currents of the strait are used, and the basins affecting the currents of this strait are identified using linear regression model. Then the neural network inputs are defined in two different cases. The first case of the time series of known basins is considered as the input of the neural network. In the other case, combinations of known time series are considered as the input of the neural network using the Jones schema. By comparing these two cases, it is concluded that the neural network model using the Jones scheme has a good performance in predicting the surface currents of this strait. In order to further investigate the neural network model, the flow data is divided into 16 different categories; so that in each category the difference between the minimum and maximum speed is 0.03 and the average of each category is considered as the output of the neural network. To determine the network inputs, the same is done for the two cases previously mentioned. In this case, the results show that the neural network model predicts surface currents with an accuracy of R = 0.85.

Significant Wave Height (Hs) prediction is used in the analysis of marine systems including marine structural engineering and sediment transport. The Gulf of Mexico faces tropical storms shaped hurricane annually that affects the height of waves in this region, Therefore it is important to have precise estimation of the significant wave height. In this paper, we review previous studies and train artificial neural network model, to predict the effect of different wind speed powers and shear velocity in predicting the wave height in the next hours. The results showed that the presence of different wind speed powers increases the accuracy of predicting Hs relative to the wind shear speed. To increase the prediction accuracy, autocorrelation of the wave height data recorded in this region was used and a suitable model was presented. In this model was calculated power 2.3 of wind speed to predict the height of the next 3, 6 and 8 hours and power 1.9 to predict the height of the next 12 hours. Finally, the prediction results were compared with previous studies, and indicated the higher prediction accuracy.

Given the extensive use of common mammography tests for screening and diagnosis of breast cancer, there are concerns over the increased dose absorbed by the patient due to the sensitivity of the breast tissue. Thus, knowing the Mean Glandular Dose (MGD) before radiation to the patient through its estimation can be helpful. For this reason, the MultiLayer Perceptron (MLP) neural network model was trained with Levenberg-Marquardt (LM) backpropagation training algorithm and the Entrance Surface Air Kerma (ESAK) was estimated. After running the program, it was found that 35 neurons is the most optimal value, offering a regression coefficient of 95.7%, where the Mean Squared Error (MSE) for all data was 0.437 mGy, accounting for 4.8% of the range of output changes, representing a prediction with 95.2% accuracy in the present research. In comparison with the Monte-Carlo simulation method, it enjoys a desirable accuracy.

Brittleness is one of the most important properties of the rock. Brittleness is a function of strength and indicates rock strength to deformation in the elastic modules. However, there is no direct and standard method for brittleness measurement but it can be done indirectly by using rock properties such as different ratios of compressive strength and tensile strength of rock to determine the concept of brittleness. The purpose of this study is to investigate the concepts of brittleness which are presented by the researchers and to use seismic inversion, multi-attribute analysis and neural network in the Whicher-Range field in Perth, Western Australia to estimate brittleness. The Perth sedimentary basin stretches about 100,000 square kilometers in the north-south direction of the western Australian margin. About half of this sedimentary basin is located 1 km deep in the sea. Whicher-Range gas field is 22 km south of Baselton and 200 km south of Perth. Two wells and one seismic section of Whicher-Range field are selected in this research. The lowest brittleness indexes in the first and second wells drilled 1 km apart, are 1.69 and 1.67 MPa using criteria. The highest values of the brittleness are 39.78 and 48.15 MPa, respectively, which are difficult for drilling. The starting point of the inversion process is to have post-stack seismic data, velocity model, well logs, and seismic horizons. The product of the density log and sonic velocity is equal to the impedance. Then the current impedance is converted from depth to time using the appropriate depth to time relation. As a result, the convolution of a suitable wavelet and reflectivity over time will produce the synthetic seismic trace. Adaptation rate between synthetic seismic trace and composite field seismic trace yields an acceptable result. Next, the initial modeling is performed using wavelet, geological model, and a model-based algorithm. Next, the acoustic impedance of the inversion process along with other attributes is used to construct the optimal combination of attributes to estimate the brittleness index. First, an attribute is selected that has the highest correlation with the target log and the least estimation error in the training step. Then the second attribute, which makes the best combination with the first attribute, has the lowest estimation error. Then each attribute step is added to its previous step combination until the resulting combination results in the lowest estimation error. The results based on this method are obtained by increasing the number of attributes and decreasing the estimation error, while the error in the validation stage until the optimal attribute combination, is ascending. In the next step, three types of neural network algorithm including probabilistic method, multi-layer feed forward and radial basis function are used to estimate the target parameter, with optimal combination of available attributes and the use of neural network algorithm training from the optimal attributes using the Hampson-Russell software. In the last step, multi-attribute analysis is compared with three neural network algorithms. The results indicate a higher correlation coefficient for probabilistic neural network than that of multi-attribute analysis for determination of the brittleness index.

Measuring volume fractions and identifying the flow regime are important challenges in the oil industry. In the present study, three different flow regimes were simulated by MCNPX code. A 137Cs source and two NaI detectors have been used in order to count the transmitted photons. The counted data had high-frequency noises. In order to tackle this problem, a Savitzky-Golay filter was applied. Therefore, four features in the time domain including STD, Skewness, Kurtosis, and Maximum Value were extracted. It was found that the extracted features are not capable of separating the flow regimes completely, without overlap. Accordingly, three different features from registered data of both detectors were extracted. After investigating all the possible statues, two ANNs were implemented to identify the flow regimes and predict the void fraction, respectively. By applying this method, all the three flow regimes were correctly distinguished and void fraction was predicted with root mean square error (RMSE) of less than 0.59.

Topography is usually resulted from the patterns of the plate tectonics and faults in relation to each other. If we can produce a model for these actions and reactions after the complete recognition of the fault charactersitics based on their slip rates in the considered region, an applied model is obtained which reconstructs the topography of the region. A more important fact is that this model which illustrates the relation between topography (as the super-structure) and faults interaction of the considered region (as the infra-structure), can be used as a criterion to recognize the undiscovered fault's structures of the study area. It can provide us a chance to determine the fault parameters such as slip rates. Iran is known as an area which is subjected to the high possibility of the earthquakes as a natural hazard. Thus earthquake studies are important to investigate this hazard. In this study, a model of the topography is constructed in the region which is prone to earthquakes. The model is compared with the digital terrain models (DTM) of the area resulted from the satellite image data sets. This comparison provides us a structural control on the faults of the region. Our case study is modeling the relationship between faulting and the topography in the North Alborz. Results of the study let us obtain criteria for understanding and prediction of the fault structures that have created the topography. In order to achieve this goal, we consider a DTM of Alborz region. With variation of the parameters of the faults and creating various fault models, an estimate of the elevation model of the region is constructed for the life time of the faults. We consider the variation domains for five parameters of the activity period, slope, horizontal slip rates in length and slope directions and the vertical slip rate of the main faults of the region. The optimum values of these parameters are obtained based on the neural network optimization method. Then, the topography of the region is modeled based on the numerical results of the method for the unknown fault parametrs. We use the algorithm of the method with some selected variation ranges for the values of the parameters for each selected fault of the region. For the slope parameter of the faults, a range of 30 to 90 degrees, for the horizontal slip rate on the length direction and the slope direction of the fault, a range of -3 to mm per year and for the vertical slip rate, a range of -0.5 to mm per year have been considered as variation ranges of the unknown parameters in the algorithm. These variation ranges are considered based on the previous studies on geodynamic setting of the region. The parameters of the nine main acitive faults in the North Alborz region make the assumed fault system of the region and the modeled topography of the region to be generated. Comparison of the modeled topography with the real elevation model of the region can be evaluated as how changing the data of fault parameters and the possible reconfiguration of these parameters, such as the slip rate, time activity and the slope fault can obtain more suitable results in the modeling. In other words, the method can be used to determine which configuration and fault structure in study region will lead us to a more consistent model and provides us the possibility of modeling the topography of the region based on the fault structures. The numerical results show the differences from -85 m to 236 m between the model result and the real elevation model. In addition, the root mean square error between the DTM and the model is 61.1 m which is an acceptable result, due to the fact that only five parameters are variable and just nine faults are calculated while ignoring the effects of erosion and sedimentation of soil in the final format of the earth topography. More accurate results can be obtained by increasing the number of faults and their parameters in the model.

The ionosphere as the upper part of Earth’s atmosphere consists of electrons and atoms affecting the signal propagation in the radio frequency domain. Nowadays, Global Navigation Satellite Systems (GNSS), like GPS, are widely used for various applications. The majority of navigation satellite receivers operate on a single frequency and experience an error due to the ionospheric delay. They compensate for the ionospheric delay using an ionospheric model which typically only corrects for 50% of the delay. An alternative approach is to map the ionosphere with a network of real-time measurements. Global Positioning System (GPS) networks prepare chance to study the dynamics and continuous variations in the ionosphere by complementary ionospheric measurements, which are usually obtained by different techniques such as ionosondes, incoherent scatter radars and satellites. The ionospheric delay is characterized by the Total Electron Content (TEC) along the signal path from the satellite to the receiver. Elimination (or reduction) of the ionosphere effects is possible using dual-frequency receivers by a very useful combination of dual-frequency data known as geometry free combination (L4) as follows: where the L4 signature is derived from L1(1.57542 GHz) and L2(1.22760 GHz) phase observables. Single-frequency users, however cannot take advantage of this combination. So, they have to use a proper ionospheric model to correct the ionospheric delay. Ionosondes (up to an altitude of 1000 km) can determine TEC while GPS measurements give completely information about the topside ionosphere. In this paper, the suitability of Neural Networks (NNs) in order to predict the Total Electron Content (TEC) obtained from Iranian Permanent GPS Network (IPGN) during the low-solar-activity period 2006 has been investigated. TEC has many non-linear variations while the neural network has a significant ability to model and approximate it (Williscroft and Poole, 1996; Hernandez-Pajares et al., 1997; Xenos et al., 2003; Sarma and Mahdu, 2005; Leandro and Santos, 2007). The input space included the day number (DN,seasonal variation), hour (HR,diurnal variation), sunspot number (SSN,measure of the solar activity) and magnetic index (measure of magnetic activity). To make the data continuous, the first two parameters were each split into sine and cosine components, two cyclic as follows: where DNS, DNC, HRS and HRC are the sine and cosine components of DN and HR, respectively. In this paper, the TEC values have been estimated using the PPP (Precise Point Positioning) module of the Bernese over Iranshahr (27˚N, 60˚E). Optimum situation of the neural network include of single hidden layer and eight neurons of inputs layer and fifty neurons of hidden layer and one neuron of output layer. To this end, the single hidden layer feed-forward network with a back propagation algorithm has been designed. An analysis was done by comparing predicted NN TEC(TEC values predicted by the NN model) with TEC values from the IRI2007 version of the International Reference Ionosphere, validating GPS TEC(TEC values calculated from the GPS measurements) with the maximum electron density obtained from ionosonde and calculating the performance of the NN model during equinoxes and solstices. The results show high correlation with GPS TEC and NN TEC. Their Root-Mean-Square Error(RMSE) and coefficient of determination (R2) are 1.5273 TECU and 0.9334 respectively. RMSE is defined as: where N is the number of data points. In table 3, the absolute error (Eabs) is defined as the magnitude of the difference between the NN predicted TEC and the GPS TEC, while the relative error(Erel) is the ratio of the absolute error to GPS TEC and can be represented as a percentage (Habarulema et al, 2007).These errors were calculated as follows: is the absolute error and is the relative error respectively. The difference (100- % gives the relative correction, which indicates the approximate TEC prediction accuracy for the NN model (Leandro and Santos, 2007). An average error of ~11.41% means that the NN can predict about 88.58% of the GPS TEC on average. Results show that the neural network works better rather than the IRI model for IRAN.

Identification and tracking of solar coronal loops is key to understanding solar magnetic field. Slow and fast Magnetohydrodynamic oscillation of tracked loops from sequence 171Å extreme ultra – violet images was detected. The method was demonstrated using 171Å images taken by SDO/AIA on 14 August 2010 and 20 January 2012. Two dimensional continuous wavelet transform (CWT) was used to clarify images of loops and to eliminate additional noises. Applying OCCULT method، the loops were labeled and their widths were determined. The Zernike moments of loops as an invariant property (transformation، scale and rotation) of loops were calculated. Then، the Probabilistic Neural Network (PNN) was used for identification of the same loops from sequence images. Slow longitudinal Magnetohydrodynamic oscillations were extracted by averaging pixels intensities perpendicular to tube axis. Fast oscillations in distinguishable tubs were observed observable by averaging pixels’ intensities in direction of tube axis. Also، periods، phase velocity، and damping time of oscillations were computed

In this paper, the method of identifying of the similar solar flux tubes from image sequence EUVI/STEREO is presented. Using the oriented coronal curved loop tracing, the loops of an image are labeled. Based on local maxima intensities, the width of loops is determined. The Zernike moments of each loop are calculated and fed to probabilistic network classifier. Also, 588 loops from STEREO are studied.

Cs-137 is one of the fission products that is usually released in environment after nuclear accidents. This contamination remains in environment for a long time due to long half life of Cs-137 (30 years) and can enter easily into the human food chain. A two-compartmental model was implemented to describe caesium intake and its distribution in Dicentrarchus Labrax, using a proposed differential equation model. The model included two compartments, the first compartment was the blood and the second one was the tissue. The activity of Cs-137 was undertaken in each compartment by means of a numerical method and the activity of Cs-137 was considered as an input of compartmental equations. We obtained the transfer coefficients between fish tissues by comparing the radiation curves with the actual data. In the light of the differences with the transfer coefficients, the calculation by the COMKAT software was found to be about 2%. Then, we provided the activity curves of Cs-137 and their charactristics (feature extractions) by changing the transfer coefficients and they were utilized to train the neural network. The network was trained for six data groups, and the results of the network testing had about 99% correct response, therefore it can be employed to estimate the transfer coefficients in fish tissue, the salinity range, and the activity of Cs-137 in water.