فصلنامه مدل سازی در مهندسی
پیاپی 72 (بهار 1402)

Until now, there has been no certainty based on the success or failure of using feature selection methods to increase the efficiency of SER systems. This article discusses feature selection for data augmentation in a speech emotion recognition system. The experiments were performed on four databases: EMO-DB, eNTERFACE05, SAVEE, and IEMOCAP. Simulations are performed in Python software and in addition, data analysis was performed on all four databases for four emotions of sadness, anger, happiness, and neutral. This paper discusses feature selection intending to create a GAN to augment data in a speech emotion recognition system. It will demonstrate that artificial data generated by GANs can not only augment data but also can be used to feature selection to improve classification performance. We used a GAN to augment data and selected two feature-selective networks including Fisher and LDA algorithm in two steps. SVM was also used to classify emotions. With the feedback taken from the classification network, we could bring the SER system to the optimal point of sample number and feature vector dimensions. The PCA is more effective on correlated data. The LDA algorithm works better on low-dimensional data. Fisher's method is better at reducing size than PCA. The results showed that the use of both LDA and Fisher methods in the GANs can filter the features in smaller dimensions while preserving the emotional information for classification. The results were compared with recent research and the proposed method was able to achieve 86.32% accuracy in the EMO-DB database.

Breast cancer is the most common cancer in women. The need to diagnose this disease in the early stages increases the chance of treatment. Individuals and reduction of mortality with artificial intelligence approach in medicine. In implementing this applied and supervised study, a histopathological microscopic two data set, including respectively 124 and 576 patients with invasive breast cancer was used. Data preprocessing and image quality improvement, then image segmentation with U-Net network to separate cancer cells from healthy breast tissue and remove pert data, then by combining deep neural networks to extract effective features and by method The majority of data is based on the classification and screening system for the diagnosis of invasive breast cancer carcinoma. Performance in diagnosis and classification Breast cancer is one of the features of this study compared to other studies. According to the results obtained, this study is a step towards helping physicians and specialists in increasing the accuracy and sensitivity of breast cancer screening at the most optimal time, to the lesions. Triad the high risk to appropriate secondary care and increase patients' chances of survival with timely treatment.

In this study, the possibility of energy recovery from oil pipelines as a source of clean energy is discussed. At present, a significant amount of energy is lost in oil pressure reducing stations. The energy regeneration scheme proposed in this study investigates the possibility of converting the currently wasted energy into the electrical energy by using a Francis turbine coupled with a synchronous generator system. The dynamic hydraulic model of the existing pipeline system starting from the compressor stations to the pressure regulating stations is obtained. Also the state space equations of the energy regeneration system including the control valves, the turbine and the synchronous generator connected to a local load are modeled. The propositional-integral-derivative (PID) and linear quadratic regulator (LQR) are designed to control the flow, pressure, speed and voltage in the energy regeneration system. The results of two controllers are then compared under different conditions. The obtained model for Razan-Shazand transmission line crude oil are simulated and the results are interpreted. The results show the possibility of a significant energy regeneration and the appropriate performance of the designed system with the proposed controllers.

Human voice contains characteristics such as: ethnicity, gender, feelings, age and other information, and speaker recognition identifies people based on their voice. Although researchers have worked in this area over the years and provide methods to improve the speaker recognition accuracy, there are still challenges. In this paper, a new speaker recognition method is proposed based on Gabor filter bank and convolutional neural networks. At first, spectrogram of the speech signal is formed and then, effective Gabor filter bank is designed so that these filters are suitable for extracting effective features of the speech signal. In the next step, spectrogram of the signal is passed through the Gabor filter bank to extract the speech signal features. Finally, speaker recognition is done using a convolutional neural network. Two datasets Aurora2 and TIMIT are used to evaluate the proposed method. Results show that the accuracy of the proposed method is competitive with the state-of-the-art methods.

The implementation of residential demand response has been a serious challenge for power system operators. The advent of smart grids has provided required communication infrastructures to exchange online signals with smart homes. In this paper, a new model for residential demand response is presented. The model divides the household loads in three categories as uncontrollable, controllable with thermostat, and controllable without thermostats. The proposed demand response model is a mixed real-time pricing (RTP) and incentive-based program capable of reflecting hourly prices while motivating residents to participate in the requested demand side management scheme. The proposed model was validated for a residential complex including 50 smart homes using GAMS software. Results verify the efficiency of the model to shed the peak load and shift the starting time of home appliances appropriately. A sensitivity analysis was conducted to investigate the impact of welfare violation cost and incentive rate on numerical results as well.

Analyzing the opinions of social media users can lead to a correct understanding of their attitude on different topics. The emotions found in these comments, feedback, or criticisms provide useful indicators for many purposes and can be divided into negative, positive, and neutral categories. Sentiment analysis is one of the natural language processing's tasks used in various areas. Some of social media users' opinions is are multimodal and share a combination of multiple media, including text, image and imoji, which provide a useful structure for extracting and better understanding emotions. This paper presents a hybrid transfer learning method using 5 pre-trained models and hybrid convolutional networks for multimodal sentiment analysis. In this method, 2 pre-trained convolutional network-based models are used to extract the properties of images, and 3 other pre-trained models are used to extract the properties of texts and embed words. The extracted features are used in hybrid convolutional networks, and the visual attention mechanism is used to focus on the most important emotional areas of the images and the multi-head attention mechanism is used to highlight the emotional words. The results of the classification of images and texts are combined using the voting technique, and finally the late fusion is used to determine the polarity and the final label. The results of empirical experiments of the proposed model on a standard data set show 96% accuracy.

Active power filter (APF) can significantly compensate the current harmonics produced by nonlinear loads. To do this feature, harmonic detection beside harmonic generation play vital role. Direct power controller (DPC) has great harmony with instantaneous power compensation (PQ) algorithm as well as ability to eliminate internal current loops. In addition, deadbeat controller (DBC) has high compatibility for digital implementation, superior control performance and fast dynamic response. However, DBC suffers from time delay regarded to control action calculation and digital implementation. In this paper, a new deadbeat-based DPC method is proposed firstly to control and generates the harmonic currents of APF. Several simulations achieved in MATLAB/SIMULINK beside different experimental tests obtained from a DSP-based active power filter are conducted to illustrate and prove the effectiveness and superior performance of the proposed control system. As a result, the THD of grid current decreases from 22% to 3.2% under steady state where dynamic response under with/without compensation validate significant transient response lower than 5ms.

In Junctionless transistors, the source-channel-drain doping is of the same type and level, hence, the process of making Junctionless transistors is easier than inverting mode transistor. Despite this benefit, reducing the transconductance of Junctionless transistors due to reduced carrier velocity makes the operation of this type of transistor difficult for analog, radio frequency and high frequency noise usages. An effective method that increases the trans-conductance of Junctionless transistors without reducing efficiency is using a heterogeneous structure in the channel. In the present article, using Si and Si1-xGex materials in the channel is proposed and modeled so as to enhance the transconductance of Junctionless transistor. The special structure of the proposed transistor, called JL-Si / Si1-xGex, eliminates the intervalley scattering between valleys of ∆2 and ∆4. This increases the velocity of the electron and consequently enhances the transconductance. The outcomes of the modelling of the proposed JL-Si / Si1-xGex heterostructure transistor indicate the maximum transconductance of 2.5 mS / um, which increases 50% compared to similar silicon transistor. Moreover, calculations which are extracted from modelling demonstrate that the proposed JL-Si / Si1-xGex transistor has a unity gain cutoff frequency of 750 GHz, minimum noise figure of 65.0 dB, and an available gain of 28.5 dB. The parameters of cut-off frequency, minimum noise figure and available gain of the proposed JL-Si / Si1-xGex transistor have been improved by 34%, 62.5% and 53%, respectively, compared to the JL-Si transistor with similar dimensions. The proposed device can be suitable candidate for RFIC applications.

As regards to daily developments in micro/nano technology and very special intricate missions in micro/nano scale, many researches focused on manipulation and movement in correspond scales. Micro-grippers are end effectors of micro/nano manipulation systems which perform some duties such as holding, picking up, moving and cutting. In this research a micro-gripper has been introduced which consists of two perpendicular Micro-beams. Each Micro-beam has an elastic layer and two piezoelectric layer. One of these piezoelectric layers is used for actuation and another one for sense and feedback control. First static equations of Micro-beam presented. Kinematics of Micro-gripper including direct and inverse kinematics and static displacement has been presented. Eventually Micro-gripper workspace for a Micro-beam’s ultimate displacement has been specified. The results have been verified by FEM software. The difference between the analytical results of displacements in different directions with the corresponding results of finite element software output is about 2 percent. Therefore, the results indicate the accuracy and efficiency of the proposed method and the current research can be a framework and foundation for the analysis of similar systems with more complex geometry.

In this paper, the air flow around a root airfoil model of horizontal axis wind turbine, namely SG6040 is numerically simulated using finite volume method by three turbulence modelling k-ω, k-ω SST, and Spalart-Allmaras in two Reynolds Number of 137,122 and 149,969 and at 0 to 10 degrees angles of attack. The main objective of this paper is to provide a proper numerical model for simulating the root airfoils of wind turbines with small rotor sizes. Also, the effect of angles of attack and Reynolds number on aerodynamic coefficients were investigated and the results were compared with each other. The results show that the k-ω SST turbulence model presents better results than other models and is more consistent with experimental results. Also, by increasing Reynolds number, turbulence models seem to have the same trend and the greatest differences occur in the low Reynolds numbers. The results obtained in this case study can play a significant role in designing wind turbine blades.

In this study, a comparison among the empirical mode decomposition, ensemble empirical mode decomposition and Morlet continuous wavelet transform in fault diagnosis of bearings are performed. A Morlet wavelet support vector machine with one against one strategy that was optimized by a genetic algorithm was used for fault classification. A scale selection criterion based on the maximum relative energy to Renyi entropy ratio is proposed to determine the optimal decomposition scale for wavelet analysis. A comparison between the performances of optimized and non-optimized of support vector machines were also carried out. Vibration signals were collected by a test rig for different fault of a bearing such as normal case, bearing with inner and outer race fault, and bearing with ball fault and combine fault. After the processing of vibration signals their frequency components, several statistical features were extracted from each frequency component as input of wavelet support vector machine for the fault classification of ball bearings. For reducing of time and process of fault diagnosis, optimum feature sets of statistical parameters are selected by Utans method. K-fold cross validation method is used for evaluation of classifier. The results show that continuous wavelet transform with Morlet base has higher accuracy with respect to other methods in fault classification of bearings.

Today, with advances, researchers are looking for a way to prevent human and financial damage from collisions. In this research, the energy absorption of nanocomposite structures with different percentages of clay and nan osilica nanoparticles along with diethylene triamine has been investigated using epoxy resin polymer and nanoparticles. And the samples under compressive axial loading have absorbed different amounts of energy. Therefore, the purpose of this study is to investigate the different effects of nanomaterials and different percentages. The design of this experiment has been done by Design-Expert software with a parametric study and Box Behnken Design (BBD). The range of changes in nanomaterials was 0-0.4 % and also diethylene triamine with weight percentages of 1, 3, and 5 % was used. The results show that the materials used in the study had a positive effect on energy absorption and the failure of the structure and the sample with weight percentages of 0.4% nano clay, 0.4% nano-silica, and 3% diethylenetriamine is the most optimal energy absorption mode. Has had a special.

Automatic pavement crack detection is essential for assessing road maintenance and ensuring safe driving. Traditional crack detection has problems such as low efficiency and lack of complete detection. This study aims to solving the problems of traditional crack detection methods and using deep learning models. We proposed a method based on object detection algorithms for pavement crack detection and discussed the latest YOLOv5 series models for pavement crack detection while explaining the theoretical concepts. Finally, a crack detection model and effective pavement management is presented. The proposed model can determine the type, position and geometric characteristics of cracks accurately and at a higher speed in comparison with other methods. For this purpose, the images that had been taken from the asphalt of Mashhad roads were used to train and evaluate the model. Images were labeled for both linear crack and surface crack. Proposed model is developed using five YOLOv5 series algorithms and transfer learning and were evaluated for accuracy and speed of prediction. The models’ accuracy is between 77 to 98% and the prediction speed is between 17.4 to 105 milliseconds, which indicates the optimal performance of the models. The v5s model with 92.8% accuracy and a speed of 23.9 ms is selected as the final model for real prediction of cracks in one of the main thoroughfares of Mashhad. Based on the dimensions and the type of predicted crack and the use of the proposed decision tree, the maintenance approach for each part was determined.

In this research, utilizing the concepts of thermohydraulic theory, optimization of cost due to the scale formation in heat exchangers. The objective function involves the minimization of the energy consumption and production losses. First, via ASPEN we simulated shortcuts of exchangers in order to examine the effects of temperature and velocity on the scale formation. The project optimization was done by GAMS and due to the presence of nonlinear terms in the scale rate equations using CONOPT solver. For better results, Heat Exchanger 4131 and 4132 in Khorasan petrochemical Co. urea unit were used.The investigation shows that for the twin exchangers under study, first, the scale formation reduces cooling water flow by up to 40% in E4132. Second, the optimal values for minimizing cost of scale formation in the exchangers occurs at a cooling water flow rate of 152442.75 kg/h for the exchanger 4132 to reduce the cost by 15%.

Nowadays, with the expansion of various methods of obtaining information from different systems and digital cameras, the use of image processing has been widely used. Due to the increase in the huge amount of data in the industry and the need to control and monitor this amount of data, the use of image processing methods has been widely used in various industries. The analysis of these data at different levels, including the control of production line processes, is one of the most up-to-date methods in the quality control system of the statistical process of products. Therefore, using new approaches for data monitoring and quality control of manufactured products can be considered as a fruitful method. The aim of the current research is to provide a model to control the quality of manufactured products based on the image processing technique approach and using fuzzy transformations for image compression and image data processing. In order to validate the results of this research, MATLAB software was used and the developed model was compared with previous studies. The simulation results show that the proposed GLR control chart is effective in reducing the error and increasing the detection accuracy. So that in the first image after applying the changes in brightness, the warning showed that this represents the proper performance of the proposed control chart against the changes.

Undoubtedly, the best tool to help military decision-makers and commanders choose the appropriate strategy is to formulate models that are as close to the real world as possible. This is possible when these problems are presented in conditions of uncertainty. In this research, a multi-period dynamic interdiction problem in fuzzy stochastic conditions is investigated. In general, interdiction problem, scarce resources are allocated to degrade the enemy, whose behavior is formatted by the network optimization problem. In this problem, the defense forces in the role of interdictor try to minimize the maximum flow during the period so that at each stage the interdictor and the enemy are fully aware of the performance of the other side. Edge capacities in this model are considered as fuzzy stochastic variables. To solve the proposed model, first, the fuzzy stochastic dynamic interdiction problem is transformed into the deterministic dynamic interdiction problem with the help of the concepts of probability measure, credibility measure and chance constraint programming. Then, by creating the crisp two-level problem created by duality, it is transformed into a single-level problem, and then it is solved by using the generalization of Banders decomposition algorithm. Finally, the validity of the problem is evaluated by providing a numerical sample.