Transactions on Machine Intelligence2821-16931420181201A Model for Extracting the Velocity Features of Fluid Movement in Flotation Cells of the Shahr-e Babak Copper Complex Using the PIV Method18219020600810.47176/TMI.2018.182ENM.Eghbali FardMSc Student, Department of Electrical Engineering, Vali-e-Asr University of Rafsanjan, Rafsanjan, IranM. H.GholizadehAssistant Professor, Department of Electrical Engineering, Vali-e-Asr University of Rafsanjan, Rafsanjan, Iran0000-0003-4337-5067H.GhayoumizadehAssistant Professor of Biomedical Engineering, Vali-e-Asr University of Rafsanjan0000-0002-5390-3938H.Fatehi MarjAssistant Professor, Department of Electrical Engineering, Vali-e-Asr University of Rafsanjan, Rafsanjan, Iran0000-0002-1634-8197Journal Article20180406In this study, the flotation process is examined as a conventional method for processing low-grade copper sulfide ores. The performance of this process is influenced by various factors, including the type and dosage of collector, frother, pH regulator, activator, and depressant. Identifying effective reagents and determining optimal conditions are crucial for enhancing efficiency, and modeling and simulation of this process can further aid in its improvement. Traditionally, experienced operators assess and control the process based on the visual characteristics of the froth. However, with advancements in technology, machine vision has emerged as a valuable tool for monitoring and controlling flotation circuits. In this context, a machine vision system was installed on a flotation cell in the rougher circuit of the Shahr-e Babak Copper Flotation Plant to monitor the process under various conditions. The primary visual feature extracted from the captured images is the bubble velocity on the froth surface, which directly impacts flotation performance. In this research, the Particle Image Velocimetry (PIV) technique was employed to determine bubble velocities without the need for additional particles or lasers. Simulation results demonstrate that the proposed method exhibits high accuracy in providing relevant velocity features, making it a potent tool for monitoring and optimizing the performance of flotation cells.https://www.tmachineintelligence.ir/article_206008_9cbd8c93230ba86ce6c3d13ba9604ed2.pdfTransactions on Machine Intelligence2821-16931420181201Designing a Fuzzy Controller Based on Disturbance Observer and Smith Predictor for Linear Uncertain Time-Delay Systems19120320600910.47176/TMI.2018.191ENH.AsmaeM.Sc. Student of Electrical Engineering, Control, K. N. Toosi University of Technology, Tehran, IranA.KaramiPh.D. Student of Electrical Engineering, Kermanshah Branch, Islamic Azad University, Kermanshah, IranJournal Article20180602In control systems, conventional disturbance observer (DOB) structures often fail to perform optimally when the system involves delays, leading to inefficiencies in disturbance rejection. This paper addresses this limitation by first employing the conventional Smith predictor method to compensate for the negative effects of delay within the control loop. Following this, a conventional disturbance observer is implemented to estimate external disturbances. However, this approach has certain constraints, particularly in handling unpredictable or complex disturbances. To overcome these challenges, this study explores an alternative approach: the Composite Disturbance Observer (CDOB), which is based on the concept of Network Disturbance (ND). In this framework, system delay is treated as an external disturbance, and since the transformed system becomes delay-free, the conventional DOB structure can be applied effectively. A key advantage of the proposed CDOB method is its ability to estimate and compensate for disturbances without requiring prior knowledge of the delay or assuming specific conditions such as periodic disturbances, which are often considered in related studies. Furthermore, a fuzzy PID controller is utilized for process control, offering adaptive tuning capabilities to enhance performance. The proposed approach is validated through comprehensive simulations, which demonstrate the superior performance of the CDOB in mitigating the adverse effects of delay while ensuring robust disturbance rejection. The results highlight the effectiveness of this method in improving system stability and response under various operating conditions.https://www.tmachineintelligence.ir/article_206009_3b8cceb40f8ed51ff1355bca1f8651e1.pdfTransactions on Machine Intelligence2821-16931420181201Evaluation of Grid Performance on Image Quality in Digital Mammography Systems20421122032310.47176/TMI.2018.204ENH.Khodajou ChoukamiPhD, Department of Biomedical Engineering, Faculty of Electrical Engineering, Sharif University of Technology, Tehran, Iran0000-0002-8056-0379B.Vosooghi VahdatPhD, Department of Biomedical Engineering, Faculty of Electrical Engineering, Sharif University of Technology, Tehran, Iran0000-0002-5987-1703A.HosseiniPhD, Faculty of Energy Engineering, Sharif University of Technology, Tehran, IranS. M.RazaviMSc, Nik Parto Nuclear Medical Center, Sadeghieh, Tehran, IranJournal Article20180503Mammography is a crucial imaging technique for early breast cancer detection, where image quality plays a vital role in accurate diagnosis. One of the primary factors degrading image quality is scattered radiation, which reduces contrast and obscures fine details. Anti-scatter grids are widely recognized as the most effective tool for mitigating this issue. Initially developed for screen-film mammography, grids have since been integrated into digital mammography systems. However, despite their widespread adoption, their geometric performance in digital systems has not been thoroughly investigated. This study aims to fill this knowledge gap by evaluating the effectiveness of grids in digital mammography. To achieve this, a comprehensive simulation of a mammography system was conducted based on the international standard IEC 60627:2013, utilizing the latest version of the MCNPX 2.7 Monte Carlo code. The signal-to-noise ratio improvement factor (SNRIF) was used as the key metric to assess the impact of grids on image quality. Various grid parameters, including grid ratio, lead strip thickness, and line density, were analyzed to determine their influence on scattered radiation reduction. The simulation results indicate that grids with thinner lead strips, higher grid ratios, and lower line densities significantly enhance image quality in digital mammography. These findings provide valuable insights for optimizing grid design and improving image contrast, ultimately contributing to more accurate breast cancer diagnosis.https://www.tmachineintelligence.ir/article_220323_ec486fcb2e63432ca207ca1d168c92d5.pdfTransactions on Machine Intelligence2821-16931420181201Real-time Driver Drowsiness Detection Using Artificial Immune System21221922032810.47176/TMI.2018.219ENH.SoltaniZadehAssistant Professor, Department of Electronics, Faculty of Electrical Engineering and Computer Science, Semnan University, Semnan, Iran0000-0002-2210-675XS.BourbourTelecommunications Department, Faculty of Electrical Engineering and Computer Science, Semnan University, Semnan, IranJournal Article20180703Driver drowsiness is considered one of the primary causes of traffic accidents. Drowsiness detection systems are typically categorized into two types: monitoring-based and vehicle motion-based systems, each with its own advantages and disadvantages. Monitoring-based methods, which utilize driver performance sensors, are considered more practical than other methods due to their non-intrusive nature. In this study, images of open and closed eyes are initially provided to the designed system for training purposes. Then, using the trained system, prolonged eye closure is detected. The images used for the training phase were collected from 5 individuals, with 40 pictures from each person, including images of the left and right eyes. The PCA algorithm is first used to extract features, and the data is then fed to various classification systems. For the testing phase, the same number of new images were used. Four classification methods Euclidean Distance, Max Likelihood, Neural Networks, and Artificial Immune System (AIS) as the proposed method were compared and evaluated. The results showed that the first two methods, despite not requiring a training phase, needed more testing time. Neural networks, despite shorter testing times, required significant training time. On the other hand, the Artificial Immune System required short times for both the training and testing phases, with no significant difference in the recognition accuracy across the methods.https://www.tmachineintelligence.ir/article_220328_9c51632666d941890fe8d21df76df4c5.pdfTransactions on Machine Intelligence2821-16931420181201Skin Cancer Detection from Dermoscopic Images with Emphasis on Shape, Color, and Texture Feature Extraction Using a Two-Stage Classification Algorithm22022722032910.47176/TMI.2018.220ENGh.ShakourianMSc student, Department of Biomedical Engineering, Faculty of Electrical and Computer Engineering, Babol Noshirvani University of Technology, Babol, IranH.Montazery KordyAssistant Professor, Department of Biomedical Engineering, Babol Noshirvani University of Technology, Babol, Iran0000-0002-2010-4945Journal Article20180615Malignant melanoma is one of the most aggressive and life-threatening forms of skin cancer, with a high potential for metastasis if not diagnosed and treated early. The definitive treatment for melanoma is possible when it is accurately detected by a trained specialist in a timely manner. Early detection can lead to a simple excision of the tumor, which can often result in complete cure. However, current diagnostic procedures typically involve a biopsy of the lesion, an invasive and often painful procedure that can cause discomfort to the patient. Given these challenges, this study aims to develop a more efficient, non-invasive method for the early detection of melanoma, leveraging advanced machine learning and image processing techniques. The proposed method utilizes a set of features based on the shape, color, and texture of dermoscopic images, which are extracted to capture critical characteristics of the lesion. These features are then analyzed through a two-stage classification process, designed to categorize the lesion into one of three categories: common nevus, atypical nevus, and melanoma. The method was tested on the PH2 dataset, a well-known dermatological dataset containing images of skin lesions. Results demonstrate that the two-stage classification model achieved an accuracy of approximately 90%, significantly outperforming traditional single-stage classification models for multi-class lesion classification. This innovative approach holds promise for enhancing the accuracy of melanoma detection and reducing the need for invasive procedures, ultimately improving patient outcomes.https://www.tmachineintelligence.ir/article_220329_b98c5413a421a7a962e149f1aff2807d.pdfTransactions on Machine Intelligence2821-16931420181201Designing an Appropriate Neural Network to Assist in More Accurate Stock Portfolio Predictions22824222033010.47176/TMI.2018.228ENM. A.AlipourMaster’s in Computer Engineering, Software Orientation, Islamic Azad University, North Tehran Branch, IranSh.ParsinejadMaster’s in Information Technology Engineering, E-Commerce Orientation, University of Guilan, Rasht, IranJournal Article20180709The increasing complexity and volatility of financial markets demand sophisticated methodologies for stock portfolio predictions. Recent advancements in deep learning and machine learning have provided new avenues for enhancing prediction accuracy in this domain. This literature review synthesizes current research findings on neural network designs specifically tailored for stock portfolio predictions, identifies knowledge gaps, and suggests potential future research directions. In this study, by designing feedforward artificial neural networks (MLP) and feedback networks (NARX), we examined the behavior of these two artificial neural network models for predicting stock portfolio prices. Subsequently, an Adaptive Neuro-Fuzzy Inference System (ANFIS) was developed and employed to forecast the stock prices of Ford Motor Company. The output of this system was compared with that of the designed artificial neural networks. The results of the simulation indicate that the designed neural network outperforms the ANFIS model in terms of prediction accuracy, exhibiting a lower error rate. The designed artificial neural network was able to predict the next day's closing stock price with minimal error on a daily basis. Additional findings from this project revealed that the ANFIS model performs effectively with limited data, making it suitable for scenarios where comprehensive data collection is either costly or impractical.https://www.tmachineintelligence.ir/article_220330_45323021e2db8ba1f29b24c3e750b208.pdf