1. Development of a methodology to calibrate a pedestrian microsimulation model : doctoral dissertationChiara Gruden, 2022, doktorska disertacija Opis: Walking, as a mode of transport, is becoming widespread, in a world, where urban conglomerates are broadening and becoming denser. Modern lifestyle trends on a side, and eco-friendly policies on the other, push people into walking habits, increasing the need for a suitable, attractive, accessible, connected and safe walking infrastructure. To reach such a result, it is necessary to understand, what are the needs of the users of this infrastructure, taking into consideration the behavioral specificities and the safety needs of pedestrians. In this process pedestrian microsimulation models, surrogate safety techniques, and technologies able to measure specific traits of pedestrian dynamics play a central role. The firsts allow to reproduce repeatedly in a virtual environment a specific infrastructure and to study the response of pedestrians. Nevertheless, to be accurate and efficient, they need to go through long and tedious calibration and validation processes, that are often seen as an important limitation by technicians. Surrogate safety techniques are methods, that are based on the concept, that it is possible to predict the safety level of a location, using near accidents. The main advantage of such techniques is that they are proactive. Till this moment, these techniques have been mainly applied to on-field measurements and are primarily centered on motorized road users. Less interest has been shown for vulnerable road users, especially for pedestrians, who have been less extensively studied. Finally, an element that could highly affect pedestrian safety is their reaction time. Nevertheless, its measurement has long been a big issue. Eye-tracking technology could be one of the solutions, allowing to analyze the directions and objects fixated by pedestrians. These listed issues are also the topics that are addressed by this research work. Focusing on the study of the action of pedestrians while crossing the road on an unsignalized crosswalk set on a roundabout entry leg, the dissertation thesis aims at studying the crossing time, reaction time and surrogate safety aspects typical of pedestrians at the recalled location. The main purpose of the research work is to develop a methodology to calibrate pedestrian Social Force Model at a selected location, using a specifically formulated neural network as a tool to fine-tune model's behavioral parameters. Eight parameters have been chosen to be fine-tuned, five of those are related to pedestrian behavior and three of them are related to car-following behavior. After the selection of input parameters, a feedforward network has been formulated. Its application in the framework of the whole calibration process has brought to considerably positive results, finding a combination of input parameters that improved the performance of the microsimulation model of 37 % in comparison to the default one. The outputs of the calibrated model have been used to calculate three measures of surrogate safety, and also in this case results demonstrated an improvement in the calculation of surrogate safety measures when using the calibrated outcomes in comparison to their calculation on the “default” model outputs. Finally, reaction time measurement and prediction have been addressed by the thesis, in order to be able to describe pedestrian crossing action in its completeness. Quantitative eye-tracking outputs have been the starting point for the calculation of pedestrian reaction time at different locations, and they allowed to create a database of behavioral, geometric, regulatory and flow characteristics, which was the foundation for the formulation of a new prediction model for pedestrian reaction time. The prediction model, which consists of a cascade-correlation neural network, gave a good response to the learning and generalization steps, turning a 74 % correlation between the measured reaction time values and the predicted ones, and being able to follow the variability of these values. Ključne besede: pedestrian, microsimulation model, calibration, neural network, surrogate safety indicators, reaction time. Objavljeno v DKUM: 03.10.2022; Ogledov: 284; Prenosov: 42 Celotno besedilo (5,93 MB) |
2. ARM-Based Video Intercom System with Next-Gen Human Presence Detection using Deep Learning : magistrsko deloMario Gavran, 2022, magistrsko delo Opis: This master's thesis presents an advanced video system with human presence detection based on deep learning and an ARM microcontroller. The objective of the thesis is to develop a system that works as a smart video intercom, which could be installed, e.g. on the entrance door, and autonomously alert the owner that a guest is in front of the door. The main goal is to use an AI algorithm, namely the neural network model on a constrained device, such as an ARM microcontroller, as their main advantage is lower power consumption and cost.
The thesis also describes commonly used methods to reduce the power and memory footprint and to implement and accelerate the deep learning algorithms more effectively. Further, the most notable deep learning hardware and some general platforms are described in more detail.
The thesis also presents the development of a human presence detection system based on an ARM microcontroller, VGA camera, and LCD, where Tensorflow Lite Micro, an open-source C++ framework for deploying deep learning models to embedded platforms and a pre-trained neural network model for person presence detection are used. Ključne besede: TensorFlow Lite Micro, Video intercom system, ARM Cortex-M microcontroller, Human presence detection, Neural network Objavljeno v DKUM: 08.07.2022; Ogledov: 249; Prenosov: 43 Celotno besedilo (7,86 MB) |
3. Development of a Model for Predicting Brake Torque Using LSTM and TCN Models : magistrsko deloTomaž Roškar, 2020, magistrsko delo Opis: The main purpose of this thesis is to compare two state-of-the-art machine learning models, LSTM (Long Short-Term Memory) and TCN (Temporal Convolutional Network), on an AVL List GmbH case use, where the goal is to predict vehicle brake torque. Dataset used for model testing consists of multiple features which are preprocessed using several preprocessing methods. For model implementation Python’s libraries Keras and TensorFlow are used. Results from this thesis show that TCN is able to outperform LSTM. TCN achieves lower RMSE on the test dataset and is significantly faster in training and evaluation. Ključne besede: brake torque, machine learning, neural network, LSTM, TCN, RNN, CNN Objavljeno v DKUM: 24.09.2020; Ogledov: 780; Prenosov: 0 |
4. Estimating piping potential in earth dams and levees using generalized neural networksXinhua Xue, Xingguo Yang, Xin Chen, 2014, izvirni znanstveni članek Opis: Internal erosion and piping in embankments and their foundations is the main cause of failures and accidents to embankment dams. To estimate the risks of dam failure phenomenon, it is necessary to understand this phenomenon and to develop scientifically derived analytical models that are simpler, easier to implement, and more accurate than traditional methods for evaluation of piping potential. In this study, a generalized regression neural network (GRNN) technique has been applied for the assessment of piping potential, as well, due to its ability to fit complex nonlinear models. The performance of GRNN has been cross validated using the k-fold cross validation method technique. The GRNN model is found to have very good predictive ability and is expected to be very reliable for evaluation of piping potential. Ključne besede: piping, generalized neural network, cross validation, BP neural network Objavljeno v DKUM: 14.06.2018; Ogledov: 855; Prenosov: 59 Celotno besedilo (463,20 KB) Gradivo ima več datotek! Več... |
5. Web based education tool for neural network robot controlJure Čas, Darko Hercog, Riko Šafarič, izvirni znanstveni članek Opis: This paper describes the application for teleoperations of the SCARA robot via the internet. The SCARA robot is used by students of mehatronics at the University of Maribor as a remote educational tool. The developed software consists of two parts i.e. the continuous neural network sliding mode controller (CNNSMC) and the graphical user interface (GUI). Application is based on two well-known commercially available software packages i.e. MATLAB/Simulink and LabVIEW. Matlab/Simulink and the DSP2 Library for Simulink are used for control algorithm development, simulation and executable code generation. While this code is executing on the DSP-2 Roby controller and through the analog and digital I/O lines drives the real process, LabVIEW virtual instrument (VI), running on the PC, is used as a user front end. LabVIEW VI provides the ability for on-line parameter tuning, signal monitoring, on-line analysis and via Remote Panels technology also teleoperation. The main advantage of a CNNSMC is the exploitation of its self-learning capability. When friction or an unexpected impediment occurs for example, the user of a remote application has no information about any changed robot dynamic and thus is unable to dispatch it manually. This is not a control problem anymore because, when a CNNSMC is used, any approximation of changed robot dynamic is estimated independently of the remote's user. Ključne besede: LabVIEW, Matlab/Simulink, neural network control, remote educational tool Objavljeno v DKUM: 19.07.2017; Ogledov: 1093; Prenosov: 82 Celotno besedilo (792,88 KB) Gradivo ima več datotek! Več... |
6. Prediction of technological parameters of sheet metal bending in two stages using feed-forward neural networkJernej Šenveter, Jože Balič, Mirko Ficko, Simon Klančnik, 2016, izvirni znanstveni članek Opis: This paper describes sheet metal bending in two stages as well as predicting and testing of the final bend angle by means of a feed-forward neural network. The primary objective was to research the technological parameters of bending sheet metal in two stages and to develop an intelligent method that would enable the predicting of those technological parameters. The process of bending sheet metal in two stages is presented by demonstrating the various technological parameters and the test tool used to carry out tests and measurements. The results of the tests and measurements were of decisive guidance in the evaluation of individual technological parameters. Developed method for prediction of the final bend angle is based on a feed-forward neural network that receives signals at the input level. These signals then travel through the hidden level to the output level, where the responses to input signals are received. The input to the neural network is composed of data that affect the selection of the final bend angle. Only five different inputs are used for the total neural network. By choosing the desired final bend angle by means of the trained neural network, bending sheet metal in two stages is optimised and made more efficient. Ključne besede: bending in two stages, intelligent system, neural network, prediction of the final bend angle Objavljeno v DKUM: 12.07.2017; Ogledov: 940; Prenosov: 428 Celotno besedilo (900,30 KB) Gradivo ima več datotek! Več... |
7. Prediction of dimensional deviation of workpiece using regression, ANN and PSO models in turning operationDavid Močnik, Matej Paulič, Simon Klančnik, Jože Balič, 2014, izvirni znanstveni članek Opis: As manufacturing companies pursue higher-quality products, they spend much of their efforts monitoring and controlling dimensional accuracy. In the present work for dimensional deviation prediction of workpiece in turning 11SMn30 steel, the conventional deterministic approach, such as multiple linear regression and two artificial intelligence techniques, back-propagation feed-forward artificial neural network (ANN) and particle swarm optimization (PSO) have been used. Spindle speed, feed rate, depth of cut, pressure of cooling lubrication fluid and number of produced parts were taken as input parameters and dimensional deviation of workpiece as an output parameter. Significance of a single parameter and their interactive influences on dimensional deviation were statistically analysed and values predicted from regression, ANN and PSO models were compared with experimental results to estimate prediction accuracy. A predictive PSO based model showed better predictions than two remaining models. However, all three models can be used for the prediction of dimensional deviation in turning. Ključne besede: artificial neural network, dimensional dviation, particle swarm optimization, regression Objavljeno v DKUM: 12.07.2017; Ogledov: 839; Prenosov: 143 Celotno besedilo (1,17 MB) Gradivo ima več datotek! Več... |
8. Intelligent system for prediction of mechanical properties of material based on metallographic imagesMatej Paulič, David Močnik, Mirko Ficko, Jože Balič, Tomaž Irgolič, Simon Klančnik, 2015, izvirni znanstveni članek Opis: This article presents developed intelligent system for prediction of mechanical properties of material based on metallographic images. The system is composed of two modules. The first module of the system is an algorithm for features extraction from metallographic images. The first algorithm reads metallographic image, which was obtained by microscope, followed by image features extraction with developed algorithm and in the end algorithm calculates proportions of the material microstructure. In this research we need to determine proportions of graphite, ferrite and ausferrite from metallographic images as accurately as possible. The second module of the developed system is a system for prediction of mechanical properties of material. Prediction of mechanical properties of material was performed by feed-forward artificial neural network. As inputs into artificial neural network calculated proportions of graphite, ferrite and ausferrite were used, as targets for training mechanical properties of material were used. Training of artificial neural network was performed on quite small database, but with parameters changing we succeeded. Artificial neural network learned to such extent that the error was acceptable. With the oriented neural network we successfully predicted mechanical properties for excluded sample. Ključne besede: artificial neural network, factor of phase coherence between the surfaces, fracture toughness, image processing, mechanical properties, metallographic image, ultimate tensile strength, yield strength Objavljeno v DKUM: 12.07.2017; Ogledov: 1095; Prenosov: 403 Celotno besedilo (2,02 MB) Gradivo ima več datotek! Več... |
9. Characterization of Slovenian coal and estimation of coal heating value based on proximate analysis using regression and artificial neural networksDarja Kavšek, Adriána Bednárová, Miša Biro, Roman Kranvogl, Darinka Brodnjak-Vončina, Ernest Beinrohr, 2013, izvirni znanstveni članek Opis: Chemical composition of Slovenian coal has been characterised in terms of proximate and ultimate analyses and the relations among the chemical descriptors and the higher heating value (HHV) examined using correlation analysis and multivariate data analysis methods. The proximate analysis descriptors were used to predict HHV using multiple linear regression (MLR) and artificial neural network (ANN) methods. An attempt has been made to select the model with the optimal number of predictor variables. According to the adjusted multiple coefficient of determination in the MLR model, and alternatively, according to sensitivity analysis in ANN developing, two descriptors were evaluated by both methods as optimal predictors: fixed carbonand volatile matter. The performances of MLR and ANN when modelling HHV were comparable; the mean relative difference between the actual and calculated HHV values in the training data was 1.11% for MLR and 0.91% for ANN. The predictive ability of the models was evaluated by an external validation data set; the mean relative difference between the actual and predicted HHV values was 1.39% in MLR and 1.47% in ANN. Thus, the developed models could be appropriately used to calculate HHV. Ključne besede: Slovenian coal, higher heating value, HHV, regression, artificial neural network Objavljeno v DKUM: 03.04.2017; Ogledov: 28689; Prenosov: 356 Celotno besedilo (749,77 KB) Gradivo ima več datotek! Več... |
10. Prediction of the hardness of hardened specimens with a neural networkMatej Babič, Peter Kokol, Igor Belič, Peter Panjan, Miha Kovačič, Jože Balič, Timotej Verbovšek, 2014, izvirni znanstveni članek Opis: In this article we describe the methods of intelligent systems to predict the hardness of hardened specimens. We use the mathematical method of fractal geometry in laser techniques. To optimize the structure and properties of tool steel, it is necessary to take into account the effect of the self-organization of a dissipative structure with fractal properties at a load. Fractal material science researches the relation between the parameters of fractal structures and the dissipative properties of tool steel. This paper describes an application of the fractal dimension in the robot laser hardening of specimens. By using fractal dimensions, the changes in the structure can be determined because the fractal dimension is an indicator of the complexity of the sample forms. The tool steel was hardened with different speeds and at different temperatures. The effect of the parameters of robot cells on the material was better understood by researching the fractal dimensions of the microstructures of hardened specimens. With an intelligent system the productivity of the process of laser hardening was increased because the time of the process was decreased and the topographical property of the material was increased. Ključne besede: fractal dimension, fractal geometry, neural network, prediction, hardness, steel, tool steel, laser Objavljeno v DKUM: 17.03.2017; Ogledov: 1437; Prenosov: 97 Celotno besedilo (632,41 KB) Gradivo ima več datotek! Več... |