| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Search the digital library catalog Help

Query: search in
search in
search in
search in
* old and bologna study programme


1 - 10 / 21
First pagePrevious page123Next pageLast page
Using neural networks in the process of calibrating the microsimulation models in the analysis and design of roundabouts in urban areas
Irena Ištoka Otković, 2011, dissertation

Abstract: The thesis researches the application of neural networks in computer program calibration of traffic micro-simulation models. The calibration process is designed on the basis of the VISSIM micro-simulation model of local urban roundabouts. From the five analyzed methods of computer program calibration, Methods I, II and V were selected for a more detailed research. The three chosen calibration methods varied the number of outgoing traffic indicators predicted by neural networks and a number of neural networks in the computer program calibration procedure. Within the calibration program, the task of neural networks was to predict the output of VISSIM simulations for selected functional traffic parameters - traveling time between the measurement points and queue parameters (maximum queue and number of stopping at the roundabout entrance). The Databases for neural network training consisted of 1379 combinations of input parameters whereas the number of output indicators of VISSIM simulations was varied. The neural networks (176 of them) were trained and compared for the calibration process according to training and generalization criteria. The best neural network for each calibration method was chosen by using the two-phase validation of neural networks. The Method I is the calibration method based on calibration of a traffic indicator -traveling time and it enables validation related to the second observed indicator – queue parameters. Methods II and V connect the previously described calibration and validation procedures in one calibration process which calibrates input parameters according to two traffic indicators. Validation of the analyzed calibration methods was performed on three new sets of measured data - two sets at the same roundabout and one set on another location. The best results in validation of computer program calibration were achieved by the Method I which is the recommended method for computer program calibration. The modeling results of selected traffic parameters obtained by calibrated VISSIM traffic model were compared with: values obtained by measurements in the field, the existing analysis methods of operational roundabouts characteristics (Lausanne method, Kimber-Hollis, HCM) and modeling by the uncalibrated VISSIM model. The calibrated model shows good correspondence with measured values in real traffic conditions. The efficiency of the calibration process was confirmed by comparing the measured and modeled values of delays, of an independent traffic indicator that was not used in the process of calibration and validation of traffic micro-simulation models. There is also an example of using the calibrated model in the impact analysis of pedestrian flows on conflicting input and output flows of vehicles in the roundabout. Different traffic scenarios were analyzed in the real and anticipated traffic conditions.
Keywords: traffic models, traffic micro-simulation, calibration of the VISSIM model, computer program calibration method, neural networks in the calibration process, micro-simulation of roundabouts, traffic modeling parameters, driving time, queue parameters, delay
Published: 02.06.2011; Views: 3774; Downloads: 264
.pdf Full text (13,21 MB)

The use of artificial neural networks for colour prediction in textile printing
Darko Golob, Jure Zupan, Đurđica Parac-Osterman, 2008, published scientific conference contribution

Abstract: An attempt of using artificial neural networks for the prediction of dzes in textile printing paste preparation is presented. An existing collection of printed samples served as the basis for neural network training. It consists of 1340 samples printed using either a single dze or a combination of two dzes. First the proper combination of dzes was determined, because in most cases onlz two dzes are combined in the printing paste. Then the necessarz concentration of each dze was predicted. The reflectance value, and the colourvalues L*, a*, b* serve as input data and the known combination and concentrations of dzes for each sample were the targets. Some variations of neural network were tested, as well as various numbers of neurons in the hidden lazer. In addition, the influence of the training set organisation was examined, together with the number of learning epochs on the learning success.
Keywords: artificial neural networks, textile printing, colour recipe prediction
Published: 31.05.2012; Views: 1140; Downloads: 33
URL Link to full text

Combined feedforward and feedback control of end milling system
Franc Čuš, Uroš Župerl, Jože Balič, 2011, original scientific article

Keywords: machining, force control, neural networks
Published: 01.06.2012; Views: 840; Downloads: 18
URL Link to full text

A generalized neural network model of ball-end milling force system
Uroš Župerl, Franc Čuš, Bogomir Muršec, Anton Ploj, 2005, original scientific article

Abstract: The focus of this paper is to develop a reliable method to predict 3D cutting forces during ball-end milling process. This paper uses the artificial neural networks (ANNs) approach to evolve an generalized model for prediction of cutting forces, based on a set of input cutting conditions. A set of ten input milling parameters that have a major impact on the cutting forces was chosen to represent the machining conditions. The training of the networks is performed with experimental machining data. This approach greatly reduces the time-consuming mathematical work normally required for obtaining the cutting force expressions. The estimation performance of the network is evaluated through a detailed simulation study. The accuracy of an analytical model, which is a feasible alternative to the network, is compared to that of the network. With similar system parameter estimates for both methods, the network is found to be considerably more accurate than the analytical model. The results of model validation experiments on machining Ck45 are also reported. Experimental results demonstrate that this method can accurately estimate feed cutting force within an error of 4%. The results also indicate that when the combination of sigmoidal and gaussian transfer function were applied, the prediction accuracy of neural network is as high as 98%.
Keywords: end-milling, cutting forces, cutting parameters, generalized neural networks, modeling
Published: 01.06.2012; Views: 1604; Downloads: 63
URL Link to full text

Computer-based workpiece detection on CNC milling machine tools using optical camera and neural networks
Simon Klančnik, Jernej Šenveter, 2010, original scientific article

Abstract: In this paper, system for optical determining the workpiece origin on the CNC machine is presented. Similar high sophisticated systems are commercially available but in most cases they are very expensive and so their purchase is economically unjustified. The purpose of our research is to develop an inexpensive system for non-contact determination of the workpiece origin, which is also sufficiently precise for practical use. The system is implemented on a three-axis CNC milling machine Lakos 150 G, which is primarily designed for good machinability materials. Calibration procedure using feed-forward neural networks was developed. With this method the calibration procedure is simplified and the mathematical derivation of camera model is avoided. Learned neural network represents the camera calibration model. After neural network learning is complete, we can begin using the system for determining the workpiece origin. This developed system was through a number of tests proved to be reliable and suitable for use in practice. In the paper, working of system is illustrated with a practical example, which confirms the effectiveness of the implemented system in actual use on machine.
Keywords: neural networks, image processing, milling, workpiece detection
Published: 01.06.2012; Views: 1179; Downloads: 26
URL Link to full text

Tool cutting force modeling in ball-end milling using multilevel perceptron
Uroš Župerl, Franc Čuš, 2004, original scientific article

Abstract: This paper uses the artificial neural networks (ANNs) approach to evolve an efficient model for estimation of cutting forces, based on a set of input cutting conditions. A neural network algorithms are developed for use as a direct modeling method, to predict forces for ball-end milling operation. Supervised neural networks are used to successfully estimate the cutting forces developed during end milling process. The training of the networks is preformed with experimental machining data. The predictive capability of using analytical and neural network approaches are compared using statistics, which showed that neural network predictions for three cutting force components were for 4% closer to the experimental measurements, compared to 11% using analytical method. Exhaustive experimentation is conduced to develop the model and to validate it. The milling experiments prove that this model can predict accurately the cutting forces in three Cartesian directions.The force model can be used for simulation purposes and for defining threshold values in cutting tool condition monitoring system.
Keywords: ball end milling, cutting forces, modelling, artificial intelligence, neural networks
Published: 01.06.2012; Views: 1347; Downloads: 66
URL Link to full text

Neural-network-based numerical control for milling machine
Jože Balič, 2004, original scientific article

Abstract: We describe a device which uses a neural network to generate part-programs for milling, drilling and similar operations on machining centres, on the basis of 2D, 2.5D or 3D geometric models of prismatic parts, without operator intervention. The neural network consists of networks for prediction of milling strategy, for prediction of surface quality and for the optimisation of technological parameters in milling. We introduce the surface complexity index (SCI) for identifying surfaces which are very difficult to machine. The SCI takes the surface roughness and machining strategy into account. Teaching and testing of the NN is described. The device, which can be retrofitted to a CNC controller, can be trained from a set of typical parts and will then generate new NC part-programs. A case study of a tool used in the automotive supplier industry shows how a milling strategy is proposed, according to set constraints.
Keywords: intelligent CNC control, retrofit CNC, intelligent CAD/CAM, neural networks, machining centres
Published: 01.06.2012; Views: 838; Downloads: 56
URL Link to full text

Implementation of massive artificial neural networks with CUDA
Domen Verber, 2012, independent scientific component part or a chapter in a monograph

Keywords: CUDA, artificial neural networks, implementation
Published: 10.07.2015; Views: 585; Downloads: 74
URL Link to full text

Search done in 0.15 sec.
Back to top
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica