6209

5806

كارشناسي ارشد

تكنولوژي نساجي

اصفهان: دانشگاه صنعتي اصفهان، دانشكده نساجي

1390

[هشت]،118ص.: مصور،جدول،نمودار

ص.ع.به فارسي و انگليسي

داريوش سمناني، محمد مرشد

مصطفي يوسفي

20/6/090

عبدالكريم حسيني راوندي، محمدرضا احمدزاده

مهندسي نساجي

ID5806

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Prediction of Mechanical Properties of Nanofibers by Artificial Intelligence Maryam Bakhshayesh nargesflower32@yahoo com Data of Submission 3 03 2011 Department of Textile Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiSuperviser Dariush Semnani d semnani@cc iut ac irMohammad Morshed m sh110@cc iut ac irAbstractElectrospun nanofibers are important in many applications such as filtration tissue engineering reinforcement incomposites and micro nano electro mechanical systems In the most applications of nanofibers measuring thetenacity of single nanofibers is essentinal Because of the small size of nanofibers touching without damagingthem is difficult micro manipulation systems can be used to move the fibers and attach the fiber ends to grips However such systems may not be suitable for certain types of nanofibers Often these systems require ScanningElectron Microscope have limitations in that sample may be conductive These experimental systems are veryexpensive time consuming and limited to the number of tests Durigned in this research method is much easier cheaper without damaging the nanofibers in the shortest time can be extended to all nanofibres with acceptableerror for nanofibers provide strength Considering that the structural properties of fibers such as degree of crystallinity percent arrangement size ofcrystals crystal plates and the polymer chain length The properties of fibers affects their strength This study evaluated the effect of structural properties of fibers and their mechanical properties Then themechanical properties of nanofibres predicted by artificial intelligence Degree of crystallinity percentagearrangement size of crystals crystal plates away from the X ray diffraction and crystallinity index from IRspectrum of poly fibers for eight samples have been obtained for poly acrilonitril then stress their strength havebeen measured The data along with 40 percent conventional fiber nano fibers as training data to the neuralnetwork with backward error was used and the rest 60 percent of nano fibers as test data to the neural network isapplied Measured data presented to a BP neural networks Finally the results of predicted and actual test resultsof nano fibers parallel batch strength have been compared together Nod optimized number of hidden layers and training parameters of neural network genetic algorithm was usedand the best network with three layers function training trainbr 22 Tdadnd the hidden layer learning rate of0 04277 Fixed Momentom 0 6483 the parameter Adjusted mu 0 0646 reduction factor mu0 9909 and increasedFaktv mu 4 515 was selected The average percentage error of the network test 8 9046 and the percentage error ofEducation 0 0068 is The lowest percentage of error obtained with this training 0 89009 was related to theweights were saved Then using the weights between input and hidden layer can be concluded that the distancebetween crystal plates in 1020 for the strength of other parameters has more impact and also the lowest indexarrangement had the effect With weights between output and hidden layers can be concluded that the elongationstress affected more input parameters Key Words Nanofiber X ray diffraction FTIR Genetic algorithm Artificial Intelligence

داريوش سمناني، محمد مرشد

مصطفي يوسفي

عبدالكريم حسيني راوندي، محمدرضا احمدزاده