|
|
An efficiency comparison of simulation methods for artificial neural network training and inverse analysis
Nezval, Michal ; Novák, Drahomír (referee) ; Lehký, David (advisor)
The thesis deals with inverse analysis which is based on combination of artificial neural network and stochastic methods. The goal is to compare an efficiency of new simulation method Hierarchical Subset Latin Hypercube Sampling to classical Monte Carlo method and standard Latin Hypercube Sampling method used for neural network training. The efficiency is compared for a different neural network structures. The inverse analysis is then applied for engineering tasks – identification of limit state fiction parameters related to pitched-roof frame and material parameters of concrete specimen subjected to three-point bending. Finally an efficiency of Hierarchical Subset Latin Hypercube method comparing to Monte Carlo and Latin Hypercube Sampling methods is discussed.
|
|
|
Probabilistic analysis of reinforced concrete bridge made of I-73 girders
Nezval, Michal ; Šomodíková, Martina (referee) ; Lehký, David (advisor)
Many older bridges on the highways and roads in the Czech republic and all arend the world have been designed by codes, which are distinct of currently valid codes. Considering current codes, it is possible to set load bearing capacity in few different ways. In presented diploma thesis nonlinear finite element analysis using deterministic methods is compared with fully probabilistic nonlinear finite element analysis. Girder I-73 is analysed. Also the influence of degradation processes and following corrosion of reinforcement is taken into account, when load capacity is predicted for residual service life.
|
|
|
Probabilistic analysis of reinforced concrete bridge made of I-73 girders
Nezval, Michal ; Šomodíková, Martina (referee) ; Lehký, David (advisor)
Many older bridges on the highways and roads in the Czech republic and all arend the world have been designed by codes, which are distinct of currently valid codes. Considering current codes, it is possible to set load bearing capacity in few different ways. In presented diploma thesis nonlinear finite element analysis using deterministic methods is compared with fully probabilistic nonlinear finite element analysis. Girder I-73 is analysed. Also the influence of degradation processes and following corrosion of reinforcement is taken into account, when load capacity is predicted for residual service life.
|
|
|
An efficiency comparison of simulation methods for artificial neural network training and inverse analysis
Nezval, Michal ; Novák, Drahomír (referee) ; Lehký, David (advisor)
The thesis deals with inverse analysis which is based on combination of artificial neural network and stochastic methods. The goal is to compare an efficiency of new simulation method Hierarchical Subset Latin Hypercube Sampling to classical Monte Carlo method and standard Latin Hypercube Sampling method used for neural network training. The efficiency is compared for a different neural network structures. The inverse analysis is then applied for engineering tasks – identification of limit state fiction parameters related to pitched-roof frame and material parameters of concrete specimen subjected to three-point bending. Finally an efficiency of Hierarchical Subset Latin Hypercube method comparing to Monte Carlo and Latin Hypercube Sampling methods is discussed.
|
guest ::
