David Lehký

ANN inverse analysis based on stochastic small-sample training set simulation

A new approach of inverse analysis is proposed to obtain parameters of a computational model in order to achieve the best agreement with experimental data. The inverse analysis is based on the coupling of a stochastic simulation and an artificial neural network (ANN). The identification parameters play the role of basic random variables with a scatter reflecting the physical range of potential values. A novelty of the approach is the utilization of the efficient small-sample simulation method Latin Hypercube Sampling (LHS) used for the stochastic preparation of the training set utilized in training the artificial neural network. Once the network has been trained, it represents an approximation consequently utilized to solve the key task: To provide the best possible set of model parameters for the given experimental data. The efficiency of the approach is shown using numerical examples from civil engineering computational mechanics.

Solving Inverse Structural Reliability Problem Using Artificial Neural Networks and Small-Sample Simulation

A new general inverse reliability analysis approach based on artificial neural networks is proposed. An inverse reliability analysis is a problem of obtaining design parameters corresponding to a specified reliability (reliability index or theoretical failure probability). Design parameters can be deterministic or they can be associated with random variables. The aim is to generally solve not only single design parameter cases but also multiple parameter problems with given multiple reliability constraints. Inverse analysis is based on the coupling of a stochastic simulation of the Monte Carlo type (the small-sample simulation method Latin hypercube sampling) and an artificial neural network. The validity and efficiency of this approach is shown using numerical examples with single as well as multiple reliability constraints and with single as well as multiple design parameters, and with independent basic random variables as well as random variables with prescribed statistical correlations.

FraMePID-3PB software for material parameter identification using fracture tests and inverse analysis

Abstract Knowledge regarding the values of fracture-mechanical parameters is critical for the virtual failure modeling of elements and structures made of concrete. A key parameter in nonlinear fracture mechanics modeling is the specific fracture energy of concrete, and its variability. Three-point bending tests on notched-beam specimens are fundamental experiments for the determination of fracture-mechanical parameters. In the present paper, two basic approaches are applied to determine fracture-mechanical parameter values from these tests: (i) the effective crack model/work-of-fracture method, and (ii) inverse analysis using artificial neural networks and stochastic simulations. First, the paper describes suitable methods for the determination of fracture-mechanical parameters. Second, the FraMePID-3PB software tool, which has been developed in order to automate the whole time consuming process of inverse analysis, is described. Finally, the verification of methodology and software is presented using two illustrative examples.

Neural network ensemble-based parameter sensitivity analysis in civil engineering systems

The use of artificial neural networks for parameter sensitivity analysis in civil engineering systems is an emerging research focus of increased interest. Existing methods are generally based on a single neural network, but are inadequate as a basis for parameter sensitivity analysis because of the instability of a single neural network. To address this deficiency, this study develops a neural network ensemble-based parameter sensitivity analysis paradigm. This paradigm features use of a set of preselected superior neural networks to make decisions about parameter sensitivity by synthesizing sensitivity analysis results of individual neural networks. The proposed paradigm is employed to address two classic civil engineering problems: (1) identification of critical parameters in the fracture failure of notched concrete beams and (2) recognition of the most significant parameters in the lateral deformation of deep foundation pits. The results show that tensile strength and modulus of elasticity are the critical parameters in the fracture failure of the notched concrete beam, and elasticity modulus of soil, Poisson’s ratio and soil cohesion are the most significant influential factors in the lateral deformation of the deep foundation pit. The proposed method provides a common paradigm for analysing the sensitivity of influential parameters, shedding light on the underlying mechanisms of civil engineering systems.

Neural Network Based Damage Detection of Dynamically Loaded Structures

The aim of the paper is to describe a methodology of damage detection which is based on artificial neural networks in combination with stochastic analysis. The damage is defined as a stiffness reduction (bending or torsion) in certain part of a structure. The key stone of the method is feed-forward multilayer network. It is impossible to obtain appropriate training set for real structure in usage, therefore stochastic analysis using numerical model is carried out to get training set virtually. Due to possible time demanding nonlinear calculations the effective simulation Latin Hypercube Sampling is used here. The important part of identification process is proper selection of input information. In case of dynamically loaded structures their modal properties seem to be proper input information as those are not dependent on actual loading (traffic, wind, temperature). The methodology verification was carried out using laboratory beam.

Neural Network Based Identification of Material Model Parameters to Capture Experimental Load-deflection Curve

A new approach is presented for identifying material model parameters. The approach is based on coupling stochastic nonlinear analysis and an artificial neural network. The model parameters play the role of random variables. The Monte Carlo type simulation method is used for training the neural network. The feasibility of the presented approach is demonstrated using examples of high performance concrete for prestressed railway sleepers and an example of a shear wall failure.

Fracture parameters of concrete C40/50 and C50/60 determined by experimental testing and numerical simulation via inverse analysis

The use of stochastic non-linear computational mechanics in real-world applications faces a fundamental obstacle: the lack of knowledge regarding the stochastic properties of material parameters involved in the problem. The paper describes the results of an experimental program which was focused on the determination of fracture-mechanical parameters and their stochastic models of different concrete types after different hardening durations. The investigated types of concrete, C40/50 and C50/60, are used industrially for the production of pre-fabricated concrete elements (e.g. double T-shaped beam elements for roofing). The fracture-mechanical parameters were determined by wedge splitting tests on cubic specimens with notch. Moreover, numerical simulations were carried out for the identification of material parameters by using artificial neural networks. All results obtained from the individual tests will here be presented, compared and discussed. Finally, recommendations for stochastic models of selected parameters of the analysed concrete will be given.

Reliability calculation of time-consuming problems using a small-sample artificial neural network-based response surface method

An important step when designing and assessing the reliability of existing structures and/or structural elements is to calculate the reliability level described by failure probability or reliability index. Since calculating the structural response of complex systems such as bridges is usually a time-consuming task, the utilization of approximation methods with a view to reducing the computational effort to an acceptable level is an appropriate solution. The paper introduces a small-sample artificial neural network-based response surface method. An artificial neural network is used as an approximation (a so-called response surface) of the original limit state function. In order to be as effective as possible with respect to computational effort, a stratified Latin hypercube sampling simulation method is utilized to properly select training set elements. Subsequently, the artificial neural network-based response surface is utilized to calculate failure probability. To increase the accuracy of the determined failure probability, the response surface can be updated close to the failure region. This is performed by finding a new anchor point, which lies close to the design point of the limit state function. The new anchor point is then used to prepare the updated training set. The efficiency of the proposed method is tested for different training set sizes using a nonlinear limit state function taken from the literature, and the reliability assessment of three concrete bridges, one with explicit and two with implicit limit state functions in the form of finite element method models.

Inverse Response Surface Method in Reliability-Based Design

Abstract The paper introduces an inverse response surface method utilized when performing reliability-based design optimization of time-consuming problems. Proposed procedure is based on a coupling of the adaptive response surface method and the artificial neural network-based inverse reliability method. The validity and accuracy of the method is tested using examples with explicit nonlinear limit state functions. Obtained results as well as important aspects of the method are discussed.

Reliability-based design: Artificial neural networks and double-loop reliability-based optimization approaches

Abstract Two advanced optimization approaches to solving a reliability-based design problem are presented. The first approach is based on the utilization of an artificial neural network and a small-sample simulation technique. The second approach considers an inverse reliability task as a reliability-based optimization task using a double-loop optimization method based on small-sample simulation. Both techniques utilize Latin hypercube sampling with correlation control. The efficiency of both approaches is tested using three numerical examples of structural design – a cantilever beam, a reinforced concrete slab and a post-tensioned composite bridge. The advantages and disadvantages of the approaches are discussed.