DeterminingK and related stress-field parameters from displacement fields

A general algorithm is presented that computes the stress-field parameters for opening-mode crack problems in a least-squares sense from full-field moiré or speckle-displacement fringe patterns. The algorithm can be used in the presence of rigid-body rotation and does not require absolute fringe numbering. Extensive numerical experiments were conducted with the algorithm to determine the sensitivity of the method to experimental errors. Small random position errors in locating the fringe maxima were found to have a negligible influence on the stress-intensity-factor calculation when the number of data points was about ten times greater than the number of unknown stress-field parameters. It was also found that systematic position errors due to an incorrectly specified crack-tip location could be minimized by assuming various crack-tip locations in the vicinity of the actual crack tip and selecting the best fit results. Bothu andv fields were found to be equally suitable for determination of the stress-intensity factor. Paper was presented at 1983 SESA Spring Meeting held in Cleveland, OH on May 15–19, 1983.     

An artificial neural network for double exposure PIV image analysis

 This note presents a back propagation neural network for PIV image analysis. Unlike the conventional auto-correlation method that identifies one pair of image out of the picture, the proposed network distinguishes all the image pairs in the measurement area and provides different labels for each pair. Experimental investigations show good agreement with the auto-correlation process for the uniform flow measurement, and a 78.1% success ratio for the stagnation flow. Received: 9 January 1997/Accepted: 10 September 1997     

A new artificial neural network tracking technique for particle image velocimetry

 The analysis of Particle Image Velocimetry (PIV) data requires effective algorithms to track efficiently the particles suspended in the fluid flow. The artificial neural network algorithm method described here presents a new approach to solve this problem. Contrary to the classic cross correlation method, this new method does not require a large number of particles per frame, it can handle flows with large velocity gradients, and is suited for tracking images with multiple exposures as well as tracking through consecutive images. The algorithm was tested on synthetic and available experimental data to provide a thorough performance analysis. Received: 28 May 1996/Accepted: 25 December 1996     

Statistical regression and artificial neural network analyses of impinging jet experiments

The purpose of this paper is to focus on the experimentally obtained results of impinging jet applications by the help of two different analysis methods. Circular round pipes (D = 7.9, 10.8, 13.8 and 23.1 mm) have been used as the impinging jets. The heat transfer is calculated with Nusselt number (Nu). The variable parameters are the dimensionless jet-to-impingement plate distance (z/D), Reynolds number (Re) and dimensionless temperature measurement points on the heated surface (x/L, y/L). Some important analysis methods such as artificial neural network (ANN), statistical regression, and uncertainty analysis are applied to the obtained data. It is shown that the ANN application is not simply a classification analysis; it is actually an application of the convergence of functions. As a result, by considering random data, 4.57% convergence level is obtained regarding the pipe diameter. The software STATISTICA 5.0 is used to estimate new empirical correlations nonlinearly. The smallest regression coefficient for the correlations is 0.87, while the highest value is 0.99. The result of the uncertainty analyses showed that the total uncertainties are in the agreeable range; 8% for Nu, and 2.89% for Re. Dr. Nevin Celik is a Post Doctoral Fellowship in University of Minnesota since August 2007.     

环形激光陀螺人工神经网络温度漂移建模

研究了系统在常温条件下开机后,壳体密闭条件下环形激光陀螺(RLG)漂移的温度特性建模问题。一系列不同环境温度条件下的RLG自升温实验表明,由于系统壳体的保温和隔温功能,系统内部温度场变化缓慢、均匀,此时漂移的变化主要与内部温度场的温度值变化相关。利用一组25℃-55℃范围内静态漂移测试数据作为学习样本,建立起基于BP和RBF神经网络的温度漂移补偿模型,并利用四组不同测试条件下测得的静态数据对文中模型进行检验。结果表明,若采用均方根误差(RMS)指标进行评价,则得到的温度漂移模型可以有效补偿RLG的漂移输出趋势项,使陀螺的稳定性指标提高20%-40%,且BP网络建模补偿精度优于常规最小二乘中的一阶线性分段拟合,RBF网络建模优于二阶抛物线分段拟合。     

BP混沌混合神经网络在光纤陀螺温度漂移预测中的应用

提出使用BP混沌混合神经网络建立FOG温度漂移模型的方法.该方法在BP算法中采用了改进型Logistic-Map映射生成的混沌变量,能够避免陷入局部最小,可迅速达到全局最优.应用该方法分析某型FOG温度漂移实测数据,结果表明其具有良好的预测效果.     

In-plane displacement measurement of composite material by objective speckle

The objective speckle was used to measure the in-plane displacement around the hole in a coupon specimen of composite material under tensile loading. Mirror transplantation method was adopted to ensure the high reflectivity of specimen surface and to obtain high quality double exposure specklegram. The adjustable spatial frequency of the whole-field pattern of the displacement made it possible to measure a broad range of strain, from elastic to plastic. The results obtained by objective speckle agree well with those by Moire method.     

A new dynamic subgrid-scale model using artificial neural network for compressible flow

The subgrid-scale (SGS) kinetic energy has been used to predict the SGS stress in compressible flow and it was resolved through the SGS kinetic energy transport equation in past studies. In this paper, a new SGS eddy-viscosity model is proposed using artificial neural network to obtain the SGS kinetic energy precisely, instead of using the SGS kinetic energy equation. Using the infinite series expansion and reserving the first term of the expanded term, we obtain an approximated SGS kinetic energy, which has a high correlation with the real SGS kinetic energy. Then, the coefficient of the modelled SGS kinetic energy is resolved by the artificial neural network and the modelled SGS kinetic energy is more accurate through this method compared to the SGS kinetic energy obtained from the SGS kinetic energy equation. The coefficients of the SGS stress and SGS heat flux terms are determined by the dynamic procedure. The new model is tested in the compressible turbulent channel flow. From the a posterior tests, we know that the new model can precisely predict the mean velocity, the Reynolds stress, the mean temperature and turbulence intensities, etc.     

Identification of elastic property and loading fields from full-field displacement measurements

A method is introduced to identify simultaneously elastic properties and loading fields from a measured displacement field. Since the mechanical behavior of micro-electro-mechanical systems (MEMS) is governed by surface effects, this type of identification tool is thought to be of major interest. However, increasing the number of parameters to retrieve affects the redundancy necessary for an accurate identification. A finite-element formulation of a distance between measured and statically admissible (SA) displacement fields is shown to be equivalent to a standard least-squares distance to kinematically admissible (KA) fields if the used modeling is suitable. Any deviation from this equivalence is then the signature of a modeling error. Balancing the distance to KA and SA displacement fields allows one to retrieve unknown modeling parameters. This method is detailed on heterogeneous Euler–Bernoulli beams submitted to an unknown loading field and applied to experimental displacement fields of micro-cantilevers obtained with an electrostatic set-up. An elastic property field and a parameterized loading field are then identified, and the quality of the identification is assessed.     

Transitional intermittency detection by neural network

 A neural network has been used to predict the flow intermittency from velocity signals in the transition zone in a boundary layer. Unlike many of the available intermittency detection methods requiring a proper threshold choice in order to distinguish between the turbulent and non-turbulent parts of a signal, a trained neural network does not involve any threshold decision. The intermittency prediction based on the neural network has been found to be very satisfactory. Received: 15 December 1997/Accepted: 30 December 1998     

Prediction of nonlinear viscoelastic behavior of polymeric composites using an artificial neural network

Creep tests at constant stress are performed for the carbon-fiber reinforced epoxy composite at various temperatures and initial stresses. A nonlinear viscoelastic constitutive model is developed, and its material parameters are determined by fitting it to creep test data. Model results are found to agree very well with the experimental data at low temperature and low stress conditions. However, the agreement deteriorates at high temperatures, particularly in the vicinity of the glass transition temperature.An alternative model based on an artificial neural network (ANN) is developed to predict the stress relaxation of the polymer matrix composite. The ANN model is trained and validated with 9000 experimental data sets obtained from stress relaxation tests performed at various constant strain (initial stress) and constant temperature conditions. Training of the ANN employs a scaled conjugate gradient method. The optimal brain surgeon algorithm is employed to optimize the topology. The optimal ANN configuration has 88 processing elements (3 in the input layer, 45 in the first hidden layer, 39 in the second hidden layer, and 1 in the output layer) and 410 links. The predictions of the ANN model are found to be more accurate over a wider range of stress and temperature conditions than those of the explicit nonlinear viscoelastic model, in particular near the glass transition temperature.     

Simultaneous on-line measurement of orthogonal displacement fields by Moiré interferometry

Experimental Techniques - A modified four-beam optical arrangement for moiré interferometry provides simultaneous determination of the in-plane U and V displacement fields, and is suitable for...     

Active control of flow past an elliptic cylinder using an artificial neural network trained by deep reinforcement learning

The active control of flow past an elliptical cylinder using the deep reinforcement learning (DRL) method is conducted. The axis ratio of the elliptical cylinder $\Gamma$ varies from 1.2 to 2.0, and four angles of attack $\alpha=0^\circ, 15^\circ, 30^\circ$, and $45^\circ$ are taken into consideration for a fixed Reynolds number $Re=100$. The mass flow rates of two synthetic jets imposed on different positions of the cylinder $\theta_1$ and $\theta_2$ are trained to control the flow. The optimal jet placement that achieves the highest drag reduction is determined for each case. For a low axis ratio ellipse, i.e., $\Gamma=1.2$, the controlled results at $\alpha=0^\circ$ are similar to those for a circular cylinder with control jets applied at $\theta_1=90^\circ$ and $\theta_2=270^\circ$. It is found that either applying the jets asymmetrically or increasing the angle of attack can achieve a higher drag reduction rate, which, however, is accompanied by increased fluctuation. The control jets elongate the vortex shedding, and reduce the pressure drop. Meanwhile, the flow topology is modified at a high angle of attack. For an ellipse with a relatively higher axis ratio, i.e., $\Gamma\ge1.6$, the drag reduction is achieved for all the angles of attack studied. The larger the angle of attack is, the higher the drag reduction ratio is. The increased fluctuation in the drag coefficient under control is encountered, regardless of the position of the control jets. The control jets modify the flow topology by inducing an external vortex near the wall, causing the drag reduction. The results suggest that the DRL can learn an active control strategy for the present configuration.     

Finite element analysis of V-ribbed belts using neural network based hyperelastic material model

A three-dimensional finite element model was built to study V-ribbed belt pulley contact mechanics. The model consists of a pulley and a segment of V-ribbed belt in contact with the pulley. A material model for the belt, including the rubber compound and the reinforcing cord is developed. Rubber is modeled as hyperelastic material. The hyperelastic strain energy function is approximated by neural network trained by rubber test data. Reinforcing cord is modeled as elastic rebar. The material model developed is implemented in the commercial finite element code ABAQUS to simulate the V-ribbed belt-pulley system. A study is then conducted to investigate the effect of belt pulley system parameters on the contact mechanics. The effects of temperature and aging on belt materials are also investigated. The information gained from the analysis can be applied to optimize V-ribbed belt and pulley design.     

Spatial artificial neural network model for subgrid-scale stress and heat flux of compressible turbulence

The subgrid-scale(SGS) stress and SGS heat flux are modeled by using an artificial neural network(ANN) for large eddy simulation(LES) of compressible turbulence. The input features of ANN model are based on the first-order and second-order derivatives of filtered velocity and temperature at different spatial locations. The proposed spatial artificial neural network(SANN)model gives much larger correlation coefficients and much smaller relative errors than the gradient model in an a priori analysis. In an a posteriori analysis, the SANN model performs better than the dynamic mixed model(DMM) in the prediction of spectra and statistical properties of velocity and temperature, and the instantaneous flow structures.     

Prediction of CHF in concentric-tube open thermosiphon using artificial neural network and genetic algorithm

In this paper, an artificial neural network (ANN) for predicting critical heat flux (CHF) of concentric-tube open thermosiphon has been trained successfully based on the experimental data from the literature. The dimensionless input parameters of the ANN are density ratio, ρ _l/ρ _v; the ratio of the heated tube length to the inner diameter of the outer tube, L/D _i; the ratio of frictional area, d _i/(D _i + d _o); and the ratio of equivalent heated diameter to characteristic bubble size, D _he/[σ/g(ρ _l−ρ _v)]^0.5, the output is Kutateladze number, Ku. The predicted values of ANN are found to be in reasonable agreement with the actual values from the experiments with a mean relative error (MRE) of 8.46%. New correlations for predicting CHF were also proposed by using genetic algorithm (GA) and succeeded to correlate the existing CHF data with better accuracy than the existing empirical correlations.     

Nonlinear constitutive models from nanoindentation tests using artificial neural networks

This paper presents a new approach using Artificial Neural Networks (ANNs) models to simulate the response during nanohardness tests of a variety of materials with nonlinear behavior. The ANNs continuous input and output variables usually include material parameters, indentation deflection, and resisting force. Different ANN models, including dimensionless input/output variables, are generated and trained with discrete finite-element (FE) simulations with different geometries and nonlinear material parameters. Only the monotonic loading part of the load–displacement indentation response is used to generate the trained ANN models. This is a departure from classical indentation simulations or tests where typically the unloading portion is used to determine the stiffness and hardness. The experimental part of this study includes nanoindentation tests performed on a silicon (Si) substrate with and without a nanocrystalline copper (Cu) film. The new ANN models are used to back-calculate (inverse problem) the in situ nonlinear material parameters for different copper material systems. The results are compared with available data in the literature. The proposed FE–ANN modeling approach is very effective and can be used in calibrating and predicting the in situ inelastic material properties using the monotonic part of the indentation response and for depths above 50 nm where the overall resisting force represents a continuum response.