Neural network modeling to evaluate the dynamic flow stress of high strength armor steels under high strain rate compression

An artificial neural network (ANN) constitutive model is developed for high strength armor steel tempered at 500 °C, 600 °C and 650 °C based on high strain rate data generated from split Hopkinson pressure bar (SHPB) experiments. A new neural network configuration consisting of both training and validation is effectively employed to predict flow stress. Tempering temperature, strain rate and strain are considered as inputs, whereas flow stress is taken as output of the neural network. A comparative study on Johnson–Cook (J–C) model and neural network model is performed. It was observed that the developed neural network model could predict flow stress under various strain rates and tempering temperatures. The experimental stress–strain data obtained from high strain rate compression tests using SHPB, over a range of tempering temperatures (500–650 °C), strains (0.05–0.2) and strain rates (1000–5500/s) are employed to formulate J–C model to predict the high strain rate deformation behavior of high strength armor steels. The J-C model and the back-propagation ANN model were developed to predict the high strain rate deformation behavior of high strength armor steel and their predictability is evaluated in terms of correlation coefficient (R) and average absolute relative error (AARE). R and AARE for the J–C model are found to be 0.7461 and 27.624%, respectively, while R and AARE for the ANN model are 0.9995 and 2.58%, respectively. It was observed that the predictions by ANN model are in consistence with the experimental data for all tempering temperatures.     

Phenomenological modelling of flow behaviour of 20MnMoNi55 reactor pressure vessel steel at cryogenic temperature with different strain rates

In the present study a phenomenological constitutive model is developed to describe the flow behaviour of 20MnMoNi55 low carbon reactor pressure vessel (RPV) steel at sub-zero temperature under different strain rates. A set of uniaxial tensile tests is done with the variation of strain rates and temperature ranging from 10⁻⁴ s⁻¹ to 10⁻¹ s⁻¹ and -80 °C to −140 °C respectively. From the experimental data, family of flow curves at different temperatures and strain rates are generated and fitted exponentially. The strain rate and temperature dependence of the coefficients of the exponential flow curves are extracted from these curves and characterised through a general phenomenological constitutive coupled equation. The coefficients of this coupled equation are optimised using genetic algorithm. Finite element simulation of tensile tests at different strain rates and temperatures are done using this coupled equation in material model of Abaqus FEA software and validated with experimental results. The novelties of proposed model are: (a) it can predict precisely the flow behaviour of tensile tests (b) it is a simple form of equation where fitting parameters are both function of strain rate ratio and temperature ratio, (c) it has ability to characterize flow behaviour with decreasing subzero temperatures and increasing strain rates.     

Determination of penetration depth at high velocity impact using finite element method and artificial neural network tools

Determination of ballistic performance of an armor solution is a complicated task and evolved significantly with the application of finite element methods(FEM) in this research field.The traditional armor design studies performed with FEM requires sophisticated procedures and intensive computational effort,therefore simpler and accurate numerical approaches are always worthwhile to decrease armor development time.This study aims to apply a hybrid method using FEM simulation and artificial neural network(ANN) analysis to approximate ballistic limit thickness for armor steels.To achieve this objective,a predictive model based on the artificial neural networks is developed to determine ballistic resistance of high hardness armor steels against 7.62 mm armor piercing ammunition.In this methodology,the FEM simulations are used to create training cases for Multilayer Perceptron(MLP) three layer networks.In order to validate FE simulation methodology,ballistic shot tests on 20 mm thickness target were performed according to standard Stanag 4569.Afterwards,the successfully trained ANN(s) is used to predict the ballistic limit thickness of 500 HB high hardness steel armor.Results show that even with limited number of data,FEM-ANN approach can be used to predict ballistic penetration depth with adequate accuracy.     

Mechanical behavior and failure mechanism of polyurea nanocomposites under quasi-static and dynamic compressive loading

Polyurea is an elastomeric material that can be applied to enhance the protection ability of structures under blast and impact loading.In order to study the compressive mechanical properties of SiC/polyurea nanocomposites under quasi-static and dynamic loading,a universal testing machine and split Hop-kinson pressure bar(SHPB)apparatus were used respectively.The stress-strain curves were obtained on polyurea and its composites at strain rates of 0.001-8000 s-1.The results of the experiment suggested that increase in the strain rates led to the rise of the flow stress,compressive strength,strain rate sensitivity and strain energy.This indicates that all of the presented materials were dependent on strain rate.Moreover,these mechanical characters were enhanced by incorporating a small amount of SiC into polyurea matrix.The relation between yield stress and strain rates were established using the power law functions.Finally,in order to investigate the fracture surfaces and inside information of failed specimens,scanning electron microscopy(SEM)and micro X-ray computed tomography(micro-CT)were used respectively.Multiple voids,crazes,micro-cracks and cracking were observed in fracture surfaces.On the other hand,the cracking propagation was found in the micro-CT slice images.It is essential to understand the deformation and failure mechanisms in all the polyurea materials.     

Numerical investigation of the failure mechanism of cubic concrete specimens in SHPB tests

Cylindrical specimens are commonly used in Split Hopkinson pressure bar (SHPB) tests to study the uniaxial dynamic properties of concrete-like materials.In recent years,true tri-axial SHPB equipment has also been developed or is under development to investigate the material dynamic properties under tri-axial impact loads.For such tests,cubic specimens are needed.It is well understood that static material strength obtained from cylinder and cube specimens are different.Conversion factors are obtained and adopted in some guidelines to convert the material strength obtained from the two types of specimens.Previous uniaxial impact tests have also demonstrated that the failure mode and the strain rate effect of cubic specimens are very different from that of cylindrical ones.However,the mechanical background of these findings is unclear.As an extension of the previous laboratory study,this study performs numerical SHPB tests of cubic and cylindrical concrete specimens subjected to uniaxial impact load with the validated numerical model.The stress states of cubic specimens in relation to its failure mode under different strain rates is analyzed and compared with cylindrical specimens.The detailed analyses of the numerical simulation results show that the lateral inertial confinement of the cylindrical specimen is higher than that of the cubic specimen under the same strain rates.For cubic specimen,the corners are more severely damaged because of the lower lateral confinement and the occurrence of the tensile radial stress which is not observed in cylindrical specimens.These results explain why the dynamic material strengths obtained from the two types of specimens are different and are strain rate dependent.Based on the simulation results,an empirical formula of conversion factor as a function of strain rate is proposed,which supplements the traditional conversion factor for quasi-static material strength.It can be used for transforming the dynamic compressive strength from cylinders to cubes obtained from impact tests at different strain rates.     

基于独立分量分析和神经网络的钢结构损伤识别方法

为了有效剔除钢结构振动信号中的噪声,提取用于损伤识别的特征量,采用独立分量分析方法分离统计独立信号,同时得到表征结构损伤状态的混合矩阵,然后将混合矩阵作为特征量输入至神经网络进行训练,最后将训练好的神经网络作为分类器进行结构损伤识别.在冲击载荷作用下,针对钢框架结构模型进行了不同损伤部位的振动实验,结果表明:基于独立分量分析和神经网络的损伤识别方法具有较高的识别率和可重复性,而且实现简单,在结构损伤识别领域具有较大的应用潜力.     

Dynamic necking of a near α titanium alloy at high strain rates:Experiments and modelling

The tensile behaviour of near α Ti3Al2.5 V alloy, conceived for applications in aerospace and automotive engineering, is characterized from quasi-static to high strain rates. The material is found to present noticeable strain rate sensitivity. The dynamic true strain rate in the necking cross-section reaches values up to ten times higher than the nominal strain rate. It is also observed that beyond necking the dynamic true stress-strain curves present limited rate dependence. The experimental results at different strain rates are used to determine a suitable constitutive model for finite element simulations of the dynamic tensile tests. The model predicts the experimentally macroscopic force-time response, true stress-strain response and effective strain rate evolution with good agreement.     

多重结构神经网络在平台罗经故障诊断中的应用

提出了基于层次分类诊断模型的多重结构神经网络 (MNN)在平台罗经故障诊断中的应用方法 ,并对其温控子系统的人工神经网络 (ANN)进行了Matlab实现     

基于遗传算法和神经网络的混合自学习策略

通过对神经网络自学习方法和基因遗传算法的研究 ,采用自适应基因遗传算法对神经网络进行训练 ,形成基于遗传算法和神经网络的混合自学习策略 ,克服了神经网络自学习方法的不足 ,提高了系统的自学习能力     

恒电量微扰法研究在含Cl-介质Ce3+对航空铝材的缓蚀作用

本文用恒电量微扰法研究了Ce3+对铝合金在0．1mol·L-1NaCl介质中的缓蚀机理及抗孔蚀性。结果表明：铝合金孔蚀诱导过程及其表面形成林转化膜过程具有相同的电化学模式,电化学腐蚀速度控制步骤取决于钝化膜（转化膜）／溶液界面过程。Ce3+对铝合金孔蚀的诱导期和发展期均有抑制作用,形成铈转化膜后,铝基表面阻抗大大提高,耐孔蚀性增强。     

基于灰色神经网络武器装备研制费用预测模型

提出了基于灰色系统理论与神经网络的武器装备研制费用组合预测模型,该模型首先采用灰色GM(0,N)模型对研制费用进行预测,利用LMBP神经网络对预测误差进行了模拟与修正,实例验证该方法具有较高的预测精度.     

一座小型建筑空调动态负荷的ANN预测研究

人工神经网络(Artificial Neural Network,简称ANN)是一种崭新非线性建模和预测方法,具有良好的非线性品质和极高的拟合精度。在前人研究的基础上,针对一座小型建筑物建立了其空调动态负荷预测的BP模型。为克服常规BP训练算法的缺陷,提出了一种改进的遗传搜索算法,结果证明是一种高效实用的算法。     

武警总队应急车辆故障诊断系统研究

将人工神经网络和基于案例推理技术结合用于车辆故障诊断系统中,构建了ANN与CBR结合模型,阐述了各子系统的基本功能及相互关系,并对关键技术进行了详细解释。实例诊断表明,ANN和CBR方法的结合有效地弥补了它们各自的缺陷,提高了应急条件下车辆的维修保障能力。     

BP神经网络在极移预报中的应用

为了提高地球定向参数极移的预报精度,建立了一个极移数据预报模型。利用傅里叶分析研究插值基础序列的周期特性,验证了基础序列重采样的可行性,提取插值基础序列数据的趋势项,利用多输入-单输出反向传播(Back Propagation,BP)神经网络建模预报不同跨度的残差序列,合并趋势项和残差序列得到最终的极移预报。预报结果表明,选取合适的插值基础序列得到的预报极移精度较高,此BP神经网络能够有效地应用于地球定向参数极移的预报。     

Research on the influences of confining pressure and strain rate on NEPE propellant: Experimental assessment and constitutive model

In order to study the influences of confining pressure and strain rate on the mechanical properties of the Nitrate Ester Plasticized Polyether (NEPE) propellant, uniaxial tensile tests were conducted using the self-made confining pressure system and material testing machine. The stress-strain responses of the NEPE propellant under different confining pressure conditions and strain rates were obtained and analyzed. The results show that confining pressure and strain rate have a remarkably influence on the mechanical responses of the NEPE propellant. As confining pressure increases (from 0 to 5.4 MPa), the maximum tensile stress and ultimate strain increase gradually. With the coupled effects of confining pressure and strain rate, the value of the maximum tensile stress and ultimate strain at 5.4 MPa and 0.0667 s⁻¹ is 2.03 times and 2.19 times of their values under 0 MPa and 0.00333 s⁻¹, respectively. Afterwards, the influence mechanism of confining pressure on the NEPE propellant was analyzed. Finally, based on the viscoelastic theory and continuous damage theory, a nonlinear constitutive model considering confining pressure and strain rate was developed. The damage was considered to be rate-dependent and pressure-dependent. The constitutive model was validated by comparing experimental data with predictions of the constitutive model. The whole maximum stress errors of the model predictions are lower than 4% and the corresponding strain errors are lower than 7%. The results show that confining pressure can suppress the damage initiation and evolution of the NEPE propellant and the nonlinear constitutive model can describe the mechanical responses of the NEPE propellant under various confining pressure conditions and strain rates. This research can lay a theoretical foundation for analyzing the structural integrity of propellant grain accurately under working pressure loading.     

Formation of adiabatic shearing band for high-strength Ti-5553 alloy: A dramatic thermoplastic microstructural evolution

By using split Hopkinson pressure bar, optical microscopy and electronic microscopy, we investigate the influence of initial microstructures on the adiabatic shear behavior of high-strength Ti–5Al–5V–5Mo–3Cr (Ti-5553) alloy with lamellar microstructure and bimodal microstructure. Lamellar alloy tends to form adiabatic shearing band (ASB) at low compression strain, while bimodal alloy is considerably ASB-resistant. Comparing with the initial microstructure of Ti-5553 alloy, we find that the microstructure of the ASB changes dramatically. Adiabatic shear of lamellar Ti-5553 alloy not only results in the formation of recrystallized β nano-grains within the ASB, but also leads to the chemical redistribution of the alloying elements such as Al, V, Cr and Mo. As a result, the alloying elements distribute evenly in the ASB. In contrast, the dramatic adiabatic shear of bimodal alloy might give rise to the complete lamination of the globular primary α grain and the equiaxial prior β grain, which is accompanied by the dynamic recrystallization of α lamellae and β lamellae. As a result, ASB of bimodal alloy is composed of α/β nano-multilayers. Chemical redistribution does not occur in ASB of bimodal alloy. Bimodal Ti-5553 alloy should be a promising candidate for high performance armors with high mass efficiency due to the processes high dynamic flow stress and excellent ASB-resistance.     

基于ANN模式识别的某型导弹测试车配电系统故障诊断

人工神经网络诊断特点与基于模式识别的诊断特点非常相似。将ANN模式识别技术应用于某型导弹测试车配电系统故障诊断。根据测试车配电系统的故障特点,设计ANN为4层BP网络,具有9个输入、10个输出,两个隐含层神经元数目分别为9和6。测试结果表明该方法能有效诊断测试车配电系统故障。     

Comparative study on transverse shrinkage,mechanical and metallurgical properties of AA2219 aluminium weld joints prepared by gas tungsten arc and gas metal arc welding processes

Aluminium alloy AA2219 is a high strength alloy belonging to 2000 series. It has been widely used for aerospace applications, especially for construction of cryogenic fuel tank. However, arc welding of AA2219 material is very critical. The major problems that arise in arc welding of AA2219 are the adverse development of residual stresses and the re-distribution as well as dissolution of copper rich phase in the weld joint.These effects increase with increase in heat input. Thus, special attention was taken to especially thick section welding of AA2219-T87 aluminium alloy. Hence, the present work describes the 25 mm-thick AA2219-T87 aluminium alloy plate butt welded by GTAW and GMAW processes using multi-pass welding procedure in double V groove design. The transverse shrinkage, conventional mechanical and metallurgical properties of both the locations on weld joints were studied. It is observed that the fair copper rich cellular(CRC) network is on Side-A of both the weldments. Further, it is noticed that, the severity of weld thermal cycle near to the fusion line of HAZ is reduced due to low heat input in GTAW process which results in non dissolution of copper rich phase. Based on the mechanical and metallurgical properties it is inferred that GTAW process is used to improve the aforementioned characteristics of weld joints in comparison to GMAW process.     

基于BP神经网络的极移预报*

为了提高地球定向参数极移的预报精度,建立了一个极移数据预报模型。利用傅里叶分析研究插值基础序列的周期特性,验证了基础序列重采样的可行性,提取插值基础序列数据的趋势项,利用多输入-单输出BP神经网络建模预报不同跨度的残差序列,合并趋势项和残差序列得到最终的极移预报。预报结果表明,选取合适的插值基础序列得到的预报极移精度较高,此BP神经网络能够有效地应用于地球定向参数极移的预报。     

A three-stage model for the perforation of finite metallic plates by long rods at high velocities

A three-stage theoretical model is presented herein to predict the perforation of a thick metallic plate struck normally by a long rod at high velocities. The model is suggested on the basis of the assumption that the perforation of a thick metallic plate by a long rod can be divided into three stages: (1) initial penetration; (2) plug formation and (3) plug slipping and separation. Various analytical equations are derived which can be employed to predict the ballistic limit, residual velocity and residual length of the long rod. It is demonstrated that the present model predictions are in good agreement with available experimental results for the perforation of finite steel targets struck normally by steel as well as tungsten alloy long rods at high velocities. It is also demonstrated that the dynamic maximum shear stress of a plate material has strong effect on plug formation and plug thickness which, in turn, exerts considerable influence on the residual velocities and lengths of a long rod at impact velocities just above the ballistic limit.