Defect Detection in Carbon-Fiber Composites using Lamb-Wave Tomographic Methods

Lamb-wave tomography (LWT) offers a powerful nondestructive technique for the health assessment of large structures as their propagation properties depend on the thickness and the mechanical properties of the material. Development of a fast and accurate algorithm for defect detection is of paramount importance in any structural-health-monitoring (SHM) system. The present study explores the prospects of LWT as a SHM technique with an accent on developing a suitable algorithm for real-time inspection. Projection data is collected by electronically scanning an array of ultrasonic sensors arranged in a modified cross-hole geometry. The data thus collected is investigated to extract energy profile of the traveling waves. Multiplicative algebraic reconstruction technique (MART) algorithms are used as a tool for tomographic reconstruction from a set of multiple independent measurements. The performance of algorithms is evaluated from the point of view of the cost of algorithm, achievable resolution, and accuracy of results. Experimental results show that MART is capable of characterizing defects in thin isotropic and composite plates within a reasonable error band (±26% normalized, ±2.6 RMS) and is suitable for application to LWT of large structures such as aircraft skins.     

Local cone-beam tomography image reconstruction on chords

We develop reconstruction algorithms for local cone-beam tomography for use with generalized scanning trajectories. The algorithms are grounded theoretically in a recently developed chord-based theory for exact image reconstruction and principles of lambda tomography. Being chord based, they are distinct mathematically and conceptually from conventional local tomography reconstruction algorithms. The salient feature of our algorithms is that they permit reconstruction of discontinuities in the profiles of the object function along chords. By consideration of all possible chords, a 3D image that describes the locations of object discontinuities can be reconstructed. Results from microlocal analysis are applied for understanding the object features that can be reconstructed stably by use of the algorithms. A computer-simulation study is conducted to demonstrate the algorithms and compare their performance with an existing algorithm.     

Cubic spline algorithms for orientation interpolation

This article presents a class of spline algorithms for generating orientation trajectories that approximately minimize angular acceleration. Each algorithm constructs a twice‐differentiable curve on the rotation group SO(3) that interpolates a given ordered set of rotation matrices at specified knot times. Rotation matrices are parametrized, respectively, by the unit quaternion, canonical co‐ordinate, and Cayley–Rodrigues representations. All the algorithms share the common feature of (i) being invariant with respect to choice of fixed and moving frames (bi‐invariant), and (ii) being cubic in the parametrized co‐ordinates. We assess the performance of these algorithms by comparing the resulting trajectories with the minimum angular acceleration curve. Copyright © 1999 John Wiley & Sons, Ltd.     

Optimized on-line control of MMA polymerization using fast multi-objective DE

Optimized on-line control (OOC) of polymerization reactors combine the optimization with the on-line operation and control. In this, re-optimized control variable trajectories, in the presence of unplanned disturbances, are obtained and implemented on-line to save the batch. Also, the available computational time for the optimization is limited as the re-optimized trajectories need to be implemented in real time on the actual system. In the present study, the OOC of such a system, i.e., bulk polymerization of methyl methacrylate (MMA) in a batch reactor, is carried out in the occurrence of heater malfunction. To solve the underlying multi-objective problem, a multi-objective variant of differential evolution with an improved mutation strategy is developed. The developed algorithm shows faster convergence with respect to other compared algorithms for a large number of benchmark problems. Finally, this algorithm is used to find the optimal temperature trajectories and the OOC with these trajectories found to be successfully countering the effect of heater malfunction.     

Improved genetic algorithm based on particle swarm optimization-inspired reference point placement

This article investigates the use of optimal reference point placement to improve performance of non-dominated sorting genetic algorithm (NSGA). Placement of reference points for many-objective optimization is inspired by wheel and Von Neumann topologies of Particle Swarm Optimization (PSO). Results obtained show that the pattern of reference point placement determines performance efficiency of NSGA. The better-performing wheel topology (called wheel reference point genetic algorithm (wRPGA), is compared to three other many-objective evolutionary algorithms: knee-driven evolutionary algorithm (KnEA), non-dominated sorting genetic algorithm III (NSGAIII) and multi-objective evolutionary algorithm based on dominance and decomposition (MOEAD/D). The selected many-objective benchmark problems are Walking Fish Group 2 (WFG2) and Deb-Thiele-Laumanns-Zitzler 2 (DTLZ2). It is also tested on a 3-objective cost function for a hypothetical model of a stand-alone microgrid. Through the simulations, the wheel configuration performed 88.9% better than the Von Neumann configuration. The wheel topology also achieved better performance with respect to inverted generational distance (IGD) compared to KnEA, NSGAIII and MOEAD/D for 7 out of 15 IEEE Congress on Evolutionary Computation (CEC) 2017 benchmark problems. wRPGA gave a good approximation of the Pareto front for the 3-objective model representing the hypothetical microgrid.     

Research on Multi-View Image Reconstruction Technology Based on Auto-Encoding Learning

Traditional three-dimensional (3D) image reconstruction method, which highly dependent on the environment and has poor reconstruction effect, is easy to lead to mismatch and poor real-time performance. The accuracy of feature extraction from multiple images affects the reliability and real-time performance of 3D reconstruction technology. To solve the problem, a multi-view image 3D reconstruction algorithm based on self-encoding convolutional neural network is proposed in this paper. The algorithm first extracts the feature information of multiple two-dimensional (2D) images based on scale and rotation invariance parameters of Scale-invariant feature transform (SIFT) operator. Secondly, self-encoding learning neural network is introduced into the feature refinement process to take full advantage of its feature extraction ability. Then, Fish-Net is used to replace the U-Net structure inside the self-encoding network to improve gradient propagation between U-Net structures, and Generative Adversarial Networks (GAN) loss function is used to replace mean square error (MSE) to better express image features, discarding useless features to obtain effective image features. Finally, an incremental structure from motion (SFM) algorithm is performed to calculate rotation matrix and translation vector of the camera, and the feature points are triangulated to obtain a sparse spatial point cloud, and meshlab software is used to display the results. Simulation experiments show that compared with the traditional method, the image feature extraction method proposed in this paper can significantly improve the rendering effect of 3D point cloud, with an accuracy rate of 92.5% and a reconstruction complete rate of 83.6%.     

Direct-Fourier reconstruction in tomography and synthetic aperture radar

We investigate the use of direct-Fourier (DF) image reconstruction in computed tomography and synthetic aperture radar (SAR). One of our aims is to determine why the convolution-backprojection (CBP) method is favored over DF methods in tomography, while DF methods are virtually always used in SAR. We show that the CBP algorithm is equivalent to DF reconstruction using a Jacobian-weighted two-dimensional periodic sinc-kernel interpolator. This interpolation is not optimal in any sense, which suggests that DF algorithms using optimal interpolators may surpass CBP in image quality. We consider use of two types of DF interpolation: a windowed sinc kernel, and the least-squares optimal Yen interpolator. Simulations show that reconstructions using the Yen interpolator do not possess the expected visual quality, because of regularization needed to preserve numerical stability. Next, we show that with a concentric-squares sampling scheme, DF interpolation can be performed accurately and efficiently, producing imagery that is superior to that obtainable by other algorithms. In the case of SAR, we show that the DF method performs very well with interpolators of low complexity. We also study DF reconstruction in SAR for trapezoidal grids. We conclude that the success of the DF method in SAR imaging is due to the nearly Cartesian shape of the sampling grid. © 1998 John Wiley & Sons, Inc. Int J Imaging Syst Technol, 9, 1–13, 1998     

Three dimension reconstruction through measure-based image selection

Three dimension (3D) reconstruction is one of the research focus of computer vision and widely applied in various fields. The main steps of 3D reconstruction include image acquisition, feature point extraction and matching, camera calibration and production of dense 3D scene models. Generally, not all the input images are useful for camera calibration because some images contain similar and redundant visual information. These images can even reduce the calibration accuracy. In this paper, we propose an effective image selection method to improve the accuracy of camera calibration. Then a new 3D reconstruction algorithm is proposed by adding the image selection step to 3D reconstruction. The image selection method uses structure-from-motion algorithm to estimate the position and attitude of each camera, first. Then the contributed value to 3D reconstruction of each image is calculated. Finally, images are selected according to the contributed value of each image and their effects on the contributed values of other images. Experimental results show that our image selection algorithm can improve the accuracy of camera calibration and the 3D reconstruction algorithm proposed in this paper can get better dense 3D models than the normal algorithm without image selection.     

3D非均匀直线网格GPU体绘制方法研究

计算机图形硬件技术的快速发展可以用来加速可视化过程,为此针对非均匀直线网格,给出了基于均匀辅助网格的CPU光线投射算法、基于辅助纹理的GPU光线投射算法,以及基于切片的3D纹理体绘制算法,并在Nvidia Geforce 6800GT图形卡上对这些算法进行了测试。结果表明,GPU算法远远快于CPU算法,而基于切片的3D纹理体绘制算法则快于GPU光线投射算法。     

A quasi‐implicit characteristic–based penalty finite‐element method for incompressible laminar viscous flows

In this paper, a novel characteristic–based penalty (CBP) scheme for the finite‐element method (FEM) is proposed to solve 2‐dimensional incompressible laminar flow. This new CBP scheme employs the characteristic‐Galerkin method to stabilize the convective oscillation. To mitigate the incompressible constraint, the selective reduced integration (SRI) and the recently proposed selective node–based smoothed FEM (SNS‐FEM) are used for the 4‐node quadrilateral element (CBP‐Q4SRI) and the 3‐node triangular element (CBP‐T3SNS), respectively. Meanwhile, the reduced integration (RI) for Q4 element (CBP‐Q4RI) and NS‐FEM for T3 element (CBP‐T3NS) with CBP scheme are also investigated. The quasi‐implicit CBP scheme is applied to allow a large time step for sufficient large penalty parameters. Due to the absences of pressure degree of freedoms, the quasi‐implicit CBP‐FEM has higher efficiency than quasi‐implicit CBS‐FEM. In this paper, the CBP‐Q4SRI has been verified and validated with high accuracy, stability, and fast convergence. Unexpectedly, CBP‐Q4RI is of no instability, high accuracy, and even slightly faster convergence than CBP‐Q4SRI. For unstructured T3 elements, CBP‐T3SNS also shows high accuracy and good convergence but with pressure oscillation using a large penalty parameter; CBP‐T3NS produces oscillated wrong velocity and pressure results. In addition, the applicable ranges of penalty parameter for different proposed methods have been investigated.     

气体浓度层析成像重建算法性能比较与研究

针对少射线成像算法在研究气体浓度层析成像中的重要作用,对常用的几种少射线成像算法的性能进行了详细分析和比较,同时构造了一个多目标函数用于气体浓度优化重建成像。通过实验与常用的几种算法的重建性能相比,多目标优化重建算法具有良好的鲁棒性和抗噪性。在含5%测量误差的情况下,ART 重建结果与原始浓度分布的欧氏距离达 0.1680,而多目标优化重建算法下的该值仅为 0.0325。     

体绘制及三维交互技术在地质数据可视化中的应用

对地质勘探所提供的大规模三维规则风格数据进行可视化处理。     

Fast computation for generating CGH of a 3D object by employing connections between layers

An approximate method for generating computer-generated holograms (CGH) of a 3D object with six times faster speed than the conventional algorithm is presented. In the conventional algorithm, a 3D object is sliced into many layers and treated as a collection of self-illuminated point light source. The propagation process of a light ray from every point of an object to all the points on the hologram plane is simulated and interfered with by the reference beam to form a CGH. In our proposed method, under the assumption that the depth of a 3D object is much smaller than the recording distance, we just need to calculate the oblique distance between the first layer and the hologram plane, and then the oblique distances from the other layers to the hologram plane can be obtained from a simple relation, thus the computational time is much reduced. The CGH is optically reconstructed and the quality of the reconstructed image agrees well with that from the conventional algorithm.     

分布源目标方位估计的降维最大似然估计

介绍了已有的分布源目标方位估计中的最大似然估计(MLE)算法,它是四维非线性最优化问题,文中称之为四维MLE算法,因计算量庞大,同时提出了一种降为三维的MLE算法,简化为三维非线性最优化,称之为三维MLE算法。两种算法均采用牛顿型搜索算法,来搜索未知参数的全局最优点。在单次迭代过程中,三维MLE算法比四维MLE算法减少了51次协方差矩阵求逆和87次矩阵乘法,搜索效率得到提高,并且节省了存储空间。得出了新算法克拉美-罗界的计算公式,其计算量也有所降低。计算机仿真验证,三维MLE算法和四维MLE算法的估计精度相当,新算法在减少计算量的同时并无损失性能,所以实用性和实时性都得到显著提高。     

An Algorithm for Identifying a Crack Within a Measured Displacement Field

This paper discusses an algorithm for the detection of crack damage in plate structures based on the governing differential equation (GDE) of in-plane displacement of a plate. A state-of-the-art 3D scanning laser Doppler vibrometer, which is able to provide very accurate measurements of a three-dimensional displacement field, was utilized to implement the algorithm. To evaluate the GDE from the measured displacements, a 2D Savitzky–Golay differentiating filter was used. The potential of this algorithm for detecting a crack was compared against another two algorithms which are based on displacement measurements. The first is an algorithm based on the error in smoothed to measured displacements and the other uses surface strains normalized by their mean. To investigate the efficiency of the algorithms in detecting crack damage, a number of PMMA plate specimens were fabricated with a range of crack lengths extending from a V-notch and tested under a cyclic, quasi-static, uni-axial load. For each algorithm, the crack location could be positively identified. However, the two algorithms based on the error in smoothed to measured displacements and surface strains normalized by their mean were found to be dependent on the direction of the applied load, whereas, theoretically the algorithm based on the GDE of in-plane displacement would work regardless of the direction of the applied load.     

二维建筑结构图的三维模型重建

分析了建筑结构图的工程对象类型和描述方式特征,针对建筑领域的典型实例提出了与机械三维重建不同的建筑工程三维重建方案。该方案按照最常用的结构制图标准,以DXF格式的结构图纸为处理对象,以AutoCAD二次开发为主要技术路线,选择了标准层平面图为主要对象视图。讨论了建筑工程对象的数据结构设计和重建微观策略,并给出重建结果的实例。     

Developing exact and Tabu search algorithms for simultaneously determining AGV loop and P/D stations in single loop systems

There are some issues which have to be addressed when designing an automated guided vehicles system (AGVS) such as flow-path layout, traffic management, the number and the location of pick up and delivery points, vehicle routing and so on. One of the AGVS guide path configurations discussed in the previous researches includes a single-loop which is the subject of this paper. In unidirectional single loop systems, determining the loop for the motion of an AGV, and the location of pick up and delivery (P/D) stations in the cells, are prominent points which, when considered simultaneously, lead to better results than determining each one independently. However, in the literature it is proved that the problem of separately determining the shortest feasible loop is a NP-complete problem. In this paper, by considering a from-to chart and a block layout as the input of problem, we try to determine: (1) a single loop, with at least one shared edge with each cell, (2) the direction of the flow and (3) the location of P/D stations on the loop, all at the same time, in a way that the total travel distance on the loop be minimised. In this regard, first a new exact algorithm is presented and then three heuristic algorithms are developed utilising a Tabu search (TS) method. Solving randomly generated test problems shows that our exact algorithm is capable of solving small size problems; also all three TS algorithms work efficiently in solving problems that could not be solved by exact algorithms.     

Experimental exploration on loss surface of deep neural network

The loss function of the deep neural network is high dimensional, nonconvex and complex. So far, the geometric properties of the loss surface of the neural network have not been well understood. Different from most theoretical studies on the loss surface, this article makes the experimental exploration on the loss surface of the deep neural network, including trajectories of various adaptive optimization algorithms, the Hessian matrix of the loss function of the deep neural network, the curvature of the loss surface along the trajectories of the various adaptive optimization algorithms. It is found that the gradient direction of the adaptive optimization algorithms is almost perpendicular to the direction of the maximum curvature of the loss surface, while the gradient directions of the stochastic gradient descent (SGD) algorithm do not have such a rule. The Hessian matrix of the loss surface along the trajectory of the optimization algorithm is degraded, which is inconsistent with the hypothetical that nonsingular of the Hessian matrix in many theoretical studies of deep learning. Besides, this article proposes a new ensemble learning method of the neural network based on the scaling invariance of the ReLu neural network and mode connectivity.     

机器人辅助无框架定位脑外科手术系统

研制了机器人辅助无框架定位脑外科手术系统,该系统可以在CT和MRI图像上实施病灶以及头颅的半自动分割,并将分割结果重构为三维模型,此模型可以进一步应用于虚拟手术规划及手术实施。它提供的三维组织模型清晰准确,手术路径精确稳定,从而提高了手术的效率和质量,降低了手术的费用。该系统在多例临床手术中获得了良好的效果。     

基于图像处理的外螺纹三维模型重构

提出了一种基于多角度序列图像特征实现外螺纹的三维模型重建的方法。首先在旋转平台上采集多角度序列螺纹件图像,然后对每帧图像进行特征点提取,将序列图像的特征点进行三维变换和插值,最终生成三维模型。实验结果表明,此算法能精确高精度地实现外螺纹三维模型重构。