Comparison of regression and neural network models for prediction of inspection profiles for aging aircraft

Currently under phase 2 development by the Federal Aviation Administration (FAA), the Safety Performance Analysis System (SPAS) contains alert indicators of aircraft safety performance that can signal potential problem areas for inspectors. The Service Difficulty Reporting (SDR) system is one component of SPAS and contains data related to the identification of abnormal, potentially unsafe conditions in aircraft and/or aircraft components/equipment. SPAS contains performance indicators to assist safety inspectors in diagnosing an airline's safety profile compared with others in the same peer class. This paper details the development of SDR prediction models for the DC-9 aircraft by analyzing sample data from the SDR database that have been merged with aircraft utilization data. Both multiple regression and neural networks are used to create prediction models for the overall number of SDRs and for SDR cracking and corrosion cases. These prediction models establish a range for the number of SDRs outside which safety advisory warnings would be issued. It appears that a data grouping strategy to create aircraft profiles is very effective at enhancing the predictive accuracy of the models. The results from each competing modeling approach are compared and managerial implications to improve the SDR performance indicator in SPAS are provided.     

Performance-based design and seismic reliability analysis using designed experiments and neural networks

Seismic design involves many uncertainties that arise from the earthquake motions, structural geometries, material properties, and analytical models. Taking into account all major uncertainties, reliability analysis is applied to estimate probability of failure in each of a set of performance requirements. The probability estimation is best conducted through Monte Carlo simulations with variance reduction techniques. However, this may involve many performance function evaluations, each requiring a non-linear dynamic analysis, which may be very computationally demanding. In order to improve computational efficiency, this paper explores Design of Computer Experiments and Neural Networks for representation of structural behavior. The neural networks are directly employed for reliability assessment and design optimization. Performance-based seismic design is formulated as an optimization problem, with design parameters optimally calculated. Two case studies are presented to demonstrate efficiency and applicability of the methodology: a bridge bent with or without seismic isolation and a steel pipe pile foundation.     

Surface roughness predictive modeling: neural networks versus regression

Surface roughness plays an important role in product quality and manufacturing process planning. This research focuses on developing an empirical model for surface roughness prediction in finish turning. The model considers the following working parameters: work-piece hardness (material), feed, cutter nose radius, spindle speed and depth of cut. Two competing data mining techniques, nonlinear regression analysis and computational neural networks, are applied in developing the empirical models. The values of surface roughness predicted by these models are then compared with those from some of the representative models in the literature. Metal cutting experiments and tests of hypothesis demonstrate that the models developed in this research have a satisfactory goodness of fit. It has also presented a rigorous procedure for model validation and model comparison. In addition, some future research directions are outlined.     

Self-organizing neural systems based on predictive learning

The ability to predict future events based on the past is an important attribute of organisms that engage in adaptive behaviour. One prominent computational method for learning to predict is called temporal-difference (TD) learning. It is so named because it uses the difference between successive predictions to learn to predict correctly. TD learning is well suited to modelling the biological phenomenon of conditioning, wherein an organism learns to predict a reward even though the reward may occur later in time. We review a model for conditioning in bees based on TD learning. The model illustrates how the TD-learning algorithm allows an organism to learn an appropriate sequence of actions leading up to a reward, based solely on reinforcement signals. The second part of the paper describes how TD learning can be used at the cellular level to model the recently discovered phenomenon of spike-timing-dependent plasticity. Using a biophysical model of a neocortical neuron, we demonstrate that the shape of the spike-timing-dependent learning windows found in biology can be interpreted as a form of TD learning occurring at the cellular level. We conclude by showing that such spike-based TD-learning mechanisms can produce direction selectivity in visual-motion-sensitive cells and can endow recurrent neocortical circuits with the powerful ability to predict their inputs at the millisecond time-scale.     

基于神经网络的地形匹配方法研究

提出了基于神经网络实现多特征融合的地形匹配算法，充分利用地形的各种不同的统计特征和几何特征，构造了一种地形匹配网络模型．通过对实时图和基准图的分析，给出了计算网络节点之间的权值函数，建立了网络系统能量方程，通过求系统的最小能量得到最佳匹配位置．由于网络能融合地形的不同统计特征和几何特征，所以算法大大提高了系统的抗干扰能力和定位精度，适合于实时图容易发生畸变的地形匹配领域．实验结果表明，定位精度和抗干扰能力均优于传统的地形匹配方法．     

基于神经网络混合建模的结构振动滑模控制

将神经网络和标称系统混合建模方法引入到离散滑模控制当中,得到神经网络滑模控制,然后对结构振动进行控制,振动结构为具有不确定性参数的柔性附件,并受到随机外扰作用。离散滑模控制的滑模面是以标称系统为基础,由最优二次型价值函数求解黎卡提方程确定。利用标称模型和神经网络混合建模方法来减小系统的不确定性,达到提高滑模控制在实际控制系统中的控制效果。其中利用前馈神经网络来对不确定部分进行建模。最后通过对滑模控制和神经网络滑模控制进行仿真,结果表明,本文所提出的神经网络滑模控制对具有不确定性参数和随机外扰的柔性结构系统振动的控制效果要优于滑模控制。     

Autoprogressive training of neural network constitutive models

A new method, termed autoprogressive training, for training neural networks to learn complex stress–strain behaviour of materials using global load–deflection response measured in a structural test is described. The richness of the constitutive information that is generally implicitly contained in the results of structural tests may in many cases make it possible to train a neural network material model from only a small number of such tests, thus overcoming one of the perceived limitations of a neural network approach to modelling of material behaviour; namely, that a voluminous amount of material test data is required. The method uses the partially-trained neural network in a central way in an iterative non-linear finite element analysis of the test specimen in order to extract approximate, but gradually improving, stress–strain information with which to train the neural network. An example is presented in which a simple neural network constitutive model of a T300/976 graphite/epoxy unidirectional lamina is trained, using the load–deflection response recorded during a destructive compressive test of a [(±45)₆]_S laminated structural plate containing an open hole. The results of a subsequent forward analysis are also presented, in which the trained material model is used to simulate the response of a compressively loaded [(±30)₆]_S structural laminate containing an open hole. Avenues for further improvement of the neural network model are also suggested. The proposed autoprogressive algorithm appears to have wide application in the general area of Non-Destructive Evaluation (NDE) and damage detection. Most NDE experiments can be viewed as structural tests and the proposed methodology can be used to determine certain damage indices, similar to the way in which constitutive models are determined. © 1998 John Wiley & Sons, Ltd.     

基于卷积神经网络和Transformer网络的鸟声识别

针对传统鸟声识别算法中特征提取方式单一、分类识别准确率低等问题,提出一种结合卷积神经网络和Transformer网络的鸟声识别方法。该方法综合考虑网络局部特征学习和全局上下文依赖性构造,从原始鸟声音频信号中提取短时傅里叶变换(Short Time Fourier Transform,STFT)语谱图特征,将其输入到卷积神经网络(ConvolutionalNeural Network,CNN)中提取局部频谱特征信息,同时提取鸟声信号的对数梅尔特征及一阶差分、二阶差分特征用于合成梅尔频率倒谱系数(Mel Frequency Cepstrum Coefficient,MFCC)混合特征向量,将其输入到Transformer网络中获取全局序列特征信息,最后融合所提取的特征可得到更丰富的鸟声特征参数,通过Softmax分类器得到鸟声识别结果。在Birdsdata和xeno-canto鸟声数据集上进行实验,平均识别准确率分别达到了97.81%和89.47%。实验结果表明该方法相较于其他现有的鸟声识别模型具有更高的识别准确率。     

基于神经网络的图像边缘检测方法

提出了一种基于神经网络的图像边缘检测新方法.该方法首先基于邻域灰度极值提取边界候选图像,然后以边界候选象素及其邻域象素的二值模式作为样本集,输入边缘检测神经网络进行训练.边缘检测神经网络采用BP网络,为加快网络的训练速度,采用了滚动训练和权值随机扰动的方法.实验表明,该方法提高了神经网络的学习效率,获得的边缘图像封闭性好,边缘描述真实.     

基于神经网络趋势分析

文章在分析研究了国内外现状的基础上 ,利用神经网络的非线性处理特性 ,提出了通过神经网络预测常见机械零件剩余寿命的方法 ,用实例验证了其有效性     

Modelling of plasma etching using a generalized regression neural network

Plasma etching was modelled by using a generalized regression neural network (GRNN). The etching process was characterized with a statistical experimental design. Three etch responses were modelled, which include two etch rates of aluminium and silica and etching profile. GRNN prediction ability was optimized as a function of training factor. Three types of models were constructed depending on the type of prepared data. Type I model corresponds to the model constructed with the original, non-classified data. Type II and III models were built for the classified data without and with the control of data interface, respectively. Compared to type I models, type II models for two etch rates demonstrated more than 25% improvement. By the control of data interface, type III models exhibited more than 15% improvement over type II models. Classification-based models in conjunction with data control thus illustrated much improved prediction of GRNN over those for non-classified models.     

基于Kohonen神经网络算法的网络入侵聚类算法的测试研究

提出一种基于Kohonen网络的网络入侵聚类研究的方法,在阐述基本理论、原理和算法步骤基础上,利用Matlab软件平台对提出的网络入侵算法进行测试研究,并同其他方法进行仿真对比,发现Kohonen神经网络算法的网络入侵聚类在训练准确率、测试准确率和运行时间3个方面都优于PNN算法,其准确率可以达到98.1％.     

基于BP神经网络的导线脱冰跳跃高度预测模型

冰区导线脱冰振动会引起绝缘间隙减小,严重时甚至导致闪络和跳闸等电气事故.首先利用数值方法模拟得到各种参数条件下导线的脱冰动力响应,获得导线的最大脱冰跳跃高度.进而基于数值模拟结果和BP神经网络构建导线脱冰跳跃高度预测模型,将线路的导线分裂数、导线型号、档距、高差等结构参数以及初始应力、覆冰厚度和脱冰率等载荷参数作为输入...     

Adaptable product configuration system based on neural network

In the rapidly diversifying and globalising market, product configuration is implemented in a dynamic environment with continuous change of configuration knowledge. The adaptability of the product configuration system, which is defined as the capability to adjust product configurator, human resources and organisational resources to fit a new environment, is becoming more and more crucial. To keep the adaptability, this research suggests an adaptable product configuration (APC) system which transforms the development of the configuration system in a dynamic environment from a straightforward process to a closed circle. In the existing research on product configuration, most issues are addressed separately by different approaches and most approaches lack a systematic view which considers the interaction among product configurator, resources and environment. The circle of APC is therefore divided into several isolated stages and involves intensive human work, consumes a lot of organisational resources and results in a long response time. To successfully implement APC, this research adopts an artificial neural network and a specific rule extraction mechanism to develop a product configuration system. The neural network is able to automatically acquire configuration knowledge from historical transaction data and then directly apply it without further knowledge programming. Rule extraction mechanism has the capability to interpret the behaviour of the trained neural network and make it comprehensible and adjustable. Finally, knowledge acquisition, representation and application in product configuration are incorporated into the same connectionist methodology. And consequently, the APC circle is accelerated and the adaptability of the product configuration system is improved. A case study of computer configuration is presented.     

基于脉冲耦合神经网络的红外图像增强

在分析了脉冲耦合神经网络的工作机理和行为特性后,指出可以利用神经元的点火-熄灭特性对图像进行增强.为了区分神经元的点火方式,提出一种根据链接矩阵判定神经元点火方式的方法,并利用自然点火和捕获点火建立了能使图像得到增强的非线性映射.文中对算法参数的设置及其对增强图像的影响做了详细地讨论,实验结果表明该算法不仅能使图像的对比度和亮度得到适当的增强,而且能够有效地抑制图像中的椒盐噪声,尤其适用于对比度和亮度都较低的红外图像.     

Prediction of plasma etching using a randomized generalized regression neural network

A new empirical technique to construct predictive models of plasma etch processes is presented. This was accomplished by combining a generalized regression neural network (GRNN) and a random generator (RG). The RG played a critical role to control neuron spreads in the pattern layer. The proposed R-GRNN was evaluated with experimental plasma etch data. The etching of silica thin films was characterized by a 2³ full factorial experiment. The etch responses examined include aluminium etch rate, silica etch rate, profile angle, and DC bias. Additional test data were prepared to evaluate model appropriateness. Compared to conventional GRNN, the R-GRNN demonstrated much improved predictions of more than 40% for all etch responses. This was illustrated over statistical regression models. As a result, the proposed R-GRNN is an effective way to considerably improve the predictive ability of conventional GRNN.     

近红外光谱结合神经网络判别茶鲜叶产地

为快速、无损的判别鲜叶产地,维护恩施玉露的地理标志产品属性,采集恩施市芭蕉乡、白果乡和咸丰县茶鲜叶近红外光谱,经光谱预处理后,对校正集66个样品光谱数据进行主成分分析,然后建立BP神经网络预测模型,对验证集鲜叶样品的产地进行了预测,建立了8(输入节点)-4(隐含层节点)-1(输出节点)三层网络模型,验证集样品判别准确率为100%.近红外光谱技术结合神经网络能够快速、准确地判别茶鲜叶产地.     

Stereo vision calibration based on GMDH neural network

In order to improve the accuracy and stability of stereo vision calibration, a novel stereo vision calibration approach based on the group method of data handling (GMDH) neural network is presented. Three GMDH neural networks are utilized to build a spatial mapping relationship adaptively in individual dimension. In the process of modeling, the Levenberg-Marquardt optimization algorithm is introduced as an interior criterion to train each partial model, and the corrected Akaike's information criterion is introduced as an exterior criterion to evaluate these models. Experiments demonstrate that the proposed approach is stable and able to calibrate three-dimensional (3D) locations more accurately and learn the stereo mapping models adaptively. It is a convenient way to calibrate the stereo vision without specialized knowledge of stereo vision.     

Role of neural network models for developing speech systems

This paper discusses the application of neural networks for developing different speech systems. Prosodic parameters of speech at syllable level depend on positional, contextual and phonological features of the syllables. In this paper, neural networks are explored to model the prosodic parameters of the syllables from their positional, contextual and phonological features. The prosodic parameters considered in this work are duration and sequence of pitch (F ₀) values of the syllables. These prosody models are further examined for applications such as text to speech synthesis, speech recognition, speaker recognition and language identification. Neural network models in voice conversion system are explored for capturing the mapping functions between source and target speakers at source, system and prosodic levels. We have also used neural network models for characterizing the emotions present in speech. For identification of dialects in Hindi, neural network models are used to capture the dialect specific information from spectral and prosodic features of speech.