基于神经网络优化Cu-W-Ni电镀工艺

本文用神经网络BP优化电镀Cu-W-Ni工艺。BP预测数据与同参数正交实验结果相同,优化后的CuW-Ni镀层质量好。说明BP神经网络有很好的非线性映射能力和泛化能力,与传统的实验方法比较,优化复杂的电镀工艺参数更具有优越性。     

Predicting indoor PM2.5/PM10 concentrations using simplified neural network models

Journal of Mechanical Science and Technology - Neural network models were presented for prediction of indoor concentrations of particulate matters (PMs). Indoor PM concentrations are generally...     

Prediction of the surface roughness and material removal rate in chemical mechanical polishing of single-crystal SiC via a back-propagation neural network

Chemical mechanical polishing (CMP) is a common method for realising the global planarisation and polishing of single-crystal SiC and other semiconductor substrates. The strong oxidant hydroxyl radicals (·OH) generated by the Fenton reaction can effectively oxidise and corrode the SiC substrate, and are thus used to improve the material removal rate (MRR) and surface roughness (Ra) after polishing of SiC during CMP. Therefore, it is necessary to study the material removal mechanism in detail. Based on the modified Preston equation, the effects of the CMP process parameters on the MRR and Ra after polishing of SiC and their relationship were studied, and a prediction model of the CMP process parameters, MRR, and Ra after polishing was also established based on a back-propagation neural network. The MRR initially increased and then decreased, and the Ra after polishing initially decreased and then increased, with increasing FeSO₄ concentration, H₂O₂ concentration, and pH value. The MRR continuously increased with increasing abrasive particle size, abrasive concentration, polishing pressure, and polishing speed. However, the Ra continuously decreased with increasing abrasive particle size and abrasive concentration, increased with increasing polishing pressure, and initially decreased and then increased with increasing polishing speed. The established prediction model could accurately predict the relationship between the process parameters, MRR and Ra after polishing in CMP (relative prediction error of less than 10%), which could provide a theoretical basis for CMP of SiC.     

Acoustic diagnosis of a pump by using neural network

A fundamental study for developing a fault diagnosis system of a pump is performed by using neural network. Acoustic signals were obtained and converted to frequency domain for normal products and artificially deformed products. The neural network model used in this study was 3-layer type composed of input, hidden, and output layer. The normalized amplitudes at the multiples of real driving frequency were chosen as units of input layer. And the codes of pump malfunctions were selected as units of output layer. Various sets of teach signals made from original data by eliminating some random cases were used in the training. The average errors were approximately proportional to the number of untaught data. Neural network trained by acoustic signals can detect malfunction or diagnose fault of a given machine from the results.     

A method for rapidly evaluating reliability and predicting remaining useful life using two-dimensional convolutional neural network with signal conversion

Journal of Mechanical Science and Technology - Real-time monitoring and rapid evaluation of bearing operating conditions, especially for the reliability evaluation and remaining useful life (RUL)...     

A neural network based approach for background noise reduction in airborne acoustic emission of a machining process

Tool wear prediction has become an indispensable technique to prevent downtime in manufacturing and production processes. Airborne emission from a machining process using a low-cost microphone may provide a vital signal of tool health. However, the effect of background noise results in anomaly in data that may lead to wrong prediction of tool health. The paper presents an adaptive approach using neural networks for background noise filtration in acoustic signal for a turning process. Acoustic signal of a turning process is mixed with background noise from four different machines and introduced at different RPMs and feed-rate at a constant depth of cut. A comparison of Backpropagation neural network (BPNN), Self-organizing map and k-means clustering algorithm for noise filtration is investigated in this paper. In this regard, back-propagation neural network showed better performance with an average accuracy for all the four sources. It shows 100 % accuracy for grinding machine signal, 94.78 % accuracy for background signal from 3-axis milling machine, 45.57 % and 12.69 % for motor and 4-axis milling machine, respectively. Signal reconstruction is then done using Discrete cosine transform (DCT). The proposed technique shows a promising future for noise filtration in airborne acoustic data of a machining process.     

Intelligent condition monitoring of a gearbox using artificial neural network

This paper concentrates on a new procedure which experimentally recognises gears and bearings faults of a typical gearbox system using a multi-layer perceptron neural network. Feature vector which is one of the most significant parameters to design an appropriate neural network was innovated by standard deviation of wavelet packet coefficients. The gear conditions were considered to be normal gearbox and slight- and medium-worn and broken-teeth gears faults and a general bearing fault which were five neurons of output layer with the aim of fault detection and identification. A downscaled 2-layer multi-layer perceptron neural-network-based system with great accuracy was designed to carry out the task. In this research, vibration signals were recognised as the most reliable source to extract the feature vector which were synchronised by piecewise cubic hermite interpolation (PCHI) and pre-processed using the standard deviation of wavelet packet coefficients.     

Optimisation of the electrical discharge machining process using a GA-based neural network

In this paper, the optimisation of the EDM process parameters from the rough cutting stage to the finish cutting stage has been reported. A trained neural network was used to establish the relationship between the process parameters and machining performance. Genetic algorithms with properly defined objective functions were then adapted to the neural network to determine the optimal process parameters. Examples with specifications intentionally assigned the same values as those recorded in the database or selected arbitrarily have been fed into the developed GA-based neural network in order to verify the optimisation ability throughout the machining process. Accordingly, the optimised results indicate that the GA-based neural network can be successfully used to generate optimal process parameters from the rough cutting stage to the finish cutting stage.     

Multiparameter optimization for the nonlinear performance improvement of centrifugal pumps using a multilayer neural network

To increase efficiency at the design point of a centrifugal pump, this study adopted an artificial neural network in the construction of an accurate nonlinear function between the optimization objective and the design variables of impellers. Modified particle swarm optimization was further applied to refine the mathematical model globally. The database, which consisted of 200 sets of impellers, were generated from the Latin hypercube sampling method, and their corresponding efficiencies were obtained automatically from numerical simulation. Design variables were the distributions of blade angles, and results established that the difference between the numerical performance curve and the experimental results was acceptable. Optimization with a two-layer feedforward network improved the pump efficiency at the design point by 0.454 %. Flow complexity improved as the blade curvature increased. The application of the multilayer neural network could provide a meaningful reference to single- and multi-objective optimization of complex and nonlinear pump performance.

     

基于模糊BP网络的加工方法链决策

为实现计算机辅助工艺计划(CAPP)系统中零件特征加工方法的决策,分析了影响零件加工方法的因素,提出利用模糊BP神经网络算法对零件特征加工方法链进行优化决策.根据BP网络能够学习任何模式的映射关系的特点,建立了从输入到输出的网络决策模型,并对网络结构、参数确定和样本选取等问题进行详细阐述,运用合理的学习算法来训练网络,最后通过实例验证了该网络的有效性.结果表明,利用模糊BP神经网络进行零件加工方法的选择是有效的和可靠的.     

基于Adaline神经网络的家用电器谐波分析

为了分析常用家用电器用电对公共电网的谐波污染,应用Adaline神经网络对其进行自适应谐波分析。改进的增强型Adaline神经网络将频率作为待定权值,同时估计被测信号的频率、幅值和相位。幅值相位权值的学习采用变步长方法,频率权值的学习采用动量项方法,提高了收敛速度。修正的频率调整公式和频率延迟调整策略简化了频率学习率的设置。基于实测电压信号的对比研究验证了改进算法的收敛性能和分析精度。通过数据采集实验装置得到计算机、电视机、洗衣机、微波炉等家用电器的用电波形,并用改进的Adaline方法对波形信号进行谐波分析。实验结果表明,计算机的电流谐波总畸变率超过60%,微波炉的电流谐波总畸变率在40%以上,电视机和洗衣机的电流谐波总畸变率在10%以上。     

Construction of a prediction model for the structural stability of a surface grinder using backpropagation neural network

Learning and prediction capability of the backpropagation neural network (BPNN) have been used to build the prediction model for the structural stability of a surface grinder. The Lagrange energy method is applied to derive the dynamic equation of the lumped parameter model of the surface grinder. The major factors influencing the structural stability of the system can be determined after the ratio of kinetic energy of the sub-structure and the ratio of potential energy of the sub-structure interface are obtained. An orthogonal rotatable central composite design is adopted to dispose the treatment combinations of the major factors. The BPNN model is constructed by the treatment combinations of the training patterns and verified by the treatment combinations of the test patterns. In this paper, a 3-layer BPNN model with a 10-neuron hidden layer which converged after 4,072 learning cycles is selected to predict the structural stability of a surface grinder within the planned ranges. The percentage residuals of both training patterns and test patterns are all within 3.41%, thus the prediction accuracy of the BPNN model is excellent so that the engineering demands are well satisfied.     

基于神经网络的压力容器焊缝磁性无损检测的研究

研究了压力容器焊缝磁性无损检测系统。结合压力容器检测的具体要求，设计了检测装置，探讨神经网络的方法在压力容器焊缝缺陷的特征识别应用。应用表明基于神经网络的压力容器焊缝磁性无损检测技术具有重要的应用前景。     

基于神经网络的压力容器焊缝磁性无损检测的研究

研究了压力容器焊缝磁性无损检测系统。结合压力容器检测的具体要求,设计了检测装置,探讨神经网络的方法在压力容器焊缝缺陷的特征识别应用。应用表明基于神经网络的压力容器焊缝磁性无损检测技术具有重要的应用前景。     

Research on localization algorithm of naval vessel for submarine based on bearing-range measurements

Combined with naval vessel practical antisubmarine equipment of towed linear array sonar,a mathematical model of naval vessel localization for submarine based on bearing measurement was built,and local...     

实验数据RBF神经网络模型中噪声的处理方法

实验数据的非线性建模,是对各种仪器、设备的性能进行校正和补偿的基础.讨论了神经网络非线性建模时数据中的噪声成分造成的过拟合现象以及对模型精度的影响,针对RBF网络给出了2种提高建模精度的方法建模数据预处理法和网络参数优化法.在数据预处理方法中,根据建模样本的特点,分别采用滑动平均法和灰色模型法对原始建模数据进行修正,并分析了它们的适用场合;对于后一种方法,选择径向基函数分布宽度和学习目标进行优化.以精密平台为例进行了实验,通过对其定位误差的测量、建模和预测,证明了上述各种方法的有效性,特别是后一种方法,可以得到非常高的建模精度.     

前馈神经网络在有源噪声控制中的应用及分析

在有源噪声控制中,可以通过离线建模的方式得到次级通道传递函数S(z)的估计(S)(z),然而在实际环境中S(z)会发生改变,使得(S)(z)相对于S(z)出现大的偏差,可能导致系统不能正常工作,因此需要对(S)(z)进行在线调整.人工神经网络具有很强的自适应能力,可以实现对S(z)的估计和在线调整.     

应用小波神经网络处理CCD图像噪声

提出了一种用于数字图像中CCD噪声处理的小波神经网络滤波器.分析了CCD噪声模型,找出了导致CCD噪声模型复杂的原因:CCD相机响应功能的非线性.在对自适应噪声平滑(ANS)滤波器分析的基础上,考虑了影响滤波效果的两大问题:滤波窗口和图像强度.将小波神经网络非线性逼近CCD噪声曲线,按照噪声参数对图像进行区域划分并分配相应的权值.然后,结合相应的非线性滤波器进行针对性滤波,综合输出.实验结果表明:本文改进的滤波器滤波效果明显,信噪比得到进一步提高(24.65).利用小波神经网络良好的非线性函数逼近性,结合ANS滤波器,构造出了小波神经网络非线性ANS(WNN-NANS)滤波器,该滤波器在去除噪声的同时很好地保留了边缘细节,同时提高了信噪比.     

Calculation of PID controller parameters by using a fuzzy neural network

In this paper, we use the fuzzy neural network (FNN) to develop a formula for designing the proportional-integral-derivative (PID) controller. This PID controller satisfies the criteria of minimum integrated absolute error (IAE) and maximum of sensitivity (Ms). The FNN system is used to identify the relationship between plant model and controller parameters based on IAE and Ms. To derive the tuning rule, the dominant pole assignment method is applied to simplify our optimization processes. Therefore, the FNN system is used to automatically tune the PID controller for different system parameters so that neither theoretical methods nor numerical methods need be used. Moreover, the FNN-based formula can modify the controller to meet our specification when the system model changes. A simulation result for applying to the motor position control problem is given to demonstrate the effectiveness of our approach.