Application of a Taguchi-based neural network prediction design of the film coating process for polymer blends

In this paper, an approach for developing the prediction model for polymer blends using a back-propagation neural network (BPNN) combined with the Taguchi quality method is presented in an attempt to improve the deficiencies in current neural networks associated with the design of network architecture, including the selection of one optimal set of learning parameters to accomplish faster convergence during training and the desired accuracy during the recall step. The objective of the prediction model is to explore the relationships between the control factor levels and surface roughness in the film coating process. In addition, the feasibility of adopting this approach is demonstrated in the study optimizing the learning parameters of the BPNN structure to forecast the target characteristics of the product or process with various control conditions in the manufacturing system.     

Dynamics and cutting stability of the dynamically loaded worktable subjected to simply supported conditions

This paper aims to explore the dynamic characteristics and cutting stability of a surface grinder. In simulated grinding, the dynamically loaded worktable is described by the Euler–Bernoulli beam theory. In the model, the elastic worktable has both ends simply supported on a movable and massless rigid base. The analysis of the dynamics and stability of the worktable is complex due to the operating worktable being dynamically loaded in variable positions. With the Lagrange energy method combined with the assumed mode expansion method, the system dynamic equations are derived and a state space model for the dynamically loaded worktable subjected to simply supported conditions is developed. In this study, the maximum negative real part of the overall dynamic compliance and the limiting depth of cut are used as indicators to assess the structural static and dynamic performance of the worktable in various positions. The effects of worktable damping, contact stiffness, and damping between the tool and the workpiece on the system performance are studied. Based on the regenerative chatter and stability theory, the 3D stability lobes diagram is analyzed to optimize the maximum depth of cut at the highest available spindle speed. The cutting stability is verified by comparing the results obtained in the time domain analysis with the stability lobe diagram. The procedure illustrated in this study to improve the dynamics performance of a surface grinder can also be implemented in a similar fashion for many machine tool applications.     

基于广义回归神经网络的时间序列预测研究

介绍了广义回归神经网络的基本理论,提出了应用BIC准则确定输入神经元数目的方法．将其应用于大型旋转机械振动状态时间序列的单步和多步预测,与传统的采用误差反向传播学习算法的三层前馈感知器网络(BP神经网络)的预测结果进行对比。结果表明,该网络的预测性能优于后者,即使样本数据稀少,也能获得满意的预测结果。     

Development of a friction energy capacity approach to predict the surface coating endurance under complex oscillating sliding conditions

In the case of surface coatings application it is crucial to establish when the substrate is reached to prevent catastrophic consequences. In this study, a model based on local dissipated energy is developed and related to the friction process. Indeed, the friction dissipated energy is a unique parameter that takes into account the major loading variables which are the pressure, sliding distance and the friction coefficient. To illustrate the approach a sphere/plane (Alumina/TiC) contact is studied under gross slip fretting regime. Considering the contact area extension, the wear depth evolution can be predicted from the cumulated dissipated energy density. Nevertheless, some difference is observed between the predicted and detected surface coating endurance. This has been explained by a coating spalling phenomenon observed below a critical residual coating thickness. Introducing an effective wear coating parameter, the coating endurance is better quantified and finally an effective energy density threshold, associated to a friction energy capacity approach, is introduced to rationalize the coating endurance prediction. The surface treatment lifetime is then simply deduced from an energy ratio between this specific energy capacity and a mean energy density dissipated per fretting cycle. The stability of this approach has been validated under constant and variable sliding conditions and illustrated through an Energy Density–Coating Endurance chart.     

Application of hybrid neural network in discharge coefficient prediction of triangular labyrinth weir

Discharge coefficient (C_d) is an important parameter of triangular labyrinth weir. It is of great significance to predict the discharge coefficient accurately. In this research, in order to more accurately predict the C_d, in view of the traditional BP neural network is easy to fall into the local minimum in the training process, genetic algorithm (GA) and particle swarm optimization (PSO) are employed to optimize the traditional BP neural network's initial weights and thresholds. Nonlinear regression analysis (NLR) is also added to compare with these intelligent methods and four discharge coefficient prediction models are built, namely the NLR, the BPNN, the GA-BPNN and the PSO-BPNN. After the completion of the model construction, in order to objectively evaluate the performance of these models, the prediction results of these models are compared with the experiment results, and the determination coefficient (R²), the mean absolute error (MAE) and the root mean square error (RMSE) are introduced as the performance indicators to quantify the model performance. The results show that the accuracy and stability of the NLR are much worse than that of the intelligent models. The prediction results of the GA-BPNN and the PSO-BPNN are quite accurate with a higher decision coefficient than the BPNN. Moreover, the MAEs and the RMSEs of the GA-BPNN and the PSO-BPNN were significantly reduced by 25 and 40% compared with BPNN, respectively, and the maximum prediction errors were 4.4% and 2.6%, severally. Meanwhile, the width of error uncertainty band of GA-BPNN and PSO-BPNN is narrower than BPNN. By comparing GA-BPNN and PSO-BPNN with the discharge coefficient prediction models of triangular labyrinth weir in previous literatures, it is found that the mean absolute percentage error (MAPE) values of GA-BPNN and PSO-BPNN are 1.504% and 1.225% respectively, which are lower than other existing models. At the same time, the other performance indexes are better than most existing models, indicating that the genetic algorithm and PSO algorithm are more effective than the traditional BP algorithm in adjusting BP neural network parameters, easier to find the global optimal value, and improve the prediction accuracy and applicability of the model.     

Springback prediction of high-strength sheet metal under air bending forming and tool design based on GA�CBPNN

Based on orthogonal test for air bending of high-strength steel sheets, 125 values of sheet thickness (t), tool gap (c), punch radius (r), ratio of yield strength to Young??s modulus (?? _y /E), and punch displacement (e) are used to model the springback for air bending of high-strength sheet metal using the genetic algorithm (GA) and back propagation neural network (BPNN) approach, where the positive model and reverse model of springback prediction are established, respectively, with GA and BPNN. Adopting the ??object-positive model?Creverse model?? learning method, air bending springback law is studied with positive model and punch radius is predicted by reverse model. Manifested by the experiment for air bending forming of a workpiece used as crane boom, the prediction method proposed yields satisfactory effect in sheet metal air bending forming and punch design.     

基于机器学习的回流焊焊点形貌预测

提出一种基于机器学习预测回流焊焊点形貌的方法,通过该方法建立一个针对钽电容回流焊焊点形貌的预测模型,该模型为现有实验方式提供了新的思路。通过峰值温度、降温速率和焊膏厚度3种影响因素以及焊点厚度、焊点宽度和焊料爬高3种评价焊点形貌的评价标准,分别基于BPNN和LightGBM算法建立钽电容回流焊焊点形貌预测模型。对比实验证明,通过LightGBM算法建立的预测模型优于通过BPNN建立的预测模型,并通过实际测试帮助实验人员减少实验次数,节约大量时间成本。     

利用声发射信号时频特征在线监测慢走丝线切割加工表面粗糙度

针对慢走丝线切割加工(WEDM-LS)时因温度高、切缝窄等因素造成的加工过程难以监测的问题,利用声发射检测技术对占空比可调脉冲的慢走丝线切割加工过程进行在线监测。首先利用小波包能谱算法将AE信号分解到8个独立的频段上:分别为W1~W8,且频率依次降低;然后提取各频段上的能量特征,研究其与加工工件表面粗糙度值之间的相关性。试验结果表明:W8频段的能量与表面粗糙度值之间具有较高的相关性,该频段的能量与脉冲放电能量均随着脉冲信号占空比的增大而增大,且加工表面粗糙度值也随之逐渐增大。最后通过回归分析得到了反应材料表面粗糙度值与W8频段能量占比关系的数学预测模型,该模型的预测结果与实际测得的表面粗糙度值误差仅为3.51%。说明该模型具有较高的预测精度,可用于加工表面粗糙度的在线监测。     

基于RCSA的深孔内圆磨床主轴端点频响函数预测

基于响应耦合子结构分析法预测了深孔内圆磨床主轴端点的频响函数。首先对磨床主轴进行子结构划分,计算各子结构自由状态下的频响函数矩阵,然后顺序刚性耦合各子结构的频响函数矩阵,对轴承支撑点使用结构修改法修改轴承约束下的已耦合子结构频响函数矩阵,直至耦合到最后一个子结构,得到主轴端点的频响函数。以某深孔内圆磨床为研究对象,分别基于该方法和有限元法,对其主轴端点的频响函数进行预测,并对其进行实验测试。实验及分析结果表明：该方法预测精度高于有限元分析方法预测精度、计算速度快,便于深孔内圆磨床主轴系统的结构优化。     

高速磨床整机动态特性研究

建立了某高速凸轮轴磨床整机的三维有限元模型,利用反求方法确定了结合部的基础参数,对整机进行了模态分析,初步确定了整机动态性能的薄弱环节,研究了结合部刚度参数对整机低阶模态的影响,提出了结构改进方案。建立径向基函数近似模型表征床身—工作台系统结合部刚度与其固有频率之间的隐式函数关系,通过对该系统进行模态实验测试,并与优化方法相结合,确定了该高速磨床床身工作台系统的结合部刚度参数。结果表明,该方法对整机建模和动力学性能的分析简单有效,增加垫板在床身上的约束可以有效地改善该高速磨床的动态特性。     

基于灰关联神经网络的铣削表面粗糙度预测

不同的铣削加工工艺参数会影响加工表面形貌和表面粗糙度。考虑灰关联分析与神经网络法的各自优点,提出了一种新的基于灰关联神经网络模型进行表面粗糙度预测的模型。首先利用灰关联分析,将各因子与预测目标作关联性的排序,且把不必要的因子剔除,接着进行神经网络的训练及预测。将所提的预测模型运用到铣削加工的表面粗糙度预测中,构建出表面粗糙度预测系统,最后采用两样本T分配假设检验,以此验证该预测系统的有效性与可行性。     

线性液动压抛光流场的剪切特性研究

以动压润滑理论为基础,推导了线性液动压抛光的流场剪切力数学模型。借助FLUENT软件研究了流场底面的剪切力分布,通过单因素控制变量试验探究了各参数对剪切力的影响规律,灵敏度从高到低依次为：抛光间隙、抛光液黏度、抛光辊子转速、抛光辊子半径。以正交试验结果为训练集,建立了基于支持向量回归（SVR）的剪切特性预测模型,相关系数为98.35%、均方误差为3.44×10-3。最后计算了剪切力理论值。对比发现,数值模拟和理论计算误差在15%以内;不同参数组合下剪切力分布趋势相同;SVR预测模型可信度高。     

基于接触单元的磨床螺栓连接面有限元建模与模型修正

以接触单元和弹簧－－阻尼单元建立了磨床螺栓连接件的结合面动力学有限元模型。并以前3阶试验模态及有限元模型理论计算的固有频率均方差最小为目标函数,对包含接触单元和弹簧－－阻尼单元的结合面特征参数进行优化,在此基础上修正并取得更为符合实际的结合面动力学模型,这对机床结构的动态优化设计具有重要意义。     

磨齿机在机检测系统的测头综合预行程误差建模

因为测头预行程误差的存在,现有研究大都考虑单项或双项影响因素进行误差补偿。然而多次的实验统计表明,由于接触式测头的各向异性,导致信号传输迟滞、检测速度、测球半径、测杆长度、测头重力及测球表面测点法矢等因素都会对检测信号的触发时机产生影响,因而存在测头综合预行程误差,故而很难进行精确补偿。借助BP神经网络的高效逼近算法,利于求解输入为多项误差影响因素场合的测量误差输出问题,有效提高在机检测精度。根据自主研发卧式磨齿机L300G在机检测原理,以多项误差影响因素为输入节点,以测头综合预行程误差为输出节点,建立基于BP神经网络的测头综合预行程误差预测模型。完成误差补偿后,开展磨齿机标准样板齿轮在机检测实验。结果表明：误差补偿前后,齿向精度均为4级;误差补偿后,齿形精度提高2个等级,为4级精度,与格里森检测结果相吻合。结果验证了模型的正确性,有望在国产低成本磨齿机的高精度在机检测系统中推广使用。     

平面磨床床身结构分析与优化

针对某型平面磨床床身进行有限元分析,并对床身结构进行改进,然后在此基础上进行优化。优化后,结构固有频率及刚度变化不大,而床身质量却大大下降,取得明显效果。     

平面磨床床身结构分析与优化设计

平面磨床床身动静刚度直接影响机床的加工质量。通过建立某型平面磨床床身有限元分析模型并进行静态和模态分析,对床身结构进行改进,然后在此基础上进行结构优化,优化后床身质量下降15.6%,而结构的固有频率和刚度变化不大,取得明显的优化效果。     

Prediction of TC11 single-track geometry in laser metal deposition based on back propagation neural network and random forest

Laser metal deposition process usually involves the nonlinear interaction of multiple factors, such as process parameters and ambient temperature. In this study, random forest (RF) and multilayer back propagation neural network (BPNN) algorithms were employed to investigate the coupling relationship between process parameters and single-track geometry in laser metal deposition for TC11 alloy. With laser power, scanning speed, and powder feeding rate as inputs and track width and height as outputs, 30 different groups of experimental results were adopted as training groups. Their geometries were also predicted. The maximum relative errors of track width and height predictions based on BPNN model were 0.007 % and 0.029 %, respectively, which were lower than those based on RF model. Then, the two models were used to predict the geometry under four new sets of process parameters. Experimental results showed that the maximum error of BPNN model is lower than that of RF model. BPNN model also showed potential to improve cladding quality and efficiency.

     

基于混合算法的薄壁件铣削加工工艺参数优化

结合神经网络法和遗传算法的优点,提出了一种以倒传递神经网络法为基础的加工工艺参数优化方法,对薄壁件铣削加工工艺参数进行优化。将田口实验所得数据经倒传递神经网络进行训练与测试,来建立薄壁件铣削加工的信噪比预测器,并通过最大化信噪比,将铣削过程变异降至最低,进而找出最佳加工工艺参数组合。通过数值模拟与加工实验,验证了所提方法在薄壁件铣削加工工艺参数优化中的有效性。     

新型自准直导轨磨床床身-基础系统的热态特性分析

为消除环境温度变化对导轨磨床加工精度的影响，提出了一种新型导轨磨床(自准直导轨磨床)的床身-基础系统，并将自准直导轨磨床床身-基础系统的热态特性与普通导轨磨床床身-基础系统的热态特性进行了对比分析。理论分析和实验结果表明，自准直导轨磨床的床身-基础系统具有很好的精度保持性。     

基于机器学习的水液压高速开关阀退化趋势预测

高速开关阀以其结构简单、响应速度快、抗污染能力强、稳定性好等优点得到了广泛的应用。水液压高速开关阀的工作介质黏性低,更容易因性能退化发生故障。提出了一种基于机器学习的水液压高速开关阀性能退化状态识别及退化趋势预测方法。搭建了高速开关阀性能测试试验台,将电流信号的变化作为高速开关阀的性能退化指标。根据高速开关阀性能退化程度,将其退化状态定义为正常期、退化期和严重退化期3个阶段。采用BP神经网络(BPNN)方法对高速开关阀的退化状态进行了识别,并采用粒子群优化长短期记忆模型(PSO-LSTM)方法对高速开关阀的退化趋势进行了预测。使用高速开关阀的性能退化试验数据对提出模型的有效性进行了检验,结果表明该方法具有较高的预测精度。