Prediction of drill failure using features extraction in time and frequency domains of feed motor current

In this paper, a drill prefailure prediction method based on the feed motor current is proposed. The characteristic parameters of drill failure (CPDF) are defined in the time and frequency domains to express the features of the feed motor current at drill failure. In the time domain, the CPDFs represent the increase of average value and the standard deviation of the feed motor current at drill failure. In the frequency domain, the CPDFs represent the magnitude of vibration at the spindle rotational frequency and at two times this frequency of the feed motor current. The CPDFs are used as inputs to the neural network. The output of the neural network is defined as the drill state index (DSI). Drill failure is predicted by monitoring the number of times the DSI exceeds the threshold value of DSI. Experiments showed that the proposed algorithm could accurately identify impending failure before drill breakage regardless of cutting conditions and machine tool types.     

基于改进LSTM-AdaBoost的铣刀磨损量预测

针对铣刀磨损量预测时精度低的问题,提出一种基于黑寡妇算法（BWO）优化的长短期记忆神经网络（LSTM）与AdaBoost集成学习算法相结合的铣刀磨损量预测方法。在铣刀磨损振动信号中提取时域、频域以及时频域多域特征。通过BWO算法优化LSTM的核心参数,并将优化后的LSTM网络与AdaBoost算法进行结合,构建铣刀磨损量预测模型。最后用PHM Society 2010铣刀全寿命周期的振动数据进行实验。研究结果表明：所提方法能够有效地预测出铣刀磨损量变化值,优化后模型的平均绝对误差百分比为3.436%、均方根误差为6.471、决定系数 R2 为0.935。该方法能够获得准确率更高的铣刀磨损量预测值,预测效率更高。     

Tool condition monitoring in drilling using vibration signature analysis

This paper presents a study on monitoring tool wear and failure in drilling using vibration signature analysis techniques. Discriminant features, which are sensitive to drill wear and breakage, were developed in both time and frequency domains. These features were found to be relatively insensitive to cutting conditions, and sensor location. In the time domain, a monitoring feature based on calculating the kurtosis value of both the transverse and thrust vibrations, was found to be rather effective for on-line detection of drill breakage. On the other hand, in the frequency domain, a cepstrum ratio, derived from the spectra of the vibrations monitored in both directions, was also found effective in detecting breakage events. The effect of different types of wear on the vibration power spectra, in both the transverse and the thrust directions, was also investigated. A signature feature, namely the instantaneous ratio of the absolute mean value (RAMV_i), was developed in this study and used as a threshold for controlled capture of the vibration signal. The ability of the monitoring features to detect drill wear and breakage was verified experimentally. The drilling tests were performed using 3 and 6 mm diameter high speed steel twist drills, and cast iron workpieces. The results confirmed the effectiveness and robustness of the proposed monitoring features.     

基于小波包分解和奇异谱分析方法的立铣加工颤振监测

在立铣加工过程中,颤振是加工过程失稳的一个最重要的原因。颤振将会严重影响工件表面质量和材料去除率,加剧刀具磨损和恶化工作环境。虽然大部分颤振监测系统可以监测到颤振发生,但颤振发生时已经对工件和刀具产生了严重的损伤,因此,需要提前监测到颤振特征。在颤振发生过程中,振动信号具有在时域中不断增大,在频域中能量频移的特性。考虑这两个振动信号特征,提出了一种颤振特征提取方法。提取颤振发生频带中振动信号的能量比和奇异谱熵系数作为两个颤振特征,并通过人工神经网络模型实现切削颤振的识别。文中提出的颤振监测系统包括特征提取和分类,能够精确辨识立铣加工中的稳定、过渡和颤振状态。     

基于主成分分析和BP神经网络的微铣刀磨损在线监测

为提高微铣刀磨损在线监测系统的预测精度,尝试通过主成分分析法对微铣削振动信号的时域和频域特征进行降维,将降维后的特征输入改进型BP神经网络模型,实现微铣刀磨损特征分类。结果表明,提出的微铣刀在线监测方法能够准确识别微铣刀的各种磨损状态,此外,和其它分类算法相比,提出的基于遗传算法的BP神经网络模型在分类精度和计算效率方面具有综合优势,对微铣刀磨损的其它在线监测方法具有一定的指导意义和借鉴价值。     

Drill wear monitoring using neural networks

The primary objective of this research is to monitor drill wear on-line. In this paper, drill wear monitoring is carried out by measuring the thrust force and torque signals. In order to identify the tool wear conditions based on the signal measured, a neural network, using a cumulative back-propagation algorithm, is adopted. This paper also describes the experimental procedure used and presents the results obtained for establishing the neural network. The inputs to the neural network are the mean values of thrust force and torque, spindle rotational speed, feedrate and drill diameter. The neural network is trained to estimate the average drill wear. It is confirmed experimentally that the tool wear can be accurately estimated by the trained neural network. The accuracy of tool wear estimation using the neural network is superior to that using other regression models.     

基于多传感数据融合的变速运行齿轮异常振动故障诊断

变速运行齿轮异常振动故障诊断性能过差会增加汽车维护成本,缩短齿轮使用寿命。为了及时识别齿轮故障,保证汽车变速器总成具有良好的振动特性,提出基于多传感数据融合的变速运行齿轮异常振动故障诊断方法。通过分析多传感器数据融合技术,掌握变速运行齿轮异常振动故障诊断的理论框架,并以此为基础,参考传感器融合模块、特征级并行多神经网络局部诊断模块和终端分类模块,结合变分模态分解、多通道加权融合和单隐层前馈神经网络训练算法,从信号采集、信号特征提取和信号特征分类3个步骤实现变速运行齿轮异常振动故障诊断。实验结果表明：在齿轮发生轻度磨损时,磨损振动信号的幅值在20~40 mV之间,磨损振动信号的频率在0~4 000 Hz区间;中度磨损时,信号的幅值在30~55 mV之间,信号频率在3 000~7 000 Hz区间;重度磨损时,信号幅值在50~70 mV之间,信号频率在6 000~12 000 Hz区间,且各阶段诊断结果均与故障程度的实际转折点吻合。由此可知在各样本数量均相同的情况下,提出的故障诊断方法预测值与真实值均相同,故障程度和故障类型的诊断性能均较好。     

基于HPSO优化BP神经网络的刀具磨损状态识别

针对BP神经网络容易陷入局部极值导致识别精度低的问题,文章提出了一种基于混合粒子群算法(HPSO)的BP神经网络优化算法。在刀具磨损监测实验过程中,采集刀具切削的声发射(AE)信号,利用小波包分解算法对AE信号进行滤波,并进行特征提取。将频带能量特征和切削参数分别作为主特征和辅助特征,并对其对归一化处理。采用混合粒子群优化算法(HPSO)对BP神经网络预测模型进行优化,利用优化后的模型对测试样本进行模式识别,结果表明,优化后的HPSO-BP模型能够有效地降低神经网络陷入局部极值的情况,提高刀具磨损识别精度。     

基于改进长短期记忆网络的铣刀磨损量预测研究

针对铣刀磨损量预测精度低的问题，提出一种高精度铣刀磨损量预测方法。该方法通过遗传算法（GA）寻出长短期记忆网络（LSTM）的最优参数，并将参数输入LSTM实现改进模型GA-LSTM。采用时域、频域及时频域方法提取特征，应用皮尔逊相关系数法筛选出与铣刀磨损量高度相似的特征向量，输入GA-LSTM模型进行训练，并对测试数据进行预测。实验结果表明:与传统的机器学习方法BPNN或深度学习方法FE-LSTM、CNN相比，GA-LSTM的均方根误差分别下降了41.3%、39.0%、51.5%，平均相对误差分别下降了48.3%、40.8%、56.7%，模型的预测识别精度有较大提高，实现了铣刀磨损量的有效预测。     

基于主成分分析的多域特征融合轴承故障诊断

针对复杂工况下难以区分轴承故障状态的问题,提出一种基于主成分分析的多域特征融合轴承故障诊断方法。采集轴承振动加速度信号,提取轴承时域新量纲一化特征、频域幅值谱特征和时频域经验模态分解特征共13维特征用于完整表征轴承状态;利用主成分分析方法对所提取特征融合与降维,降低诊断模型复杂度与数据分析难度;最后,选择合适的卷积神经网络进行分类,通过石化机组故障诊断实验平台进行验证。结果表明：多域融合特征相对于单域特征诊断效果更好,卷积神经网络分类模型相对于其他经典分类模型诊断准确率更高,融合诊断分类方法整体诊断准确率达到86%。     

Modeling of tool wear in drilling by statistical analysis and artificial neural network

The useful life of a cutting tool and its operating conditions largely control the economics of the machining operations. Hence, it is imperative that the condition of the cutting tool, particularly some indication as to when it requires changing, to be monitored. The drilling operation is frequently used as a preliminary step for many operations like boring, reaming and tapping, however, the operation itself is complex and demanding.

Back propagation neural networks were used for detection of drill wear. The neural network consisted of three layers input, hidden and output. Drill size, feed, spindle speed, torque, machining time and thrust force are given as inputs to the ANN and the flank wear was estimated. Drilling experiments with 8 mm drill size were performed by changing the cutting speed and feed at two different levels. The number of neurons in the hidden layer were selected from 1, 2, 3, …, 20. The learning rate was selected as 0.01 and no smoothing factor was used. The estimated values of tool wear were obtained by statistical analysis and by various neural network structures. Comparative analysis has been done between statistical analysis, neural network structures and the actual values of tool wear obtained by experimentation.     

A back-propagation algorithm applied to tool wear monitoring

In this paper, a distributed neural network has been applied to a pattern recognition problem for classification of tool wear in a turning operation to discriminate between a worn-out tool and a fresh tool. A multilayered perceptron with back-propagation algorithm has been used. The network was trained offline using 30 patterns each of six inputs. Using the weights obtained during training, fresh patterns were tested. Results for six fresh patterns are presented.     

Using neural network for tool condition monitoring based on wavelet decomposition

This paper presents a neural network application for on-line tool condition monitoring in a turning operation. A wavelet technique was used to decompose dynamic cutting force signal into different frequency bands in time domain. Two features were extracted from the decomposed signal for each frequency band. The two extracted features were mean values and variances of the local maxima of the absolute value of the composed signal. In addition, coherence coefficient in low frequency band was also selected as a signal feature. After scaling, these features were fed to a back-propagation neural network for the diagnostic purposes. The effect on tool condition monitoring due to the presence of chip breaking was studied. The different numbers of training samples were used to train the neural network and the results were discussed. The experimental results show that the features extracted by wavelet technique had a low sensitivity to changes of the cutting conditions and the neural network has high diagnosis success rate in a wide range of cutting conditions.     

Frequency and time domain analyses of sensor signals in drilling—II. Investigation on some problems associated with sensor integration

Sensor integration has received considerable attention recently for monitoring machining processes. This is because it is similar to the action of an experienced machinist, who uses his different sensory devices such as hearing, sight, etc. to monitor the cutting operation. Different neural network paradigms have been attempted by researchers for this purpose. In this investigation, a multisensor approach to drill wear monitoring was studied. Four sensors, namely, thrust, torque, and strains on the machine table in two orthogonal directions perpendicular to the drill axis, were used. As shown in Part I [A. Noori-Khajavi and R. Komanduri, Int. J. Mach. Tools Manufact. 35, 000-000 (1995)] three sensor signals, namely, thrust, torque, and strain on the machine table in the X-direction, showed good correlation in the frequency domain with drill wear. In addition, the signal-to-noise ratio analysis at different states of drill wear in the frequency domain showed that as the drill wear increased, the noise also increased. In this paper, it will be shown that when sensor signals are noisy and are integrated using a neural network, such a system could actually result in the deterioration of the correct estimation of drill wear. Consequently, there appears to be no need for the integration of the sensor signals under the conditions used.     

基于免疫神经网络的数控机床故障诊断研究

针对BP神经网络在数控机床故障预测中出现的收敛速度慢和训练容易陷入局部极值问题,提出一种基于人工免疫算法优化BP神经网络(IMBP)的数据机床故障诊断算法。介绍了常见的数控机床故障类型和分类,阐述了人工免疫算法和BP神经网络以及人工免疫优化BP神经网络算法的工作流程。利用免疫算法的全局搜索性能先对神经网络权值和阈值进行全局优化,加快了BP算法训练过程的收敛速度,减少训练过程所需要的时间。通过仿真性能测试分析,结果表明:与BP、GABP和IMBP 3种算法对比,比BP神经网络算法的数控机床故障预测诊断提高了18.3%,比GABP神经网络算法提高了12.05%,提高了数控机床故障诊断精度。     

基于C-GRU的电机轴承故障诊断

针对单一卷积神经网络模型在轴承故障诊断工作对于训练样本需求过多的不足,根据采集到的电机轴承振动数据为时序数据的特点,结合门控循环单元在处理时序数据所具有的优势,采用了基于卷积神经网络和门控循环单元(C-GRU)的电机轴承故障诊断算法.将CNN在特征提取的优点与GRU处理时序数据的优点有机结合起来,在选择合适的网络结构和...     

基于深度卷积神经网络的轴承故障诊断技术研究

为了提高轴承故障诊断的准确度,采用深度卷积神经网络算法来实现轴承故障分类。首先根据轴承振动故障特征频率建立轴承故障数据库,接着对轴承的振动信号按不同切片长度和固定宽度进行周期提取,建立特征向量矩阵,然后建立深度卷积神经网络的故障诊断模型,在网络设计时,差异化设置卷积核与池化尺寸,优化神经网络训练的核心参数,最后获得稳定的卷积神经网络模型。经过实例仿真,基于深度卷积神经网络的轴承故障分类准确率高,标准差小。     

Machine ensemble approach for simultaneous detection of transient and gradual abnormalities in end milling using multisensor fusion

In a fully automated manufacturing environment, instant detection of the cutting tool condition is essential for the improved productivity and cost effectiveness. This paper studies a tool condition monitoring system (TCM) via machine learning (ML) and machine ensemble (ME) approach to investigate the effectiveness of multisensor fusion technique when machining 4340 steel with multilayer coated and multiflute carbide end mill cutter. In this study, 135 different features are extracted from multiple sensor signals of force, vibration, acoustic emission and spindle power in the time and frequency domain by using data acquisition and signal processing module. Then, a correlation-based feature selection technique (CFS) evaluates the significance of these features along with machining parameters collected from machining experiments. Next, an optimal feature subset is computed for various assorted combinations of sensors. Finally, machine ensemble methods based on majority voting and stacked generalization are studied for the selected features to classify not only flank wear but also breakage and chipping. It has been found in this paper that the stacked generalization ensemble can ensure the highest accuracy in tool condition monitoring. In addition, it has been shown that the support vector machine (SVM) outperforms other ML algorithms in most cases tested.     

基于BP神经网络和D-S证据理论的刀具磨损监测方法

将BP神经网络和D-S证据理论相结合的方法运用于刀具磨损监测中,采用小波包分解法对刀具磨损过程中产生的声发射信号进行特征提取,构建特征向量,利用BP神经网络识别判断刀具磨损状态;通过BP神经网络的输出结果和训练误差计算D-S证据理论的基本概率赋值,并用D-S证据理论对BP神经网络的识别结果进行决策级融合。实验结果表明:该方法避免了神经网络识别时的误诊,提高了整个刀具磨损监测系统识别的准确性和可靠性。     

基于深度信念网络和多信息融合的复杂机电装备故障诊断方法

针对复杂机电装备故障诊断中存在的数据量大、提取故障特征困难等问题,结合深度学习理论强大的感知与自我学习能力,提出一种基于深度信念网络和多信息融合的复杂机电装备故障诊断方法。将多个传感器的原始时域信号数据输入深度信念网络进行训练,通过反向微调学习对深度信念网络进行整体微调,提高分类准确性,同时在训练过程采用ReLu激活函数和加入Batch Normalization,减少过拟合出现概率的同时提高了网络收敛的速度。将此方法运用到复杂数控加工中心刀具的故障诊断中,结果表明该方法相比传统BPNN算法和采用Sigmoid激活函数的深度神经网络算法准确率更高。