Working condition recognition of sucker rod pumping system based on 4-segment time-frequency signature matrix and deep learning

High-precision and real-time diagnosis of sucker rod pumping system(SRPS)is important for quickly mastering oil well operations.Deep learning-based method for classifying the dynamometer card(DC)of oil wells is an efficient diagnosis method.However,the input of the DC as a two-dimensional image into the deep learning framework suffers from low feature utilization and high computational effort.Additionally,different SRPSs in an oil field have various system parameters,and the same SRPS generates different DCs at different moments.Thus,there is heterogeneity in field data,which can dramatically impair the diagnostic accuracy.To solve the above problems,a working condition recognition method based on 4-segment time-frequency signature matrix(4S-TFSM)and deep learning is presented in this paper.First,the 4-segment time-frequency signature(4S-TFS)method that can reduce the computing power requirements is proposed for feature extraction of DC data.Subsequently,the 4S-TFSM is constructed by relative normalization and matrix calculation to synthesize the features of multiple data and solve the problem of data heterogeneity.Finally,a convolutional neural network(CNN),one of the deep learning frameworks,is used to determine the functioning conditions based on the 4S-TFSM.Experiments on field data verify that the proposed diagnostic method based on 4S-TFSM and CNN(4S-TFSM-CNN)can significantly improve the accuracy of working condition recognition with lower computational cost.To the best of our knowledge,this is the first work to discuss the effect of data heterogeneity on the working condition recognition performance of SRPS.

The W transform and its improved methods for time-frequency analysis of seismic data

The conventional linear time-frequency analysis method cannot achieve high resolution and energy focusing in the time and frequency dimensions at the same time,especially in the low frequency region.In order to improve the resolution of the linear time-frequency analysis method in the low-frequency region,we have proposed a W transform method,in which the instantaneous frequency is introduced as a parameter into the linear transformation,and the analysis time window is constructed which matches the instantaneous frequency of the seismic data.In this paper,the W transform method is compared with the Wigner-Ville distribution(WVD),a typical nonlinear time-frequency analysis method.The WVD method that shows the energy distribution in the time-frequency domain clearly indicates the gravitational center of time and the gravitational center of frequency of a wavelet,while the time-frequency spectrum of the W transform also has a clear gravitational center of energy focusing,because the instantaneous frequency corresponding to any time position is introduced as the transformation parameter.Therefore,the W transform can be benchmarked directly by the WVD method.We summarize the development of the W transform and three improved methods in recent years,and elaborate on the evolution of the standard W transform,the chirp-modulated W transform,the fractional-order W transform,and the linear canonical W transform.Through three application examples of W transform in fluvial sand body identification and reservoir prediction,it is verified that W transform can improve the resolution and energy focusing of time-frequency spectra.

Rapid identification of time-frequency domain gravitational wave signals from binary black holes using deep learning

Recent developments in deep learning techniques have provided alternative and complementary approaches to the traditional matched-filtering methods for identifying gravitational wave(GW)signals.The rapid and accurate identification of GW signals is crucial to the advancement of GW physics and multi-messenger astronomy,particularly considering the upcoming fourth and fifth observing runs of LIGO-Virgo-KAGRA.In this study,we used the 2D U-Net algorithm to identify time-frequency domain GW signals from stellar-mass binary black hole(BBH)mergers.We simulated BBH mergers with component masses ranging from 7 to 50 M_(⊙)and accounted for the LIGO detector noise.We found that the GW events in the first and second observation runs could all be clearly and rapidly identified.For the third observing run,approximately 80% of the GW events could be identified.In contrast to traditional convolutional neural networks,the U-Net algorithm can output time-frequency domain signal images corresponding to probabilities,providing a more intuitive analysis.In conclusion,the U-Net algorithm can rapidly identify the time-frequency domain GW signals from BBH mergers.

High-resolution seismic inversion method based on joint data-driven in the time-frequency domain

Seismic inversion can be divided into time-domain inversion and frequency-domain inversion based on different transform domains.Time-domain inversion has stronger stability and noise resistance compared to frequencydomain inversion.Frequency domain inversion has stronger ability to identify small-scale bodies and higher inversion resolution.Therefore,the research on the joint inversion method in the time-frequency domain is of great significance for improving the inversion resolution,stability,and noise resistance.The introduction of prior information constraints can effectively reduce ambiguity in the inversion process.However,the existing modeldriven time-frequency joint inversion assumes a specific prior distribution of the reservoir.These methods do not consider the original features of the data and are difficult to describe the relationship between time-domain features and frequency-domain features.Therefore,this paper proposes a high-resolution seismic inversion method based on joint data-driven in the time-frequency domain.The method is based on the impedance and reflectivity samples from logging,using joint dictionary learning to obtain adaptive feature information of the reservoir,and using sparse coefficients to capture the intrinsic relationship between impedance and reflectivity.The optimization result of the inversion is achieved through the regularization term of the joint dictionary sparse representation.We have finally achieved an inversion method that combines constraints on time-domain features and frequency features.By testing the model data and field data,the method has higher resolution in the inversion results and good noise resistance.

Application of sparse time-frequency decomposition to seismic data

The Gabor and S transforms are frequently used in time-frequency decomposition methods. Constrained by the uncertainty principle, both transforms produce low-resolution time-frequency decomposition results in the time and frequency domains. To improve the resolution of the time-frequency decomposition results, we use the instantaneous frequency distribution function（IFDF） to express the seismic signal. When the instantaneous frequencies of the nonstationary signal satisfy the requirements of the uncertainty principle, the support of IFDF is just the support of the amplitude ridges in the signal obtained using the short-time Fourier transform. Based on this feature, we propose a new iteration algorithm to achieve the sparse time-frequency decomposition of the signal. The iteration algorithm uses the support of the amplitude ridges of the residual signal obtained with the short-time Fourier transform to update the time-frequency components of the signal. The summation of the updated time-frequency components in each iteration is the result of the sparse timefrequency decomposition. Numerical examples show that the proposed method improves the resolution of the time-frequency decomposition results and the accuracy of the analysis of the nonstationary signal. We also use the proposed method to attenuate the ground roll of field seismic data with good results.

TVAR Time-frequency Analysis for Non-stationary Vibration Signals of Spacecraft

Predicting the time-varying auto-spectral density of a spacecraft in high-altitude orbits requires an accurate model for the non-stationary random vibration signals with densely spaced modal frequency. The traditional time-varying algorithm limits prediction accuracy, thus affecting a number of operational decisions. To solve this problem, a time-varying auto regressive （TVAR） model based on the process neural network （PNN） and the empirical mode decomposition （EMD） is proposed. The time-varying system is tracked on-line by establishing a time-varying parameter model, and then the relevant parameter spectrum is obtained. Firstly, the EMD method is utilized to decompose the signal into several intrinsic mode functions （IMFs）. Then for each IMF, the PNN is established and the time-varying auto-spectral density is obtained. Finally, the time-frequency distribution of the signals can be reconstructed by linear superposition. The simulation and the analytical results from an example demonstrate that this approach possesses simplicity, effectiveness, and feasibility, as well as higher frequency resolution.

Intelligibility evaluation of enhanced whisper in joint time-frequency domain

Some factors influencing the intelligibility of the enhanced whisper in the joint time-frequency domain are evaluated. Specifically, both the spectrum density and different regions of the enhanced spectrum are analyzed. Experimental results show that for a spectrum of some density, the joint time-frequency gain-modification based speech enhancement algorithm achieves significant improvement in intelligibility. Additionally, the spectrum region where the estimated spectrum is smaller than the clean spectrum, is the most important region contributing to intelligibility improvement for the enhanced whisper. The spectrum region where the estimated spectrum is larger than twice the size of the clean spectrum is detrimental to speech intelligibility perception within the whisper context.

Effects of Gabor transform parameters on signa time-frequency resolution

In this paper, it is described that the time-frequency resolution of geophysical signals is affected by the time window function attenuation coefficient and sampling interval and how such effects are eliminated effectively. Improving the signal resolution is the key to signal time-frequency analysis processing and has wide use in geophysical data processing and extraction of attribute parameters. In this paper, authors research the effects of the attenuation coefficient choice of the Gabor transform window function and sampling interval on signal resolution. Unsuitable parameters not only decrease the signal resolution on the frequency spectrum but also miss the signals. It is essential to first give the optimum window and range of parameters through time-frequency analysis simulation using the Gabor transform. In the paper, the suggestions about the range and choice of the optimum sampling interval and processing methods of general seismic signals are given.

改进Mask RCNN的盾构隧道渗漏水检测方法

渗漏水是盾构隧道结构存在潜在损伤或缺陷的重要表征,快速、准确检测出渗漏水位置,对隧道安全运营和维护具有重要意义。现有的方法大多采用光学影像对隧道渗漏水进行检测,受隧道内空间和光线条件限制,难以获得高质量病害图片。因此,本文提出了一种基于激光点云数据与改进Mask RCNN相结合的渗漏水检测方法。首先对激光点云反射强度进行修正;然后生成灰度图像并建立渗漏水病害数据集;最后在Mask RCNN算法中引入空洞卷积和变形卷积,实现了隧道渗漏水病害的快速检测。利用某地铁采集的数据进行验证,结果表明,本文提出的改进Mask RCNN算法相较于原始算法和FCN算法检测精度均有明显提升,在盾构隧道渗漏水识别方面性能表现较好。

基于空间注意力机制的Mask R-CNN致密储层岩石薄片图像鉴定

针对陆相致密储层岩石薄片鉴定识别难、制片成本高、时间消耗长和人为主观强等难题,选取鄂尔多斯盆地临兴区块上古生界和松辽盆地三肇凹陷扶余油层为靶区,提出一种基于深度学习的致密油储层岩石薄片人工智能鉴定方法,引入图像预处理技术去除岩石薄片图像噪声并统一图像像素大小,构建空间几何增广机制,基于空间注意力机制改进Mask R-CNN算法,并将上述方法应用于实例靶区进行有效性验证。结果表明:图像预处理技术能够在保障图像特征的前提下,有效提高图像质量,减少噪声干扰;空间几何图像增广机制能够在在一定程度上增加可用样本的数量;基于空间注意力机制的Mask R-CNN算法可以同时完成复杂岩石薄片成分的分割与智能识别工作,分割精度在不同数据集情况下的平均精度为89.2%,整体识别准确率为93%,适用于致密油储层岩石薄片特征鉴定。

Mask在包装材料气体阻隔性能检测中的应用研究

目的研究和评价mask在材料气体阻隔性检测中的应用及其检测数据的重复性、准确性和数据稳定性。方法选用覆盖高阻隔、中阻隔、低阻隔等阻隔性能范围的5种样品,使用3种不同面积的mask和仪器测试腔原有面积对样品进行测试,并对测试结果重复性、稳定性和准确性进行分析评价。结果高阻隔材料PET硬片使用面积12.56 cm^(2)的mask测试时,可以得到较为稳定的检测结果,而在使用更小面积(1.77、5 cm^(2))的mask时,测试结果的相对标准偏差、相对极差和测试数据偏差都较差,不推荐使用。KOP/CPP在使用1.77 cm^(2)的mask测试时,测试结果相对标准偏差和测试数据偏差都略大于10%。PET/CPP在使用1.77 cm^(2)的mask测试时,其测试数据偏差略大于10%。BOPE/LDPE和TPU使用1.77 cm^(2)的mask测试可以得到良好的检测结果。结论Mask是解决试样材料特性、设备量程限制、试样尺寸等测试困难的优秀解决方案。对于中、低阻隔材料的透气性测试,使用mask可获得具有良好可信度和稳定性的测试数据。而在进行氧气透过率的测试时应尽量选择大的测试面积。小面积mask不适用于高阻隔材料的气体阻隔性测试。

改进Mask R-CNN的馆藏报纸图像内容分割

开展馆藏报纸图像内容分割研究,能提升文字识别准确率,对促进机器识别取代人工操作、提高图书馆数字化工作效率具有重要意义。文章根据报纸图像呈现的特征,提出一种基于改进MaskR-CNN的算法,实现报纸图像内容分割。首先,通过优化锚框比例和损失函数,对原始MaskR-CNN算法进行改进。其次,采用数据增强、调整训练参数开展样本训练。最后,通过实验的方式对改进后的MaskR-CNN算法训练模型和原始算法训练模型进行比较,并采用AP_bbox和AP_segm评价指标对实验结果进行评估,改进后的算法训练模型AP_bbox为0.935,AP_segm为0.943,均超过原始算法训练模型。实验结果表明,改进后的MaskR-CNN算法能够实现报纸图像内容有效检测与分割。

基于FPGA加速的Mask R-CNN稻瘟病高通量自适应识别模型研究

针对基于图像的稻瘟病现场检测技术依赖先验知识且受制于算力与田间网络状况,无法实现自适应实时检测的问题,提出一种可利用现场可编程门阵列(Field programmable gate array,FPGA)加速的Mask R-CNN(Mask region-based convolutional neural network)稻瘟病高通量自适应快速识别模型。首先将骨干网络改进为MobileNetV2,利用其倒残差模块降低计算量,提高模型并行处理能力;随后增加用于稻瘟病多尺度特征融合的特征金字塔网络模块,使模型具备多尺度自适应处理能力;最后由全卷积网络(Fully convolutional network,FCN)分支输出稻瘟病病斑的实例分割,同时使用交叉熵损失函数完成稻瘟病的定位与分类。稻瘟病实测数据集对模型的验证结果表明:当输入为全高清图像时,模型平均推理时间减少至85 ms,相较GPU服务器、同级别GPU边缘计算平台,速度分别提高86.2%、63.0%。在交并比为0.6时,准确率可达98.0%,病斑捕获能力平均提升21.2%。提出的Mask R-CNN自适应快速识别模型能够在田间恶劣网络状况下实现稻瘟病的快速现场检测,具有更好的抗噪能力和鲁棒性能,为水稻病害实时检测、察打一体提供了高效实时的片上系统方案。

基于改进Mask R-CNN的青菜杂质检测研究

绿叶蔬菜的智能包装加工是实现绿叶蔬菜智能化生产、降低生产成本的重要部分,对绿叶蔬菜在包装加工时的杂质检测是其重要前提。以青菜为研究对象,提出一种基于Mask R-CNN的青菜杂质检测模型。首先采集标注掺杂枯树叶、枯菜叶和碎纸片3种常见杂质的青菜图像1370多张,并通过数据增强的方法扩充建立含有2740张青菜杂质图像的数据集。为减少背景对杂质检测的影响,通过在Mask R-CNN模型中加入协调注意力机制,同时添加全连接层和Dropout层,增强模型特征提取能力,减少过拟合现象,并使用迁移学习方法对模型进行微调。结果表明改进后的Mask R-CNN算法对青菜杂质识别的平均精度均值为99.19%,检测速度为8.45 FPS,检测效果良好,可以满足青菜杂质的检测需求。

基于Mask R-CNN的油田井场指针仪表识别方法研究

针对无人机巡检流程中采集到井场仪表图像模糊以及油田仪表定位模型实时性较差的问题,提出一种改进后最大后验概率模型去模糊方法和基于Mask R-CNN的指针式仪表定位算法。首先,通过采用变步长LMS滤波器的方法优化图像的先验信息,根据输入数据的统计特性调整滤波器参数,生成初步的仪表图像恢复结果,从而提升了最大后验概率的去模糊效果;其次,在Mask R-CNN网络结构的基础上,选用MobileNetV3作为主干特征提取网络减少参数量,再加入注意力机制模块保证准确率以完成仪表定位。最后,实验证明,仪表图像评价指标高于其他算法,该文提出的仪表定位算法减少了48.25 M参数量,FPS值达到37.3 frame/s,准确率为94.02%。

基于Protel的Solder Masks与Paste Masks辨析

在不同的Protel 99 SE教材中,对于Solder Masks和Paste Masks的理解存在两种截然相反的意见,而且大多数教材对于Masks层没有给出详尽的解释,这些弊端都给初学者带来较大的困惑。通过对印制电路板掩膜层制造工艺的详细描述、单层显示演示以及Protel 99 SE帮助文件的诠注全面而充分地论证了Solder Masks和Paste Masks的实质含意。认为两者的实质是"阻焊膜与防锡膏膜",没有"互补关系"。

基于Mask R-CNN的电力关键设备运行状态检测

为判断电力关键设备两端负载电压与干路总电压之间的数值关系,提出基于Mask R-CNN的电力关键设备运行状态检测方法。在Mask R-CNN网络结构中,判断电量信号的时域范围与频域范围。然后根据负荷阻抗特征计算交流参数的取值结果,再联合已获取的电信号参量,求解连续相关函数,从而检测电力关键设备运行状态。实验过程中,设备两端负载电压、内阻消耗电压之和与干路总电压之间的差值未超过1.5 V,说明该方法能够证明电力关键设备两端负载电压、内阻消耗电压之和等于干路总电压数值的猜想成立,可以根据该验证检测电力关键设备运行状态是否正常。

基于改进Mask R-CNN的笼养死鸭识别方法

针对规模化笼养肉鸭舍内死鸭识别采用人工作业方式时,存在作业效率低、劳动强度大、养殖成本高等问题,以层叠式笼养肉鸭为研究对象,提出了一种基于深度学习的笼养死鸭识别方法。为了采集数据,首先面向立体层叠式养殖环境设计了一款适用于肉鸭舍的自主巡检装备。针对笼养肉鸭舍铁丝网遮挡严重的问题,基于机器视觉对笼网进行修复,基于OpenCV对图像进行增强处理。构建了一种基于Mask R-CNN的死鸭识别模型,采用Swin Transformer对模型进行优化,解决了Mask R-CNN网络缺乏整合全局信息能力的问题。对比分析了SOLO v2、Mask R-CNN和Mask R-CNN+Swin Transformer模型识别笼内死鸭准确率。实验结果表明,在平均精度均值为90%的条件下,Mask R-CNN+Swin Transformer模型对笼内死鸭总体识别准确率可达95.8%,在自主巡检装备上的检测效果优于其他主流的目标检测算法。

基于Mask R-CNN的地质雷达岩溶预报图像识别研究

岩溶隧道开挖可能遭遇岩溶涌水、突泥等岩溶地质灾害,地质雷达能够有效预报岩溶等地质灾害。然而,传统地质雷达图像解译存在专家经验依赖性强、解译效率慢且易误判漏判等情况。本文采用可实现端到端识别的深度学习技术开展地质雷达图像目标检测与识别的研究,将基于Mask R-CNN的卷积神经网络算法应用于地质雷达岩溶预报图像异常的智能识别。在TensorFlow和Keras框架下,利用地质雷达设备采集获得的数据构建训练数据集和测试数据集,对Mask R-CNN深度学习模型进行训练,最终得到权重参数较好的地质雷达岩溶预报图像的双曲异常检测模型。试验结果及应用案例表明,Mask R-CNN目标检测方法在地质雷达岩溶预报图像的目标检测与识别上取得了良好的效果,有效提高了地质雷达图像的智能化识别效率。

Time-Frequency Signal Processing for Gas-Liquid Two Phase Flow Through a Horizontal Venturi Based on Adaptive Optimal-Kernel Theory

A time-frequency signal processing method for two-phase flow through a horizontal Venturi based on adaptive optimal-kernel （AOK） was presented in this paper.First,the collected dynamic differential pressure signal of gas-liquid two-phase flow was preprocessed,and then the AOK theory was used to analyze the dynamic differ-ential pressure signal.The mechanism of two-phase flow was discussed through the time-frequency spectrum.On the condition of steady water flow rate,with the increasing of gas flow rate,the flow pattern changes from bubbly flow to slug flow,then to plug flow,meanwhile,the energy distribution of signal fluctuations show significant change that energy transfer from 15-35 Hz band to 0-8 Hz band;moreover,when the flow pattern is slug flow,there are two wave peaks showed in the time-frequency spectrum.Finally,a number of characteristic variables were defined by using the time-frequency spectrum and the ridge of AOK.When the characteristic variables were visu-ally analyzed,the relationship between different combination of characteristic variables and flow patterns would be gotten.The results show that,this method can explain the law of flow in different flow patterns.And characteristic variables,defined by this method,can get a clear description of the flow information.This method provides a new way for the flow pattern identification,and the percentage of correct prediction is up to 91.11%.