• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    Method and system to automatically detect faults in a rotary axis. The present invention relates to a method and system for automatically detecting failures in a rotary axis. The invention comprises the steps of: acquiring, by means of at least one sensor, a vibratory-type signal of the rotary axis; processing, by means of a processor, in the time domain and in the domain of the frequency, the signal acquired by the sensor, obtaining as a result of said processing, energy measurements of the acquired signal; compare, in the processor, the energy measurements with previously established energy standards; and finally determine if there is any failure in the rotary axis based on the comparison between the energy measurements and the previously established patterns. (Machine-translation by Google Translate, not legally binding)
    公开号:ES2549652A1
    申请号:ES201430606
    申请日:2014-04-24
    公开日:2015-10-30
    发明作者:Juan Carlos García Prada；Cristina CASTEJÓN SISAMÓN；María Jesús GÓMEZ GARCÍA；Jesús MENESES ALONSO
    申请人:Universidad Carlos III de Madrid；Danobat S Coop Ltda；Alstom Transporte SA；SKF Espanola SA；
    IPC主号:
    专利说明:

    image 1
    image2
    image3
    image4
    image5
    image6
    image7
    image8
    signals, are introduced into a neural network (8) trained with previous samples to compare and classify the signals depending on the state of the axis. The failures detected are automatically identified with very high success rates.
    To apply the monitoring methods of gears and bearings to the case of railway shafts, which are very critical elements, studies on various functionalities have been carried out according to different embodiments of the present invention to conclude which are the most optimal indicators of the presence of a defect in an axis, such as a crack or a crack.
    The functional ones usually studied in the state of the art are, on the one hand, statistical parameters (such as mean, median, standard deviation, bias, kurtosis, effective value, minimum, maximum, peak value). peak, the peak value, the form factor or the crest factor) and on the other, representations of the signal that allow it to be studied both in the time and frequency domain.
    Regarding the representations in the frequency domain, they have traditionally been used to examine the frequency peaks at which the amplitude is modified or that are displaced by the presence of a defect. This is due to the fact that the presence of a crack in a structural element reduces its stiffness, so that the natural frequencies are reduced, and the vibration modes change. The study of natural frequencies has always been the first step to find an indicator of damage in a mechanical element, since it can be measured easily and quickly and that is poorly contaminated by experimental noise. Performing an exhaustive analysis by frequency bands, the differences in energy are detected in certain bands that indicate the presence of a defect, but the main disadvantage of this technique is that when a defect appears, changes in natural frequencies are very small, and can be blocked due to experimental errors.
    In the present invention, however, the vibration modes are not observed, but the response signals that appear during the normal movement of the machine associated with the rotary axis, which will have the disadvantage of noise.
    As for the functional ones that describe a signal in the time and frequency domain, the wavelet transform stands out, which is the one chosen by the present invention.
    10
    image9
    image10
    image11
    image12
    different tests have been carried out to verify which one is the most appropriate for the particular case of detection of axle failures through the execution of different tests. In the test cases, according to an embodiment of the invention, the energy of the wavelet packets obtained from a mother wavelet Daubechies 6 and with a decomposition level equal to 3 have been used as standards. That is, they are presented to the neural network as input 8 energy levels corresponding to the 8 packages (23), such as those shown in Fig. 7. Tests have been carried out with two types of multilayer RNA integrated in the easy-to-apply MATLABTM toolbox, which will allow study the applicability of this type of classification system to the particular case of railway axes. These are the FeedForward type network and the pattern recognition network:
    -The RNA Feed Forward (unidirectional) is a multilayer neural network with activation functions in its hidden sigmoid type layer. It is called Feed-Forward because the information is propagated forward and there is no feedback.
    This neural network is trained with the 8 energy values of the wavelet packets of the signals obtained with the system (patterns). In the specific examples shown below, we have worked with 1000 signals, corresponding to healthy axes and defective axes (16% depth of crack with respect to the diameter of the axis). 75% of these signals have been used for the training of the network and the remaining 25% (400 standard vectors of dimension 8) have been presented to the network already trained as a validation system. In this network configuration, 8 neurons have been used in the input layer (corresponding to the size of the standard vector), 5 neurons in the hidden layer and a neuron in the output layer corresponding to the classification result. In this first approach, the answer was chosen: 0 for a healthy axis and 1 for a defective axis.
    In the experimental results the convergence of the error can be verified. In 26 iterations the algorithm converges to the set value, however, the best previous validation value is achieved in iteration 20 with an error of 0.0158. However, once the network has already been trained, by presenting new patterns to be classified, a success rate of 84.25% is obtained. In addition, the algorithm presented has a success rate of 100% when presented
    fifteen
    image13
    权利要求:
    Claims (1)
    [1]
    image 1
    image2
    类似技术:
    公开号 | 公开日 | 专利标题
    ES2549652A1|2015-10-30|Method and system to automatically detect failures in a rotary axis |
    BR112018009546A2|2018-11-06|method to check a tire, and check line of a tire
    CN105659080B|2018-11-09|System and method for crossing noise filtering from from the acoustic energy of valve
    JP2011245277A5|2014-11-13|Subject information acquisition apparatus and signal processing method
    BR112017009025A2|2018-02-06|methods for calibrating optical computing devices or optical sensors.
    WO2012115912A3|2012-11-22|Design based device risk assessment
    JP2018020062A5|2020-01-16|
    JP2016075585A5|2017-11-16|
    JP2014004736A5|2015-01-15|
    JP2018068748A5|2019-10-17|
    JP2010268009A5|2011-07-07|
    WO2015196109A3|2016-02-11|Acceptance, commissioning, and ongoing benchmarking of a linear accelerator | using an electronic portal imaging device |
    JP2008545127A|2008-12-11|Method and apparatus for inspecting a measurement object using invasive radiation
    ES2625761B1|2018-04-10|METHOD FOR AUTOMATION IN DOSIMETRIC CALIBRATION, RECONSTRUCTION AND VERIFICATION OF COMPREHENSIVE TREATMENTS OF INTEGRATED RADIOTHERAPY IN AN ENVIRONMENT AND SYSTEM FOR PRACTICE OF THE SAME
    JP2017080345A5|2018-10-04|
    NO20151465A1|2015-10-28|DEVICE AND PROCEDURE FOR CORROSION DETECTION AND PRE-EVALUATION EVALUATION BY INTEGRATED CALCULATING ELEMENTS
    WO2019238152A3|2020-02-27|Inspection method and system
    JP2018205033A5|2020-07-02|
    JP2020057144A5|2021-10-21|
    WO2018088760A3|2018-08-16|Method and device for adjusting quality determination conditions for test body
    Sahoo et al.2018|Bearing Fault Detection and Classification Using ANC-Based Filtered Vibration Signal
    JP3857193B2|2006-12-13|Silicon wafer crack detection method
    UA109586U|2016-08-25|METHOD OF DIAGNOSIS OF THE Severity of BRONCHO-OBSTRUCTIVE SYNDROME IN CHILDREN WITH CONDITIONAL DYSPLASIA
    JP2011254124A5|2013-06-13|
    TW201538968A|2015-10-16|Quality assessment of directed self-assembling method
    同族专利:
    公开号 | 公开日
    EP3136255A1|2017-03-01|
    WO2015162331A1|2015-10-29|
    ES2549652B1|2016-10-07|
    US20170052060A1|2017-02-23|
    EP3136255A4|2018-01-17|
    CN106796579A|2017-05-31|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US6382029B1|2000-03-07|2002-05-07|The United States Of America As Represented By The Secretary Of Commerce|Apparatus and method for utilizing electromagnetic acoustic transducers to non-destructively analyze in-service conductive materials|
    US20040022436A1|2001-03-16|2004-02-05|Paul Patti|System and method for data analysis of x-ray images|
    US20020194159A1|2001-06-08|2002-12-19|The Regents Of The University Of California|Parallel object-oriented data mining system|
    US6857553B1|2002-04-17|2005-02-22|The United States Of America As Represented By The United States Department Of Energy|Method and apparatus for in-process sensing of manufacturing quality|
    WO2005031502A2|2003-09-22|2005-04-07|Kim Hyeung-Yun|Methods for monitoring structural health conditions|
    US20060069520A1|2004-09-28|2006-03-30|Dimitry Gorinevsky|Structure health monitoring system and method|
    US20110023610A1|2009-08-03|2011-02-03|Georgia Tech Research Corporation|Methods and systems for classifying the type and severity of defects in welds|
    CN102797671A|2011-05-25|2012-11-28|中国石油大学|Fault detection method and device of reciprocating compressor|
    CN103294849B|2013-05-08|2015-11-18|大连交通大学|Based on the Fault Diagnosis of AC Electric Machine model construction method of RBF neural|
    CN103235649A|2013-05-09|2013-08-07|上海工程技术大学|Three-dimensional wireless input device based on micro-electro-mechanical systems accelerometer|EP3489641A1|2017-11-27|2019-05-29|Goodrich Actuation Systems Limited|Improved system for detecting a mechanical fault in a rotating shaft|
    CN108416294A|2018-03-08|2018-08-17|南京天数信息科技有限公司|A kind of fan blade fault intelligent identification method based on deep learning|
    WO2019175984A1|2018-03-13|2019-09-19|三菱電機株式会社|Failure level calculation device and failure level calculation system|
    CN109506907B|2018-11-06|2020-05-22|西安交通大学|Method for identifying loosening state of bolt connection structure based on vibration response information|
    CN109765055B|2019-01-31|2020-11-10|杭州安脉盛智能技术有限公司|Rolling bearing fault detection method and system based on EWT, spectrum effective value and KNN|
    EP3832417A1|2019-12-05|2021-06-09|Elmos Semiconductor SE|Method for the detection of a state of a system|
    CN111273101A|2020-02-25|2020-06-12|江门市天一达电气科技有限公司|Fault arc detection method and device and storage medium|
    CN112395968B|2020-11-11|2021-08-27|北京京航计算通讯研究所|Mechanical rotating part fault diagnosis method and device based on neural network|
    CN113566953B|2021-09-23|2021-11-30|中国空气动力研究与发展中心设备设计与测试技术研究所|Online monitoring method for flexible-wall spray pipe|
    法律状态:
    2016-10-07| FG2A| Definitive protection|Ref document number: 2549652 Country of ref document: ES Kind code of ref document: B1 Effective date: 20161007 |
    优先权:
    申请号 | 申请日 | 专利标题
    ES201430606A|ES2549652B1|2014-04-24|2014-04-24|Method and system to automatically detect faults in a rotary axis|ES201430606A| ES2549652B1|2014-04-24|2014-04-24|Method and system to automatically detect faults in a rotary axis|
    CN201580027222.5A| CN106796579A|2014-04-24|2015-04-24|For the method and system of the defect in automatic detection rotary shaft|
    US15/306,472| US20170052060A1|2014-04-24|2015-04-24|Method and system for automatically detecting faults in a rotating shaft|
    PCT/ES2015/070348| WO2015162331A1|2014-04-24|2015-04-24|Method and system for automatically detecting defects on a rotating shaft|
    EP15782634.8A| EP3136255A4|2014-04-24|2015-04-24|Method and system for automatically detecting defects on a rotating shaft|
    [返回顶部]