WO2014160760A2 - Procédé et appareil pour utiliser des signatures d'interaction post-processus d'assemblage pour détecter des échecs d'assemblage - Google Patents

Procédé et appareil pour utiliser des signatures d'interaction post-processus d'assemblage pour détecter des échecs d'assemblage Download PDF

Info

Publication number
WO2014160760A2
WO2014160760A2 PCT/US2014/031837 US2014031837W WO2014160760A2 WO 2014160760 A2 WO2014160760 A2 WO 2014160760A2 US 2014031837 W US2014031837 W US 2014031837W WO 2014160760 A2 WO2014160760 A2 WO 2014160760A2
Authority
WO
WIPO (PCT)
Prior art keywords
manufacture
articles
interaction
successful
failed
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/US2014/031837
Other languages
English (en)
Other versions
WO2014160760A3 (fr
Inventor
Jianjun Wang
Thomas A. Fuhlbrigge
Jonas HAULIN
Gregory Rossano
David Alan Bourne
Biao Zhang
Alberto Rodriguez GARCIA
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
ABB Technology AG
Original Assignee
ABB Technology AG
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by ABB Technology AG filed Critical ABB Technology AG
Priority to CN201480028747.6A priority Critical patent/CN105229548B/zh
Publication of WO2014160760A2 publication Critical patent/WO2014160760A2/fr
Publication of WO2014160760A3 publication Critical patent/WO2014160760A3/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0221—Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Program-control systems
    • G05B19/02—Program-control systems electric
    • G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
    • G05B19/41875—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by quality surveillance of production
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/30—Nc systems
    • G05B2219/32—Operator till task planning
    • G05B2219/32201—Build statistical model of past normal proces, compare with actual process
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Definitions

  • This invention relates to automated assembly and more particularly to the detection of assembly failures after the assembly is completed.
  • Automated assembly such as for example and without limitation robotic assembly, is used in various industries such as automotive, electronics etc.
  • robotic assembly for electronics is described in WO2011153156 published on December 8, 2011.
  • In-process failure detection methods exist, such as those described in the paper by Rodriguez, et al, entitled Failure Detection in Assembly: Force Signature Analysis (from the IEEE CASE conference in 2010) . Such methods use data from a sensor that is often corrupted by the dynamics of the assembly process. Post assembly inspection can be performed by automated vision inspection and/or a destructive tension if the concern is the mechanical strength of the assembled parts.
  • a method for detecting the success of an automated process that produces an article of manufacture includes but is not limited to:
  • a method for detecting the success of an automated process that produces an article of manufacture includes but is not limited to:
  • a method for detecting the success of an automated process that produces an article of manufacture by testing a statistically significant number of successful and failed articles of manufacture produced using the automated process includes but is not limited to:
  • Fig. 1 shows a robot holding an assembled product.
  • Fig. 2 shows the test platform that is used in the post assembly test system in testing the assembled product for assembly failures.
  • Figs. 3 and 4 show one example of the parts to be assembled into a product that is tested by the post assembly test system.
  • Figs. 5a to 5c show examples of a faulty assembly for the parts that are that are shown in Figs. 3 and 4 and Fig. 5d shows a correct assembly for those parts.
  • Figs. 6a and 6b show the motion used by the post assembly test system to detect an assembly failure for the parts shown in Figs. 3 and 4.
  • Figs. 7a and 7b show flowcharts for two phases in the use of the Support Vector Machine (SVM) for assembly failure detection.
  • SVM Support Vector Machine
  • Fig. 8 shows a flowchart for the post processing of the collected force signatures.
  • Fig. 9 shows the hyperplanes that can be used to classify the collected force signatures.
  • Figs. 10 and 10b show a flowchart for optimizing the motion used to test the manufactured articles to improve the correlation of the difference between the interaction signals of successful articles and the interaction signals of failed articles.
  • Figs. 1 and 2 there is shown one embodiment for the post assembly test system.
  • an assembled product 16 is held by a gripper 14 mounted on the tip of a robot 10.
  • Robot 10 may, for example, be an articulated 6-axis robot, a Cartesian gantry robot, a robot having less than 6 axes such as a SCARA robot, or a robot having more than 6 axes such as a multi-arm robot.
  • controller 12 As is shown in Fig. 1, test platform 18, which is to contact the assembled product 16, is mounted on a workbench 20.
  • Test platform 18 can be anything that interacts with the assembled product, that is, article of manufacture, 16. While Fig. 1 shows the test platform 18 mounted on workbench 20 it is well known that the test platform 18 can be held by a robot, not shown in Fig. 1, and robot 10 can bring the assembled product 16 to the robot holding test platform 18 or the robot holding the test platform 18 can bring the test platform 18 into contact with the assembled product 16.
  • the controller 12 of the present invention may include a computer readable medium having computer- readable instructions stored thereon which, when executed by a processor, carry out the operations herein described.
  • the computer-readable medium may be any tangible medium that can contain, store, communicate, propagate, or transport the user-interface program instruction for use by or in connection with the instruction execution system, apparatus, or device and may by way of example but without limitation, be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation tangible medium. More specific examples (a non-exhaustive list) of the computer-readable tangible medium include: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM) , an erasable programmable read-only memory (EPROM or Flash memory) , a portable compact disc read- only memory (CD-ROM), an optical storage device, or a magnetic storage device.
  • RAM random access memory
  • ROM read-only memory
  • EPROM or Flash memory erasable programmable read-only memory
  • CD-ROM compact disc read- only memory
  • CD-ROM compact disc read- only memory
  • controller 12 can perform the operations shown in the flowcharts of Figs. 7, 8 and 10, the embodiment shown in Fig. 1 can also include a separate computation device that communicates with controller 12 to perform those operations.
  • the test platform 18 is constructed with at least a lower layer 26 and a top layer 27.
  • the top layer 27 is made of hard materials, while the lower layer 26 is made of compliant materials such as rubber and foam.
  • compliant materials such as rubber and foam.
  • the purpose of such a design is to provide a compliant and nondestructive contact of the test platform 18 with the assembled product 16.
  • a compliant layer may not be needed due to existence of compliance elsewhere in the system or due to the nature of the assembly operation.
  • Figs. 3 and 4 as the combination of a printed circuit board 24 that has mounted on it various circuit elements, a generally rectangular socket 30 that has therein other circuit elements and a aluminum shield can 36 to cover the generally rectangular socket 30.
  • Socket 30 includes four raised side-walls 32a to 32d that form four corners 34a to 34d.
  • the shield can 28 includes a generally rectangular shaped planar face 36. Side-walls 38a to 38d project upwardly from each edge and form corners 40a to 40d. Shield can 28 is sized to be snap-fit over socket 30. The assembly of the shield can 28 to socket 30 may be performed in a manner well known to those in the robot assembly art by one or more robots (not shown) which are other than robot 10.
  • Figs . 5a to 5d show the good and the faulty assemblies of the shield can 28 to the socket 30. More particularly, Fig. 5a shows that the shield can 28 is missing, that is, the shield can 28 was not assembled to the socket 30.
  • the major fault of that assembly is that one or two of the four corners 40a to 40d of the shield can 28 are not pressed enough into the socket 30.
  • Fig. 5b shows this fault for one corner out of position and
  • Fig. 5c shows this fault for two corners out of position.
  • Fig. 5d shows a good assembly.
  • the shield can 28 covers socket 30 and all four corners 40 to 40d of shield can 28 are in position on the socket 30.
  • Figs. 6a and 6b show the test motion to detect the assembly failure for the parts illustrated in Figs. 3 and 4 using the setup in Figs. 1 and 2.
  • the test motion is a rocking motion that presses each of the four corners 40a-40d of the socket 30 against the top layer 27 of the test platform 18.
  • the fault is a corner or corners of the shield can 28 when assembled to socket 30 are out of position such as shown in Figs. 5b and 5c
  • the pressing of each of the four corners 40a to 40d of the socket 30 against layer 27 can result in a fixing of that faulty as sembly .
  • test motion is executed by the robot 10 and is preprogrammed during the system setup. Due to the high repeatability of the robot 10, this test motion is the same for each assembled product 16. If the gripping of each assembled product 16 is repeatable, then the test condition of the assembly failure has very little variation. As a result, the contact force induced during the rocking and pressing motion is free of other side effects such as the dynamics in the actual assembly process.
  • the contact force signature obtained from the test motion by use of a force sensor can be processed using many existing algorithms for the failure detection.
  • An exemplar algorithm is the industry-standard statistical classification tool, the Support Vector Machine (SVM).
  • SVM Support Vector Machine
  • the SVM can classify the contact force signature data into two categories: successful and unsuccessful failures .
  • step 714 recording at step 714 the category of each the tested series of assembled products 16 as either a success or a failure.
  • Post processing at step 716 the force signatures collected at step 712.
  • the flowchart 800 for the post processing is shown in Fig. 8 and has the following steps :
  • step 802 resampling at step 802 the recorded force signature if the sampling time is uneven and smoothing to remove the noise (various well known techniques such as for example box car averaging are available to perform this function) ;
  • step 806 normalizing at step 806 the force signature data such that the highest value is 1 and the lowest is 0 across all signatures.
  • the step 718 is performed after the post processing step 716 is performed.
  • the feature vector of the post processed force signatures is extracted.
  • PCA Principle Component Analysis
  • step 720 the SVM is trained based on the feature vector as is described in C. Burges, "A tutorial on Support Vector Machines for Pattern Recognition", Data Mining and Knowledge Discovery 2, 121-167, 1998 (“Burges”) .
  • the trained SVM can be saved and used in the testing phase for each assembled product.
  • the testing phase uses the same (or similar) system setup that was used during the training phase.
  • a statistically significant number of N assembled products 16 means that enough samples are taken so that force signatures can be categorized with a predetermined level of accuracy. For example, as describe in the reference by C. Burges, an error rate or "actual risk" (p. 156) of a trained SVM can be calculated. If the "actual risk" value of an SVM is too high, based on a predetermined threshold, more samples could be taken, and the SVM retrained until the risk value is acceptable. Alternatively, other SVM attributes could also be used to determine whether or not the number of samples is statistically significant.
  • the flowchart 722 for the testing phase is shown in Fig. 7b where steps 724, 726 and 728 are identical to steps 716, 718 and 720, respectively of the training phase and therefore do not need to be further described.
  • the contact force signature during the testing phase is recorded and fed into the SVM.
  • the feature vector is then input to the SVM which at step 730 runs the trained SVM.
  • the output of the trained SVM predicts whether the assembly of each tested product was either successful or a failure.
  • a tested product is discarded when the output of the trained SVM predicts that the tested product was not correctly assembled.
  • different actions could be performed on the product, such as retrying the assembly step or setting the product aside for later manual rework.
  • the resampling of the force signature data can use the sim le linear interpolation technique:
  • F'(t) is the force data at a new sampling point
  • F(t a )and F(t b ) are recorded force data at the original sampling time t a and t b .
  • noise removal algorithms include the low pass filter (Eq. 2) and the weighted moving average (Eq. 3) .
  • F is the force data after the noise is removed.
  • f k is the recorded force signature
  • g k is the reference signature.
  • the argument k at the maximum of the cross correlation function (f*g)k i- s the misalignment of the force signature f k with respect to the reference signature g k .
  • the aligned force signature is obtained by simply shifting the time index by this misalignment :
  • ⁇ and o are the mean and standard deviation of force signature f k .
  • the post processing of the force signature can have different steps than those shown in Fig. 8, depending on the quality of the recorded force signature .
  • PCA Principle Component Analysis
  • ' x is the post processed force signature.
  • P be the linear coordinate transformation matrix of dimension m*m, after the transformation the original data ' x changes to the new data ' :
  • the transformation matrix P that maximizes the variances is related to the eigenvectors of the following covariance matrix of the original dataset:
  • V CV D (10) Where V is the matrix of eigenvectors, D is a diagonal matrix of eigenvalues of C arranged in decreasing order.
  • the transformation matrix P in Eq (7) then equals matrix
  • the first few principal components of a recorded force signature can contain enough information to be selected as the feature vector. For example, a test has shown that the first five (5) principal components are good candidates for the feature vector.
  • a classifier After obtaining the feature vector for each recorded force signature, a classifier is ready to be trained.
  • the training data includes a series of feature vectors and the classes each feature vector belongs to. In the case of the assembly failure detection, the classes are success and failure.
  • a trained classifier can predict which class a new feature vector is in.
  • One good classifier is the linear support vector machine (SVM) . Linear SVM tries to divide the feature space by a hyperplane such that the two classes lie on the opposite side of the hyperplane.
  • SVM linear support vector machine
  • hyperplanes that might classify the data.
  • One reasonable choice as the best hyperplane is the one that represents the largest separation, or margin, between the two classes.
  • This hyperplane known as the maximum- margin hyperplane, has the optimal stability with respect to the noise.
  • the linear SVM algorithm is to find such a hyperplane in the feature space .
  • 'x is the feature vector of dimension p
  • the 'y is either 1 or -1, indicating the class to which the feature vector 'x belongs. Any hyperplane in the feature space can be written as the set of points satisfying
  • the two classes are on the opposite side of the hyperplane, thus the hyperplane satisfies
  • a trained SVM is parameterized by and b as shown in eq 12.
  • the prediction of SVM during the test phase can use the following decision logic to predict the class to which the new test belongs to:
  • assembled product 16 is only one example of the assembled products that the method and system described herein can be used with to detect assembly failure in the assembled product after the product is assembled.
  • the measure of what is a successful assembly or manufacture or non-successful, assembly or manufacture for those products will depend on the product .
  • the displacement sensor In the shield can example described above, the failed assemblies typically have raised corners or edges. When they are pressed against a compliant object, the compliant object will deform more compared to the case of successfully assembled ones. A displacement sensor can therefore be used to obtain the interaction signature between the articles and the compliant object.
  • the procedure and algorithms described above can also be followed to train the SVM and use the SVM to detect the success or failure of the article when the displacement sensor is used.
  • the interaction signature from the displacement sensor can be a measurement along one or more axes position and/or a reorientation around those axes .
  • the interaction signature can be a combination of the interaction signatures from a force sensor and a displacement sensor .
  • FIGs. 10a and 10b there is shown a flowchart 1000 for optimizing the motion used to test the manufactured articles to improve the correlation of the difference between the interaction signals of successful articles and the interaction signals of failed articles.
  • Blocks 1002 and 1004 shown in Fig. 10a are two operations that are repeated for each of N products.
  • the test motion is performed for each of the N products and the force signature resulting from performing that motion is collected.
  • the success or failure category of the assembly for each of the N products is record.
  • the flow proceeds to a second group of two operations that are performed for each of the N products.
  • the force signature for each of the N products is post processed.
  • the feature vector is extracted for each of the N products from the post process force signature for each of the N products .
  • test motion is changed for the N products.
  • the change of the test motion can for example and without limitation be with regard to the test motions shown in Figs . 6a and 6b that a different location of the socket 30 is pressed against the test platform top layer 27 by the rocking motion.
  • the test motion parameters such as for speed, angle etc. are changed or are read from different force sensors that are in place in different positions and orientations of the test platform 18.
  • blocks 1014 and 1016 are two operations that are repeated for each of N products.
  • the operations performed at blocks 1014 and 1016 are identical to the operations performed at blocks 1002 and 1004, respectively, except that they are for the changed test motion for each of the N products .
  • the flow proceeds to a second group of two operations at blocks 1018 and 1020 that are performed for each of the N products.
  • the operations performed at blocks 1018 and 1020 are identical to the operations performed at blocks 1006 and 1008, respectively, except they are for the feature vector extracted for the changed test motion for each of the N products.
  • this trained SVM is for the N products that have had the changed test motion at block 1012.
  • the goal of flow 1000 is to minimize the correlation of the difference between the interaction signals of successful articles and the interaction signatures of failed articles . If the correlation has to be improved, then the test motion has to be optimized and the flow returns to block 1012 where the test motion is again changed. If the correlation difference does not have to be improved, then the flow proceeds to block 1026 and the training of the SVM is completed. The minimization of the correlation difference is ended when the difference meets a predetermined criteria for that difference.
  • the predetermined criteria can for example and without limitation be, the value of the improved correlation is lower than a preset threshold (which means the optimized test motion can generate a clear difference on the force sensor signature between a successful and failed assembly) ; or after all the different test motions are preformed no decrease on the value of the correlation (which means the optimized test motion is best among all the test motions to generate the difference on force sensor signature between a successful and a failed assembly) ; or the overall test number of products is over a preset number; or the overall time to produce the N products for optimization is over a preset time.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • General Engineering & Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • Quality & Reliability (AREA)
  • Automatic Assembly (AREA)
  • Manipulator (AREA)
  • General Factory Administration (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

L'invention porte sur une technique pour détecter le succès d'un processus automatisé qui produit un article de manufacture. Un nombre statistiquement important d'articles réussis et défectueux sont produits par le processus automatisé. Chacun de ces articles fait l'objet d'une interaction avec une plateforme de test pour mesurer des signatures d'interaction qui indiquent des articles réussis et défectueux. Une corrélation de la différence entre les signatures d'interaction est calculée. Une signature d'interaction est ensuite obtenue pour un article fabriqué par le processus après les articles fabriqués antérieurement. La nouvelle signature d'interaction est analysée en termes de la différence de corrélation calculée afin de catégoriser automatiquement l'article de manufacture supplémentaire comme étant soit réussi soit défectueux. Une technique est également décrite pour optimiser le mouvement utilisé pour tester les articles fabriqués afin d'améliorer la corrélation de la différence entre les signaux d'interaction d'articles réussis et les signaux d'interaction d'articles défectueux.
PCT/US2014/031837 2013-03-27 2014-03-26 Procédé et appareil pour utiliser des signatures d'interaction post-processus d'assemblage pour détecter des échecs d'assemblage Ceased WO2014160760A2 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201480028747.6A CN105229548B (zh) 2013-03-27 2014-03-26 使用组装后过程交互印记来检测组装故障的方法和装置

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US201361805864P 2013-03-27 2013-03-27
US61/805,864 2013-03-27

Publications (2)

Publication Number Publication Date
WO2014160760A2 true WO2014160760A2 (fr) 2014-10-02
WO2014160760A3 WO2014160760A3 (fr) 2014-11-13

Family

ID=50694025

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2014/031837 Ceased WO2014160760A2 (fr) 2013-03-27 2014-03-26 Procédé et appareil pour utiliser des signatures d'interaction post-processus d'assemblage pour détecter des échecs d'assemblage

Country Status (2)

Country Link
CN (1) CN105229548B (fr)
WO (1) WO2014160760A2 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023184252A1 (fr) * 2022-03-30 2023-10-05 京东方科技集团股份有限公司 Procédé d'analyse de corrélation de données de processus de production, dispositif et support de stockage

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011153156A2 (fr) 2010-06-02 2011-12-08 Abb Research Ltd Ensemble robotique à fixation partielle

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6738450B1 (en) * 2002-12-10 2004-05-18 Agilent Technologies, Inc. System and method for cost-effective classification of an object under inspection
EP1866819A2 (fr) * 2005-02-25 2007-12-19 Biogen Idec MA Inc. Surveillance d'equipement de traitement
US8090676B2 (en) * 2008-09-11 2012-01-03 Honeywell International Inc. Systems and methods for real time classification and performance monitoring of batch processes
JP4730451B2 (ja) * 2009-03-16 2011-07-20 富士ゼロックス株式会社 検出データ処理装置及びプログラム
CN102879881B (zh) * 2012-10-31 2015-03-04 中国科学院自动化研究所 元件夹持装置

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011153156A2 (fr) 2010-06-02 2011-12-08 Abb Research Ltd Ensemble robotique à fixation partielle

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
C. BURGES: "A Tutorial on Support Vector Machines for Pattern Recognition", DATA MINING AND KNOWLEDGE DISCOVERY, vol. 2, 1998, pages 121 - 167, XP002087854, DOI: doi:10.1023/A:1009715923555
L. SMITH, A TUTORIAL ON PRINCIPAL COMPONENTS ANALYSIS, 2002
RODRIGUEZ ET AL., FAILURE DETECTION IN ASSEMBLY: FORCE SIGNATURE ANALYSIS, 2010

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023184252A1 (fr) * 2022-03-30 2023-10-05 京东方科技集团股份有限公司 Procédé d'analyse de corrélation de données de processus de production, dispositif et support de stockage

Also Published As

Publication number Publication date
CN105229548B (zh) 2018-04-06
CN105229548A (zh) 2016-01-06
WO2014160760A3 (fr) 2014-11-13

Similar Documents

Publication Publication Date Title
CN110678727A (zh) 判定装置、判定方法和判定程序
US20200320402A1 (en) Novelty detection using deep learning neural network
CN103198322B (zh) 基于机器视觉的磁瓦表面缺陷特征提取及缺陷分类方法
US20200193219A1 (en) Discrimination device and machine learning method
CN107942940A (zh) 一种基于指令域分析的数控机床的进给轴装配故障的检测方法和装置
US11961255B2 (en) Object detection device and object detection computer program
US11094082B2 (en) Information processing apparatus, information processing method, robot system, and non-transitory computer-readable storage medium
US20190370982A1 (en) Movement learning device, skill discriminating device, and skill discriminating system
CN119188759B (zh) 用于自动化流水线的机械臂自动控制系统及方法
CN109590805A (zh) 一种车削刀具工作状态的确定方法及系统
WO2014160760A2 (fr) Procédé et appareil pour utiliser des signatures d'interaction post-processus d'assemblage pour détecter des échecs d'assemblage
Phan et al. Development of an autonomous component testing system with reliability improvement using computer vision and machine learning
Kwiatkowski et al. The good grasp, the bad grasp, and the plateau in tactile-based grasp stability prediction
CN107696034A (zh) 一种针对工业机器人的错误自主恢复方法
Krabbe et al. Autonomous optimization of fine motions for robotic assembly
Sun et al. An adaptable automated visual inspection scheme through online learning
WO2021077044A1 (fr) Système et procédé pour l'identification unique d'articles
Chen et al. AIoT-enabled defect detection with minimal data: A few-shot learning approach combining prototypical and relational networks for smart manufacturing
Mankad et al. PCB classification using convolutional neural network
Sun et al. Further development of adaptable automated visual inspection—part I: concept and scheme
Fattah et al. Anomaly detection for industrial robot prognostics and health management
KR102437913B1 (ko) 3차원 형상측정 결과 및 cad 설계 데이터의 정량적 비교를 통한 형상분류 시스템
CN118883451B (zh) 用于全自动ai镜检分析仪的样本自动装卸方法及装置
CN119963509B (zh) 一种倒装芯片封装结构及封装方法
CN114787877B (zh) 用于计算用于评估对象探测算法的质量度量的方法

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 201480028747.6

Country of ref document: CN

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14723600

Country of ref document: EP

Kind code of ref document: A2

122 Ep: pct application non-entry in european phase

Ref document number: 14723600

Country of ref document: EP

Kind code of ref document: A2