Robot reducer fault detection system based on big data analysis

Abstract

The present invention discloses a robot reducer fault detection system based on big data analysis, including a detection unit, a fault detection system, a fault prediction and evaluation model, an early warning maintenance module and a display panel, and the present invention relates to the technical field of big data analysis. The robot reducer fault detection system based on big data analysis is provided with a fault detection system, and uses a data processing and analysis module to convert vibration data into a linear image and judge whether there is a fault problem after image processing, and then generates a data compensation instruction after processing and calculating the change image data generated by rotation and determining the specific value, so as to complete the operation of detecting the fault of the reducer, and after data processing, determine the corresponding fault node to implement the detection operation of the data fed back by the fault, and after realizing the data compensation, optimize the data applied to the reducer in the subsequent application, and monitor the product fault faster and in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and in particular to a robot reducer fault detection system based on big data analysis. Background Art

[0002] Reference patent name: A mechanical equipment fault diagnosis system (patent publication number: CN110517245A, patent publication date: 2019-11-29), including a fault detection subsystem, an information transmission subsystem and an early warning subsystem. The fault detection subsystem detects mechanical equipment faults based on infrared images and mechanical equipment vibrations. The information transmission subsystem is used to send fault information to the early warning subsystem. The early warning subsystem issues an alarm after receiving the fault information. Different from the previous manual detection of mechanical equipment, the present invention realizes automatic diagnosis and alarm of mechanical equipment faults. The infrared image of the mechanical equipment during a fault will be different from the infrared image of the mechanical equipment during normal operation, thereby realizing mechanical equipment fault detection based on infrared images. In addition, the vibration of the mechanical equipment will be abnormal during a fault, and fault detection is performed in combination with the vibration of the mechanical equipment, which improves the detection accuracy and scope of application.

[0003] Based on the description in the above-mentioned documents, during the use of the existing reducer, its meshing transmission causes the output shaft and even the entire product to vibrate. However, currently, the deviation caused by the vibration of the product during operation has not been identified and detected in detail. At the same time, the processing and judgment of the data are relatively lagging, so that the product is still used even when it has a fault, resulting in large losses or safety hazards. For this reason, the present invention provides a robot reducer fault detection system based on big data analysis. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a robot reducer fault detection system based on big data analysis, which solves the problem that the deviations caused by the vibration of the product during operation are currently not identified and detected in detail, and the processing and judgment of data are relatively lagging, so that the product is still used even when it has faults, resulting in large losses or safety hazards.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a robot reducer fault detection system based on big data analysis, comprising:

[0006] The detection unit includes a vibration signal sensor installed on the reducer and a CDD camera facing the output shaft of the reducer, which is used to detect the vibration data of the reducer and the image data of the changes generated by the rotation;

[0007] A fault detection system is used to receive vibration data and image data, and complete fault analysis and judgment operations based on data analysis and processing;

[0008] The data acquisition and receiving module acquires and receives the vibration data of the reducer and the image data of the changes caused by the rotation, and classifies the data;

[0009] The data processing and analysis module converts the vibration data into a linear image and processes the image to determine whether there is a fault problem. It then processes and calculates the change image data generated by the rotation and determines the specific value to generate a data compensation instruction;

[0010] The data result output module outputs the judgment result of the fault problem to the early warning maintenance module and the display panel for maintenance reminder and data display, and transmits the fault and compensation data to the fault prediction and evaluation model for training;

[0011] The fault prediction and evaluation model introduces the compensation data after the reducer fault detection into the model, completes the training optimization, and is used for the actual reducer fault detection;

[0012] The early warning maintenance module is used to inform personnel of the existence of faults and the need for maintenance through a buzzer or text display;

[0013] The display panel shows the data during fault detection and can be controlled and consulted by personnel.

[0014] Preferably, the operation of the data processing and analysis module to convert the vibration data into a linear image is:

[0015] A1. The detected forward and reverse vibration data of the reducer are sorted according to the detection order of time to obtain a vibration data set M;

[0016] A2. Establish a coordinate axis for the change of the floating frequency of the vibration data, establish the X-axis with the change value of time, and establish the Y-axis with the value of the vibration signal, and mark the nodes according to the corresponding vibration signal value per second;

[0017] A3, and connect the vibration signal value nodes located above the X-axis in a curve in the order of time, so as to form linear image data about the floating frequency of the vibration data.

[0018] Preferably, the data processing and analysis module processes the linear image as follows:

[0019] B1. Perform image segmentation on the linear image, extend the segmented linear image in a direction parallel to the Y axis at the time node per second, and sort the linear image in chronological order to form a linear image data set labeled N;

[0020] B2. Complete the matching operation of the linear image data set through the image coverage comparison algorithm, and mark the linear image data set as N{N1, N2, ..., Nn}, first use the linear image data subset N1 as the initial cover image, and move the linear image data subset N1 to the adjacent linear image data subset N2 for feature comparison;

[0021] B3, and when the features of the linear image data subset N1 do not match the features of the linear image data subset N2, the linear image data subset N1 is merged with the linear image data subset N2 to form a new linear image data subset N 1+2 , then the linear image data subset N 1+2 The linear image data subset N of subsequent equal time nodes 3+4 The features are compared again, and the operation is repeated until the features of the linear image data subsets at two equal time nodes are consistent. The frequency domain regular feature of the vibration data is marked as L. If the linear image data subsets cannot match the same features when merged to half the number, it is directly determined that the product has a fault problem.

[0022] Preferably, the feature matching operation for the linear image data subset N1 and the linear image data subset N2 in B3 is:

[0023] b31, firstly, the linear features in the images of the linear image data subset N1 and the linear image data subset N2 are extracted by grayscale processing;

[0024] b32, then extract the node data of the linear feature parts of the linear image data subset N1 and the linear image data subset N2, and correspond the starting node on the linear image data subset N1 to the starting node on the linear image data subset N2, and compare whether the curves of the two reaching their corresponding end nodes overlap;

[0025] b33. If they overlap, the linear image data subset N1 is the frequency domain regular feature of the vibration data. If they do not overlap, continue the operation of B3 until the frequency domain regular feature L of the vibration data is obtained or the linear image data subsets cannot be matched to the same feature when merged to half the number.

[0026] Preferably, the frequency domain regularity feature L of the vibration data is expressed as:

[0027]

[0028] Among them, F(L) represents the generation of frequency domain regularity feature L of vibration data, represents the frequency domain regularity feature L of vibration data generated after matching the features of linear image data subsets at two adjacent and equal time nodes, and N1=N2 represents that the linear image features of image data subset N1 and linear image data subset N2 are the same, and so on. It means that the same linear image feature cannot be matched, that is, the entire linear image data is the frequency domain regular feature L, that is, the result data located in this way means that the product has a fault problem.

[0029] Preferably, the data processing and analysis module processes the frequency domain regularity feature L of the vibration data by:

[0030] f(L) = maxF(L);

[0031] Where f(L) is the frequency value calculation of the frequency domain regular feature L of the vibration data, and maxF(L) is the maximum value calculation of the frequency domain regular feature L of the vibration data;

[0032] The obtained maximum frequency value is compared with the frequency value threshold of the product in the database. If the maximum frequency value is greater than or equal to the frequency value threshold set for the product, it is determined that the product has a fault problem.

[0033] Preferably, the data processing and analysis module processes the changed image data generated by the rotation by:

[0034] C1. First, the video data collected by the camera is processed into frames, and a frame image data set generated by the rotation of the output shaft of the reducer is obtained, which is marked as P;

[0035] C2, performing grayscale processing on the image data in the frame number image data set P, and introducing the end face image before the rotation starts and overlaying the image in the frame number image data set P;

[0036] C3. At this time, the difference result caused by the fault is calculated based on the image result generated by the overlay, marked as Q, and a data compensation instruction is generated through the difference result Q.

[0037] Preferably, the frame number processing operation of the video data in C1 is:

[0038] c11, performing a video segmentation operation on the video data image in a manner of one frame per second, and classifying the image into a frame number image data set P in sequence;

[0039] c12, then filter the image data in the frame image data set P, align the four corner points of adjacent images, and make the background of the upper layer image data transparent, leaving the feature part to compare with the feature part of the lower layer;

[0040] c13. If the characteristic part of the upper layer image data completely overlaps with the characteristic part of the lower layer, the upper layer image data is filtered out. Conversely, if the characteristic part of the upper layer image data differs from the characteristic part of the lower layer, the image data of both the upper layer and the lower layer are retained in the frame number image data set P.

[0041] Preferably, the image overlay operation of the end face image in C2 and the frame image data set P is:

[0042] c21, at this time, the end face image before the rotation starts is used as the basic feature, and the data subsets in the frame number image data set P are sequentially compared with the end face image;

[0043] c22, still keeping the four corner points of the end face image and the image of the frame number image data subset aligned, while making the image data of the data subset in the frame number image data set P transparent, and judging whether there is a difference between the end face image and the frame number image data subset through the two characteristic edge boundaries after covering;

[0044] c23. If the two feature edge boundaries after covering overlap, filter the subset of the frame number image data; if the two feature edge boundaries after covering differ, retain and process the calculated difference value.

[0045] Preferably, the calculation steps for the difference result Q in C3 are:

[0046] c31, connect the intersection nodes of the last two features covered in c22, and set a segmentation line marked as S at 90 degrees based on the center point of the connecting line segment, and the segmentation line S extends to the intersection of the two feature edge boundaries as the endpoint;

[0047] c32. Based on the diameter data of the characteristic part, the center points of the two characteristic parts are obtained by extending from the two end points of the dividing line S into the characteristic part, and the deviation data generated by the vibration is obtained by connecting the center points;

[0048] c33. Then compare the deviation data with the vibration setting deviation value of the product in the database and calculate the difference result Q. When the deviation data is greater than the vibration setting deviation value of the product, it is fault data, and the calculation formula of the difference result Q is:

[0049] Q = (s-2r)*T;

[0050] Where s is the length of the segmentation line S in the image data, that is, the length between the two ends of the segmentation line S and the outermost intersection of the feature edge boundary, r is the radius length of the end face feature in the output shaft end face image, and T is the ratio between the actual product and the image.

[0051] The present invention provides a robot reducer fault detection system based on big data analysis. Compared with the prior art, it has the following beneficial effects:

[0052] (1) The robot reducer fault detection system based on big data analysis is provided with a fault detection system. The data processing and analysis module is used to convert the vibration data into a linear image and determine whether there is a fault problem after image processing. Then, the change image data generated by the rotation is processed and calculated to determine the specific value and then generate a data compensation instruction. In this way, the reducer fault detection operation can be completed. After data processing, the corresponding fault node is determined to implement the detection operation on the data fed back by the fault. After the data compensation is realized, the data subsequently applied to the reducer is optimized, and the product fault is monitored faster and more timely.

[0053] (2) The robot reducer fault detection system based on big data analysis converts vibration data into linear images through the data processing and analysis module and determines whether there is a fault problem after image processing. According to the segmentation and combination matching comparison of linear image data, the frequency domain regularity features are extracted or the product fault is directly determined. Based on the frequency domain regularity features, the fault problem of the product at a certain time point can also be obtained according to the time node, which realizes the data conversion process of analyzing and judging the fault more quickly and improves the efficiency of detection.

[0054] (3) The robot reducer fault detection system based on big data analysis processes the changed image data generated by the rotation through the data processing and analysis module, keeps the four corner points of the end face image and the image of the frame image data subset aligned, and makes the data subset image data in the frame image data set P transparent. The difference between the end face image and the frame image data subset is judged by the two characteristic edge boundaries after covering and the difference result is calculated. The fault prediction and evaluation model is introduced for training and optimization, so as to more accurately improve the accuracy of subsequent applications of the fault detection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a functional block diagram of the fault detection system of the present invention;

[0056] Figure 2 It is a schematic diagram of the difference result calculation operation of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] See also Figure 1-Figure 2 , the present invention provides two technical solutions:

[0059] Embodiment 1: A robot reducer fault detection system based on big data analysis includes:

[0060] The detection unit includes a vibration signal sensor installed on the reducer and a CDD camera facing the output shaft of the reducer, which is used to detect the vibration data of the reducer and the image data of the changes generated by the rotation;

[0061] A fault detection system is used to receive vibration data and image data, and complete fault analysis and judgment operations based on data analysis and processing;

[0062] The data acquisition and receiving module acquires and receives the vibration data of the reducer and the image data of the changes caused by the rotation, and classifies the data;

[0063] The data processing and analysis module converts the vibration data into a linear image and processes the image to determine whether there is a fault problem. It then processes and calculates the change image data generated by the rotation and determines the specific value to generate a data compensation instruction;

[0064] The data result output module outputs the judgment result of the fault problem to the early warning maintenance module and the display panel for maintenance reminder and data display, and transmits the fault and compensation data to the fault prediction and evaluation model for training;

[0065] The fault prediction and evaluation model introduces the compensation data after the reducer fault detection into the model, completes the training optimization, and is used for the actual reducer fault detection;

[0066] The early warning maintenance module is used to inform personnel of the existence of faults and the need for maintenance through a buzzer or text display;

[0067] The display panel shows the data during fault detection and can be controlled and consulted by personnel.

[0068] The data collected by the vibration signal sensor and CDD camera are transmitted to the computer or display panel through the wireless LAN, and the fault prediction and evaluation model completes the training and data optimization by providing the data collected from the product and the optimized compensation data. In this way, when applied to different reducer products for fault detection, the corresponding fault problems can be found more accurately and quickly.

[0069] Among them, by setting up a fault detection system, the data processing and analysis module is used to convert the vibration data into a linear image and determine whether there is a fault problem after image processing, and then the change image data generated by the rotation is processed and calculated to determine the specific value and then generate a data compensation instruction. In this way, the operation of detecting the reducer fault can be completed. After data processing, the corresponding fault node is determined to implement the detection operation on the data fed back by the fault, and after the data compensation is realized, the data subsequently applied to the reducer is optimized, and product faults are monitored faster and more immediately.

[0070] In the embodiment of the present invention, the operation of the data processing and analysis module to convert the vibration data into a linear image is as follows:

[0071] A1. The detected forward and reverse vibration data of the reducer are sorted according to the detection order of time to obtain a vibration data set M;

[0072] A2. Establish a coordinate axis for the change of the floating frequency of the vibration data, establish the X-axis with the change value of time, and establish the Y-axis with the value of the vibration signal, and mark the nodes according to the corresponding vibration signal value per second;

[0073] A3, and connect the vibration signal value nodes located above the X-axis in a curve in the order of time, so as to form linear image data about the floating frequency of the vibration data.

[0074] In the embodiment of the present invention, the data processing and analysis module processes the linear image as follows:

[0075] B1. Perform image segmentation on the linear image, extend the segmented linear image in a direction parallel to the Y axis at the time node per second, and sort the linear image in chronological order to form a linear image data set labeled N;

[0076] B2. Complete the matching operation of the linear image data set through the image coverage comparison algorithm, and mark the linear image data set as N{N1, N2, ..., N n}, first use the linear image data subset N1 as the initial cover image, and move the linear image data subset N1 to the adjacent linear image data subset N2 for feature comparison;

[0077] B3, and when the features of the linear image data subset N1 do not match the features of the linear image data subset N2, the linear image data subset N1 is merged with the linear image data subset N2 to form a new linear image data subset N 1+2 , then the linear image data subset N 1+2 The linear image data subset N of subsequent equal time nodes 3+4The features are compared again, and the operation is repeated until the features of the linear image data subsets at two equal time nodes are consistent. The frequency domain regular feature of the vibration data is marked as L. If the linear image data subsets cannot match the same features when merged to half the number, it is directly determined that the product has a fault problem.

[0078] In the embodiment of the present invention, the feature matching operation for the linear image data subset N1 and the linear image data subset N2 in B3 is:

[0079] b31, firstly, the linear features in the images of the linear image data subset N1 and the linear image data subset N2 are extracted by grayscale processing;

[0080] b32, then extract the node data of the linear feature parts of the linear image data subset N1 and the linear image data subset N2, and correspond the starting node on the linear image data subset N1 to the starting node on the linear image data subset N2, and compare whether the curves of the two reaching their corresponding end nodes overlap;

[0081] b33. If they overlap, the linear image data subset N1 is the frequency domain regular feature of the vibration data. If they do not overlap, continue the operation of B3 until the frequency domain regular feature L of the vibration data is obtained or the linear image data subsets cannot be matched to the same feature when merged to half the number.

[0082] In the embodiment of the present invention, the frequency domain regularity feature L of the vibration data is expressed as:

[0083]

[0084] Among them, F(L) represents the generation of frequency domain regularity feature L of vibration data, represents the frequency domain regularity feature L of vibration data generated after matching the features of linear image data subsets at two adjacent and equal time nodes, and N1=N2 represents that the linear image features of image data subset N1 and linear image data subset N2 are the same, and so on. It means that the same linear image feature cannot be matched, that is, the entire linear image data is the frequency domain regular feature L, that is, the result data located in this way means that the product has a fault problem.

[0085] And the data from top to bottom in the formula are all result data, except All other data except can be matched to the same linear image features.

[0086] In the embodiment of the present invention, the data processing and analysis module processes the frequency domain regularity feature L of the vibration data as follows:

[0087] f(L) = maxF(L);

[0088] Where f(L) is the frequency value calculation of the frequency domain regular feature L of the vibration data, and maxF(L) is the maximum value calculation of the frequency domain regular feature L of the vibration data;

[0089] The obtained maximum frequency value is compared with the frequency value threshold of the product in the database. If the maximum frequency value is greater than or equal to the frequency value threshold set for the product, it is determined that the product has a fault problem.

[0090] Among them, the vibration data is converted into a linear image through the data processing and analysis module, and the image is processed to determine whether there is a fault problem. According to the segmentation and combination matching and comparison of the linear image data, the frequency domain regularity characteristics are extracted or the product fault is directly determined. Based on the frequency domain regularity characteristics, the fault problem of the product at a certain time point can also be obtained according to the time node, which realizes the data conversion process of analyzing and judging the fault more quickly and improves the efficiency of detection.

[0091] Embodiment 2 is different from Embodiment 1 in that: in the embodiment of the present invention, the data processing and analysis module processes the changed image data generated by the rotation as follows:

[0092] C1. First, the video data collected by the camera is processed into frames, and a frame image data set generated by the rotation of the output shaft of the reducer is obtained, which is marked as P;

[0093] C2, performing grayscale processing on the image data in the frame number image data set P, and introducing the end face image before the rotation starts and overlaying the image in the frame number image data set P;

[0094] C3. At this time, the difference result caused by the fault is calculated based on the image result generated by the overlay, marked as Q, and a data compensation instruction is generated through the difference result Q.

[0095] In the embodiment of the present invention, the frame number processing operation of the video data in C1 is:

[0096] c11, performing a video segmentation operation on the video data image in a manner of one frame per second, and classifying the image into a frame number image data set P in sequence;

[0097] c12, then filter the image data in the frame image data set P, align the four corner points of adjacent images, and make the background of the upper layer image data transparent, leaving the feature part to compare with the feature part of the lower layer;

[0098] c13. If the characteristic part of the upper layer image data completely overlaps with the characteristic part of the lower layer, the upper layer image data is filtered out. Conversely, if the characteristic part of the upper layer image data differs from the characteristic part of the lower layer, the image data of both the upper layer and the lower layer are retained in the frame number image data set P.

[0099] In the embodiment of the present invention, the image overlay operation of the end face image in C2 and the frame image data set P is:

[0100] c21, at this time, the end face image before the rotation starts is used as the basic feature, and the data subsets in the frame number image data set P are sequentially compared with the end face image;

[0101] c22, still keeping the four corner points of the end face image and the image of the frame number image data subset aligned, while making the image data of the data subset in the frame number image data set P transparent, and judging whether there is a difference between the end face image and the frame number image data subset through the two characteristic edge boundaries after covering;

[0102] c23. If the two feature edge boundaries after covering overlap, filter the subset of the frame number image data; if the two feature edge boundaries after covering differ, retain and process the calculated difference value.

[0103] In the embodiment of the present invention, the calculation steps for the difference result Q in C3 are:

[0104] c31, connect the intersection nodes of the last two features covered in c22, and set a segmentation line marked as S at 90 degrees based on the center point of the connecting line segment, and the segmentation line S extends to the intersection of the two feature edge boundaries as the endpoint;

[0105] c32. Based on the diameter data of the characteristic part, the center points of the two characteristic parts are obtained by extending from the two end points of the dividing line S into the characteristic part, and the deviation data generated by the vibration is obtained by connecting the center points;

[0106] c33. Then compare the deviation data with the vibration setting deviation value of the product in the database and calculate the difference result Q. When the deviation data is greater than the vibration setting deviation value of the product, it is fault data, and the calculation formula of the difference result Q is:

[0107] Q = (s-2r)*T;

[0108] Where s is the length of the segmentation line S in the image data, that is, the length between the two ends of the segmentation line S and the outermost intersection of the feature edge boundary, r is the radius length of the end face feature in the output shaft end face image, and T is the ratio between the actual product and the image.

[0109] Among them, the changed image data generated by the rotation is processed by the data processing and analysis module, the four corner points of the end face image and the image of the frame image data subset are kept aligned, and the data subset image data in the frame image data set P is made transparent. The difference between the end face image and the frame image data subset is judged by the two characteristic edge boundaries after covering and the difference result is calculated. The fault prediction and evaluation model is introduced for training and optimization, so as to more accurately improve the accuracy of subsequent applications of the fault detection system.

[0110] Comparative experiment

[0111] Now, the existing reducer fault detection system and the reducer fault detection system of the present invention are used to perform fault detection operations on multiple groups of reducers with known fault conditions and without fault problems, and the data after the detection are recorded, and the time when the existing reducer fault detection system and the reducer fault detection system of the present invention finally complete the fault detection operation and the accuracy rate of whether the fault can be detected are recorded, and the results are shown in Table 1:

[0112] Table 1 Experimental results comparison table

[0113] Mean time to detect faults The accuracy of the results Existing reducer fault detection system 112s 88.5％ Reducer fault detection system of the present invention 57s 99％

[0114] In summary, the reducer fault detection system of the present invention takes a shorter time to detect the data fault of the reducer, and the detection results of the reducer fault detection system of the present invention are more accurate. At the same time, the existing reducer fault detection system can also be used for the data fault detection operation of the reducer.

[0115] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0116] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0117] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A robot reducer fault detection system based on big data analysis, characterized by: include: The detection unit includes a vibration signal sensor installed on the reducer and a CDD camera facing the output shaft of the reducer, which is used to detect the vibration data of the reducer and the image data of the changes generated by the rotation; A fault detection system is used to receive vibration data and image data, and complete fault analysis and judgment operations based on data analysis and processing; The data acquisition and receiving module acquires and receives the vibration data of the reducer and the image data of the changes caused by the rotation, and classifies the data; The data processing and analysis module converts the vibration data into a linear image and processes the image to determine whether there is a fault problem. It then processes and calculates the change image data generated by the rotation and determines the specific value to generate a data compensation instruction; The data result output module outputs the judgment result of the fault problem to the early warning maintenance module and the display panel for maintenance reminder and data display, and transmits the fault and compensation data to the fault prediction and evaluation model for training; Fault prediction and assessment model, which introduces the compensation data after reducer fault detection into the model, completes training optimization, and is used for actual reducer fault detection; The early warning maintenance module is used to inform personnel of the existence of faults and the need for maintenance through a buzzer or text display; Display panel, which displays the data during fault detection and can be controlled and consulted by personnel; The data processing and analysis module converts vibration data into linear images by: A1. The detected forward and reverse vibration data of the reducer are sorted according to the detection order of time to obtain a vibration data set M; A2. Establish a coordinate axis for the change of the floating frequency of the vibration data, establish the X-axis with the change value of time, and establish the Y-axis with the value of the vibration signal, and mark the nodes according to the corresponding vibration signal value per second; A3, and connecting the vibration signal value nodes located above the X-axis in a curve in the order of time, so as to form linear image data about the floating frequency of the vibration data; The data processing and analysis module performs the following operations on the linear image: B1. Perform image segmentation on the linear image, extend the segmented linear image in a direction parallel to the Y axis at the time node per second, and sort the linear image in chronological order to form a linear image data set; B2. Complete the matching operation of the linear image data set through the image coverage comparison algorithm, and mark the linear image data set as N, where N∈{N1, N1, …, N n }, first use the linear image data subset N1 as the initial cover image, and move the linear image data subset N1 to the adjacent linear image data subset N2 for feature comparison; B3, and when the features of the linear image data subset N1 do not match the features of the linear image data subset N2, the linear image data subset N1 is merged with the linear image data subset N2 to form a new linear image data subset N 1+2 , then the linear image data subset N 1+2 The linear image data subset N of the subsequent equal time nodes 3+4 The feature comparison is repeated again until the features of the linear image data subsets at two equal time nodes are consistent. The frequency domain regularity feature of the vibration data is determined to be marked as L. If the same feature cannot be matched when the linear image data subsets are merged to half of the number, it is directly determined that the product has a fault problem. The feature matching operation for the linear image data subset N1 and the linear image data subset N2 in B3 is: b31, firstly, the linear features in the images of the linear image data subset N1 and the linear image data subset N2 are extracted by grayscale processing; b32, then extract the node data of the linear feature parts of the linear image data subset N1 and the linear image data subset N2, and align the start node on the linear image data subset N1 with the start node on the linear image data subset N2, and compare whether the curves of the two reaching their corresponding end nodes overlap; b33, if they overlap, the linear image data subset N1 is the frequency domain regular feature of the vibration data; if they do not overlap, continue the operation of B3 until the frequency domain regular feature L of the vibration data is obtained or the linear image data subsets cannot match the same feature when they are merged to half the number; The expression of the frequency domain regularity characteristic L of the vibration data is: Among them, F(L) represents the generation of the frequency domain regularity feature L of the vibration data, represents the frequency domain regularity feature L of vibration data generated after matching the features of linear image data subsets at two adjacent and equal time nodes, and N1=N2 represents that the linear image features of linear image data subset N1 are the same as those of linear image data subset N2, and so on. It means that the same linear image feature cannot be matched, that is, the entire linear image data is the frequency domain regular feature L, that is, the result data is located in this way, that is, the product has a fault problem.

2. The robot reducer fault detection system based on big data analysis according to claim 1 is characterized in that: The data processing and analysis module processes the frequency domain regularity feature L of the vibration data as follows: f(L) = maxF(L); Where f(L) is the frequency value calculation of the frequency domain regular feature L of the vibration data, and maxF(L) is the maximum frequency value of the frequency domain regular feature L in the vibration data; The obtained maximum frequency value is compared with the frequency value threshold of the product. If the maximum frequency value is greater than or equal to the frequency value threshold set for the product, it is determined that the product has a fault problem.

3. The robot reducer fault detection system based on big data analysis according to claim 1 is characterized in that: The data processing and analysis module processes the changed image data generated by the rotation as follows: C1. First, the video data collected by the camera is processed into frames to obtain a frame image data set generated by the rotation of the output shaft of the reducer, and it is marked as P; C2, graying the image data in the frame number image data set P, and overlaying the end face image before the rotation starts with the image in the frame number image data set P; C3. Calculate the difference result caused by the fault according to the image result generated by the overlay, mark it as Q, and generate a data compensation instruction through the difference result Q.

4. The robot reducer fault detection system based on big data analysis according to claim 3 is characterized in that: The frame number processing operation of the video data in C1 is as follows: c11, performing a video segmentation operation on the video data image in a manner of one frame per second, and classifying the image into a frame number image data set P in sequence; c12, then filter the image data in the frame image data set P, align the four corner points of adjacent images, and make the background of the upper layer image data transparent, leaving the feature part to compare with the feature part of the lower layer; c13. If the characteristic part of the upper layer image data completely overlaps with the characteristic part of the lower layer, the upper layer image data is filtered out. Conversely, if the characteristic part of the upper layer image data differs from the characteristic part of the lower layer, the image data of both the upper layer and the lower layer are retained in the frame number image data set P.

5. The robot reducer fault detection system based on big data analysis according to claim 3 is characterized in that: The image overlay operation of the end face image in C2 and the frame image dataset P is: c21, at this time, the end face image before the rotation starts is used as the basic feature, and the data subsets in the frame number image data set P are sequentially compared with the end face image; c22, still keeping the four corner points of the end face image and the frame number image data subset aligned, while making the image data of the data subset in the frame number image data set P transparent, and judging whether there is a difference between the end face image and the frame number image data subset through the two characteristic edge boundaries after covering; c23. If the two feature edge boundaries after covering overlap, filter the subset of the frame number image data; if the two feature edge boundaries after covering differ, retain and process the calculated difference value.

6. The robot reducer fault detection system based on big data analysis according to claim 3 is characterized in that: The calculation steps for the difference result Q in C3 are: c31, connect the nodes corresponding to the intersection of the feature boundaries after the image is covered in c22 operation to form a connecting line segment, set a segmentation line at 90 degrees from the center point of the connecting line segment, and mark it as S, and the segmentation line S extends to the intersection of the two feature edge boundaries as the endpoint; c32. Based on the diameter data of the characteristic part, the center points of the two characteristic parts are obtained by extending from the two end points of the dividing line S into the characteristic part, and the deviation data caused by the vibration is obtained by connecting the center points; c33. The difference result Q is obtained by calculating the deviation data, and the difference result Q is compared with the deviation threshold of the product vibration setting extracted from the database. The deviation data corresponding to the difference result Q greater than the product vibration setting deviation threshold is the fault data, and the calculation formula of the difference result Q is: Q = (s-2r)*T; Where s is the length between the two ends of the segmentation line S located in the image data and intersecting the outside of the feature edge boundary, r is the radius length of the end face feature in the output shaft end face image, and T is the ratio between the actual product and the image.

Images