CN118379564A - Abrasion stage division method and system based on data mechanism fusion - Google Patents
Abrasion stage division method and system based on data mechanism fusion Download PDFInfo
- Publication number
- CN118379564A CN118379564A CN202410814654.1A CN202410814654A CN118379564A CN 118379564 A CN118379564 A CN 118379564A CN 202410814654 A CN202410814654 A CN 202410814654A CN 118379564 A CN118379564 A CN 118379564A
- Authority
- CN
- China
- Prior art keywords
- wear
- abrasive particle
- abrasion
- random variable
- image
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 81
- 238000005299 abrasion Methods 0.000 title claims abstract description 69
- 230000007246 mechanism Effects 0.000 title claims abstract description 39
- 230000004927 fusion Effects 0.000 title claims abstract description 29
- 208000035874 Excoriation Diseases 0.000 title abstract description 55
- 239000002245 particle Substances 0.000 claims abstract description 184
- 230000002159 abnormal effect Effects 0.000 claims abstract description 31
- 238000007781 pre-processing Methods 0.000 claims abstract description 25
- 238000012545 processing Methods 0.000 claims description 13
- 230000036541 health Effects 0.000 claims description 11
- 238000011426 transformation method Methods 0.000 claims description 10
- 239000006061 abrasive grain Substances 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 8
- 238000000638 solvent extraction Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 abstract description 14
- 230000008569 process Effects 0.000 abstract description 14
- 238000004458 analytical method Methods 0.000 abstract description 9
- 238000012423 maintenance Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000003449 preventive effect Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 239000010687 lubricating oil Substances 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003921 oil Substances 0.000 description 1
- 238000000399 optical microscopy Methods 0.000 description 1
- 238000003921 particle size analysis Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Landscapes
- Length Measuring Devices By Optical Means (AREA)
Abstract
The invention provides a wear phase dividing method and system based on data mechanism fusion, which belong to the technical field of mechanical equipment wear state monitoring, and are used for acquiring an abrasive particle ring image and preprocessing the abrasive particle ring image; estimating the distribution of image data indexes of the abrasive particle ring by using the maximum entropy and the nuclear density, and dividing the abrasion stage by combining a threshold value; preprocessing the scattered abrasive particle image, and filling abrasive particles; and numbering each abrasive particle after filling, calculating the roundness and elongation of the abrasive particles, and judging the abnormal wear state according to the threshold value of each characteristic parameter of the abrasive particles corresponding to the abnormal wear state. According to the method, analysis of the abrasion mechanism is fused in the time domain, normal abrasive particles and abnormal abrasive particles are distinguished by using roundness indexes of single abrasive particles, comprehensive description of the whole abrasion state evolution process is realized, and accurate and rapid division of the abrasion stage is realized.
Description
Technical Field
The invention belongs to the technical field of wear state monitoring of mechanical equipment, and particularly relates to a wear stage dividing method and system based on data mechanism fusion.
Background
The wear phase of the present mechanical equipment can be generally divided into the following three phases, namely, a running-in phase (running-in phase), a stable wear phase and a severe wear phase. On-line monitoring is generally required for the wear phase.
Existing online wear particle (hereinafter referred to as "abrasive particle") monitoring systems often provide limited information on the shape and type of abrasive particles based on the concentration of abrasive particles. For the identification of the wear state, the shape, size, surface morphology and other characteristics of the abrasive particles are all important factors for root cause analysis, and the lack of wear mechanism analysis can affect the judgment of the abnormal wear state. In addition, the traditional phase dividing method, namely the three-line method, is only suitable for simple abrasive particle data, has certain limitations, and cannot accurately divide the abrasion phase when complex abrasive particle data are processed.
The division of the wear phases of the mechanical equipment is very important for preventing malfunctions, improving the reliability of the equipment and prolonging the life of the equipment. By timely and accurately identifying the wear state of the mechanical equipment, corresponding maintenance measures can be taken to avoid equipment damage or increased downtime caused by failure to timely discover wear. Oil detection is one of the main methods for monitoring the state of wear, and wear particles directly generated by friction pairs contain valuable wear information, so that lubricating oil carrying the wear particles can be used for monitoring the state of a machine. By means of a data analysis technology, whether abnormal wear or other problems exist in the equipment is identified, the wear state of mechanical equipment is monitored and identified in real time, predictive maintenance can be achieved, and therefore downtime and maintenance cost are reduced, and reliability and production efficiency of the equipment are improved.
The concentration of wear particles or the equivalent circular diameter of wear particles in combination with the identification of dynamic full life wear conditions is currently used mostly. By monitoring the change of the particle concentration by means of sensor monitoring, optical microscopy or using laser particle size analysis, etc., it is possible to identify whether the equipment has wear problems and the extent of wear. The equivalent circle diameter means a diameter of a circle having an area equal to that of the wear particle calculated as the equivalent diameter of the wear particle. By monitoring the change of the equivalent circle diameter of the particles, the abrasion condition of the equipment can be known, and the larger the equivalent circle diameter is, the more serious the abrasion degree of the equipment is.
However, existing wear staging methods are often based on monitoring data such as wear particle concentration and size, lack of analysis of wear mechanism, resulting in wear staging with some hysteresis and difficulty in tracing. In addition, the "three-wire method" currently used for dividing the wear state is still based on the assumption that the wear monitoring data is gaussian distributed, and cannot cope with complex wear particle data.
Disclosure of Invention
The invention provides a wear phase division method based on data mechanism fusion, which solves the problems that the traditional three-wire method cannot cope with complex data and the wear phase division is inaccurate, and realizes the accurate and quick division of the wear phase.
The method comprises the following steps:
Step S101: acquiring an abrasive particle ring image, and preprocessing the abrasive particle ring image;
step S102: estimating distribution of image data indexes of the abrasive particle ring by using maximum entropy and nuclear density, calculating a threshold value, and dividing a wearing stage;
Wherein, step S1021: selecting a bounded support domain;
Step S1022: normalizing the random variable;
step S1023: determining first M generalized moments of the random variable;
step S1024: approximating a probability density function of the random variable using the kernel density function;
Step S1025: defining a discrete random variable, estimating probability distribution of the discrete random variable according to the maximum entropy principle, and ensuring that a probability density function is obtained under a given constraint condition;
Step S1026: selecting an optimal kernel density function, and approximating a target probability density function by a kernel density estimation method to maximize the maximum entropy of the kernel density;
Step S1027: according to a coordinate transformation method, obtaining a transformed functional relation based on a random variable Z and an original random variable X, converting the random variable Z back to the original random variable X through the functional relation to obtain a maximum entropy approximation of kernel density, and converting a probability density function back to an original domain;
Step S1028: dividing the wear phase based on the wear characteristics of the abrasive particles;
step S103: preprocessing the scattered abrasive particle image, and filling abrasive particles; and numbering each abrasive particle after filling, calculating the roundness and elongation of the abrasive particles, and judging the abnormal wear state according to the threshold value of each characteristic parameter of the abrasive particles corresponding to the abnormal wear state.
It should be further noted that step S101 further includes:
Step S1011: graying treatment is carried out on a background image and an abrasive particle image of the abrasive particle ring image;
Step S1012: performing background reduction treatment on the abrasive particle gray level graph;
step S1013: solving an optimal threshold value by an iterative threshold segmentation method;
Step S1014: performing binarization processing on the abrasive particle image with the background image removed;
step S1015: calculating the total gray value TGSV of the abrasive particles;
step S1016: and carrying out noise reduction processing on TGSV data based on a wavelet transformation method.
It should be further noted that, in step S1015, the formula for calculating the total gray value TGSV of the abrasive particles is as follows:
Wherein, The denominator is the area of the annular field of view in the image of the abrasive particles, and the total gray value is used to describe the distribution density of the abrasive particles in the annular field of view of the sensor.
It should be further noted that the expression of the one-dimensional noisy signal in step S1016 is as follows:
Wherein, In the case of a signal containing a noise,Is a low-frequency real signal which is a real signal,In the case of a high-frequency noise signal,Is a coefficient;
reconstructing the signal using an inverse wavelet transform to obtain At this time, consider:
The noise-reduced signal is obtained.
It should be further noted that step S1021 further includes: defining a bounded support domainSo thatLimiting the value range of the random variable X;
The method of normalizing the random variable in step S1022 is to define a new random variable Z, introduce coordinate transformation, and let Then；
In step S1023, assume M moment functions asWhereinThen the ith generalized momentExpressed as:
Wherein, As a function of the moment of which it is desired,Is a probability distribution of Z.
It should be further noted that in step S1024, a data point is givenA kernel function is built centered on the data point to estimate the density of the point in the vicinity, expressed as:
Wherein, Is the data point in the sample and,Is a bandwidth parameter;
the maximum entropy function is maximized in step S1025 as:
The constraint conditions are as follows:
Wherein, As discrete random variablesIs a function of the probability distribution of (1),It is possible to obtain a value that is,Is a given generalized moment.
In step S1028, it is further noted thatProbability density function within an interval；
The defined threshold is:
: s1 and S2 boundary points;
: s2 and S3 demarcation points;
: s3 and S4 demarcation points;
and dividing the abrasion health condition into four stages according to the obtained abrasion state evaluation threshold value.
In step S103, the roundness of the abrasive grains is calculated based on the following expression:
Wherein, Is the coverage area of a single abrasive particle, L is the maximum length of the abrasive particle.
It should be further noted that, in step S103, the elongation is defined as a ratio of a maximum Feret diameter to a minimum Feret diameter of the projected profile of the wear particle, where the maximum Feret diameter and the minimum Feret diameter are respectively defined as a maximum value and a minimum value of a distance between two parallel lines tangent to the lobe on the projected profile of the particle; the elongation SL is expressed as:
。
the application also provides a wearing phase dividing system based on data mechanism fusion, which comprises: the device comprises an image preprocessing module, an abrasive particle abrasion dividing module and an abrasion state judging module;
the image preprocessing module is used for preprocessing the abrasive particle ring image and the scattered abrasive particle image;
The abrasive particle abrasion dividing module is used for estimating distribution of image data indexes of the abrasive particle ring by utilizing maximum entropy and nuclear density, and dividing abrasion stages by combining a threshold value;
The abrasive particle abrasion dividing module is also used for selecting a bounded supporting domain; establishing random variable normalization; determining first M generalized moments of the random variable; approximating a probability density function of the random variable using the kernel density function; defining a discrete random variable, estimating probability distribution of the discrete random variable according to the maximum entropy principle, and ensuring that the most unbiased probability density function estimation is obtained under a given constraint condition; selecting an optimal kernel density function, and approximating a target probability density function by a kernel density estimation method to maximize the maximum entropy of the kernel density; according to a coordinate transformation method, obtaining a transformed functional relation based on a random variable Z and an original random variable X, converting the random variable Z back to the original random variable X through the functional relation to obtain a maximum entropy approximation of kernel density, and converting a probability density function back to an original domain; dividing the wear phase based on the particle wear characteristics;
The abrasion state judging module is used for preprocessing the scattered abrasive particle image and filling abrasive particles; and numbering each abrasive particle after filling, calculating the roundness and elongation of the abrasive particles, and judging the abnormal wear state according to the threshold value of each characteristic parameter of the abrasive particles corresponding to the abnormal wear state.
From the above technical scheme, the invention has the following advantages:
The abrasion phase division method and system based on data mechanism fusion solve the problems that the traditional three-wire method cannot cope with complex data and the abrasion phase division is inaccurate, and realize accurate and rapid division of the abrasion phase. The application can preprocess the abrasive particle ring image and the scattered abrasive particle image, can eliminate noise in the image, enhance image details and enable the subsequent feature extraction and recognition to be more accurate.
The application can accurately estimate the distribution of the image data indexes of the abrasive particle ring by using the maximum entropy and the nuclear density estimation method, thereby dividing the abrasion stage more accurately. This division is important for predicting the remaining life of the equipment and for planning maintenance.
The system effectively identifies the abnormal wear state, and can accurately judge the abnormal wear state of the equipment by numbering the filled abrasive particles and calculating the roundness, elongation and other characteristic parameters of the abrasive particles and combining the abrasive particle characteristic parameter threshold corresponding to the abnormal wear state. This helps to discover potential problems with the device in time, avoiding failure of the device due to excessive wear. The application also can greatly improve the efficiency and accuracy of abrasion detection by automatically processing and analyzing the abrasive particle images. By monitoring and analyzing the abrasive particle images in real time, the wear tendency and potential problems of the equipment can be found in advance, thereby implementing preventive maintenance measures. This helps to reduce the failure rate of the device, improve the reliability and lifetime of the device.
Drawings
In order to more clearly illustrate the technical solutions of the present invention, the drawings that are needed in the description will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of abrasive particle ring image preprocessing and data preprocessing;
FIG. 2 is a flow chart of a wear phase partitioning method based on maximum entropy and kernel density estimation;
FIG. 3 is a graph of the wear phase partitioning result based on the maximum entropy and kernel density estimation method;
FIG. 4 is a schematic view of an image of the abrasive particle ring at 6520 min;
fig. 5 is an enlarged view of a portion a of fig. 4;
FIG. 6 is a wear mechanism identification flow chart;
FIG. 7 is a pre-warning chart of gearbox wear failure.
Detailed Description
According to the abrasion stage division method based on data mechanism fusion, which is provided by the invention, the total gray value TGSV (the ratio of the coverage area of all abrasive particles in an image to the area of a view field) is used as a data index, and the wavelet transformation method is adopted to carry out noise reduction treatment on TGSV data. And estimating the probability distribution of the whole data by using a maximum entropy method, and further improving the accuracy of distribution estimation by using a kernel density estimation method. On this basis, the threshold is determined using a three-wire method, divided into four wear phases. In order to realize the judgment of the abnormal abrasion state, the analysis of the abrasion mechanism is fused in the time domain, and the roundness index Ci of the single abrasive particle is utilized to distinguish the normal abrasive particle from the abnormal abrasive particle, so that the comprehensive description of the whole abrasion state evolution process is realized. Thus, the wear rule of the mechanical equipment in the use process can be better analyzed, and corresponding measures are taken to prolong the service life and improve the safety of the mechanical equipment.
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to the drawings, which are shown in the flowchart of a wear-stage classification method based on data mechanism fusion in an embodiment, the method includes:
step S101: and acquiring an abrasive particle ring image, and preprocessing the abrasive particle ring image.
In one exemplary embodiment, after the wear particles are imaged using the abrasive particle sensor, an abrasive particle ring image is formed, an RGB image of the abrasive particle ring image is preprocessed, the abrasive particle ring image is converted into a binary image, gray value data is extracted, and noise reduction processing is performed on the gray value data. The overall flowchart of the image preprocessing and the data preprocessing of the abrasive particle ring is shown in fig. 1.
It can be seen that by observing the image of the abrasive particle ring, the present embodiment can analyze information such as distribution characteristics, size, shape, color, etc. of the abrasive particles, thereby judging the wear state and failure type of the mechanical device. The implementation can apply the image processing technology and the artificial intelligence technology to the analysis of the abrasive particle ring image, and improve the accuracy and the efficiency of the analysis. For example, the image of the abrasive particles may be preprocessed using digital image processing techniques, including smoothing, filtering, thresholding, etc., to effectively reduce noise in the image of the abrasive particles, and to separate the abrasive particles from the background of the image, which may provide more information describing the characteristics of the abrasive particles.
In one embodiment of the present invention, based on step S101, a possible embodiment thereof will be given below for non-limiting illustration.
Step S1011: and carrying out gray-scale treatment on the background image and the abrasive particle image of the abrasive particle ring image.
Graying of the present embodiment is a process of converting a color image into a gray image. The graying process is realized by converting the background image of the abrasive particle ring image and the RGB value of each pixel point of the abrasive particle image into a gray value. The gray value is typically an integer between 0 and 255, 0 representing black and 255 representing white.
Step S1012: and performing background reduction treatment on the abrasive particle gray scale image.
The background process here is to remove background information from the gray scale of the abrasive particles to better extract the abrasive particles. By subtracting a background gray-scale image from the original gray-scale image. The background gray-scale map may be obtained by various methods, for example, taking an image as a background when the apparatus is not in operation, or using image processing techniques to estimate and construct a background image.
Step S1013: and (5) obtaining an optimal threshold value through an iterative threshold segmentation method.
The iterative thresholding method of the present embodiment is a method of automatically determining an image binarization threshold. When the method is carried out, an initial threshold value is selected, the abrasive particle ring image is divided into a foreground part and a background part according to the threshold value, the gray average value of pixels of the two parts is calculated, and the intermediate value of the two average values is used as a new threshold value. This process is repeated until the difference between the new threshold and the last threshold is less than a preset threshold. The threshold thus obtained is the optimal threshold for the subsequent image binarization processing.
Step S1014: and carrying out binarization processing on the abrasive particle image with the background image removed.
In the binarization process of the abrasive grain image, pixels with gray values higher than the optimal threshold value are usually set to be white (or 1) to represent abrasive grains; the pixel with gray value below the optimal threshold is set to black (or 0) to represent the background.
Step S1015: the total gray value TGSV of the abrasive particles is calculated.
Wherein, the calculation formula of the total gray value TGSV of the abrasive particles is as follows:
the total gray value is used for describing the distribution density of the abrasive particles in the annular view field of the sensor, and is a main data index for judging the subsequent abrasion stage.
Step S1016: and carrying out noise reduction processing on TGSV data based on a wavelet transformation method.
It should be noted that, the basic expression of the one-dimensional noisy signal is as follows:
Wherein, In the case of a signal containing a noise,Is a low-frequency real signal which is a real signal,In the case of a high-frequency noise signal,Is a coefficient.
Selecting proper wavelet base, forPerforming n-layer wavelet decomposition until noise signal contained in low frequency part is small enough to be considered to contain only real signal component, and reconstructing signal by wavelet inverse transformation to obtainAt this time, it can be considered that:
The noise-reduced signal is obtained.
Step S102: and estimating the distribution of the image data indexes of the abrasive particle ring by using the maximum entropy and the nuclear density, calculating a threshold value, and dividing the abrasion stage.
In the implementation, the abrasion monitoring data are assumed to be in Gaussian distribution by combining a traditional three-wire method, and three thresholds are set to divide the abrasion process into four stages. The method utilizes maximum entropy and kernel density representation to accurately estimate the distribution of data indexes, and on the basis, a threshold value is determined again, so that the abrasion stage is divided. A flowchart of a wear-phase partitioning method based on maximum entropy and kernel density estimation is shown in fig. 2.
Step S1021: bounded support domains are selected.
The present embodiment defines a bounded support domainSo thatLimiting the range of values of the random variable X.
Step S1022: and (5) normalizing the random variable.
It should be noted that, defining new random variable Z, introducing coordinate transformation to makeThenFacilitating further analysis.
Step S1023: the first M generalized moments of the random variable are determined.
According to an embodiment of the application, assume M moment functions asWhereinThen the ith generalized momentCan be expressed as:
Wherein, As a function of the moment of which it is desired,Is a probability distribution of Z.
Step S1024: the probability density function of the random variable is approximated using a kernel density function.
In this embodiment, a Kernel Density Function (KDF) is used to approximate the Probability Density Function (PDF) of the random variable Z. Here given a data pointThe kernel function is built centered on the data point to estimate the density of the point in the vicinity, which can be expressed as:
Wherein, Is the data point in the sample and,Is a bandwidth parameter. For controlling the smoothness of the estimation.
Step S1025: and defining a discrete random variable, estimating probability distribution of the discrete random variable according to the maximum entropy principle, and ensuring that the most unbiased probability density function estimation is obtained under a given constraint condition.
Specifically, the present embodiment defines a discrete random variableThe probability distribution of the probability distribution is estimated according to the principle of Maximum Entropy (ME), so that the most unbiased probability density function estimation is obtained under the given constraint condition. It is desirable to maximize the maximum entropy function:
The constraint conditions are as follows:
Wherein, As discrete random variablesIs a function of the probability distribution of (1),It is possible to obtain a value that is,Is a given generalized moment.
Step S1026: and selecting an optimal kernel density function, and approximating the target probability density function by a kernel density estimation method so as to maximize the maximum entropy of the kernel density.
In this embodiment, an optimal kernel density function is selected, and the kernel density estimation method approximates the target probability density functionSo that the maximum entropy of the kernel densityMaximization.
In the embodiment, a cross verification method is adopted to select proper kernel functions and bandwidth parameters, a data set is divided into a training set and a verification set, kernel density estimation is carried out on each kernel density function and bandwidth parameter combination, and errors on the verification set are calculated. The combination of kernel density function and bandwidth parameter is selected that minimizes the error.
Step S1027: according to the coordinate transformation method, a transformed functional relation based on the random variable Z and the original random variable X is obtained, the random variable Z is converted back to the original random variable X through the functional relation, the maximum entropy approximation of the kernel density is obtained, and the probability density function is converted back to the original domain.
Step S1028: the wear phase is divided based on the abrasive wear characteristics.
The present embodiment makes the following based on the Probability Density Function (PDF) obtained in step S1025Probability density function within an interval。
As shown in fig. 3, the wear state evaluation threshold is defined as:
: s1 and S2 boundary points;
: s2 and S3 demarcation points;
: s3 and S4 demarcation points.
Dividing the abrasion health condition into four stages according to the abrasion state evaluation threshold value; respectively correspond to the abrasion health condition area S1, the abrasion health condition area S2, the abrasion health condition area S3 and the abrasion health condition area S4.
Step S103: preprocessing the scattered abrasive particle image, and filling abrasive particles; and numbering each abrasive particle after filling, calculating the roundness and elongation of the abrasive particles, and judging the abnormal wear state according to the threshold value of each characteristic parameter of the abrasive particles corresponding to the abnormal wear state.
In an exemplary embodiment, an image of abrasive particles imaged by the abrasive particle sensor at 7060min may be selected, as shown in fig. 4 and 5, in which large particles are already apparent, but the overall TSGV index does not reach the early warning value, which may cause missed judgment in some abnormal states. Thus, a fusion mechanism is needed to further analyze the characteristics of individual abrasive particles. The wear mechanism identification flow is shown in fig. 6.
The method comprises the following specific steps:
step S1031: extracting preprocessing and characteristic parameters of the abrasive particle images; the specific process is shown in fig. 1.
Step S1032: and filling the abrasive particles.
In this embodiment, hole filling is performed on the initially segmented binary image to complement the image blank particle pixels due to specular reflection on the smooth abrasive particle surface.
Step S1033: each abrasive grain after the filling treatment is numbered, and is distinguished and counted.
Step S1034: the roundness (Ci) and elongation (SL) of the abrasive grains were calculated.
Roundness (Ci) here is a parameter describing how close a particle shape is to a perfect circle. The roundness ranges from 0 to 1, where a value of 1 indicates that the abrasive grain shape is a perfect circle. The calculation expression of the roundness is:
Wherein, Is the coverage area of a single abrasive particle, L is the maximum length of the abrasive particle.
Elongation (SL), defined as the ratio of the maximum Feret diameter to the minimum Feret diameter of the projected profile of the wear particle, can be used to reflect the elongation of the particle. Wherein the maximum Feret diameter and the minimum Feret diameter are defined as the maximum and minimum, respectively, of the distance between two parallel lines (which cannot pass through the particle) tangential to the lobe on the projection profile of the particle. The closer SL is to 1, the closer the particles are to perfect circles. The elongation is calculated as:
Step S1035: and judging the abnormal wear state according to the threshold value of each characteristic parameter of the abrasive particles in the abnormal wear state.
The abnormal wear state abrasive grain number characteristic parameter thresholds are shown in table 1:
table 1 characteristic parameter thresholds for three types of typical abrasive particles
It can be seen that by combining the wear mechanism with the change in the number of abrasive particles shown in fig. 3 to characterize the evolution of the wear phase during operation of the gearbox, a full description of the overall process of the wear state of the gearbox is achieved, as shown in fig. 7. The method not only monitors and describes the abrasion process based on the change of the abrasive particle quantity, but also considers the specific content of the abrasion mechanism, and provides a richer and comprehensive view for the abrasion state assessment. Compared with the method which only relies on concentration (TGSV) for prediction, the method based on the comprehensive mechanism can perform early warning at least 500 minutes in advance.
It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic, and should not limit the implementation process of the embodiment of the present invention.
The following is an embodiment of a wear-leveling system based on data mechanism fusion provided by the embodiment of the present disclosure, where the system belongs to the same inventive concept as the wear-leveling method based on data mechanism fusion of the above embodiments, and details of the wear-leveling system based on data mechanism fusion, which are not described in detail in the embodiment of the wear-leveling system based on data mechanism fusion, may refer to the embodiment of the wear-leveling method based on data mechanism fusion.
The system comprises: the device comprises an image preprocessing module, an abrasive particle abrasion dividing module and an abrasion state judging module.
The image preprocessing module is used for preprocessing the abrasive particle ring image and the scattered abrasive particle image.
The abrasive particle abrasion dividing module is used for estimating distribution of image data indexes of the abrasive particle ring by using maximum entropy and nuclear density and dividing abrasion stages by combining a threshold value.
The abrasive particle abrasion dividing module is also used for selecting a bounded supporting domain; establishing random variable normalization; determining first M generalized moments of the random variable; approximating a probability density function of the random variable using the kernel density function; defining a discrete random variable, estimating probability distribution of the discrete random variable according to the maximum entropy principle, and ensuring that the most unbiased probability density function estimation is obtained under a given constraint condition; selecting an optimal kernel density function, and approximating a target probability density function by a kernel density estimation method to maximize the maximum entropy of the kernel density; according to a coordinate transformation method, obtaining a transformed functional relation based on a random variable Z and an original random variable X, converting the random variable Z back to the original random variable X through the functional relation to obtain a maximum entropy approximation of kernel density, and converting a probability density function back to an original domain; the wear stages are divided based on the particle wear characteristics.
The abrasion state judging module is used for preprocessing the scattered abrasive particle image and filling abrasive particles; and numbering each abrasive particle after filling, calculating the roundness and elongation of the abrasive particles, and judging the abnormal wear state according to the threshold value of each characteristic parameter of the abrasive particles corresponding to the abnormal wear state.
The abrasion phase division system based on data mechanism fusion adopts an improved probability density function estimation method, solves the problems that the traditional three-wire method cannot cope with complex data and causes inaccurate abrasion phase division, and realizes accurate and rapid division of the abrasion phase. Furthermore, the application incorporates analysis of the wear mechanism. The roundness and the elongation are used for representing the shape of a single abrasion particle so as to distinguish normal abrasive particles from abnormal abrasive particles, so that the judgment of abnormal abrasion states is realized, and the judgment precision of the abrasion states is ensured.
The abrasion phase dividing system based on data mechanism fusion can preprocess the abrasive particle ring image and the scattered abrasive particle image, can eliminate noise in the image, enhance image details and enable subsequent feature extraction and identification to be more accurate. The application can accurately estimate the distribution of the image data indexes of the abrasive particle ring by using the maximum entropy and the nuclear density estimation method, thereby dividing the abrasion stage more accurately. This division is important for predicting the remaining life of the equipment and for planning maintenance. The system effectively identifies the abnormal wear state, and can accurately judge the abnormal wear state of the equipment by numbering the filled abrasive particles and calculating the roundness, elongation and other characteristic parameters of the abrasive particles and combining the abrasive particle characteristic parameter threshold corresponding to the abnormal wear state. This helps to discover potential problems with the device in time, avoiding failure of the device due to excessive wear. The application also can greatly improve the efficiency and accuracy of abrasion detection by automatically processing and analyzing the abrasive particle images. By monitoring and analyzing the abrasive particle images in real time, the wear tendency and potential problems of the equipment can be found in advance, thereby implementing preventive maintenance measures. This helps to reduce the failure rate of the device, improve the reliability and lifetime of the device.
The wear-phase partitioning system based on data mechanism fusion, to which the present application relates, is a unit and algorithm step of each example described in connection with the embodiments disclosed herein, can be implemented in electronic hardware, computer software, or a combination of both, and to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
Those skilled in the art will appreciate that the various aspects of the wear phase partitioning method based on data mechanism fusion to which the present application relates may be implemented as a system, method or program product. Accordingly, various aspects of the disclosure may be embodied in the following forms, namely: an entirely hardware embodiment, an entirely software embodiment (including firmware, micro-code, etc.) or an embodiment combining hardware and software aspects may be referred to herein as a "circuit," module "or" system.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. The abrasion phase division method based on data mechanism fusion is characterized by comprising the following steps:
Step S101: acquiring an abrasive particle ring image, and preprocessing the abrasive particle ring image;
step S102: estimating distribution of image data indexes of the abrasive particle ring by using maximum entropy and nuclear density, calculating a threshold value, and dividing a wearing stage;
Wherein, step S1021: selecting a bounded support domain;
Step S1022: normalizing the random variable;
step S1023: determining first M generalized moments of the random variable;
step S1024: approximating a probability density function of the random variable using the kernel density function;
Step S1025: defining a discrete random variable, estimating probability distribution of the discrete random variable according to the maximum entropy principle, and ensuring that a probability density function is obtained under a given constraint condition;
Step S1026: selecting an optimal kernel density function, and approximating a target probability density function by a kernel density estimation method to maximize the maximum entropy of the kernel density;
Step S1027: according to a coordinate transformation method, obtaining a transformed functional relation based on a random variable Z and an original random variable X, converting the random variable Z back to the original random variable X through the functional relation to obtain a maximum entropy approximation of kernel density, and converting a probability density function back to an original domain;
Step S1028: dividing the wear phase based on the wear characteristics of the abrasive particles;
step S103: preprocessing the scattered abrasive particle image, and filling abrasive particles; and numbering each abrasive particle after filling, calculating the roundness and elongation of the abrasive particles, and judging the abnormal wear state according to the threshold value of each characteristic parameter of the abrasive particles corresponding to the abnormal wear state.
2. The method of wear-phase classification based on data mechanism fusion according to claim 1, wherein step S101 further comprises:
Step S1011: graying treatment is carried out on a background image and an abrasive particle image of the abrasive particle ring image;
Step S1012: performing background reduction treatment on the abrasive particle gray level graph;
step S1013: solving an optimal threshold value by an iterative threshold segmentation method;
Step S1014: performing binarization processing on the abrasive particle image with the background image removed;
step S1015: calculating the total gray value TGSV of the abrasive particles;
step S1016: and carrying out noise reduction processing on TGSV data based on a wavelet transformation method.
3. The method for dividing the wear phase based on the data mechanism fusion according to claim 2, wherein the formula for calculating the total gray value TGSV of the abrasive particles in step S1015 is as follows:
Wherein, The denominator is the area of the annular field of view in the image of the abrasive particles, and the total gray value is used to describe the distribution density of the abrasive particles in the annular field of view of the sensor.
4. The method for partitioning wear stages based on data mechanism fusion according to claim 2, wherein the expression of the one-dimensional noisy signal in step S1016 is as follows:
Wherein, In the case of a signal containing a noise,Is a low-frequency real signal which is a real signal,In the case of a high-frequency noise signal,Is a coefficient;
reconstructing the signal using an inverse wavelet transform to obtain At this time, consider:
The noise-reduced signal is obtained.
5. The method of wear staging based on data mechanism fusion according to claim 1 or 2, characterized in that step S1021 further comprises: defining a bounded support domainSo thatLimiting the value range of the random variable X;
The method of normalizing the random variable in step S1022 is to define a new random variable Z, introduce coordinate transformation, and let Then；
In step S1023, assume M moment functions asWhereinThen the ith generalized momentExpressed as:
Wherein, As a function of the moment of which it is desired,Is a probability distribution of Z.
6. The method of wear staging based on data mechanism fusion according to claim 1 or 2, characterized in that in step S1024 a data point is givenA kernel function is built centered on the data point to estimate the density of the point in the vicinity, expressed as:
Wherein, Is the data point in the sample and,Is a bandwidth parameter;
the maximum entropy function is maximized in step S1025 as:
The constraint conditions are as follows:
Wherein, As discrete random variablesIs a function of the probability distribution of (1),It is possible to obtain a value that is,Is a given generalized moment.
7. The method of wear staging based on data mechanism fusion according to claim 1 or 2, characterized in that in step S1028, definingProbability density function within an interval；
The defined wear state evaluation threshold is:
: s1 and S2 boundary points;
: s2 and S3 demarcation points;
: s3 and S4 demarcation points;
Dividing the abrasion health condition into four stages according to the abrasion state evaluation threshold value; respectively correspond to the abrasion health condition area S1, the abrasion health condition area S2, the abrasion health condition area S3 and the abrasion health condition area S4.
8. The abrasion phase division method based on data mechanism fusion according to claim 1 or 2, wherein in step S103, the roundness of abrasive grains is calculated based on the following expression:
Wherein, Is the coverage area of a single abrasive particle, L is the maximum length of the abrasive particle.
9. The method of claim 8, wherein in step S103, the elongation is defined as a ratio of a maximum Feret diameter to a minimum Feret diameter of the projected profile of the wear particle, wherein the maximum Feret diameter and the minimum Feret diameter are defined as a maximum value and a minimum value, respectively, of a distance between two parallel lines tangential to a lobe on the projected profile of the particle; the elongation SL is expressed as:
。
10. A data mechanism fusion-based wear staging system, characterized in that the system is adapted to implement the steps of the data mechanism fusion-based wear staging method as claimed in any one of claims 1 to 9;
The system comprises: the device comprises an image preprocessing module, an abrasive particle abrasion dividing module and an abrasion state judging module;
the image preprocessing module is used for preprocessing the abrasive particle ring image and the scattered abrasive particle image;
The abrasive particle abrasion dividing module is used for estimating distribution of image data indexes of the abrasive particle ring by utilizing maximum entropy and nuclear density, and dividing abrasion stages by combining a threshold value;
The abrasive particle abrasion dividing module is also used for selecting a bounded supporting domain; establishing random variable normalization; determining first M generalized moments of the random variable; approximating a probability density function of the random variable using the kernel density function; defining a discrete random variable, estimating probability distribution of the discrete random variable according to the maximum entropy principle, and ensuring that the most unbiased probability density function estimation is obtained under a given constraint condition; selecting an optimal kernel density function, and approximating a target probability density function by a kernel density estimation method to maximize the maximum entropy of the kernel density; according to a coordinate transformation method, obtaining a transformed functional relation based on a random variable Z and an original random variable X, converting the random variable Z back to the original random variable X through the functional relation to obtain a maximum entropy approximation of kernel density, and converting a probability density function back to an original domain; dividing the wear phase based on the particle wear characteristics;
The abrasion state judging module is used for preprocessing the scattered abrasive particle image and filling abrasive particles; and numbering each abrasive particle after filling, calculating the roundness and elongation of the abrasive particles, and judging the abnormal wear state according to the threshold value of each characteristic parameter of the abrasive particles corresponding to the abnormal wear state.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410814654.1A CN118379564B (en) | 2024-06-24 | 2024-06-24 | Abrasion stage division method and system based on data mechanism fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410814654.1A CN118379564B (en) | 2024-06-24 | 2024-06-24 | Abrasion stage division method and system based on data mechanism fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118379564A true CN118379564A (en) | 2024-07-23 |
| CN118379564B CN118379564B (en) | 2024-09-17 |
Family
ID=91905805
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410814654.1A Active CN118379564B (en) | 2024-06-24 | 2024-06-24 | Abrasion stage division method and system based on data mechanism fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118379564B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0298593A2 (en) * | 1987-05-19 | 1989-01-11 | Kabushiki Kaisha Toshiba | Matrix material for bonding abrasive material, and method of manufacturing same |
| JPH10557A (en) * | 1996-06-12 | 1998-01-06 | Nippon Steel Corp | Abrasive grain projection evaluation method for grinding wheel |
| CN109030268A (en) * | 2018-09-07 | 2018-12-18 | 山东大学 | A kind of adjustable gas fixed double phase flow tester for testing scouring wear structure |
| CN110208124A (en) * | 2019-05-30 | 2019-09-06 | 新疆大学 | The development approach of mechanical wear system based on Abrasive Wear Mechanism |
| CN112381140A (en) * | 2020-11-13 | 2021-02-19 | 国家能源集团泰州发电有限公司 | Abrasive particle image machine learning identification method based on new characteristic parameters |
| CN115266498A (en) * | 2022-07-14 | 2022-11-01 | 山东大学 | High throughput imaging sensor, system and method for monitoring multi-grit features |
| CN115859087A (en) * | 2022-12-16 | 2023-03-28 | 重庆邮电大学 | Oil abrasive particle characteristic signal extraction method based on segmentation entropy |
| CN116228695A (en) * | 2023-02-16 | 2023-06-06 | 武汉理工大学 | Wearing state sensing system and method based on video image extraction technology |
-
2024
- 2024-06-24 CN CN202410814654.1A patent/CN118379564B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0298593A2 (en) * | 1987-05-19 | 1989-01-11 | Kabushiki Kaisha Toshiba | Matrix material for bonding abrasive material, and method of manufacturing same |
| JPH10557A (en) * | 1996-06-12 | 1998-01-06 | Nippon Steel Corp | Abrasive grain projection evaluation method for grinding wheel |
| CN109030268A (en) * | 2018-09-07 | 2018-12-18 | 山东大学 | A kind of adjustable gas fixed double phase flow tester for testing scouring wear structure |
| CN110208124A (en) * | 2019-05-30 | 2019-09-06 | 新疆大学 | The development approach of mechanical wear system based on Abrasive Wear Mechanism |
| CN112381140A (en) * | 2020-11-13 | 2021-02-19 | 国家能源集团泰州发电有限公司 | Abrasive particle image machine learning identification method based on new characteristic parameters |
| CN115266498A (en) * | 2022-07-14 | 2022-11-01 | 山东大学 | High throughput imaging sensor, system and method for monitoring multi-grit features |
| CN115859087A (en) * | 2022-12-16 | 2023-03-28 | 重庆邮电大学 | Oil abrasive particle characteristic signal extraction method based on segmentation entropy |
| CN116228695A (en) * | 2023-02-16 | 2023-06-06 | 武汉理工大学 | Wearing state sensing system and method based on video image extraction technology |
Non-Patent Citations (3)
| Title |
|---|
| SOSULSKI, K: "Preprocessed barley, rye, and Triticale as a feedstock for an integrated fuel ethanol feedlot plant", < APPLIED BIOCHEMISTRY AND BIOTECHNOLOGY>, 1 March 1997 (1997-03-01), pages 59 - 70 * |
| 徐斌;温广瑞;苏宇;张志芬;陈峰;孙耀宁;: "多层次信息融合在铁谱图像磨粒识别中的应用", 光学精密工程, no. 06, 15 June 2018 (2018-06-15) * |
| 李娟;王飞雪;: "油液监测技术在神东矿区的应用", 陕西煤炭, no. 03, 16 May 2019 (2019-05-16) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118379564B (en) | 2024-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115222741B (en) | Cable surface defect detection method | |
| Liu et al. | A two-stage approach for predicting the remaining useful life of tools using bidirectional long short-term memory | |
| Valliammal et al. | Plant leaf segmentation using non linear K means clustering | |
| CN114494256B (en) | Electric wire production defect detection method based on image processing | |
| KR102051226B1 (en) | Predictive diagnosis method and system of nuclear power plant equipment | |
| Wu et al. | Watershed-based morphological separation of wear debris chains for on-line ferrograph analysis | |
| CN115496692B (en) | Lubricating oil abrasive particle image enhancement method | |
| CN113569903A (en) | Method, system, equipment, medium and terminal for predicting abrasion of numerical control machine tool cutter | |
| CN108389216A (en) | Local auto-adaptive threshold segmentation method towards on-line ferrograph image wear Particles Recognition | |
| CN116137036B (en) | Gene detection data intelligent processing system based on machine learning | |
| CN116777917B (en) | Defect detection method and system for optical cable production | |
| CN114998343A (en) | Mold surface polishing degree detection method based on vision | |
| CN112801949A (en) | Method and device for determining discharge area in ultraviolet imaging detection technology | |
| CN116152749A (en) | Intelligent gear wear monitoring method based on digital twin | |
| CN115330799A (en) | Automatic fault diagnosis method for instrument | |
| CN118379564B (en) | Abrasion stage division method and system based on data mechanism fusion | |
| KR102028845B1 (en) | Predictive diagnosis method and system of nuclear power plant equipment | |
| CN113515554A (en) | Anomaly detection method and system for irregularly sampled time series | |
| CN111259926A (en) | Meat freshness detection method and device, computing equipment and storage medium | |
| CN110763466A (en) | Adaboost algorithm combined GABP rolling bearing diagnosis method | |
| Liu et al. | The Segmentation of Wear Particles Images Using J‐Segmentation Algorithm | |
| CN112767361B (en) | Reflected light ferrograph image segmentation method based on lightweight residual U-net | |
| KR102017162B1 (en) | Predictive diagnosis method and system of nuclear power plant equipment | |
| CN112419304B (en) | Multi-stage target detection method and device for one-dimensional data | |
| Wu et al. | A Segmentation algorithm of wear debris reflected image based on watershed and h-minima transform for on-Line visual ferrograph analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |