CN116168036A - Abnormal intelligent monitoring system for inductance winding equipment - Google Patents
Abnormal intelligent monitoring system for inductance winding equipment Download PDFInfo
- Publication number
- CN116168036A CN116168036A CN202310457838.2A CN202310457838A CN116168036A CN 116168036 A CN116168036 A CN 116168036A CN 202310457838 A CN202310457838 A CN 202310457838A CN 116168036 A CN116168036 A CN 116168036A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- cloud data
- dimensional point
- data
- clustering
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of pattern recognition, in particular to an abnormal intelligent monitoring system for an inductance winding device, which calculates a scaling factor of each three-dimensional point cloud data according to data distribution of the three-dimensional point cloud data, further obtains an overall scaling factor, obtains a voxel cube according to the overall scaling factor and a coordinate range of the three-dimensional point cloud data, clusters the three-dimensional point cloud data by taking a center point cloud as an initial clustering center in each voxel cube to obtain inductance point cloud types, and optimizes a K-means clustering process through a bending coefficient difference and a position difference between the three-dimensional point cloud data and a corresponding clustering center. The invention utilizes the data distribution characteristics to self-adaptively determine the proper voxel cube, and the algorithm is tightly combined with the field by introducing the tortuosity coefficient, so that the clustering effect is excellent, and the real-time performance and the accuracy of monitoring the abnormal state of the inductance winding equipment are improved.
Description
Technical Field
The invention relates to the technical field of pattern recognition processing, in particular to an abnormal intelligent monitoring system for inductance winding equipment.
Background
Inductor coils are widely used in different fields such as different industrial manufacturing and daily living equipment due to their special physical properties. The inductance coil is processed by the winding mechanical equipment in the production and manufacturing process, but the quality problem of the inductance coil obtained by production is caused by the factors of improper operation of technicians or improper parameter setting of the winding mechanical equipment, abrasion of parts and the like, and the inductance coil is generally uneven in coil density, so that the uneven inductance coil not only can influence the physical characteristics of the coil, but also has larger influence on the overall trimming degree of the coil. It is therefore necessary to monitor the winding abnormality of the coil.
In the prior art, the data points of the induction coil can be clustered into one type through a clustering algorithm, so that a point cloud area of the induction coil is obtained, and anomaly monitoring is carried out according to the characteristics of the point cloud area. However, because the data points of the inductance coils are distributed in a concentrated manner, a clustering algorithm adopted in the prior art can be poor in clustering effect due to improper selection of a clustering center, and therefore abnormality of inductance winding equipment cannot be accurately monitored.
Disclosure of Invention
In order to solve the technical problem that a clustering algorithm adopted in the prior art can cause poor clustering effect due to improper selection of a clustering center, the invention aims to provide an abnormal intelligent monitoring system for inductance winding equipment, and the adopted technical scheme is as follows:
the invention provides an abnormal intelligent monitoring system for inductance winding equipment, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the following method steps when executing the computer program:
acquiring three-dimensional point cloud data of the inductance coil;
calculating the scaling factor of each three-dimensional point cloud data according to the data distribution of the three-dimensional point cloud data, and further obtaining the overall scaling factor of all the three-dimensional point cloud data; obtaining the side length and the number of the voxel cubes according to the integral scaling coefficient and the coordinate range of the three-dimensional point cloud data; constructing a voxel cube of the three-dimensional point cloud data according to the side length and the number;
in the three-dimensional space, calculating the tortuosity coefficient of the corresponding three-dimensional point cloud data according to the coordinate difference of each three-dimensional point cloud data in different preset directions and the neighborhood data point in the preset neighborhood range;
in each voxel cube, clustering the three-dimensional point cloud data by taking the center point cloud of the voxel cube as an initial clustering center of the inductance point cloud class to obtain the inductance point cloud class; optimizing the clustering process through the difference and the position difference of the bending coefficients between the three-dimensional point cloud data and the corresponding clustering centers in the clustering process;
and the three-dimensional point cloud data in the inductance point cloud class form a point cloud area of the inductance coil, in the point cloud area, a form factor corresponding to each coil in the inductance coil is calculated according to the coordinate position difference between the three-dimensional point cloud data, a quality index of the inductance coil is obtained according to the form factors, and abnormal monitoring of inductance winding equipment is carried out according to the quality index.
Further, the method for obtaining the scaling factor includes:
constructing a Kd tree of the three-dimensional point cloud data, obtaining a neighborhood data point in any preset neighborhood of the three-dimensional point cloud data according to the Kd tree, and taking the average coordinate difference between each three-dimensional point cloud data and the corresponding neighborhood data point as the scaling coefficient.
Further, the method for obtaining the overall scaling factor comprises the following steps:
taking the average value of all the scaling coefficients of the three-dimensional point cloud data as the whole scaling coefficient.
Further, the method for acquiring the side length and the number includes:
obtaining the coordinate range of the three-dimensional point cloud data in each dimension; the polar difference of the coordinates in each dimension is multiplied, and the ratio of the multiplication result to the number of the three-dimensional point cloud data is used as an average coordinate range; multiplying the equipartition coordinate range by the integral scaling factor and then opening the coordinate range to the power of three to obtain the side length;
taking the ratio of the coordinate polar difference to the side length in each dimension as the initial number in the corresponding dimension; and multiplying the initial quantity in each dimension to obtain the quantity.
Further, the method for obtaining the tortuosity coefficient comprises the following steps:
taking a plane formed by an x axis and a y axis in a three-dimensional coordinate as a reference plane, wherein the reference plane comprises different preset directions, obtaining projection points of neighborhood data points of the three-dimensional point cloud data on the reference plane, and if the direction of connecting the projection points with an origin meets the preset direction, the neighborhood data points corresponding to the projection points are direction neighborhood data points of the three-dimensional point cloud data in the corresponding preset directions; calculating the accumulated sum of the differences between the three-dimensional point cloud data and the z-axis coordinates of each direction neighborhood data point in each preset direction to obtain an initial tortuosity coefficient in the corresponding preset direction; and accumulating the initial tortuosity coefficients in all preset directions to obtain the tortuosity coefficients corresponding to the three-dimensional point cloud data.
Further, the clustering process includes:
in each voxel cube, taking the center point cloud of the voxel cube as an initial clustering center of an inductance point cloud class; except for the initial clustering center of the inductance point cloud, a point randomly generated by a Gaussian function is used as the initial clustering center of the background point cloud; and according to the two types of initial clustering centers, utilizing a K-means clustering algorithm to gather the three-dimensional point cloud data into two types, wherein the types comprise an inductance point cloud type and a background point cloud type.
Further, optimizing the clustering process includes:
in each clustering process, calculating a bending coefficient difference and a Euclidean distance between each three-dimensional point cloud data and a clustering center point of a belonging category, and multiplying the Euclidean distance and the bending coefficient difference to serve as a category offset index of the corresponding three-dimensional point cloud data; accumulating the category offset indexes of all the three-dimensional point cloud data to obtain a clustering effect index corresponding to a clustering process; and stopping clustering iteration when the clustering effect index is minimum, and obtaining a clustering result.
Further, the method for obtaining the form factor comprises the following steps:
and obtaining the coordinate polar differences of the three-dimensional point cloud data corresponding to each coil in each dimension, and accumulating the coordinate polar differences in each dimension to obtain the form factor.
Further, the quality index obtaining method includes:
calculating the difference of the form factors among the coils, and accumulating to obtain an initial quality index; normalizing the initial quality index to obtain the quality index.
Further, the method for monitoring the abnormality of the inductance winding device according to the quality index comprises the following steps:
when the quality index is larger than a preset index threshold, the corresponding inductance winding equipment is considered to be abnormal; and when the quality index is smaller than or equal to a preset index threshold value, the corresponding inductance winding equipment is considered to be normal.
The invention has the following beneficial effects:
considering that the clustering algorithm adopted in the prior art can cause poor clustering effect due to unsuitable selection of a clustering center, and further abnormality of the inductance winding equipment cannot be accurately monitored, therefore, the edge length and the number of the voxel cubes are adaptively adjusted according to data distribution of three-dimensional point cloud data, the voxel cubes with suitable size of the three-dimensional point cloud data are further constructed, and the center point cloud in the voxel cubes is used as an initial clustering center of the inductance point cloud for clustering.
The bending coefficients represent bending information of corresponding three-dimensional point cloud data in different preset directions, the introduced bending coefficients and voxel cubes are used as clustering information to analyze and cluster the effects, an accurate inductance point cloud area can be obtained, the introduced bending coefficients can enable a clustering algorithm to be tightly combined with the field, the method has a better effect in the field of inductance analysis, compared with the traditional clustering algorithm, the method only uses Euclidean distance for clustering, the clustering result is more accurate, and the accuracy of monitoring the abnormal state of inductance winding equipment is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of an implementation method of an anomaly intelligent monitoring system for an inductance winding device according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description refers to the specific implementation, structure, characteristics and effects of an abnormality intelligent monitoring system for an inductance winding device according to the invention, which are provided by the invention, with reference to the accompanying drawings and the preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of an abnormality intelligent monitoring system for an inductance winding device provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of an implementation method of an anomaly intelligent monitoring system for an inductance winding device according to an embodiment of the present invention is shown. The embodiment of the invention provides an abnormal intelligent monitoring system for inductance winding equipment, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor can realize the following method steps when executing the computer program, and a flow chart corresponding to the method steps is shown in fig. 1, and specifically comprises the following steps:
step S1: and acquiring three-dimensional point cloud data of the inductance coil.
The traditional algorithm usually shoots the induction coil based on a 2D camera, but the shot image loses original three-dimensional data information, and ignores the influence of wire thickness on the quality of the induction coil, and errors exist in the process of abnormal monitoring of the induction coil when the image information is analyzed, so that the analysis processing is carried out by collecting three-dimensional data, and the accuracy of abnormal monitoring is higher. According to the embodiment of the invention, the Azure Kinect DK depth camera is utilized to shoot, collect and acquire the data points at each different position in the three-dimensional point cloud data of the inductance coil, wherein the data points have three coordinate information with different dimensions
The method is used for representing the spatial distance information of the position of the data point corresponding to the depth camera in the three-dimensional point cloud data obtained by shooting and collecting.
So far, three-dimensional point cloud data of the inductance coil are acquired.
Step S2: according to the data distribution of the three-dimensional point cloud data, calculating the scaling factor of each three-dimensional point cloud data, and further obtaining the overall scaling factor of all the three-dimensional point cloud data; obtaining the side length and the number of the voxel cubes according to the integral scaling coefficient and the coordinate range of the three-dimensional point cloud data; and constructing a voxel cube of the three-dimensional point cloud data according to the side lengths and the number.
In the actual shooting process, the three-dimensional point cloud data are affected by the operation conditions of the acquisition working environment, the acquisition working personnel, the depth camera equipment and the like in many aspects, abnormal data points can appear in the three-dimensional point cloud data obtained through acquisition, meanwhile, the three-dimensional point cloud data obtained through direct acquisition can be more, and if the three-dimensional point cloud data are clustered directly, the real-time performance of abnormal monitoring of the induction coil is slow and the accuracy is low. Thus, it is desirable to pre-select the cluster center points such that as many inductor coil data points as possible are contained in the resulting cluster result generated inductor point cloud class.
In the prior art, setting fixed cube sizes in the traditional voxel filtering method to construct voxel cubes can cause unsuitable quantity and size of the voxel cubes, the quantity of the voxel cubes is large, the calculated quantity is increased, and related three-dimensional point cloud data cannot be accurately analyzed together when the quantity of the voxel cubes is small, so that the subsequent clustering effect is poor, and therefore, the side length and quantity of the voxel cubes need to be obtained in a self-adaptive mode according to the obtained three-dimensional point cloud data. Because the inductance coil is a regular entity and the three-dimensional point cloud data has a neighborhood aggregation characteristic, the scaling factor of each three-dimensional point cloud data can be calculated according to the data distribution of the three-dimensional point cloud data, the integral scaling factor is further obtained, and the side length and the number of the voxel cubes can be obtained in a self-adaptive mode according to the integral scaling factor and the coordinate range of the three-dimensional point cloud data.
Preferably, in one embodiment of the present invention, a specific method for obtaining a scaling factor includes:
in order to facilitate searching for specific data, the scaling factor is calculated by using the three-dimensional point cloud data distribution characteristics in the Kd-tree, in consideration of the difference in coordinate information corresponding to the three-dimensional point cloud data photographed each time. Firstly, constructing a corresponding Kd tree according to three-dimensional point cloud data, wherein because three-dimensional point cloud data points have aggregation, each three-dimensional point cloud data needs to be analyzed for aggregation in a neighborhood range, and then a corresponding scaling factor is obtained, namely, a neighborhood data point in a preset neighborhood of any three-dimensional point cloud data is obtained according to the Kd tree, and the average coordinate difference between each three-dimensional point cloud data and the corresponding neighborhood data point is used as the scaling factor.
As an example, the first
The data point is any oneThe third dimension point cloud data can be obtained according to the Kd tree
The number of neighbor data points of the data points, noted as
Wherein, the first is determined according to the K neighbor query
The data points preset neighborhood data points in the neighborhood. In one embodiment of the invention, the K value in the K-nearest neighbor query process is taken to be 3.
According to the first
Average coordinate difference of each data point and its neighborhood data point to obtain the first
Scaling factor of data points
Average coordinate difference is the first
A value that is a sum and average of the absolute values of the individual dimensional differences of the coordinate information of each data point and each of its neighborhood data points. And calculating to obtain the scaling factor of each three-dimensional point cloud data, and preparing for the subsequent calculation of the whole scaling factor. I.e. the formula for the scaling factor comprises:
wherein, the liquid crystal display device comprises a liquid crystal display device,
is the first
The scaling factor of the data point is used,
respectively the first
Data point corresponding to
Coordinate information of each neighborhood data point
、
、
The coordinate values of the coordinate values,
respectively the first
Coordinate information of data points
、
、
The coordinate values of the coordinate values,
is the first
Number of data points in neighborhood since the K value of the K-nearest neighbor query procedure in one embodiment of the invention is taken to be 3, therefore
3.
The scaling factor characterizes the change of the coordinate distribution data of the corresponding three-dimensional point cloud data and the neighborhood data point, and when the change difference of the coordinate distribution data of the corresponding three-dimensional point cloud data and the neighborhood data point is larger, the corresponding scaling factor is larger.
The three-dimensional space is divided into cubes of uniform size, which are called voxel cubes. Voxel cubes function to divide a three-dimensional space and to assign each cube feature information. The Kd-tree is a data structure that characterizes the order of association between data points and facilitates searching for specific data, and the specific construction process of the Kd-tree is well known to those skilled in the art and is not further defined or described herein. The K-nearest neighbor query is well known to those skilled in the art and is not further defined or described herein.
In another embodiment of the present invention, the scaling factor may also be according to the first
The average euclidean distance between each data point and its neighboring data points is calculated, and the euclidean distance is calculated by a technical means known to those skilled in the art, which is not described herein.
Because each three-dimensional point cloud data corresponds to one scaling factor, to obtain the voxel cube side lengths and numbers in the space to which all three-dimensional point cloud data corresponds, an overall scaling factor needs to be obtained.
Preferably, in one embodiment of the present invention, an average value of scaling coefficients of all three-dimensional point cloud data is taken as an overall scaling coefficient
. I.e. the formula for the overall scaling factor comprises:
wherein, the liquid crystal display device comprises a liquid crystal display device,
for the amount of three-dimensional point cloud data,
is the first
Scaling factors for data points.
The integral scaling factor characterizes the change of the coordinate distribution data of all the three-dimensional point cloud data, and when the change difference of the coordinate distribution data of all the three-dimensional point cloud data is large, the integral scaling factor is scaled more.
In another embodiment of the present invention, the overall scaling factor may also be the median of the scaling factors of all three-dimensional point cloud data.
The overall scaling factor represents the dispersion of the coordinate distribution among the data points in the three-dimensional space, i.e. the larger the overall scaling factor, the longer the edge length of the voxel cube setup needs to be increased in order for the voxel cube to contain more three-dimensional point cloud data points. Because the three-dimensional point cloud data of the inductor coil has a distinct coordinate range, the side length and number of voxel cubes to be set can be obtained in combination with the coordinate range and the overall scaling factor.
Preferably, in one embodiment of the present invention, the coordinate range of the three-dimensional point cloud data in each dimension is obtained, and the coordinate range of the three-dimensional point cloud data in the corresponding dimension can be reflected by the coordinate range of the three-dimensional point cloud data in each dimension. The coordinate range of the three-dimensional point cloud data under each dimension is multiplied, the ratio of the multiplication result to the number of the three-dimensional point cloud data is used as an average coordinate range, the average coordinate range represents an initial voxel cube side length reference index, the average coordinate range is further combined with the integral scaling factor, the average coordinate range is multiplied with the integral scaling factor and then is divided by three to obtain the side length of the voxel cube with proper size
The method comprises the steps of carrying out a first treatment on the surface of the Taking the ratio of the coordinate polar difference in each dimension to the side length as the initial number in the corresponding dimension, and multiplying the initial number in each dimension to obtain the number of voxel cubes
。
Wherein, the liquid crystal display device comprises a liquid crystal display device,
、
、
respectively, the maximum of the three-dimensional point cloud data in each dimension
、
、
The coordinate values of the coordinate values,
、
、
respectively, the smallest of the three-dimensional point cloud data in each dimension
、
、
The coordinate values of the coordinate values,
、
、
respectively under corresponding dimensions of three-dimensional point cloud data
、
、
The coordinates of the coordinates are extremely poor,
、
、
the initial number of corresponding dimension voxel cubes is represented by the initial number,
for the total number of voxel cubes in the three-dimensional point cloud data space,
is the number of three-dimensional point cloud data.
Under the action of the integral scaling coefficient, the three-dimensional point cloud data space is divided into
A voxel cube of proper size with a side length of
The subsequent clustering process has good effect.
Step S3: and calculating the tortuosity coefficient of the corresponding three-dimensional point cloud data according to the coordinate difference of each three-dimensional point cloud data in different preset directions and the neighborhood data point in the preset neighborhood range in the three-dimensional space.
After the inductance winding device is used for winding, the surface of the inductance coil is zigzag, only Euclidean distance is used in a traditional K-means clustering algorithm in the prior art, and the clustering effect in the prior art is poor, so that a zigzag coefficient corresponding to one point of the inductance coil corresponding to any data point in the three-dimensional point cloud data space is introduced, and the zigzag coefficient is participated in calculation for representing the clustering effect after clustering, so that the clustering is more accurate. Therefore, in order to accurately obtain the tortuosity coefficient of each three-dimensional point cloud data, the tortuosity coefficient is calculated by considering the coordinate difference between the corresponding three-dimensional point cloud data in different preset directions and the neighborhood data points in the preset neighborhood range, namely the tortuosity coefficient comprises the tortuosity information of the corresponding three-dimensional point cloud data in different preset directions.
Preferably, in one embodiment of the present invention, the specific method for obtaining the tortuosity coefficient includes:
taking a plane formed by an x axis and a y axis in the three-dimensional coordinates as a reference plane, wherein the reference plane contains different preset directions, and obtaining projection points of neighborhood data points of each three-dimensional point cloud data on the reference plane, wherein the neighborhood data points in a preset neighborhood range are obtained according to K neighbor inquiry in a Kd tree and the value of the side length of a voxel cube. A neighborhood data point is adaptively determined from the side length of the voxel cube. If the direction of the connecting line of the projection point and the origin point meets the preset direction, the neighborhood data point corresponding to the projection point is the direction neighborhood data point of the three-dimensional point cloud data in the corresponding preset direction. Calculating the sum of the absolute values of the differences between the three-dimensional point cloud data and the z-axis coordinate values of the neighborhood data points in each preset direction to obtain an initial tortuosity coefficient in the corresponding preset direction; and accumulating the initial tortuosity coefficients in all preset directions to obtain the tortuosity coefficient corresponding to the three-dimensional point cloud data. And further calculating the tortuosity coefficient of each three-dimensional point cloud data to prepare for the subsequent clustering.
In another embodiment of the present invention, K selects other suitable values for calculation in the K-nearest neighbor query method when determining the neighborhood data points.
The formulation of the tortuosity coefficient includes:
first, the
The data point is any three-dimensional point cloud data,
is the first
The tortuosity coefficient of the data point,
for the total number of the preset directions,
is the first
The number of direction neighborhood data points in a preset direction,
is the first
Data points of
The coordinate values of the coordinate values,
for the first neighbor data point of the direction corresponding to the preset direction
Personal (S)
Coordinate values.
In one embodiment of the invention, the total of the preset directionsThe number is 4, the 1 st, 2 nd, 3 rd and 4 th preset directions respectively correspond to the directions of the positive half shaft and the negative half shaft, and the included angles of the positive half shaft and the positive half shaft are
Direction and opposite direction, y positive and negative half axis direction, x positive half axis and y negative half axis included angle
Direction and opposite thereto.
The initial bending coefficient represents the surface bending condition of one point of the induction coil corresponding to the three-dimensional point cloud data in the corresponding preset direction, and the bending coefficient represents the bending condition of one point of the induction coil corresponding to the three-dimensional point cloud data. When the surface meandering condition is larger at a certain point of the inductance coil, the meandering coefficient corresponding to a certain data point is larger.
In another embodiment of the present invention, the preset direction may be selected for calculation.
Thus, the tortuosity coefficients of all three-dimensional point cloud data are obtained.
Step S4: in each voxel cube, taking the center point cloud of the voxel cube as an initial clustering center of the inductance point cloud class, and clustering three-dimensional point cloud data to obtain the inductance point cloud class; and optimizing the clustering process through the difference of the bending coefficients and the difference of the positions between the three-dimensional point cloud data and the corresponding clustering centers in the clustering process.
Through the processing of step S2, the whole three-dimensional point cloud data is divided into individual local areas according to the voxel cube, so that detail analysis can be performed for each local area. In each voxel cube, because the inductance point cloud data shows aggregation, the possibility that the center point cloud in the voxel cube is the inductance point cloud data is the greatest, and therefore the center point cloud of the voxel cube is used as an initial clustering center of the inductance point cloud class to cluster the three-dimensional point cloud data, and the inductance point cloud class is obtained.
Preferably, in one embodiment of the present invention, the clustering process includes: in each voxel cube, taking the center point cloud of the voxel cube as an initial clustering center of an inductance point cloud class; except for the initial clustering center of the inductance point cloud, a point randomly generated by a Gaussian function is used as the initial clustering center of the background point cloud; and according to the two types of initial clustering centers, utilizing a K-means clustering algorithm to gather three-dimensional point cloud data into two types, wherein the types comprise an inductance point cloud type and a background point cloud type. And clustering the three-dimensional point cloud data in all the voxel cubes to obtain the inductance point cloud class. Background point cloud data is meaningless data.
In one embodiment of the present invention, a data point at a barycentric position in each voxel cube is taken as a center point cloud, and if no data point exists at the barycentric position, a data point closest to the barycenter is taken as the center point cloud.
In the prior art, the European distance is adopted to represent the clustering effect and is not attached to the field, so that the clustering effect is represented by introducing the combination of the bending coefficient difference, the bending coefficient difference represents the difference of bending conditions of corresponding surfaces of two points of the induction coil corresponding to each three-dimensional point cloud data and the clustering center of the category, and when the bending coefficient difference corresponding to the two points is smaller, the probability that the corresponding point cloud data is similar to electric clouds is high, so that the clustering effect is represented by combining the bending coefficient difference in order to obtain an accurate clustering result. The inductance point cloud data has aggregation, so that optimization of the clustering process can be realized by combining the difference of the tortuosity coefficients and the difference of the positions.
Preferably, in one embodiment of the present invention, the optimization of the clustering process includes: in step S2, the three-dimensional point cloud data space is divided into voxel cubes, wherein cluster analysis is facilitated, the first
The voxel cube is any voxel cube, and the analysis is carried out on the first voxel cube
Clustering in a voxel cube. Because the difference of the bending coefficients can represent the clustering effect, the difference of the bending coefficients is introduced to calculate the clustering effect, and the clustering effect analysis and optimization clustering algorithm is carried out by combining the difference of the bending coefficients and the Euclidean distance. Calculation ofThe difference of the bending coefficients of each three-dimensional point cloud data and the clustering center point of the belonging category and the Euclidean distance are absolute values of the difference of the bending coefficients of the two corresponding data points, the Euclidean distance and the difference of the bending coefficients are multiplied to be used as category deviation indexes of the corresponding three-dimensional point cloud data, the size of the category deviation indexes represents the corresponding clustering effect, and the smaller the corresponding category deviation indexes are, the better the corresponding clustering effect is; accumulating class offset indexes of the corresponding three-dimensional point cloud data to obtain clustering effect indexes of the corresponding clustering process
The method comprises the steps of carrying out a first treatment on the surface of the And stopping clustering iteration when the clustering effect index is minimum, and obtaining a clustering result.
It should be noted that, in the clustering iterative process, the clustering centers corresponding to different categories are obtained by the average dimension value of the coordinate information of the three-dimensional point cloud data of the corresponding category. And each clustering iteration process adopts the prior art to classify each data point, and the specific clustering method is a technical means well known to those skilled in the art, and is not described herein.
As one example, the acquisition formula of the cluster effect index includes:
wherein, the data point is three-dimensional point cloud data,
for the total number of different classes in the cluster,
is the first
The first voxel in the cube
Total number of data points for each category. Let the root of common angelica
When=1, the data points of the inductance point cloud class are processed,
a number of data points in the inductive point cloud class; let the root of common angelica
When=2, the data points of the background point cloud class are processed,
the number of data points that are the background point cloud class.
Is the first in the corresponding category
The tortuosity coefficient of the data point,
is the first
The tortuosity coefficients of the data points corresponding to the cluster centers of the individual categories,
is the first in the corresponding category
The coordinate information of the data points,
first, the
Coordinate information of data points corresponding to the clustering centers of the individual categories,
as a distance measurement function, the euclidean distance is adopted for calculation and analysis. In one embodiment of the invention, the total number of different classes in the cluster is 2, thus
2.
In the clustering iterative process, when the sum of the class offset indexes of the corresponding three-dimensional point cloud data and the corresponding central point cloud is minimum, namely the clustering effect index, the accurate clustering is finished in the corresponding voxel cube.
And clustering all other voxel cubes, wherein all three-dimensional point cloud data are divided into an inductance point cloud class and a background point cloud class.
And finally, the clustering is finished, and the inductance point cloud class is obtained.
Step S5: and in the point cloud area, calculating a form factor corresponding to each coil in the induction coil according to the coordinate position difference between the three-dimensional point cloud data, obtaining a quality index of the induction coil according to the form factor, and monitoring the abnormality of the induction winding equipment according to the quality index.
The point cloud area of each coil is provided with a three-dimensional coordinate range, the appearance condition of the corresponding coil is calculated according to the coordinate range, and appearance difference conditions among different coils are required to be obtained for monitoring abnormality of the inductance winding equipment, so that appearance factors of the introduced coils represent winding conditions of the coils.
In one embodiment of the invention, the point cloud area is divided by using an engineering CAD graph, and the point cloud area corresponding to each coil is obtained. Engineering CAD drawings are well known to those skilled in the art and are not further defined or described herein. In the inductor coil point cloud region, the inductor coils are numbered from left to right, which is
Each coil corresponds to an inductance point cloud region.
Preferably, in one embodiment of the present invention, the method for obtaining the form factor includes:
first, the
The first coil is any one coil, and the second coil is obtained
The coordinate polar differences of the three-dimensional point cloud data corresponding to the coils in all dimensions are the differences between the maximum coordinates and the minimum coordinates in the corresponding dimensions, and the coordinate polar differences in all dimensions are accumulated to obtain the first dimension
Form factor of inductance point cloud area corresponding to each coil
. I.e. the formula for the form factor comprises:
wherein, the liquid crystal display device comprises a liquid crystal display device,
、
、
respectively the first
Maximum data point in inductance point cloud area corresponding to each coil
、
、
The coordinate values of the coordinate values,
、
、
is the first
Minimum of data points in the inductive point cloud area corresponding to each coil
、
、
Coordinate values.
The appearance factor characterizes appearance characteristics of the corresponding inductance coil, and when the coordinate range of the corresponding three-dimensional point cloud data in each dimension is large, the corresponding inductance coil protrudes more in the corresponding dimension.
For monitoring anomalies of the inductive winding device, it is considered to calculate the quality index using the profile differences between the different coils, i.e. the quality index represents the sum of the differences between the different coils.
Preferably, in one embodiment of the present invention, the method for obtaining a quality index includes:
calculating the difference of the form factors among the coils, wherein the difference is the absolute value of the phase difference of the two corresponding form factors, and accumulating to obtain an initial quality index; normalizing the initial quality index to obtain quality index
. Namely, the formula of the quality index is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,
as a function of the normalization,
as the total number of turns of the inductor winding,
is the first
The form factor of the inductance point cloud area corresponding to each coil,
is the first
And the form factor of the inductance point cloud area corresponding to each coil.
The quality index characterizes the cumulative sum of the differences in the shape sizes of the respective coils, and the quality index is greater as the sum of the differences between the shape factors corresponding to the respective coils is greater.
In one embodiment of the present invention,
normalization function is adopted
The norm is normalized and the result is that,
norm normalization is well known to those skilled in the art and is not further defined or described herein.
The quality index reflects the sum of the form differences between the inductor coils, which corresponds to the status of the inductive winding device.
Preferably, in one embodiment of the present invention, the method for monitoring the abnormality of the inductance winding device according to the quality index includes:
when the quality index is larger than a preset index threshold, the corresponding inductance winding equipment is considered to be abnormal, and when the inductance winding equipment is in an abnormal working state, abnormal conditions of mutual overlapping or bulge phenomenon occur among the inductance coils, and the calculated quality index of the inductance coils is larger; otherwise, when the quality index is smaller than or equal to a preset experience threshold value, the corresponding inductance winding equipment is considered to be normal.
In one embodiment of the present invention, the preset index threshold is set to 0.6.
Thus, the monitoring of the abnormality of the inductance winding device is completed.
In summary, the invention calculates the scaling factor of each three-dimensional point cloud data according to the data distribution of the three-dimensional point cloud data, further obtains the overall scaling factor, obtains a voxel cube according to the overall scaling factor and the coordinate range of the three-dimensional point cloud data, clusters the three-dimensional point cloud data by taking the center point cloud as an initial clustering center in each voxel cube to obtain an inductance point cloud class, and optimizes the K-means clustering process by the difference of the tortuosity coefficient and the position difference between the three-dimensional point cloud data and the corresponding clustering center. The invention utilizes the data distribution characteristics to self-adaptively determine the proper voxel cube, and the algorithm is tightly combined with the field by introducing the tortuosity coefficient, so that the clustering effect is excellent, and the real-time performance and the accuracy of monitoring the abnormal state of the inductance winding equipment are improved.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
The foregoing is only illustrative of the present invention and is not to be construed as limiting thereof, but rather as various modifications, equivalent arrangements, improvements, etc., within the spirit and principles of the present invention.
Claims (10)
1. An anomaly intelligent monitoring system for an inductive winding device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements the method steps of:
acquiring three-dimensional point cloud data of the inductance coil;
according to the data distribution of the three-dimensional point cloud data, calculating the scaling coefficient of each three-dimensional point cloud data, and further obtaining the overall scaling coefficient of all the three-dimensional point cloud data; obtaining the side length and the number of the voxel cubes according to the integral scaling coefficient and the coordinate range of the three-dimensional point cloud data; constructing a voxel cube of the three-dimensional point cloud data according to the side length and the number;
in the three-dimensional space, calculating the tortuosity coefficient of the corresponding three-dimensional point cloud data according to the coordinate difference of each three-dimensional point cloud data in different preset directions and the neighborhood data point in the preset neighborhood range;
in each voxel cube, clustering the three-dimensional point cloud data by taking the center point cloud of the voxel cube as an initial clustering center of the inductance point cloud class to obtain the inductance point cloud class; optimizing the clustering process through the difference and the position difference of the bending coefficients between the three-dimensional point cloud data and the corresponding clustering centers in the clustering process;
and the three-dimensional point cloud data in the inductance point cloud class form a point cloud area of the inductance coil, in the point cloud area, a form factor corresponding to each coil in the inductance coil is calculated according to the coordinate position difference between the three-dimensional point cloud data, a quality index of the inductance coil is obtained according to the form factors, and abnormal monitoring of inductance winding equipment is carried out according to the quality index.
2. The anomaly intelligent monitoring system for an inductive winding device of claim 1, wherein the method for obtaining the scaling factor comprises:
constructing a Kd tree of the three-dimensional point cloud data, obtaining a neighborhood data point in any preset neighborhood of the three-dimensional point cloud data according to the Kd tree, and taking the average coordinate difference between each three-dimensional point cloud data and the corresponding neighborhood data point as the scaling coefficient.
3. The anomaly intelligent monitoring system for an inductive winding device of claim 1, wherein the overall scaling factor obtaining method comprises:
taking the average value of all the scaling coefficients of the three-dimensional point cloud data as the whole scaling coefficient.
4. An anomaly intelligent monitoring system for an inductive winding apparatus according to claim 1, wherein the method of obtaining the side length and the number comprises:
obtaining the coordinate range of the three-dimensional point cloud data in each dimension; the polar difference of the coordinates in each dimension is multiplied, and the ratio of the multiplication result to the number of the three-dimensional point cloud data is used as an average coordinate range; multiplying the equipartition coordinate range by the integral scaling factor and then opening the coordinate range to the power of three to obtain the side length;
taking the ratio of the coordinate polar difference to the side length in each dimension as the initial number in the corresponding dimension; and multiplying the initial quantity in each dimension to obtain the quantity.
5. The intelligent anomaly monitoring system for an inductive winding device of claim 2, wherein the method for obtaining the tortuosity coefficient comprises:
taking a plane formed by an x axis and a y axis in a three-dimensional coordinate as a reference plane, wherein the reference plane comprises different preset directions, obtaining projection points of neighborhood data points of the three-dimensional point cloud data on the reference plane, and if the direction of connecting the projection points with an origin meets the preset direction, the neighborhood data points corresponding to the projection points are direction neighborhood data points of the three-dimensional point cloud data in the corresponding preset directions; calculating the accumulated sum of the differences between the three-dimensional point cloud data and the z-axis coordinates of each direction neighborhood data point in each preset direction to obtain an initial tortuosity coefficient in the corresponding preset direction; and accumulating the initial tortuosity coefficients in all preset directions to obtain the tortuosity coefficients corresponding to the three-dimensional point cloud data.
6. An anomaly intelligent monitoring system for an inductive winding device according to claim 1, wherein the clustering process comprises:
in each voxel cube, taking the center point cloud of the voxel cube as an initial clustering center of an inductance point cloud class; except for the initial clustering center of the inductance point cloud, a point randomly generated by a Gaussian function is used as the initial clustering center of the background point cloud; and according to the two types of initial clustering centers, utilizing a K-means clustering algorithm to gather the three-dimensional point cloud data into two types, wherein the types comprise an inductance point cloud type and a background point cloud type.
7. The anomaly intelligent monitoring system for an inductive winding device of claim 6, wherein optimizing the clustering process comprises:
in each clustering process, calculating a bending coefficient difference and a Euclidean distance between each three-dimensional point cloud data and a clustering center point of a belonging category, and multiplying the Euclidean distance and the bending coefficient difference to serve as a category offset index of the corresponding three-dimensional point cloud data; accumulating the category offset indexes of all the three-dimensional point cloud data to obtain a clustering effect index corresponding to a clustering process; and stopping clustering iteration when the clustering effect index is minimum, and obtaining a clustering result.
8. The intelligent anomaly monitoring system for an inductive winding device of claim 1, wherein the method for obtaining the form factor comprises:
and obtaining the coordinate polar differences of the three-dimensional point cloud data corresponding to each coil in each dimension, and accumulating the coordinate polar differences in each dimension to obtain the form factor.
9. The anomaly intelligent monitoring system for an inductive winding device of claim 1, wherein the quality index obtaining method comprises:
calculating the difference of the form factors among the coils, and accumulating to obtain an initial quality index; normalizing the initial quality index to obtain the quality index.
10. The intelligent monitoring system for anomalies of an induction winding apparatus of claim 9, wherein the method of monitoring anomalies of an induction winding apparatus according to the quality index includes:
when the quality index is larger than a preset index threshold, the corresponding inductance winding equipment is considered to be abnormal; and when the quality index is smaller than or equal to a preset index threshold value, the corresponding inductance winding equipment is considered to be normal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310457838.2A CN116168036B (en) | 2023-04-26 | 2023-04-26 | Abnormal intelligent monitoring system for inductance winding equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310457838.2A CN116168036B (en) | 2023-04-26 | 2023-04-26 | Abnormal intelligent monitoring system for inductance winding equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116168036A true CN116168036A (en) | 2023-05-26 |
| CN116168036B CN116168036B (en) | 2023-07-04 |
Family
ID=86416781
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310457838.2A Active CN116168036B (en) | 2023-04-26 | 2023-04-26 | Abnormal intelligent monitoring system for inductance winding equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116168036B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117095771A (en) * | 2023-10-18 | 2023-11-21 | 昆山尚瑞智能科技有限公司 | High-precision spectrum measurement data optimization processing method |
| CN117252863A (en) * | 2023-11-13 | 2023-12-19 | 山东省地质测绘院 | Quick detection and analysis method for geographic information abnormal data |
| CN117474999A (en) * | 2023-12-25 | 2024-01-30 | 惠州市德立电子有限公司 | Micro chip type inductance double-wire winding abnormity positioning method and system |
| CN117710244A (en) * | 2024-02-05 | 2024-03-15 | 湖南裕工新能科技有限公司 | Vehicle-mounted assembly material alignment intelligent detection method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110658531A (en) * | 2019-08-23 | 2020-01-07 | 畅加风行(苏州)智能科技有限公司 | Dynamic target tracking method for port automatic driving vehicle |
| CN111488769A (en) * | 2019-01-28 | 2020-08-04 | 北京工商大学 | Unsupervised fusion point cloud superpixelization method based on light spot divergence size |
| CN112034244A (en) * | 2020-11-04 | 2020-12-04 | 南京航天工业科技有限公司 | ETC abnormal signal detection method and system |
| WO2022142948A1 (en) * | 2020-12-29 | 2022-07-07 | 深圳市普渡科技有限公司 | Dynamic target tracking and positioning method and apparatus, and device and storage medium |
| US20220269900A1 (en) * | 2021-02-19 | 2022-08-25 | Teraki Gmbh | Low level sensor fusion based on lightweight semantic segmentation of 3d point clouds |
| CN115908751A (en) * | 2022-11-23 | 2023-04-04 | 南京莱斯信息技术股份有限公司 | Aircraft capturing and identifying method based on three-dimensional point cloud |
-
2023
- 2023-04-26 CN CN202310457838.2A patent/CN116168036B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111488769A (en) * | 2019-01-28 | 2020-08-04 | 北京工商大学 | Unsupervised fusion point cloud superpixelization method based on light spot divergence size |
| CN110658531A (en) * | 2019-08-23 | 2020-01-07 | 畅加风行(苏州)智能科技有限公司 | Dynamic target tracking method for port automatic driving vehicle |
| CN112034244A (en) * | 2020-11-04 | 2020-12-04 | 南京航天工业科技有限公司 | ETC abnormal signal detection method and system |
| WO2022142948A1 (en) * | 2020-12-29 | 2022-07-07 | 深圳市普渡科技有限公司 | Dynamic target tracking and positioning method and apparatus, and device and storage medium |
| US20220269900A1 (en) * | 2021-02-19 | 2022-08-25 | Teraki Gmbh | Low level sensor fusion based on lightweight semantic segmentation of 3d point clouds |
| CN115908751A (en) * | 2022-11-23 | 2023-04-04 | 南京莱斯信息技术股份有限公司 | Aircraft capturing and identifying method based on three-dimensional point cloud |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117095771A (en) * | 2023-10-18 | 2023-11-21 | 昆山尚瑞智能科技有限公司 | High-precision spectrum measurement data optimization processing method |
| CN117095771B (en) * | 2023-10-18 | 2024-02-06 | 昆山尚瑞智能科技有限公司 | High-precision spectrum measurement data optimization processing method |
| CN117252863A (en) * | 2023-11-13 | 2023-12-19 | 山东省地质测绘院 | Quick detection and analysis method for geographic information abnormal data |
| CN117252863B (en) * | 2023-11-13 | 2024-02-09 | 山东省地质测绘院 | Quick detection and analysis method for geographic information abnormal data |
| CN117474999A (en) * | 2023-12-25 | 2024-01-30 | 惠州市德立电子有限公司 | Micro chip type inductance double-wire winding abnormity positioning method and system |
| CN117474999B (en) * | 2023-12-25 | 2024-04-19 | 惠州市德立电子有限公司 | Micro chip type inductance double-wire winding abnormity positioning method and system |
| CN117710244A (en) * | 2024-02-05 | 2024-03-15 | 湖南裕工新能科技有限公司 | Vehicle-mounted assembly material alignment intelligent detection method and system |
| CN117710244B (en) * | 2024-02-05 | 2024-04-26 | 湖南裕工新能科技有限公司 | Vehicle-mounted assembly material alignment intelligent detection method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116168036B (en) | 2023-07-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116168036B (en) | Abnormal intelligent monitoring system for inductance winding equipment | |
| CN109409437B (en) | Point cloud segmentation method and device, computer readable storage medium and terminal | |
| CN109887015B (en) | Point cloud automatic registration method based on local curved surface feature histogram | |
| CN100559398C (en) | Automatic deepness image registration method | |
| JP6216508B2 (en) | Method for recognition and pose determination of 3D objects in 3D scenes | |
| CN114972459B (en) | Point cloud registration method based on low-dimensional point cloud local feature descriptor | |
| CN108022262A (en) | A kind of point cloud registration method based on neighborhood of a point center of gravity vector characteristics | |
| CN109903319B (en) | Multi-resolution-based fast iteration closest point registration algorithm | |
| CN108665491A (en) | A kind of quick point cloud registration method based on local reference | |
| CN112381862B (en) | Full-automatic registration method and device for CAD (computer-aided design) model and triangular mesh | |
| CN113628263A (en) | Point cloud registration method based on local curvature and neighbor characteristics thereof | |
| CN111145129A (en) | Point cloud denoising method based on hyper-voxels | |
| CN116862955A (en) | Three-dimensional registration method, system and equipment for plant images | |
| CN117274339A (en) | Point cloud registration method based on improved ISS-3DSC characteristics combined with ICP | |
| CN109035311A (en) | A kind of curved bone fracture autoregistration and internal fixation steel plate pre-bending modeling method | |
| Zhao et al. | A new k nearest neighbours algorithm using cell grids for 3d scattered point cloud | |
| CN112802089A (en) | Point cloud skeleton line extraction method and system based on automatic estimation of bifurcation number | |
| CN112446952B (en) | Three-dimensional point cloud normal vector generation method and device, electronic equipment and storage medium | |
| CN107464257B (en) | Wide base line matching method and device | |
| An et al. | Self-adaptive polygon mesh reconstruction based on ball-pivoting algorithm | |
| CN110322415A (en) | High-precision surface three-dimensional reconstruction method based on point cloud | |
| Guan et al. | View-based 3D model retrieval by joint subgraph learning and matching | |
| CN115063615A (en) | Repeated texture image matching method based on Delaunay triangulation | |
| CN111583098B (en) | Line segment clustering and fitting method and system based on sequence image | |
| CN113932712A (en) | Melon and fruit vegetable size measuring method based on depth camera and key points |
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 | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: 518000 room 1101-1105, building 6, Hengda fashion Huigu building (East District), Fulong Road, Henglang community, Dalang street, Longhua District, Shenzhen, Guangdong Patentee after: Cenke Technology (Shenzhen) Group Co.,Ltd. Address before: Room 1101-1105, Building 6, Evergrande Fashion Huigu Building (East Area), Fulong Road, Henglang Community, Dalang Street, Longhua District, Shenzhen, Guangdong 518110 Patentee before: SHENZHEN CENKER ENTERPRISE Ltd. |