A method for intelligently processing running data of automated textile equipment

Abstract

The present invention relates to the field of data processing technology, specifically to an intelligent processing method for the operation data of automated textile equipment, comprising the following steps: during the operation of the textile equipment, a multidimensional data set output by a sensor is collected, a Gaussian kernel function is selected, the features of the multidimensional data set are extracted, feature projection values ​​are generated by matrix operations, dimension reduction is performed, a low-dimensional space mapping is obtained, and the density distribution of data points is calculated by drawing a point cloud distribution map. The present invention extracts features of the multidimensional data set, generates feature projection values ​​by using Gaussian kernel functions and matrix operations, accurately extracts key features, ensures that sufficient information is retained after data dimension reduction, and thus achieves more efficient data processing. The density calculation of data points and the identification of local extreme points are performed based on the point cloud distribution map, which improves the positioning accuracy of key data points in low-dimensional space and reduces noise interference in the processing process.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for intelligently processing running data of automated textile equipment. Background Art

[0002] The intelligent processing method of textile equipment operation data is a technical solution for automatic collection and intelligent analysis of the operation data of textile equipment in the production process. The purpose is to improve the production efficiency of textile equipment, reduce the failure rate, and optimize various parameters in the production process by real-time monitoring and processing of equipment operation data, thereby realizing intelligent management and control and improving the overall automation level of the textile industry.

[0003] However, the existing technology faces the problem of inaccurate data feature extraction when processing multidimensional data, which results in the inability to effectively retain key feature information during the dimensionality reduction process, affecting the overall efficiency of data processing. Since the existing methods lack accurate calculation methods when processing data density distribution and identifying local extreme points, it is difficult to effectively deal with the noise interference generated during the production process, resulting in difficulty in accurately locating key data points in low-dimensional space. These shortcomings limit the ability of existing technologies to improve the intelligent management of textile equipment and optimize production efficiency. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent processing method for operating data of automated textile equipment.

[0005] In order to achieve the above object, the present invention adopts the following technical solution, a method for intelligently processing operation data of automated textile equipment, comprising the following steps:

[0006] During the operation of textile equipment, the multidimensional data set output by the sensor is collected, the Gaussian kernel function is selected, the features of the multidimensional data set are extracted, the feature projection value is generated by matrix operation, the dimension is reduced, and the low-dimensional space mapping is obtained. The density distribution of the data points is calculated by drawing the point cloud distribution map, and the local extreme value points are identified to generate a low-dimensional feature space point cloud map;

[0007] Based on the low-dimensional feature space point cloud image, statistical operations are performed on the local extreme point and its neighborhood points, the contribution rate of the feature to the target classification label is calculated, the distance matrix of the local extreme point is calculated, the density weight is generated by using the distance matrix, the influence of the local extreme point is calculated, and the density weight matrix of the local extreme point is generated;

[0008] Based on the local extreme point density weight matrix, the cluster division of the feature point cloud is calculated, the cluster center is generated and the centroid position is calculated, the difference matrix is ​​calculated by the distance between the centroid position and the feature point, the feature clustering similarity is calculated using the difference matrix, and the cluster center centroid point projection value matrix is ​​generated;

[0009] Based on the projection value matrix of the centroid points of the cluster centers, the features of the cluster centers are calculated in parallel using the MapReduce framework, the vector inner product and difference comparison of the feature points are calculated, the feature weights of the cluster centers are generated, and the feature weight ranking table of the cluster centers is obtained by sorting.

[0010] Based on the cluster center feature weight ranking table, combined with the dynamic resource allocation scheme, the resource ratio of the cluster center feature points is adjusted, the competition intensity is calculated through resource allocation comparison, resource allocation is optimized, and a cluster center resource allocation optimization table is generated.

[0011] Preferably, the low-dimensional feature space point cloud map includes the two-dimensional coordinates, density values ​​and local extreme point positions of the feature points, the local extreme point density weight matrix includes the density values ​​of the local extreme points, the distance values ​​and influence values ​​of the neighborhood points, the cluster center centroid point projection value matrix includes the centroid position, feature point difference values ​​and cluster similarity values, the cluster center feature weight sorting table includes feature weight values, feature point sorting and vector inner product values, and the cluster center resource allocation optimization table includes resource allocation ratio, feature point competition intensity and optimized resource allocation value.

[0012] Preferably, the steps of obtaining the low-dimensional feature space point cloud image are specifically as follows:

[0013] Based on the multidimensional data set of sensor output collected during the operation of textile equipment, the key features of each dimension are extracted, matrix operations are performed to obtain feature projection values, matrix transformation is used to reduce the feature projection values ​​to low-dimensional space, feature mapping data is obtained, and low-dimensional space mapping data is generated;

[0014] Based on the low-dimensional space mapping data, a density calculation method is used to calculate the density of each data point, generate a density distribution map of the data point, analyze the local area in the map, identify and mark the local extreme value points, and generate a local extreme value point distribution map;

[0015] Based on the local extreme point distribution map, the distribution of the point cloud is analyzed, the relative position relationship of the data points is used to generate a point cloud map of the low-dimensional feature space, the distribution of each feature point in the low-dimensional space is obtained, and the low-dimensional feature space point cloud map is generated.

[0016] Preferably, the steps of obtaining the local extreme point density weight matrix are specifically as follows:

[0017] Based on the low-dimensional feature space point cloud image, the distribution characteristics of local extreme points and neighborhood points are analyzed, the contribution value of each point to the classification label is counted, and the contribution value is calculated in combination with the distribution of neighborhood points to generate the feature contribution value of the local extreme point;

[0018] Based on the characteristic contribution value of the local extreme point, the relative position of the local extreme point is analyzed, and a distance matrix operation is performed to calculate the distance relationship between the neighborhood points. The density weight of each local extreme point is generated in combination with the density of the neighborhood points, and a local extreme point density matrix is ​​generated;

[0019] Based on the local extreme point density matrix, the density weight of each local extreme point is integrated, the influence of each point is analyzed, and the characteristic contribution value and the density weight are combined for calculation to generate a local extreme point density weight matrix.

[0020] Preferably, the steps for obtaining the cluster center centroid projection value matrix are as follows:

[0021] Based on the local extreme point density weight matrix, the clustering characteristics of the feature points are analyzed, and the clustering areas of the feature points are calculated and divided through the feature space transformation method, the positions of the cluster center points are determined, and the clustering division results of each feature point are generated to generate the feature clustering division results;

[0022] Based on the feature clustering division result, determine the position of each cluster center, calculate the center of gravity position of the cluster center, combine the position of the feature point inside the cluster, calculate the coordinates of the center of gravity, and generate the center of gravity position of the cluster center;

[0023] Based on the centroid position of the cluster center, the difference between the centroid and the feature point is analyzed, the distance between each point and the centroid is calculated, and the distance difference values ​​are summarized to generate a projection value matrix of each cluster center centroid point.

[0024] Preferably, the steps for obtaining the cluster center feature weight ranking table are as follows:

[0025] Based on the cluster center centroid projection value matrix, the eigenvalue of the cluster center is analyzed, the vector inner product of each feature point is calculated, and the inner product result is compared with the eigenvalue of the cluster center to generate a feature vector inner product result;

[0026] Based on the inner product result of the feature vector, a difference calculation is performed to calculate the weight value of each feature point and the cluster center feature, and the weight values ​​are sorted according to the weight size to determine the important feature points of the cluster center and generate the weight value of the cluster center feature;

[0027] Based on the cluster center feature weight value, the ranking of each feature point of the cluster center is analyzed, arranged according to importance, and a ranking table is formed to generate a cluster center feature weight ranking table.

[0028] Preferably, the steps for obtaining the cluster center resource allocation optimization table are as follows:

[0029] Based on the cluster center feature weight ranking table, the resource ratio of each feature point is analyzed, and the resource allocation plan is adjusted in combination with the weight value of the feature point to generate a resource ratio adjustment plan;

[0030] Based on the resource ratio adjustment scheme, compare the resource allocation of each feature point, analyze the resource competition intensity between the feature points, form a competition intensity analysis table, and generate a resource competition intensity analysis table;

[0031] Based on the resource competition intensity analysis table, the resource allocation scheme of each feature point is integrated, the overall resource configuration is optimized, the resources of the cluster center are reallocated, and the cluster center resource allocation optimization table is generated.

[0032] Preferably, the feature space transformation method is according to the formula:

[0033]

[0034] Calculate the improved kernel function value between feature points, where K ′ (x, y) is the improved kernel function value of feature point x and feature point y in the feature space, α is the weight coefficient of the similarity between feature point mappings, φ(x) is the mapping of feature point x in the feature space, φ(y) is the mapping of feature point y in the feature space, β is the weight coefficient of feature point position offset, ψ(x) is the offset of feature point x, ψ(y) is the offset of feature point y, γ is the weight coefficient of feature point difference comparison, δ(x) is the density function value of feature point x, and δ(y) is the density function value of feature point y.

[0035] Compared with the prior art, the advantages and positive effects of the present invention are:

[0036] The present invention extracts features of multidimensional data sets, generates feature projection values ​​using Gaussian kernel functions and matrix operations, accurately extracts key features, ensures that sufficient information is retained after data dimensionality reduction, and thus achieves more efficient data processing. Based on the point cloud distribution map, the density calculation of data points and the identification of local extreme points are performed, which improves the positioning accuracy of key data points in low-dimensional space and reduces noise interference in the processing process. After calculating the density weights of local extreme points, the difference matrix is ​​further calculated through clustering and center of gravity position to effectively optimize feature clustering and ensure the accuracy and consistency of cluster centers. The feature points are calculated in parallel through the MapReduce framework, and combined with the dynamic resource allocation scheme to ensure the rationality and efficiency between resource allocation and computing tasks. In summary, the present invention improves the processing capacity of textile equipment operation data, reduces the interference of uncertain factors, and realizes the rational allocation of resources and the improvement of production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0039] In the description of the present invention, it should be understood that the terms length, width, up, down, front, back, left, right, vertical, horizontal, top, bottom, inside, outside, etc. indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of multiple is two or more, unless otherwise clearly and specifically defined.

[0040] See also Figure 1 The present invention provides a technical solution, a method for intelligently processing operation data of automated textile equipment, comprising the following steps:

[0041] During the operation of textile equipment, the multidimensional data set output by the sensor is collected, the Gaussian kernel function is selected, the features of the multidimensional data set are extracted, the feature projection value is generated by matrix operation, the dimension is reduced, and the low-dimensional space mapping is obtained. The density distribution of the data points is calculated by drawing the point cloud distribution map, and the local extreme value points are identified to generate a low-dimensional feature space point cloud map;

[0042] Based on the low-dimensional feature space point cloud, statistical operations are performed on local extreme points and their neighborhood points to calculate the contribution rate of the features to the target classification label, and the distance matrix of the local extreme points is calculated. The distance matrix is ​​used to generate density weights, the influence of the local extreme points is calculated, and the density weight matrix of the local extreme points is generated.

[0043] Based on the local extreme point density weight matrix, the clustering division of the feature point cloud is calculated, the cluster center is generated and the centroid position is calculated. The difference matrix is ​​calculated by the distance between the centroid position and the feature point. The feature clustering similarity is calculated using the difference matrix to generate the cluster center centroid point projection value matrix.

[0044] Based on the projection value matrix of the centroid of the cluster center, the MapReduce framework is used to parallelly calculate the features of the cluster center, calculate the vector inner product and difference comparison of the feature points, generate the feature weights of the cluster center, and sort them to obtain the cluster center feature weight sorting table;

[0045] Based on the weight ranking table of cluster center features and combined with the dynamic resource allocation scheme, the resource ratio of the cluster center feature points is adjusted, the competition intensity is calculated through resource allocation comparison, resource allocation is optimized, and the cluster center resource allocation optimization table is generated.

[0046] The low-dimensional feature space point cloud map includes the two-dimensional coordinates, density values ​​and local extreme point positions of the feature points. The local extreme point density weight matrix includes the density values ​​of the local extreme points, the distance values ​​and influence values ​​of the neighborhood points. The cluster center centroid projection value matrix includes the centroid position, feature point difference values ​​and cluster similarity values. The cluster center feature weight sorting table includes feature weight values, feature point sorting and vector inner product values. The cluster center resource allocation optimization table includes resource allocation ratio, feature point competition intensity and optimized resource allocation value.

[0047] In the step of generating a low-dimensional feature space point cloud map of textile equipment operation data, a multidimensional data set during the operation of the equipment is collected by sensors. The data set includes measurement results of various physical quantities, such as temperature, humidity, tension, etc. When mapping the multidimensional data to a low-dimensional space, a Gaussian kernel function is selected. The kernel function is used to map the high-dimensional data to feature projection values, and the feature projection values ​​are obtained through matrix operations, and a dimensionality reduction operation is performed. The reduced data is represented as two-dimensional or three-dimensional coordinate points, and density calculations are performed on the coordinate points. A point cloud distribution map is drawn, and the local extreme value point position is further identified by calculating the local density extreme value of each data point. The density value and the extreme value point position information together constitute a low-dimensional feature space point cloud map. The generated low-dimensional feature space point cloud map includes the two-dimensional coordinates, density values ​​and local extreme value point positions of the feature points.

[0048] In the step of generating the density weight matrix of local extreme points, the information of the local extreme points and their surrounding neighborhood points is statistically analyzed through further analysis of the low-dimensional feature space point cloud map. First, the contribution rate of each feature point to the target classification label is calculated. The spatial relationship between the neighborhood points is used to determine the specific influence range of each local extreme point, and the influence of the local extreme points is further analyzed by calculating the distance matrix between the points. In this process, it is necessary to accurately calculate the distance between all neighborhood points, and use the distance value to calculate the density weight. By taking the weighted sum of the distance values ​​of all local extreme points and neighborhood points, the density weight matrix of the local extreme points is finally generated. The matrix contains the density value of the local extreme points, the distance value of the neighborhood points, and the influence value.

[0049] In the step of generating the projection value matrix of the centroid points of the cluster centers, the feature point cloud is clustered and divided by the density weight matrix of the local extreme points. First, the density weights of all feature points are clustered and analyzed to divide different cluster areas. Each area represents a set of points with similar features. Then, the cluster centers of each cluster area are calculated, and the centroid position of each cluster center is calculated through the positional relationship of each feature point inside the cluster center. Next, the difference matrix is ​​generated using the distance value between the centroid position and each feature point. The matrix is ​​used to represent the difference between the cluster center and the feature points. By calculating the values ​​in the difference matrix, the feature points inside each cluster area are further analyzed for similarity, and finally the projection value matrix of the centroid points of the cluster center is generated. The matrix includes the centroid position, the feature point difference value and the cluster similarity value.

[0050] In the step of generating the cluster center feature weight sorting table, the features of the cluster centers are calculated in parallel using the MapReduce framework through the projection value matrix of the cluster center centroid points. First, the vector inner product calculation is performed on the feature points of each cluster center, and the projection value of each feature point is converted into a vector representation. Then, the inner product value between each feature point is calculated, and the inner product value between each feature point is compared and analyzed using the difference comparison method. The feature points are sorted according to the size of the inner product difference to generate the cluster center feature weight. The feature points are sorted according to their importance through the weight value to generate the cluster center feature weight sorting table, which includes feature weight values, feature point sorting and vector inner product values. The sorting table can provide a basis for subsequent resource allocation optimization to ensure that in the parallel processing process, those feature points that have a greater impact on the overall operating status can be given priority.

[0051] In the step of generating the cluster center resource allocation optimization table, the resource ratio of the cluster center feature points is adjusted through the cluster center feature weight sorting table combined with the dynamic resource allocation plan. First, the resource allocation ratio of the feature points in the sorting table is preliminarily set, and different resource weights are assigned according to the size of the sorting value. Then, the resource allocation comparison method is used to analyze the resource competition intensity between the feature points, and the competitiveness of each feature point in resource allocation is calculated. The resource allocation is optimized through the competitiveness data, and the calculated optimization value is compared with the initial setting value. The resources are dynamically adjusted, and finally the cluster center resource allocation optimization table is generated. The table includes the resource allocation ratio, the feature point competition intensity and the optimized resource allocation value.

[0052] In this embodiment, the steps for obtaining the low-dimensional feature space point cloud image are as follows:

[0053] Based on the multidimensional data set of sensor output collected during the operation of textile equipment, the key features of each dimension are extracted, matrix operations are performed to obtain feature projection values, matrix transformation is used to reduce the feature projection values ​​to low-dimensional space, feature mapping data is obtained, and low-dimensional space mapping data is generated;

[0054] Based on the low-dimensional space mapping data, the density calculation method is used to calculate the density of each data point, generate a density distribution map of the data point, analyze the local area in the map, identify and mark the local extreme points, and generate a local extreme point distribution map;

[0055] Based on the local extreme point distribution map, the distribution of the point cloud is analyzed, and the relative position relationship of the data points is used to generate a point cloud map of the low-dimensional feature space. The distribution of each feature point in the low-dimensional space is obtained to generate a point cloud map of the low-dimensional feature space.

[0056] In the step of acquiring the low-dimensional feature space point cloud map, the multidimensional data set output by the sensor collected during the operation of the textile equipment is first analyzed. The multidimensional data set contains a variety of physical quantities collected by the equipment during operation, such as temperature, humidity, pressure, tension, etc. Each physical quantity represents a data dimension, and the data dimensions are combined to form a multidimensional feature space. In this sub-step, when extracting the key features of each dimension, statistical analysis is performed on each physical quantity, and the key value that best represents the characteristics of the dimension is selected as the eigenvalue. In order to realize the mapping of multidimensional data to low-dimensional data, the extracted key eigenvalues ​​are first converted into feature projection values ​​through matrix operations, and all eigenvalues ​​need to be standardized so that they can be operated in the same feature space. Then, through the matrix transformation of the feature projection value, the features in the high-dimensional space are mapped to the low-dimensional space, and the commonly used matrix decomposition technology, such as singular value decomposition (SVD), is used to reduce the high-dimensional feature data to two-dimensional or three-dimensional space, and obtain the low-dimensional space mapping data of the features. The low-dimensional data after dimensionality reduction is stored in the form of a matrix. Each element in the matrix represents the coordinate position of a feature point in the low-dimensional space, and the coordinate data is the preliminary representation of the low-dimensional feature space.

[0057] Based on the low-dimensional feature space mapping data, density calculation is one of the core operations of this sub-step. Based on the coordinate data of the low-dimensional space, the density of each data point is evaluated by the density calculation method. In the calculation process, it is first necessary to select an appropriate distance measurement method, such as Euclidean distance or Manhattan distance. According to the selected distance measurement method, the distance between each data point and its neighborhood data point is calculated. In order to obtain a more accurate density value, it is also necessary to consider the size of the neighborhood radius. By setting different radius ranges, the number of data points within the range is calculated to evaluate the density of the data point. The calculation formula of the density value is generally the number of neighborhood data points of each point divided by the neighborhood volume, so as to obtain the density value of each point. After the density values ​​of all data points are calculated, a density distribution map is generated. The horizontal and vertical coordinates of the density distribution map represent the coordinates of the data point in the low-dimensional space, and the color or grayscale represents the density value of the data point. By analyzing the local area in the density distribution map, the local extreme value points can be identified and marked. The marked local extreme value points are extracted to generate a local extreme value point distribution map.

[0058] After the local extreme point distribution map is generated, further analyzing the distribution of the point cloud is the key to generating the low-dimensional feature space point cloud map. In the analysis process, the relative position relationship of the local extreme points needs to be calculated first, and the adjacent data points and their distribution range of each local extreme point are determined by the neighborhood analysis method. In order to accurately describe the distribution of each feature point in the low-dimensional space, the relative position relationship of the data points is used to generate the point cloud. The specific operation includes calculating the position offset of each feature point relative to its neighborhood points, and drawing the point cloud map according to the offset value. The generation of the point cloud map needs to comprehensively consider the density value and position offset of each data point. The low-dimensional feature space point cloud map generated finally contains not only the two-dimensional coordinates of each feature point, but also its density value and the position of the local extreme point. Each point in the point cloud map is assigned a specific color or mark to reflect its position and feature distribution in the low-dimensional space. The low-dimensional feature space point cloud map can intuitively display the spatial distribution characteristics of the data after dimensionality reduction, providing a reliable basis for subsequent data clustering and classification.

[0059] In this embodiment, the steps for obtaining the local extreme point density weight matrix are as follows:

[0060] Based on the low-dimensional feature space point cloud, the distribution characteristics of local extreme points and neighborhood points are analyzed, and the contribution value of each point to the classification label is counted. Combined with the distribution of neighborhood points, the contribution value is calculated to generate the feature contribution value of the local extreme point;

[0061] Based on the characteristic contribution value of the local extreme point, the relative position of the local extreme point is analyzed, and the distance matrix operation is performed to calculate the distance relationship between the neighborhood points. Combined with the density of the neighborhood points, the density weight of each local extreme point is generated, and the local extreme point density matrix is ​​generated;

[0062] Based on the local extreme point density matrix, the density weight of each local extreme point is integrated, the influence of each point is analyzed, and the characteristic contribution value and density weight are combined for calculation to generate the local extreme point density weight matrix.

[0063] In the sub-step of analyzing the distribution characteristics of local extreme points and neighborhood points based on the low-dimensional feature space point cloud map, counting the contribution value of each point to the classification label, and calculating the contribution value in combination with the distribution of the neighborhood points to generate the feature contribution value of the local extreme point, the distribution characteristics of the local extreme point and its neighborhood points in space are determined by in-depth analysis of the low-dimensional feature space point cloud map. The input format of the data includes the two-dimensional coordinates, density value and classification label corresponding to each feature point. First, according to the low-dimensional feature space point cloud map, the contribution value of the classification label is counted for each local extreme point and its neighborhood point. For example, if the label of a local extreme point is consistent with the labels of most of its neighborhood points, the contribution value of the local extreme point is high, otherwise the contribution value is low. The calculation of the contribution value combines the density of the point, the distribution of the neighborhood points and the consistency of the label. The generated contribution value is expressed in numerical form and recorded as one of the important features of the local extreme point. Then, the feature contribution value of each local extreme point is counted through the contribution value.

[0064] In the sub-step of generating the density matrix of local extreme points, the relative position of local extreme points is analyzed based on the feature contribution value of local extreme points, the distance matrix operation is performed, the distance relationship between the neighborhood points is calculated, and the density weight of each local extreme point is generated in combination with the density of the neighborhood points. The data format after the feature contribution value is generated by analyzing the relative position relationship of the local extreme points is the two-dimensional coordinates of the feature points, the density value, the feature contribution value and the label of the neighborhood point. Next, according to the data, the distance relationship between each local extreme point and its neighborhood points is calculated, and the distance between all neighborhood points is calculated in the form of a distance matrix, and the elements of the matrix are the distance values ​​between the two points. In order to more accurately reflect the influence range of the local extreme points, the density value of the neighborhood points needs to be included in the calculation, and the density weight of each local extreme point is generated by combining the distance value and the density value through a weighted method. The calculation formula of the density weight comprehensively considers the relationship between distance and density, so that the neighborhood points with close distance and high density contribute more to the weight of the local extreme points. The generated local extreme point density matrix includes the density weight of each local extreme point.

[0065] In the sub-step of integrating the density weight of each local extreme point based on the local extreme point density matrix, analyzing the influence of each point, combining the characteristic contribution value and the density weight for calculation, and generating the local extreme point density weight matrix, the density weight of each local extreme point is integrated by further analyzing the local extreme point density matrix, and the data format includes the density weight of the local extreme point, the characteristic contribution value, and the density and distance information of the neighborhood points. First, based on the previously calculated density weight, the density information of each local extreme point is weighted and summed, and the comprehensive influence of each local extreme point is calculated in combination with its characteristic contribution value. The calculation of influence needs to consider the proportional relationship between the characteristic contribution value and the density weight. By linearly combining these two parameters, the final influence value of each local extreme point is obtained. In order to ensure the accuracy of the calculation, it is also necessary to introduce additional weight factors to appropriately adjust the characteristic contribution value and the density weight. The local extreme point density weight matrix finally generated includes the comprehensive influence value of each local extreme point.

[0066] In this embodiment, the steps for obtaining the cluster center centroid projection value matrix are as follows:

[0067] Based on the local extreme point density weight matrix, the clustering characteristics of the feature points are analyzed, and the clustering area of ​​the feature points is calculated and divided through the feature space transformation method, the location of the cluster center point is determined, and the clustering division result of each feature point is generated to generate the feature clustering division result;

[0068] Based on the feature clustering results, determine the location of each cluster center, calculate the center of gravity of the cluster center, combine the location of the feature points inside the cluster, calculate the coordinates of the center of gravity, and generate the center of gravity of the cluster center;

[0069] Based on the centroid position of the cluster center, the difference between the centroid and the feature points is analyzed, the distance between each point and the centroid is calculated, and the distance difference values ​​are summarized to generate a projection value matrix of each cluster center centroid point.

[0070] Based on the local extreme point density weight matrix, the clustering characteristics of the feature points are analyzed, the clustering areas of the feature points are calculated and divided through the feature space transformation method, the location of the cluster center point is determined, and the clustering results of each feature point are generated. In the sub-step of generating the feature clustering results, the density weight of each feature point is first obtained by analyzing the local extreme point density weight matrix, and the data includes the spatial coordinates, density weight and label information of each feature point. The high-dimensional feature data is mapped to the low-dimensional space by the feature space transformation method, with the purpose of representing the density weight information of the feature points more intuitively in the low-dimensional space. After completing the feature space transformation, the position relationship of each feature point in the low-dimensional space is calculated by traversing point by point, the spatial distance and density difference are analyzed, and the feature points are divided into different clustering areas according to the relationship. In the specific operation process, it is necessary to calculate the distance value between each feature point and other feature points, and based on this, the feature points with similar density weights and close spatial distances are classified into one category, thereby generating the clustering results. At this time, each feature point is assigned a cluster label, which corresponds to a specific cluster area, and the center point position of each cluster is determined according to the label. The final generated feature clustering result file includes the coordinates of the feature point, cluster label and cluster center point position.

[0071] In the sub-step of determining the position of each cluster center based on the feature clustering division result, calculating the centroid position of the cluster center, combining the position of the feature points inside the cluster, calculating the coordinates of the centroid, and generating the centroid position of the cluster center, through further analysis of the feature clustering division result, the specific processing data format includes the coordinates of each feature point, the cluster label and the weight information of the feature point. First, according to the label information of each cluster, the coordinates and weight information of all feature points belonging to the same cluster are extracted. The centroid position of the cluster center is calculated by weighted averaging the coordinates of the feature points. In the specific operation, the coordinates of each feature point are multiplied by its corresponding weight, and the weighted coordinate values ​​of all feature points are added and divided by the sum of the weights to obtain the centroid coordinates of the cluster center. The centroid coordinates reflect the comprehensive position of the feature points in the entire cluster area and are highly representative.

[0072] Based on the centroid position of the cluster center, the difference between the centroid and the feature point is analyzed, the distance between each point and the centroid is calculated, and the distance difference values ​​are summarized to generate the projection value matrix of each cluster center centroid point. In the sub-step of generating the projection value matrix of the cluster center centroid point, the difference between the centroid position of the cluster center and the feature point is analyzed, and the data format includes the centroid coordinates, the feature point coordinates and the weight information of the feature point. First, the distance between each feature point and its corresponding cluster center centroid is calculated. In the specific operation, the coordinates of each feature point are vector-subtracted from the centroid coordinates to obtain the difference vector, and then the modulus of the difference vector is calculated, which is the distance between the feature point and the centroid. Next, the distance value between each feature point and the centroid is summarized, and the average distance difference value of the feature points inside the entire cluster is calculated, which is used as the projection value of the cluster. Through the calculation process, the projection value matrix of each cluster center centroid point is generated, and the matrix content includes the centroid position of each cluster, the feature point difference value and the projection value information of the cluster.

[0073] In this embodiment, the steps for obtaining the cluster center feature weight ranking table are as follows:

[0074] Based on the projection value matrix of the centroid of the cluster center, the eigenvalue of the cluster center is analyzed, the vector inner product of each characteristic point is calculated, and the inner product result is compared with the eigenvalue of the cluster center to generate the characteristic vector inner product result;

[0075] Based on the inner product result of the feature vector, the difference calculation is performed to calculate the weight value of each feature point and the cluster center feature, sort them by weight, determine the important feature points of the cluster center, and generate the weight value of the cluster center feature;

[0076] Based on the feature weight value of the cluster center, the ranking of each feature point of the cluster center is analyzed, arranged according to importance, and a ranking table is formed to generate a cluster center feature weight ranking table.

[0077] In the sub-step of analyzing the eigenvalue of the cluster center based on the projection value matrix of the centroid of the cluster center, calculating the vector inner product of each feature point, comparing the inner product result with the eigenvalue of the cluster center, and generating the inner product result of the eigenvector, the inner product of the eigenvector is calculated by analyzing the coordinate data of each feature point in the projection value matrix of the centroid of the cluster center and the eigenvalue of the cluster center. The input format of the data includes the spatial coordinates of each feature point, the centroid coordinates of the cluster center, and the eigenvalue of each feature point. In the specific operation steps, the vector inner product of each feature point and the coordinate value of the cluster center is calculated, and the calculation reflects the similarity between the feature point and the cluster center in the feature space. In the process of inner product calculation, the vector of each feature point is multiplied element by element with the vector of the cluster center, and the inner product value of the feature point is obtained after the results are summed. The calculated inner product value is compared with the eigenvalue of the cluster center, and the difference between the inner product value and the eigenvalue is analyzed to obtain the inner product result of the eigenvector of the feature point. The result is used to measure the relative importance of the position of the feature point in the cluster, and provides a basis for the subsequent weight value calculation. The final generated eigenvector inner product result file includes the inner product value, eigenvalue and difference information with the cluster center of each feature point, laying the foundation for the identification and sorting of important feature points in the cluster center.

[0078] In the sub-step of performing difference calculation based on the inner product result of the feature vector, calculating the weight value of each feature point and the feature of the cluster center, sorting them according to the weight size, determining the important feature points of the cluster center, and generating the weight value of the feature of the cluster center, the weight value of the feature point is further calculated by analyzing the difference of the inner product result of the feature vector. The input data format includes the inner product value of the feature point, the eigenvalue of the cluster center, and the difference information between the two. In the specific operation, the weight value of the feature point and the cluster center is first calculated, and by analyzing the difference between the inner product value and the eigenvalue, the difference value is multiplied by a predetermined weight coefficient using the weighted calculation method to ensure the contribution of the difference to the weight calculation. The selection of the weight coefficient is based on the importance of the eigenvalue and the overall distribution characteristics of the cluster center. After the weighted operation, the weight value of each feature point is calculated. The larger the weight value, the greater the contribution of the feature point to the cluster center. The weight values ​​of all feature points are sorted to identify the most important feature points for the cluster center. The cluster center feature weight value file finally generated includes the weight value of each feature point and its corresponding sorting information.

[0079] In the sub-step of analyzing the ranking of each feature point in the cluster center based on the feature weight value of the cluster center, arranging them according to importance, and forming a ranking table to generate the feature weight ranking table of the cluster center, the ranking order of each feature point in the cluster center is determined by analyzing the feature weight value of the cluster center. The input data format includes the weight value of the feature point and its corresponding ranking information. During the operation, the weight values ​​of the feature points are first reorganized and arranged in order from large to small. According to the arranged order, the importance position of each feature point in the cluster center is determined. Then, based on the sorting result, the cluster center feature weight ranking table is generated. The generation process of the ranking table includes corresponding and recording the weight value, coordinates and ranking position of the feature point one by one to ensure the accuracy and completeness of the data. The cluster center feature weight ranking table finally generated includes the weight value of the feature point, the ranking information and its position in the cluster.

[0080] In this embodiment, the steps for obtaining the cluster center resource allocation optimization table are as follows:

[0081] Based on the cluster center feature weight ranking table, analyze the resource ratio of each feature point, adjust the resource allocation plan based on the weight value of the feature point, and generate a resource ratio adjustment plan;

[0082] Based on the resource ratio adjustment plan, compare the resource allocation of each feature point, analyze the resource competition intensity between the feature points, form a competition intensity analysis table, and generate a resource competition intensity analysis table;

[0083] Based on the resource competition intensity analysis table, the resource allocation plan of each feature point is integrated to optimize the overall resource configuration, reallocate the resources of the cluster center, and generate the cluster center resource allocation optimization table.

[0084] In the sub-step of analyzing the resource ratio of each feature point based on the cluster center feature weight sorting table, adjusting the resource allocation plan in combination with the weight value of the feature point, and generating the resource ratio adjustment plan, the key parameters for resource allocation are extracted by analyzing the data in the cluster center feature weight sorting table, including the weight value and sorting information of each feature point. The parameters include the weight value of the feature point, the resource demand, and the total amount of resources currently available. In the specific operation, the priority of the resource demand of the feature point is first calculated according to the weight value of the feature point, that is, the feature point with a higher weight value has the priority to meet the resource demand. On this basis, the proportional allocation method is adopted to preliminarily allocate resources according to the weight proportion of each feature point. Subsequently, when adjusting the allocation plan, it is necessary to comprehensively consider the effective utilization rate of resources, and a resource utilization coefficient will be introduced to fine-tune the preliminary allocation plan of each feature point. Through this resource ratio adjustment, it is ensured that resources can be effectively allocated between different feature points, and the generated resource ratio adjustment plan file includes the adjusted resource allocation ratio of each feature point and its corresponding resource amount.

[0085] In the sub-step of comparing the resource allocation of each feature point based on the resource ratio adjustment plan, analyzing the resource competition intensity between the feature points, and forming a competition intensity analysis table, the resource competition relationship between different feature points is analyzed by comparing the resource allocation of each feature point horizontally. The input data includes the resource allocation ratio of each feature point, the adjusted resource allocation amount, and the resource demand of the feature point. In the specific operation process, the gap between the actual resource amount allocated to each feature point and its demand is first calculated. The larger the gap, the higher the competition intensity between the feature point and other feature points. Next, the resource competition intensity between each feature point is calculated by constructing a competition intensity matrix, in which each element of the matrix represents the resource competition intensity between two feature points, which is calculated by the resource demand difference between the two points and its corresponding weight value. The resource competition intensity analysis table file finally formed includes the competition intensity value of each feature point and its competition relationship information with other feature points.

[0086] In the sub-step of integrating the resource allocation scheme of each feature point based on the resource competition intensity analysis table, optimizing the overall resource configuration, reallocating the resources of the cluster center, and generating the cluster center resource allocation optimization table, the resource allocation scheme of each feature point is optimized by further analyzing the resource competition intensity analysis table to ensure the optimal configuration of the overall resources. The input data format includes the resource allocation ratio of the feature point, the competition intensity value, and the adjusted resource allocation amount. During the operation, the resource allocation scheme is first adjusted according to the competition intensity value, and more resources are given priority to the feature points with high competition intensity, thereby reducing the bottleneck of resource competition. Then, the resource allocation of each feature point is reallocated, and the total resources are reallocated between the feature points to ensure the maximization of resource utilization. At the same time, an optimization coefficient is introduced to fine-tune the resource allocation according to actual needs, and the final optimized resource allocation scheme ensures the reasonable distribution of resources between different feature points. The final generated cluster center resource allocation optimization table file includes the final resource allocation amount of each feature point and its optimized allocation ratio.

[0087] In this embodiment, the feature space transformation method is based on the formula:

[0088]

[0089] Calculate the improved kernel function value between feature points, where K′(x, y) is the improved kernel function value of feature point x and feature point y in the feature space, α is the weight coefficient of the similarity between feature point mappings, φ(x) is the mapping of feature point x in the feature space, φ(y) is the mapping of feature point y in the feature space, β is the weight coefficient of feature point position offset, ψ(x) is the offset of feature point x, ψ(y) is the offset of feature point y, γ is the weight coefficient of feature point difference comparison, δ(x) is the density function value of feature point x, and δ(y) is the density function value of feature point y;

[0090] Execution process: First, determine the mapping φ(x) and φ(y) of the feature points x and y in the feature space, calculate the dot product φ(x)·φ(y), then calculate the offsets ψ(x) and ψ(y) of the feature points x and y, sum them and multiply them by the weight coefficient β, then calculate the density function values ​​δ(x) and δ(y) of the feature points x and y, find the absolute value of the difference between the feature point mappings |φ(x)-φ(y)|, divide it by δ(x)+δ(y) to get the ratio, and then multiply it by the weight coefficient γ, finally, linearly combine the three calculation results, multiply them by the corresponding weight coefficients α, β, γ respectively, and summarize them to generate the final improved kernel function value K′(x,y).

[0091] Confirmation method of weight coefficients α, β, and γ: Through historical data analysis, according to the classification effect of the target feature point set under different parameter weights, the best weight coefficient combination is determined by cross-validation method. The specific steps include: setting the initial weight value, then performing preliminary calculations on the training set, then adjusting the weight value according to the feedback of the classification effect, repeating cross-validation until the optimal combination is obtained, and finally verifying the final weight coefficient on the test set.

[0092] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0093] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0094] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for intelligently processing operation data of automated textile equipment, characterized in that: The following steps are involved: During the operation of textile equipment, the multidimensional data set output by the sensor is collected, the Gaussian kernel function is selected, the features of the multidimensional data set are extracted, the feature projection value is generated by matrix operation, the dimension is reduced, and the low-dimensional space mapping is obtained. The density distribution of the data points is calculated by drawing the point cloud distribution map, and the local extreme value points are identified to generate a low-dimensional feature space point cloud map; Based on the low-dimensional feature space point cloud map, statistical operations are performed on the local extreme value points and their neighborhood points, the contribution rate of the features to the target classification label is calculated, the distance matrix of the local extreme value points is calculated, the density weight is generated by using the distance matrix, the influence of the local extreme value points is calculated, and the density weight matrix of the local extreme value points is generated. The specific calculation process includes: based on the low-dimensional feature space point cloud map, the distribution characteristics of the local extreme value points and the neighborhood points are analyzed, the contribution value of each point to the classification label is counted, and the contribution value is calculated in combination with the distribution of the neighborhood points to generate the local extreme value point feature contribution value; based on the local extreme value point feature contribution value, the relative position of the local extreme value point is analyzed, the distance matrix operation is performed, the distance relationship between the neighborhood points is calculated, and the density weight of each local extreme value point is generated in combination with the density of the neighborhood points to generate the local extreme value point density matrix; based on the local extreme value point density matrix, the density weight of each local extreme value point is integrated, the influence of each point is analyzed, and the feature contribution value and the density weight are combined for calculation to generate the local extreme value point density weight matrix; Based on the local extreme point density weight matrix, the cluster division of the feature point cloud is calculated, the cluster center is generated and the centroid position is calculated, the difference matrix is ​​calculated by the distance between the centroid position and the feature point, the feature clustering similarity is calculated using the difference matrix, and the cluster center centroid point projection value matrix is ​​generated; Based on the projection value matrix of the centroid points of the cluster centers, the features of the cluster centers are calculated in parallel using the MapReduce framework, the vector inner product and difference comparison of the feature points are calculated, the feature weights of the cluster centers are generated, and the feature weight ranking table of the cluster centers is obtained by sorting. Based on the cluster center feature weight ranking table, combined with the dynamic resource allocation scheme, the resource ratio of the cluster center feature points is adjusted, the competition intensity is calculated through resource allocation comparison, resource allocation is optimized, and a cluster center resource allocation optimization table is generated.

2. The method for intelligently processing the operation data of automated textile equipment according to claim 1, characterized in that: The low-dimensional feature space point cloud map includes the two-dimensional coordinates, density values ​​and local extreme point positions of the feature points, the local extreme point density weight matrix includes the density values ​​of the local extreme points, the distance values ​​and influence values ​​of the neighborhood points, the cluster center centroid point projection value matrix includes the centroid position, feature point difference values ​​and cluster similarity values, the cluster center feature weight sorting table includes feature weight values, feature point sorting and vector inner product values, and the cluster center resource allocation optimization table includes resource allocation ratio, feature point competition intensity and optimized resource allocation value.

3. The method for intelligently processing the operation data of automated textile equipment according to claim 1, characterized in that: The steps for obtaining the low-dimensional feature space point cloud image are specifically as follows: Based on the multidimensional data set of sensor output collected during the operation of textile equipment, the key features of each dimension are extracted, matrix operations are performed to obtain feature projection values, matrix transformation is used to reduce the feature projection values ​​to low-dimensional space, feature mapping data is obtained, and low-dimensional space mapping data is generated; Based on the low-dimensional space mapping data, a density calculation method is used to calculate the density of each data point, generate a density distribution map of the data point, analyze the local area in the map, identify and mark the local extreme value points, and generate a local extreme value point distribution map; Based on the local extreme point distribution map, the distribution of the point cloud is analyzed, the relative position relationship of the data points is used to generate a point cloud map of the low-dimensional feature space, the distribution of each feature point in the low-dimensional space is obtained, and the low-dimensional feature space point cloud map is generated.

4. The method for intelligently processing the operation data of automated textile equipment according to claim 1, characterized in that: The steps for obtaining the cluster center centroid projection value matrix are as follows: Based on the local extreme point density weight matrix, the clustering characteristics of the feature points are analyzed, and the clustering areas of the feature points are calculated and divided through the feature space transformation method, the positions of the cluster center points are determined, and the clustering division results of each feature point are generated to generate the feature clustering division results; Based on the feature clustering division result, determine the position of each cluster center, calculate the center of gravity position of the cluster center, combine the position of the feature point inside the cluster, calculate the coordinates of the center of gravity, and generate the center of gravity position of the cluster center; Based on the centroid position of the cluster center, the difference between the centroid and the feature point is analyzed, the distance between each point and the centroid is calculated, and the distance difference values ​​are summarized to generate a projection value matrix of each cluster center centroid point.

5. The method for intelligently processing the operation data of automated textile equipment according to claim 1, characterized in that: The steps for obtaining the cluster center feature weight ranking table are as follows: Based on the cluster center centroid projection value matrix, the eigenvalue of the cluster center is analyzed, the vector inner product of each feature point is calculated, and the inner product result is compared with the eigenvalue of the cluster center to generate a feature vector inner product result; Based on the inner product result of the feature vector, a difference calculation is performed to calculate the weight value of each feature point and the cluster center feature, and the weight values ​​are sorted according to the weight size to determine the important feature points of the cluster center and generate the weight value of the cluster center feature; Based on the cluster center feature weight value, the ranking of each feature point of the cluster center is analyzed, arranged according to importance, and a ranking table is formed to generate a cluster center feature weight ranking table.

6. The method for intelligently processing the operation data of automated textile equipment according to claim 1, characterized in that: The steps for obtaining the cluster center resource allocation optimization table are as follows: Based on the cluster center feature weight ranking table, the resource ratio of each feature point is analyzed, and the resource allocation plan is adjusted in combination with the weight value of the feature point to generate a resource ratio adjustment plan; Based on the resource ratio adjustment scheme, compare the resource allocation of each feature point, analyze the resource competition intensity between the feature points, form a competition intensity analysis table, and generate a resource competition intensity analysis table; Based on the resource competition intensity analysis table, the resource allocation scheme of each feature point is integrated, the overall resource configuration is optimized, the resources of the cluster center are reallocated, and the cluster center resource allocation optimization table is generated.

7. The method for intelligently processing the operation data of automated textile equipment according to claim 4, characterized in that: The feature space transformation method is according to the formula: ； Calculate the improved kernel function value between feature points, where For feature points and feature points The improved kernel function value in the feature space, is the weight coefficient of the similarity between feature point mappings, For feature points The mapping in feature space is For feature points The mapping in feature space is is the weight coefficient of feature point position offset, For feature points The offset of For feature points The offset of is the weight coefficient of feature point difference comparison, For feature points The density function value of For feature points The density function value of .

Images