Connector mold accessory machining parameter monitoring method based on big data

Abstract

The invention relates to the field of data processing, in particular to a connector mold accessory machining parameter monitoring method based on big data, and the method comprises the steps: obtaining and preprocessing workpiece coordinate and machining coordinate data in a connector mold accessory machining process, and constructing a node dimension anomaly index to reflect the anomaly degree of each node. Furthermore, an abnormal effective segmentation index is constructed based on the abnormal index, and the ability of the father node to identify the abnormal processing moment is evaluated. And by analyzing the descending trend similarity of the two, a trend cooperation index is constructed to evaluate the detection accuracy of the isolated tree. And finally, correcting an abnormal score in the isolated forest algorithm by using a trend synergy index to obtain a significant abnormal score, and realizing accurate monitoring of the processing parameters. According to the method, the abnormal data detection precision is improved by analyzing the abnormal degree of the nodes to construct the index and screening the effective detection nodes, accurate monitoring is realized, and the product quality is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing. In particular, it relates to a method for monitoring processing parameters of connector mold accessories based on big data. Background Art

[0002] Connector mold accessories are various components used to manufacture connector molds. These components together form a complete mold for producing connector products. They usually include a mold base, guide pillars, guide sleeves, retaining rings, screws, etc. The processing precision, surface treatment, and assembly quality of these accessories directly affect the performance and service life of the mold.

[0003] To ensure the high precision and stability of connector molds, it is particularly important to monitor processing parameters in real time. By monitoring processing parameters, the product quality can be effectively improved, and it can be ensured that the dimensional accuracy, appearance quality, and mechanical and electrical performance of the connector meet the standard requirements. At the same time, real-time monitoring can promptly detect abnormal situations during the production process, such as equipment failures, material defects, etc., thereby avoiding the production of a large number of unqualified products and reducing production costs. In addition, monitoring processing parameters helps to optimize the production process, improve production efficiency, extend the mold life, and ensure the stability and continuity of production.

[0004] However, in the actual monitoring process, although the traditional Isolation Forest algorithm is an unsupervised learning algorithm that can quickly detect anomalies in a large amount of data, its limitations cannot be ignored. For example, the weights of each isolation tree in the Isolation Forest algorithm are the same, and the detection effects of different isolation trees on abnormal data may vary, resulting in a low detection accuracy when monitoring the relevant parameters in the processing of connector mold accessories. Summary of the Invention

[0005] To solve the problem that the weights of each isolation tree in the traditional Isolation Forest algorithm are the same, and the detection effects of different isolation trees on abnormal data may vary, resulting in a low detection accuracy when monitoring the processing parameters of connector mold accessories, the present invention provides solutions in the following aspects.

[0006] A method for monitoring machining parameters of connector mold fittings based on big data, comprising: acquiring relevant parameters in the machining process of connector mold fittings and performing preprocessing, wherein the relevant parameters include: workpiece coordinate data and machining coordinate data; taking the degree of abnormality of the relevant parameters at the machining moment of the connector mold fittings as the abnormality index for constructing the node dimension in the isolation tree, so as to reflect the degree of abnormality at the machining moment of each node dimension; based on the abnormality index of the node dimension, constructing an abnormal effective segmentation index for identifying the abnormal machining moment of the connector mold fittings by the parent node corresponding to each node in the isolation tree; according to the similarity of the decline trends of the abnormality index and the abnormal effective segmentation index in the isolation tree, constructing a trend cooperation index to evaluate the accuracy of the detection effect of the isolation tree for the abnormal machining moment of the connector mold fittings; based on the trend cooperation index, correcting the calculation result of the abnormal score in the isolation forest algorithm to obtain a significant abnormal score, and realizing the monitoring of the machining parameters of the connector mold fittings.

[0007] By acquiring and preprocessing the relevant parameters in the machining process of the connector mold fittings, constructing the abnormality index of the node dimension, and accurately reflecting the degree of abnormality at the machining moment of each node dimension. Based on the abnormality index, further constructing the abnormal effective segmentation index to effectively evaluate the ability of the parent node to distinguish abnormal and normal machining moments, and screening out the nodes that are more effective for abnormal detection. According to the similarity of the decline trends of the abnormality index and the abnormal effective segmentation index, constructing the trend cooperation index to evaluate and optimize the detection effect of the isolation tree, highlighting the weight of the isolation tree with good detection effect and reducing the weight of the one with poor effect, so as to improve the overall abnormal detection accuracy, realize the accurate monitoring of the machining parameters of the connector mold fittings, and improve the product quality and performance.

[0008] Preferably, the preprocessing of the relevant parameters includes: Taking the relevant parameters of the connector mold fittings preset by the data detection system in the workshop as the standard parameters; taking the absolute value of the difference between the real-time acquired relevant parameters and the standard parameters as the offset coordinate data, and each moment corresponds to an offset coordinate data, so as to monitor the abnormal situation in the machining process.

[0009] Preferably, the abnormality index of the node dimension includes: For a certain node in a certain layer of a certain isolation tree in the isolation forest, traverse the ratio between the offset data of all machining moments in the node and the maximum value of all offset data in the corresponding layer to obtain the relative offset, use the exponential function to map the relative offset, and sum up the mapping values of all machining moments to obtain the abnormality index of the corresponding node.

[0010] By traversing the offset data at all processing times within a node, calculating the relative offset and using an exponential function for mapping, and summing up the mapped values at all processing times, the anomaly index corresponding to the node is obtained, thus accurately reflecting the anomaly degree of the processing times within the node and effectively identifying anomalies in the processing process.

[0011] Preferably, the anomaly index of the node dimension includes: For a certain node in a certain layer of a certain isolation tree in the isolation forest, traverse all the processing times included, calculate the difference between the offset coordinate data corresponding to each processing time and the mode of the dimension data selected in the corresponding layer of the offset coordinate data of all processing times in the isolation tree, and input the difference into the activation function. The output result will set negative values to zero and keep positive values unchanged. Sum up the output results of the activation function at all processing times to obtain the anomaly index of the node dimension corresponding to the node.

[0012] By calculating the difference between the offset coordinate data of each processing time and the mode, and using the activation function for processing, setting negative values to zero and keeping positive values, and finally summing up to obtain the anomaly index of the node dimension, thus effectively identifying and quantifying the anomaly degree of the processing times within the node and improving the accuracy of anomaly detection.

[0013] Preferably, the effective anomaly segmentation index includes: For a certain node in a certain layer of a certain isolation tree in the isolation forest, traverse and calculate the ratio between the maximum value of the node dimension mean within the node and the sum of the minimum value of the node dimension mean and the constant 1 to obtain the time interval ratio; Traverse and calculate the absolute value of the difference between the anomaly index of each node and the anomaly index of the dimension of another child node of the same parent node to obtain the anomaly difference between sibling nodes; Take the product of the time interval ratio and the anomaly difference between sibling nodes as the effective anomaly segmentation index of the parent node corresponding to a certain node in a certain layer of a certain isolation tree in the isolation forest.

[0014] By calculating the time interval ratio and the anomaly difference between sibling nodes and combining the two, the effective anomaly segmentation index is obtained, thus effectively evaluating the ability of the parent node to identify anomaly processing times and improving the accuracy and reliability of anomaly detection.

[0015] Preferably, the similarity based on the downward trends of the anomaly index and the effective anomaly segmentation index in the isolation tree includes: Traverse and calculate the average value of the anomaly index of the leaf nodes in each layer of the isolation tree, and sort them in ascending order according to the depth of the isolation tree. Use the sorted sequence as the node anomaly trend sequence; traverse and calculate the average value of the anomaly effective segmentation index of the parent nodes corresponding to each layer of nodes in the isolation tree, and sort them in ascending order according to the depth of the isolation tree. Use the sorted sequence as the in-layer classification effect sequence; Use the Mann-Kendall algorithm to test the downward trend of the node anomaly trend sequence and the in-layer classification effect sequence respectively, obtain the output results of the corresponding sequences, and analyze the similarity of the output results for calculating the trend synergy index.

[0016] By calculating the node anomaly trend sequence and the in-layer classification effect sequence, and using the Mann-Kendall algorithm to test their downward trends, analyze the similarity of the output results of the two sequences, so as to effectively calculate the trend synergy index, help evaluate the detection effect of the isolation tree on the abnormal processing moment, and improve the accuracy and stability of anomaly detection.

[0017] Preferably, the trend synergy index includes: Taking any isolation tree as the target isolation tree, calculate the absolute difference between the output result of the node anomaly trend sequence of the target isolation tree and the output result of the in-layer classification effect sequence; Take the reciprocal of the sum of the absolute difference and 1 as the trend synergy index of the target isolation tree.

[0018] By calculating the absolute difference between the output results of the node anomaly trend sequence and the in-layer classification effect sequence of the target isolation tree, and taking the reciprocal of the sum of it and 1 as the trend synergy index, it can quantitatively evaluate the detection effect of the target isolation tree on the abnormal processing moment.

[0019] Preferably, in the Mann-Kendall algorithm, it is preset that the test results of the node anomaly trend sequence and the in-layer classification effect sequence have a downward trend. Among them, the inputs are the node anomaly trend sequence and the in-layer classification effect sequence, and the outputs are the value of the hypothesis test of the node anomaly trend sequence and the value of the hypothesis test of the in-layer classification effect sequence.

[0020] Preferably, the significant anomaly score includes: Taking any isolation tree as the target isolation tree, use the trend synergy index of the target isolation tree as the weight of the anomaly score of the isolation tree, and perform a weighted sum of the weights of the anomaly scores of all isolation trees in the isolation forest algorithm to obtain the significant anomaly score corresponding to the processing moment.

[0021] Preferably, the implementation of monitoring the processing parameters of the connector mold accessories includes: In response to the significant anomaly score being greater than or equal to the preset anomaly threshold If so, there is a serious abnormality in the current processing process; in response to the significant abnormality score being greater than the preset abnormality threshold Less than the preset abnormality threshold If so, there is a medium abnormality in the current processing process; in response to the significant abnormality score being less than or equal to the preset abnormality threshold If so, there is a slight abnormality or no abnormality in the processing process; Take the detection result as the monitoring result of the processing parameters of the connector mold parts at the current moment, and realize the monitoring of the processing parameters of the connector mold parts based on big data.

[0022] The present invention has the following effects: 1. By analyzing the abnormality degree of the processing moment of the connector mold parts in the node of the isolation tree, the present invention constructs an abnormality index in the node dimension. Based on the abnormality index in the node dimension, the recognition accuracy of the parent node for abnormal and normal processing moments is further evaluated, and an effective abnormality segmentation index is constructed. This helps to effectively identify the parent nodes that perform better in distinguishing abnormal and normal processing moments, so as to screen out the nodes that are more effective for abnormality detection. 2. By analyzing the similarity of the decreasing trends of the abnormality index in the node dimension of the leaf node and the effective abnormality segmentation index of the parent node in ascending order of the depth of the isolation tree, the present invention constructs a trend coordination index. Using it as the weight for calculating the final abnormality score in the isolation forest algorithm can increase the weight of the isolation tree with better detection effect for abnormal data and reduce the weight of the isolation tree with poor effect, ultimately improving the detection accuracy of abnormal data, realizing accurate monitoring of the processing process of the connector mold parts, and further improving the product quality and performance. Description of the Drawings

[0023] Figure 1 It is a flowchart of the method from step S1 to step S5 in a method for monitoring the processing parameters of connector mold parts based on big data according to an embodiment of the present invention. Detailed Embodiments

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments.

[0025] Refer to Figure 1 , a method for monitoring the processing parameters of connector mold parts based on big data includes steps S1 - S5, specifically as follows: S1: Obtain the relevant parameters in the processing process of the connector mold parts and perform preprocessing. Among them, the relevant parameters include: workpiece coordinate data and machining coordinate data.

[0026] It should be noted that the workpiece coordinate data refers to the spatial coordinates of the connector mold fittings on the lathe, and the machining coordinate data refers to the length, width, and depth data that the equipment (such as a robotic arm) for machining the connector mold fittings needs to process the fittings.

[0027] Preprocess the relevant parameters, including: Take the relevant parameters of the connector mold fittings preset by the data detection system in the workshop as the standard parameters; take the absolute value of the difference between the real-time obtained relevant parameters and the standard parameters as the offset coordinate data, and each moment corresponds to an offset coordinate data, which is used to monitor abnormal situations during the machining process.

[0028] It should be noted that when machining the connector mold fittings on the lathe, the data monitoring system in the production workshop presets the workpiece coordinate data of the corresponding position of the fittings in the lathe and the machining coordinate data of the equipment used when machining the mold fittings, that is, the workpiece coordinate data and machining coordinate data under standard conditions when producing and machining the connector mold fittings.

[0029] Respectively take the absolute values of the differences between the workpiece coordinate data and machining coordinate data when machining the connector mold fittings and the preset workpiece coordinate data and machining coordinate data as the offset coordinate data of the mold fittings, denoted as That is, there will be corresponding offset coordinate data at each data acquisition moment, where 、 、 respectively represent the offsets of the workpiece coordinate data of the connector mold fittings and the preset workpiece coordinate data in the axis, axis and axis in the space coordinate system, 、 、 respectively represent the offsets of the machining coordinate data of the connector mold fittings and the preset machining coordinate data in the axis, axis and axis in the space coordinate system.

[0030] In this embodiment, the preset number of isolation trees is 100, the depth of the isolation trees is 15, and the number of moments randomly selected when constructing the isolation trees is 256. Since the dimension of the relevant parameter data collected in the connector mold fittings processed in the present invention is small, all dimensions are used as the dimensions selected when constructing the isolation trees. The value in the present invention is 6, which can be selected according to the situation. Thus, 100 isolation trees constructed using the traditional isolation forest algorithm are obtained.

[0031] In the process of constructing an isolation tree, abnormal data will be classified into the leaf nodes of the isolation tree. Therefore, when processing connector mold accessories on a lathe in a production workshop, the greater the deviation between the coordinates of the mold accessories and the preset coordinates, and the processing coordinates and the preset processing coordinates, the more the current position of the mold accessories and the processing position of the processing equipment deviate from the preset position. That is, the more abnormal the coordinate position of the mold accessories and the processing position of the processing equipment are at this time, the easier it is for the data to be classified into the leaf nodes of the isolation tree.

[0032] S2: The abnormality degree of the relevant parameters of the connector mold accessories processing time is used as the abnormality index of the node dimension in the isolated tree to reflect the abnormality degree of the processing time of each node dimension.

[0033] Abnormal index of node dimension, including: For a node in a certain layer of an isolated tree in the isolation forest, the ratio between the offset data of all processing moments in the node and the maximum value of all offset data in the corresponding layer is traversed to obtain the relative offset. The relative offset is mapped using an exponential function, and the mapped values ​​of all processing moments are summed to obtain the abnormality index of the corresponding node.

[0034] Specifically, the anomaly index satisfies the following relationship: ; In the formula, Represents the first The first Layer The abnormal index of the node dimension of the nodes, Indicates The first Layer The number of connector mold parts processing times within a node, Represented by natural numbers The exponential function with base , Indicates The first Layer The first node of the connector mold assembly The dimension value of the offset coordinate data at each processing moment, Indicates The offset coordinate data of all processing moments in an isolated tree are The maximum value for a dimension in a layer.

[0035] Reflects the node Inner connector mold parts The relative abnormality degree of each processing moment is amplified through the exponential function to amplify the abnormality degree of the abnormal processing moment.

[0036] In addition, in another embodiment, it further includes: For a certain node in a certain layer of a certain isolated tree in the isolation forest, traverse all the processing times included, calculate the difference between the offset coordinate data corresponding to each processing time and the mode of the offset coordinate data of all processing times in the isolation tree for the dimension data selected in the corresponding layer, and input the difference into the activation function. The output result will set negative values to zero and keep positive values unchanged. For all processing times sum up the output results of the activation function to obtain the anomaly index of the node dimension corresponding to the node.

[0037] It should be noted that since the abnormal data in the isolation tree is relatively small, the mode of each dimension data can reflect the common offset in the normal processing process. It represents the offset level of most normal processing times.

[0038] If the offset coordinate data of a certain processing time is greater than the mode of each dimension data, it indicates that the offset at this time exceeds the normal range and may belong to an abnormal situation. The function retains this excess part, indicating the degree of abnormal offset.

[0039] In anomaly detection, the anomaly index usually represents the magnitude of the anomaly degree and should be non - negative. The function can ensure that the output value is non - negative, which conforms to the physical meaning of the anomaly index. The function can amplify the influence of the abnormal offset. When the offset coordinate data is greater than the mode, the function retains the difference, indicating the degree of abnormal offset; when the offset coordinate data is less than or equal to the mode, the function sets the difference to zero, indicating the normal offset.

[0040] Specifically, the anomaly index satisfies the following relational expression: ; In the formula, represents the anomaly index of the node dimension of the th isolated tree in the th layer and the th node, represents the number of processing times of the connector die fittings in the th isolated tree in the th layer and the th node, represents the activation function, represents the th isolated tree in the th layer and the The dimensional value of the offset coordinate data at the th processing time of the connector mold fitting within a node, indicating the mode of the dimensional data selected at the layer for the offset coordinate data at all processing times in the

[0041] S3: Based on the anomaly index of the node dimension, for the identification of the abnormal processing time of the connector mold fitting by the parent node corresponding to each node in the isolation forest, construct an effective anomaly segmentation index.

[0042] The effective anomaly segmentation index includes: For a certain node in a certain layer of a certain isolation tree in the isolation forest, traverse and calculate the ratio between the maximum value of the node dimension mean within the node and the sum of the minimum value of the node dimension mean and the constant 1 to obtain the time interval ratio; Traverse and calculate the absolute value of the difference between the anomaly index of each node and the anomaly index of the dimension of another child node of the same parent node to obtain the sibling node anomaly difference; Take the product of the time interval ratio and the sibling node anomaly difference as the effective anomaly segmentation index of the parent node corresponding to a certain node in a certain layer of a certain isolation tree in the isolation forest.

[0043] Specifically, the effective segmentation coefficient satisfies the following relational expression: ; In the formula, represents the effective anomaly segmentation index of the parent node corresponding to the th node in the th layer of the th isolation tree in the isolation forest, represents the maximum value of the node dimension mean of the th node in the th layer of the th isolation tree in the isolation forest, represents the minimum value of the node dimension mean of the th node in the th layer of the th isolation tree in the isolation forest, represents the anomaly index of the node dimension of the th node in the th layer of the th isolation tree in the isolation forest, represents the anomaly index of the dimension of another child node sharing the same parent node as the th node in the th layer of the th isolation tree in the isolation forest.

[0044] It should be noted that in the isolation forest algorithm, the process of constructing a tree is to recursively select a random dimension and a splitting point to split the dataset. In each step, the parent node divides the data into two parts according to the selected dimension and splitting point, and puts them into the left subtree and the right subtree respectively. Therefore, for the th layer of the th node in an isolation tree, its sibling node is the other child node under the same parent node.

[0045] The constant 1 in the denominator is to prevent the denominator from being zero, which would lead to inability to calculate, and it can be selected according to the situation. It reflects the degree of data difference between the two nodes after splitting the offset coordinate data. and respectively represent the anomaly indices of the two node dimensions under the same parent node. It reflects the difference in the degree of data anomaly between the two nodes after splitting the offset coordinate data.

[0046] Furthermore, in the process of constructing an isolation tree through the offset coordinate data of the connector mold fittings, the greater the degree of coordinate offset during the processing of the mold fittings, the more abnormal the moment is, and the more easily it can be identified in advance. And the proportion of abnormal processing moments is usually small. Therefore, as the depth of the constructed isolation tree increases, the abnormal processing moments are gradually identified, and the degree of anomaly of the remaining processing moments will become smaller and smaller. That is, as the depth of the isolation tree increases, the coordinate offset of the processing moment in the leaf node will gradually become smaller, and the degree of difference in the coordinate offset during the processing of the mold fittings between the nodes in each layer of the isolation tree will gradually become smaller. Therefore, the downward trend of the anomaly degree of the leaf node is consistent with the downward trend of the data difference between the nodes in each layer.

[0047] S4: According to the similarity of the downward trends of the anomaly index and the anomaly effective splitting index in the isolation tree, construct a trend coordination index to evaluate the accuracy of the detection effect of the isolation tree on the abnormal processing moments of the connector mold fittings.

[0048] Traverse and calculate the mean value of the anomaly indices of the leaf nodes in each layer of the isolation tree, and sort them in ascending order according to the depth of the isolation tree. The sorted sequence is used as the node anomaly trend sequence; traverse and calculate the mean value of the anomaly effective splitting indices of the parent nodes corresponding to each layer of nodes in the isolation tree, and sort them in ascending order according to the depth of the isolation tree. The sorted sequence is used as the in-layer classification effect sequence. Use the Mann-Kendall algorithm to test the downward trends of the node anomaly trend sequence and the in-layer classification effect sequence respectively, obtain the output results of the corresponding sequences, and analyze the similarity of the output results to be used for calculating the trend coordination index.

[0049] In the Mann-Kendall algorithm, the test results of the preset node abnormal trend sequence and the intra-layer classification effect sequence show a downward trend. Among them, the input is the node abnormal trend sequence and the intra-layer classification effect sequence, and the outputs are the value of the hypothesis test of the node abnormal trend sequence and the value of the hypothesis test of the intra-layer classification effect sequence.

[0050] It should be noted that the Mann-Kendall algorithm is a well-known technology in the field and will not be described in detail.

[0051] The trend coordination index includes: Taking any isolated tree as the target isolated tree, calculate the absolute difference between the output result of the node abnormal trend sequence of the target isolated tree and the output result of the intra-layer classification effect sequence; Take the reciprocal of the sum of the absolute difference and 1 as the trend coordination index of the target isolated tree.

[0052] Specifically, the trend coordination index satisfies the following relational expression: ; In the formula, represents the trend coordination index of the th isolated tree in the isolation forest, represents the hypothesis test value of the node abnormal trend sequence of the th isolated tree in the isolation forest, represents the hypothesis test value of the intra-layer classification effect sequence of the th isolated tree in the isolation forest.

[0053] That is to say, reflects the similarity degree of the downward trends of the above two sequences. A smaller absolute difference means that the trends of the two sequences are similar, and a larger absolute difference indicates that the trends of the two sequences are significantly different. Adding 1 to the denominator is used to avoid the inability to calculate due to the denominator being zero, and it can be selected according to specific situations.

[0054] S5: Based on the trend coordination index, correct the calculation result of the abnormal score in the isolation forest algorithm to obtain a significant abnormal score, and realize the monitoring of the processing parameters of the connector mold fittings.

[0055] The significant abnormal score includes: Taking any isolated tree as the target isolated tree, using the trend coordination index of the target isolated tree as the weight of the isolated tree abnormal score, and performing a weighted sum of the weights of the abnormal scores of all isolated trees in the isolation forest algorithm to obtain the significant abnormal score corresponding to the processing moment.

[0056] Specifically, the significant anomaly score satisfies the following relational expression: ; In the formula, represents the significant anomaly score at the th processing moment in the isolation forest algorithm, represents the number of preset isolation trees in the isolation forest, represents the trend cooperation index of the th isolation tree in the isolation forest, represents the anomaly score calculated using the th isolation tree in the traditional isolation forest algorithm at the th processing moment.

[0057] In this embodiment, the number of preset isolation trees is 50, which can be selected according to specific circumstances.

[0058] To implement the monitoring of the processing parameters of the connector mold fittings, including: In response to the significant anomaly score being greater than or equal to the preset anomaly threshold , there is a serious anomaly in the current processing process; in response to the significant anomaly score being greater than the preset anomaly threshold and less than the preset anomaly threshold , there is a medium anomaly in the current processing process; in response to the significant anomaly score being less than or equal to the preset anomaly threshold , there is a slight anomaly or no anomaly in the processing process, where the preset anomaly threshold is greater than the preset anomaly threshold ; Taking the detection result as the monitoring result of the processing parameters of the connector mold fittings at the current moment, the monitoring of the processing parameters of the connector mold fittings based on big data is realized.

[0059] Exemplarily, let the anomaly threshold be , and the preset anomaly threshold be , which can be adjusted according to specific circumstances.

[0060] It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of this invention patent shall be subject to the appended claims.

Claims

1. A method for monitoring processing parameters of connector mold fittings based on big data, characterized in that, Including: Obtain relevant parameters in the processing of connector mold fittings and perform preprocessing. Among them, the relevant parameters include: workpiece coordinate data and machining coordinate data; Take the degree of abnormality of the relevant parameters at the processing moment of the connector mold fittings as the abnormality index for constructing the node dimension in the isolation tree, so as to reflect the degree of abnormality at the processing moment of each node dimension; Based on the abnormality index of the node dimension, construct an abnormal effective segmentation index for identifying the abnormal processing moment of the connector mold fittings by the parent node corresponding to each node in the isolation tree; According to the similarity of the decreasing trends of the abnormality index and the abnormal effective segmentation index in the isolation tree, construct a trend coordination index to evaluate the accuracy of the detection effect of the isolation tree on the abnormal processing moment of the connector mold fittings; Based on the trend coordination index, correct the calculation result of the abnormal score in the isolation forest algorithm to obtain a significant abnormal score, and realize the monitoring of the processing parameters of the connector mold fittings.

2. The method for monitoring machining parameters of connector mold fittings based on big data according to claim 1, characterized in that, Perform preprocessing on the relevant parameters, including: Take the relevant parameters of the connector grinding mold fittings preset by the data detection system in the workshop as standard parameters; take the absolute value of the difference between the real-time obtained relevant parameters and the standard parameters as the offset coordinate data, and each moment corresponds to an offset coordinate data, so as to monitor the abnormal situation in the processing process.

3. A method for monitoring the processing parameters of connector mold fittings based on big data according to claim 1, characterized in that, The abnormality index of the node dimension includes: For a certain node in a certain layer of a certain isolation tree in the isolation forest, traverse the ratio between the offset data at all processing moments in the node and the maximum value of all offset data in the corresponding layer to obtain the relative offset. Use the exponential function to map the relative offset, and sum the mapped values at all processing moments to obtain the abnormality index of the corresponding node.

4. A method for monitoring the processing parameters of connector mold fittings based on big data according to claim 1, characterized in that, The abnormality index of the node dimension includes: For a node in a certain layer of a certain isolated tree in the isolated forest, traverse all the processing times included, calculate the difference between the offset coordinate data corresponding to each processing time and the mode of the dimension data selected in the corresponding layer of the offset coordinate data of all processing times in the isolated tree, and input the difference into the activation function. The output result will set negative values to zero and keep positive values unchanged. For all processing times sum up the output results of the activation function to obtain the anomaly index of the node dimension corresponding to the node.

5. A method for monitoring the processing parameters of a connector mold fitting based on big data according to claim 1, characterized in that, The abnormal effective segmentation index includes: For a certain node in a certain layer of a certain isolation tree in the isolation forest, traverse and calculate the ratio between the maximum value of the node dimension mean in the node and the sum of the minimum value of the node dimension mean and the constant 1 to obtain the time interval ratio; Traverse and calculate the absolute value of the difference between the abnormality index of each node and the abnormality index of another child node dimension of the same parent node to obtain the abnormal difference between sibling nodes; Take the product of the time interval ratio and the abnormal difference between sibling nodes as the abnormal effective segmentation index of the parent node corresponding to a certain node in a certain layer of a certain isolation tree in the isolation forest.

6. The method for monitoring machining parameters of connector mold fittings based on big data according to claim 1, wherein The similarity of the decreasing trends of the abnormality index and the abnormal effective segmentation index in the isolation tree includes: Traverse and calculate the mean value of the abnormality index of each leaf node in each layer of the isolation tree, and sort them in ascending order according to the depth of the isolation tree. Take the sorted sequence as the node abnormal trend sequence; traverse and calculate the mean value of the abnormal effective segmentation index of the parent node corresponding to each node in each layer of the isolation tree, and sort them in ascending order according to the depth of the isolation tree. Take the sorted sequence as the in-layer classification effect sequence; Use the Mann-Kendall algorithm to test the decreasing trends of the node abnormal trend sequence and the in-layer classification effect sequence respectively, obtain the output results of the corresponding sequences, and analyze the similarity of the output results to calculate the trend coordination index.

7. A method for monitoring the processing parameters of connector mold fittings based on big data according to claim 6, characterized in that, The trend coordination index includes: Taking any single isolated tree as the target isolated tree, calculate the absolute difference between the output result of the node anomaly trend sequence and the output result of the in-layer classification effect sequence of the target isolated tree; Take the reciprocal of the sum of the absolute difference and 1 as the trend cooperation index of the target isolated tree.

8. A method for monitoring processing parameters of connector mold accessories based on big data according to claim 6, characterized in that, In the Mann-Kendall algorithm, there is a downward trend in the test results of the preset node abnormal trend sequence and the intra-layer classification effect sequence. Among them, the input is the node abnormal trend sequence and the intra-layer classification effect sequence, and the outputs are the value of the hypothesis test of the node abnormal trend sequence and the value of the hypothesis test of the intra-layer classification effect sequence.

9. A method for monitoring processing parameters of connector mold fittings based on big data according to claim 1, characterized in that The significant anomaly score includes: Taking any single isolated tree as the target isolated tree, using the trend cooperation index of the target isolated tree as the weight of the anomaly score of the isolated tree, and performing a weighted sum of the weights of the anomaly scores of all isolated trees in the isolated forest algorithm to obtain the significant anomaly score corresponding to the processing moment.

10. The method for monitoring the processing parameters of a connector mold fitting based on big data according to claim 1, characterized in that, The implementation of monitoring the processing parameters of the connector mold fittings includes: In response to the significant anomaly score being greater than or equal to the preset anomaly threshold , there is a serious anomaly in the current processing; in response to the significant anomaly score being greater than the preset anomaly threshold less than the preset anomaly threshold , there is a medium anomaly in the current processing; in response to the significant anomaly score being less than or equal to the preset anomaly threshold , there is a slight anomaly or no anomaly in the processing; Taking the detection result as the monitoring result of the processing parameters of the connector mold fittings at the current moment, and realizing the monitoring of the processing parameters of the connector mold fittings based on big data.

Images