Comprehensive automated testing method and system for electric control cabinets

Abstract

The present invention relates to the field of data cleaning technology, and specifically to a comprehensive automated testing method and system for electric control cabinets. When establishing a simulation test model to detect the electric control cabinet, the present invention effectively screens noise data and performs targeted denoising to improve efficiency and accuracy. Based on data similarity clustering analysis of parameter time series data, a frequent pattern model is constructed to significantly distinguish normal and noise data. The correlation and frequent patterns in the analysis model are quantified to quantify the probability of noise existence, providing a basis for evaluating the quality of the data set. The data quality is obtained by comprehensively considering the data change trend and fluctuation, combined with the noise probability. Accurately identify the noise data set, efficiently denoise, and obtain high-quality data. Finally, the performance test of the electric control cabinet is performed based on the high-quality data set, which not only improves the denoising efficiency, but also ensures the accuracy and reliability of the test results. The present invention provides an efficient and accurate data processing solution for electric control cabinet detection.

Description

Technical Field

[0001] The present invention relates to the technical field of data cleaning, and in particular to a comprehensive automatic testing method and system for an electric control cabinet. Background Art

[0002] The electric control cabinet, also known as the electrical control cabinet, integrates a variety of electrical components to achieve centralized control of the equipment, that is, the full automation of the electric control cabinet. In order to discover potential failure points of the equipment and improve the reliability and service life of the electric control cabinet, it is also necessary to test the performance of the electric control cabinet. The existing method usually achieves a comprehensive automated test of the electric control cabinet through software simulation detection.

[0003] At present, when testing the electric control cabinet, a large amount of operating data of the electric control cabinet is usually obtained, and then the software simulation is used for detection. Therefore, the quality of the obtained data is crucial to the entire simulation detection process. Since the data collection will be disturbed by factors such as the environment, the acquired data will have noise. When denoising the data, the existing technology usually uses indifferent denoising. However, considering the large amount of data required for simulation model detection, the use of indifferent denoising will result in low efficiency. Therefore, how to accurately and efficiently filter out the noisy data from a large amount of data has become a key issue in the simulation detection process. Summary of the invention

[0004] In order to solve the technical problem that using an indifferent denoising method to denoise a large amount of data will cause low efficiency and thus affect the accuracy of the test results, the purpose of the present invention is to provide a comprehensive automated testing method and system for an electric control cabinet. The technical solutions adopted are as follows:

[0005] For any performance parameter of the electric control cabinet, a parameter data set of the performance parameter in each test period is obtained, each of the parameter data sets including parameter time series data of multiple data cycles;

[0006] According to the similarity between each parameter time series data in each parameter data set, all parameter time series data are clustered to obtain clustering results; according to the numerical characteristics of all parameter time series data in each parameter data set and the frequent pattern mining algorithm, the parameter time series data in each cluster cluster in the clustering result are fused and analyzed to obtain the frequent pattern model corresponding to each parameter data set;

[0007] In the frequent pattern model, based on the distribution of each item in all parameter time series data, the similarity between frequent item sets, and the position of frequent item sets in the frequent pattern model, the probability of noise existence in each parameter data set is determined; according to the probability of noise existence in each parameter data set, the change trend of the value in each parameter time series data in the parameter data set, and the value fluctuation, the data quality of each parameter data set is obtained;

[0008] According to the data quality corresponding to the parameter data sets of all test periods, the noise data sets are screened from all parameter data sets and the parameter time series data in the noise data sets are denoised to obtain high-quality data sets corresponding to all test periods; the performance test of the electric control cabinet is performed based on all high-quality data sets corresponding to all performance parameters.

[0009] Furthermore, the method for obtaining the clustering result includes:

[0010] In each parameter data set, for any two parameter time series data in different data periods, based on the difference between the numerical values ​​and the difference between the slope values ​​of the numerical values, determine the difference characteristic value between the two parameter time series data;

[0011] The difference characteristic values ​​between the parameter time series data are used as the distance metric, and K-means clustering analysis is performed on all the parameter time series data in the parameter data set based on a preset K value to obtain the clustering result.

[0012] Furthermore, the method for acquiring the frequent pattern model includes:

[0013] For any parameter data set, a numerical range is determined based on the maximum and minimum values ​​in the parameter data set, the numerical range is evenly divided to obtain a preset number of interval ranges, and different interval ranges are marked with different labels;

[0014] Using the label to replace each value in each parameter time series data, to obtain a label sequence corresponding to each parameter time series data;

[0015] Obtain the frequency of occurrence of each tag in the parameter data set; in each cluster, according to the frequency of occurrence of all tags and the number of tag sequences in each cluster, obtain the importance index of each cluster, and the number of tag sequences in each cluster and the frequency of occurrence of tags are positively correlated with the importance index;

[0016] In the clustering results corresponding to the parameter data set, all label sequences in each cluster are analyzed based on the FP-Growth algorithm to obtain the frequent pattern tree corresponding to each cluster; the frequent pattern trees corresponding to all clusters are arranged in descending order based on the importance index to obtain an arrangement sequence, in which the first frequent pattern tree is used as the target tree, the frequent pattern tree after the target tree is used as the tree to be analyzed, and the part where the tree to be analyzed and the target tree have differences is added to the target tree to obtain a new target tree; the frequent pattern tree after the tree to be analyzed is used as the new tree to be analyzed, and the part where the new tree to be analyzed and the new target tree have differences is added to the new target tree, and the new target tree is continuously determined until the frequent pattern trees in the arrangement sequence are traversed and stopped, and the new target tree at this time is used as the frequent pattern model corresponding to the parameter data set.

[0017] Furthermore, the method for obtaining the noise existence probability includes:

[0018] For any parameter data set, obtain all frequent item sets in the corresponding frequent pattern model; the frequency of occurrence of each item in the frequent item set is the frequency of occurrence of the label corresponding to the item in the parameter data set;

[0019] For any frequent item set, based on the occurrence frequency of each item in the frequent item set, determine the first overall noise factor of the frequent item set, and the first overall noise factor is negatively correlated with the occurrence frequency; based on the number of bifurcation points of the path corresponding to the frequent item set in the frequent pattern model, determine the second overall noise factor of the frequent item set, and the second overall noise factor is negatively correlated with the number of bifurcation points;

[0020] The first overall noise factor and the second overall noise factor of all frequent item sets are merged and averaged to obtain a first noise index of the parameter data set, wherein the first overall noise factor and the second overall noise factor are both positively correlated with the first noise index;

[0021] Based on the similarities between all frequent item sets, determine the second noise index of the parameter data set;

[0022] According to the first noise index and the second noise index of the parameter data set, the noise existence probability of the parameter data set is obtained, and the first noise index and the second noise index are both positively correlated with the noise existence probability.

[0023] Furthermore, the method for obtaining the second noise index includes:

[0024] Among all frequent item sets, the frequent item sets with the same number of items are regarded as a class of frequent item sets;

[0025] In any type of frequent item sets, any two frequent item sets are combined to obtain all non-repeating combinations; for any combination, the independence value of the frequent item sets in the combination is determined based on the differences in the items at the same position in the two frequent item sets;

[0026] The mean of the independence values ​​corresponding to all combinations of frequent item sets of all classes is taken as the second noise index of the parameter data set.

[0027] Furthermore, the method for obtaining the data quality includes:

[0028] The parameter data set with a noise probability greater than or equal to a preset noise threshold is taken as the target data set, and the parameter data set with a noise probability less than the preset noise threshold is taken as the normal data set, and the data quality of the normal data set is set to a fixed value not less than 1;

[0029] For any target data set, based on the change trend of each parameter time series data in the target data set, determine the first quality parameter of the target data set;

[0030] In each parameter time series data corresponding to the target data set, based on the deviation between each value and the mean of all values, a deviation factor of each value is obtained, and based on the deviation factors of all values, an overall quality factor of each parameter time series data is determined, and the overall quality factor is negatively correlated with the deviation factor; the overall quality factors of all parameter time series data are averaged to obtain a second quality parameter of the target data set;

[0031] The data quality of the target data set is obtained according to the noise existence probability, the first quality parameter and the second quality parameter of the target data set. The noise existence probability, the first quality parameter and the second quality parameter are all negatively correlated with the data quality, and the value of the data quality is a normalized value.

[0032] Furthermore, the method for obtaining the first quality parameter includes:

[0033] The number of extreme value points in each parameter time series data in the target data set is obtained, and the first quality parameter of the target data set is determined based on the proportion of the sum of the number of extreme value points of all parameter time series data in the total number of values ​​of all parameter time series data.

[0034] Furthermore, the method for obtaining the high-quality data set includes:

[0035] Among all parameter data sets, the parameter data sets whose data quality is less than or equal to the preset quality threshold are regarded as noise data sets;

[0036] Based on the Kalman filtering method, all parameter time series data in each noise data set are denoised to obtain a denoised data set corresponding to each noise data set;

[0037] The parameter data sets with data quality greater than the preset quality threshold and all denoised data sets are regarded as high-quality data sets corresponding to all test periods.

[0038] Furthermore, the performance test of the electric control cabinet based on all high-quality data sets corresponding to all performance parameters includes:

[0039] A simulation test model corresponding to the electric control cabinet is established, all high-quality data sets corresponding to all performance parameters are used as the database of the simulation test model, and the performance test of the electric control cabinet is performed to obtain the test results.

[0040] The present invention also proposes a comprehensive automatic testing system for an electric control cabinet, the system comprising:

[0041] A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the steps of any one of the methods can be implemented when the processor executes the computer program.

[0042] The present invention has the following beneficial effects:

[0043] The present invention can accurately filter out noise data from a large amount of data when using a simulation test model to detect an electric control cabinet, thereby performing targeted denoising processing, effectively improving efficiency, and ensuring the accuracy of subsequent model test results. First, for any performance parameter, it is necessary to obtain a parameter data set of the performance parameter in multiple test periods, and the parameter data set contains multiple parameter time series data. Since noise data is an accidental phenomenon compared to normal data, the parameter time series data in each parameter data set can be clustered based on the similarity between the data to obtain a clustering result. At this time, the parameter time series data in each cluster cluster in each cluster result can distinguish the normal data and noise data in the parameter data set to a certain extent. Further, based on the numerical characteristics of all parameter time series data in the parameter data set and the frequent pattern mining algorithm, each cluster cluster in the cluster result is fused and analyzed to obtain a frequent pattern model. At this time, the frequent pattern model can more significantly characterize the difference between normal data and noise data. Furthermore, the frequent pattern model can be analyzed to analyze the correlation between data and the frequent patterns and the position distribution of frequent item sets in the frequent pattern model, so as to quantify the possibility of the existence of noise data, obtain the probability of noise existence in the parameter data set, and provide a reference for the subsequent evaluation of the quality of the parameter data set. Furthermore, since the noise data has characteristics in data fluctuation and trend change, the data quality of the parameter data set can be more accurately evaluated by comprehensively considering the change trend and data fluctuation of the parameter time series data and combining the probability of noise existence. This helps to accurately identify and filter out the noise data set, and then denoise it to obtain a high-quality data set. Finally, the performance test of the electric control cabinet is carried out based on the high-quality data set corresponding to all performance parameters to obtain the test results. Since the present invention can accurately filter out the noise data set, only the noise data set is denoised in a large number of data sets, which can improve the efficiency of data denoising processing and ensure the accuracy and reliability of the test results. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 A method flow chart of a comprehensive automated testing method for an electric control cabinet provided by one embodiment of the present invention;

[0046] Figure 2A method flow chart of a method for constructing a frequent pattern model provided by one embodiment of the present invention;

[0047] Figure 3 An exemplary schematic diagram of a merging sub-step of a frequent pattern model building process provided by an embodiment of the present invention;

[0048] Figure 4 A method flow chart of a method for obtaining noise existence probability provided by an embodiment of the present invention;

[0049] Figure 5 A method flow chart of a method for obtaining data quality provided by an embodiment of the present invention;

[0050] Figure 6 A system block diagram of a comprehensive automated testing system for an electric control cabinet provided by one embodiment of the present invention;

[0051] Figure 7 A schematic diagram of the system structure of a comprehensive automated testing system for an electric control cabinet provided by one embodiment of the present invention;

[0052] Figure 8 A schematic diagram of a computer-readable storage medium provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the comprehensive automated testing method and system for an electric control cabinet proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0054] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0055] The following is a detailed description of a comprehensive automated testing method and system for an electric control cabinet provided by the present invention in conjunction with the accompanying drawings.

[0056] See also Figure 1 , which shows a method flow chart of a comprehensive automated testing method for an electric control cabinet provided by an embodiment of the present invention, the method comprising the following steps:

[0057] Step S1: for any performance parameter of the electric control cabinet, a parameter data set of the performance parameter in each test period is obtained, and each parameter data set includes parameter time series data of multiple data cycles.

[0058] In the process of fully automated testing of the electric control cabinet by software simulation detection, a large amount of data needs to be obtained to ensure the accuracy of the simulation detection results. In the process of data acquisition, various external factors will cause noise data in the data set. In order to ensure the accuracy of the simulation operation, the data set needs to be denoised to improve the credibility of the simulation results. Considering the large amount of data, in this embodiment of the present invention, by analyzing the data in the data set, the data set with noise is screened out and denoised, which can effectively improve the efficiency of the denoising process and ensure the credibility of the simulation results.

[0059] Electric control cabinets are widely used in many fields such as new energy, power control, food machinery, machine tool machinery, etc. Since the data characteristics of electric control cabinets will show differences under different process flows, in order to obtain more accurate test results, the embodiments of the present invention are mainly tested under a certain process flow. The specific process flow may be, for example: When the CNC machine tool of the machining center is working, its spindle motor and feed system will produce variable loads, causing the current and power parameters in the electric control cabinet to fluctuate accordingly. Injection molding: During the heating, injection, cooling and opening and closing of the mold, the power demand of the injection molding machine will change periodically, causing regular fluctuations in the current and power parameters in the electric control cabinet.

[0060] For any performance parameter of the electric control cabinet, the sensors deployed in the electric control cabinet are used to collect parameter data sets of the electric control cabinet in multiple test periods under the current process flow, and each parameter data set should include parameter timing data of multiple data cycles. It should be noted that the performance parameters include voltage, current, etc. The test period is the period from the beginning to the end of the process flow, where the data period is set to 10 minutes and the sampling time interval is set to 0.1s. The specific values ​​of the data period and the sampling time interval can be adjusted according to the implementation scenario and are not limited here.

[0061] At this point, for a certain process flow, multiple parameter data sets corresponding to each new energy parameter of the electric control cabinet can be obtained, and each parameter data set can be subsequently analyzed to determine whether it contains noise data.

[0062] Step S2: Based on the similarities between the parameter time series data in each parameter data set, all the parameter time series data are clustered to obtain clustering results; based on the numerical characteristics of all the parameter time series data in each parameter data set and the frequent pattern mining algorithm, the parameter time series data in each cluster cluster in the clustering results are fused and analyzed to obtain the frequent pattern model corresponding to each parameter data set.

[0063] Since noise data is an accidental phenomenon compared to normal data, cluster analysis can be performed on the parameter time series data in each parameter data set based on the similarity between data to obtain clustering results. At this time, the parameter time series data in each cluster cluster in each cluster result can distinguish the normal data and noise data in the parameter data set to a certain extent. Since the correlation between normal data should be strong, based on the numerical characteristics of all parameter time series data in the parameter data set and the frequent pattern mining algorithm, each cluster cluster in the cluster result is fused and analyzed to obtain a frequent pattern model. At this time, the frequent pattern model can more significantly characterize the difference between normal data and noise data, preparing for the subsequent noise data screening process.

[0064] By clustering the parameter time series data in each parameter data set, similar parameter time series data can be classified into one category, thereby simplifying the data distribution of the parameter data set. At the same time, the clustering results can reveal the internal structure and pattern in the parameter data and help understand the relationship between the parameter time series data. Therefore, here, according to the similarity between the values ​​of each parameter time series data in each parameter data set, cluster analysis is performed on all parameter time series data to obtain the clustering results.

[0065] Preferably, in one embodiment of the present invention, the method for obtaining clustering results includes:

[0066] The purpose of clustering is to make samples in the same cluster as similar as possible, and samples in different clusters as dissimilar as possible. In this embodiment of the present invention, the similarity between the parameter time series data is mainly determined based on the numerical difference between the parameter time series data.

[0067] Therefore, in each parameter data set, for any two parameter time series data in different data periods, the difference characteristic value between the two parameter time series data is determined based on the difference between the numerical values ​​and the difference between the slope values ​​of the numerical values; the formula model of the difference characteristic value includes:

[0068] ;

[0069] in, Indicated in The difference feature value between parameter time series data 1 and parameter time series data 2 in the parameter data set; Indicated in Among the parameter data sets, the parameter time series data 1 The slope value of the value; Indicated in Among the parameter data sets, the parameter time series data 2 The slope value of the value; Indicated in Among the parameter data sets, the parameter time series data 1 numerical values; Indicated in Among the parameter data sets, the parameter time series data 2 numerical values; represents the normalization function; Indicates the total number of values ​​in the parameter time series data.

[0070] In the formula model of the difference eigenvalue, for any two different parameter time series data in the parameter data set, the difference between the values ​​at the corresponding positions is calculated. , the smaller the value, the more similar the values ​​at the same position in the two parameter time series data are. Similarly, after obtaining the slope value of each value, the difference between the slope values ​​of the values ​​at the same position is analyzed. , since the slope value can represent the change trend of the data, the smaller the difference value is, the more consistent the change trend is at the same position of the two parameter time series data. and Add them together to represent the similarity of the two parameter time series data at a certain position. The smaller the sum, the higher the similarity. Conversely, the larger the sum, the higher the difference. Finally, all the positions, that is, all the corresponding values The accumulated values ​​are added and the accumulated sum is normalized to obtain the difference characteristic value between the two parameter time series data.

[0071] Then, the difference characteristic values ​​between the parameter time series data are used as the distance metric, and K-means clustering analysis is performed on all the parameter time series data in the parameter data set based on the preset K value to obtain the clustering results. At this time, the parameter time series data in each cluster in the clustering results have relatively consistent numerical characteristics.

[0072] It should be noted that the slope value is calculated by dividing the difference between the latter value and the previous value by the time interval between the previous two values; when calculating the slope value, the slope value of the last value is set to the slope value of the previous value in its time sequence. The preset K value is set to 5, and the specific value can be adjusted according to the implementation scenario, which is not limited here; the K-means clustering algorithm is a technical means well known to those skilled in the art, and will not be described here. Here, the same position in the two parameter time series data is explained: for example, the values ​​in one parameter time series data are arranged in time sequence as (1, 2, 3, 4), and the values ​​in the other parameter time series data are arranged in time sequence as (a, b, c, d), then 1 and a are two values ​​at the same position, and similarly, 2 and b are two values ​​at the same position, 3 and c are two values ​​at the same position, and 4 and d are two values ​​at the same position.

[0073] In other embodiments of the present invention, the following method may also be used to analyze the parameter time series data in each parameter data set for similar situations, thereby achieving clustering and obtaining clustering results.

[0074] Since the Dynamic Time Warping (DTW) algorithm can be used to measure the similarity of two time series, in each parameter data set, the parameter time series data in any two different data periods are obtained. value, and The values ​​are normalized to obtain the difference characteristic value between the two parameter time series data. The formula model of the difference characteristic value includes:

[0075] ;

[0076] in, Indicated in The difference feature value between parameter time series data 1 and parameter time series data 2 in the parameter data set; Indicated in Among the parameter data sets, the time series data between parameter 1 and parameter 2 value; Represents the normalization function.

[0077] In this formula model, when the time series data between two parameters is The smaller the value, the more similar the time series data of the two parameters are. The values ​​are normalized to obtain the difference characteristic values ​​between the two parameter time series data.

[0078] Then, the difference characteristic values ​​between the parameter time series data are used as the distance metric, and K-means clustering analysis is performed on all the parameter time series data in the parameter data set based on the preset K value to obtain the clustering results. At this time, the parameter time series data in each cluster in the clustering results have relatively consistent characteristics.

[0079] At this point, in each parameter data set, all clusters can be obtained by clustering the parameter time series data. Different clusters have different data characteristics, which can distinguish normal data from noise data to a certain extent, providing a reference for subsequent processes.

[0080] Frequent pattern mining algorithms can be used to retrieve frequent item set information, and can quickly discover frequent patterns and associated information in parameter time series data, so as to more effectively identify whether there is noise data in the parameter data set. There will be significant differences between different clusters in the clustering results. Therefore, by performing frequent pattern mining and fusion analysis on different clusters, the main components and branch components can be more intuitively distinguished in the final frequent pattern model. Since noise data is less than normal data, the branch part is more likely to represent noise data.

[0081] Preferably, in one embodiment of the present invention, the method for acquiring the frequent pattern model includes:

[0082] See also Figure 2 , which shows a method flow chart of a method for constructing a frequent pattern model provided by an embodiment of the present invention, the method comprising the following steps:

[0083] Step S201: Based on the numerical features of all parameter time series data in the parameter data set, label conversion is performed on the numerical values ​​in each parameter time series data to determine the label sequence corresponding to each parameter time series data.

[0084] If each value is treated as a separate category, then when the value range is too large, for example, 0-100, then when building a frequent pattern model, the model will be too large. Therefore, for any parameter data set, the value range is determined based on the maximum and minimum values ​​in the parameter data set, and the value range is evenly divided to obtain a preset number of interval ranges, and all interval ranges are marked with different labels.

[0085] Then, each value in each parameter time series data is replaced by a label to obtain a label sequence corresponding to each parameter time series data.

[0086] Here is an example to illustrate the process: for example, the numerical range is 1-100, and the preset number is set to 10, then the interval range is 1-10, 11-20, 21-30...71-80, 81-90, 91-100. Each interval range is marked with a different letter. In this embodiment of the present invention, the letters AJ are used, that is, the interval range 0-10 corresponds to the label A, the label 11-20 corresponds to the label B, and so on. At this time, if the value in a parameter time series data sequence is (0, 15, 14, 35, 21, 50, 78), the corresponding label sequence is (A, B, B, D, C, E, H).

[0087] It should be noted that the setting of the preset number can be adjusted according to the implementation scenario and is not limited here.

[0088] Step S202: constructing a frequent pattern tree corresponding to the label sequence in each cluster in the clustering result based on the FP-Growth algorithm, merging all the frequent pattern trees, and obtaining a frequent pattern model corresponding to each parameter data set.

[0089] The FP-Growth algorithm uses the FP-Tree (frequent pattern tree) data structure to retrieve frequent item set information, which enables the algorithm to more quickly discover frequent patterns and associated information in parameter time series data, thereby more effectively identifying whether there is noise data in the parameter data set.

[0090] After processing in step S201, at this time, in the parameter data set, each parameter time series data corresponds to a label sequence, and the occurrence frequency of each label can be calculated among all labels. The occurrence frequency can represent the importance of the value corresponding to the label among all values.

[0091] Since the number of parameter time series data in a cluster can characterize the importance of the cluster, in each cluster, the importance index of each cluster is obtained according to the occurrence frequency of all labels and the number of label sequences in each cluster. The number of label sequences in each cluster and the frequency of occurrence of labels are positively correlated with the importance index. The formula model of the importance index can be, for example:

[0092] ;

[0093] in, Indicated in Among the parameter data sets, The importance index of each cluster; Indicated in Among the parameter data sets, The number of tag sequences in a cluster; Indicated in Among the parameter data sets, The total number of labels in the clusters; Indicated in Among the parameter data sets, The first The frequency of occurrence of a tag; Represents the normalization function.

[0094] In the formula model of the importance index, when the number of label sequences contained in a cluster is larger, it means that the data in the cluster is more likely to be normal data, and the more it can represent the data characteristics of the process flow, so the importance is higher; the frequency of occurrence of each label in all labels in the parameter data set is calculated. The larger the frequency of occurrence, the more times the data corresponding to the label appears in the parameter data set, and the more likely the data corresponding to the label is normal data, and the more it can represent the data characteristics of the process flow. Therefore, if the frequency of occurrence of labels in a cluster is larger and the number of label sequences is larger, it can be said that the importance of the cluster is higher. Therefore, based on the above logic, the formula model of the above importance index is constructed to obtain the importance index of each cluster.

[0095] It should be noted that in other embodiments of the present invention, only the normalized value of the sum of the occurrence frequencies of all labels in the cluster in the parameter data set may be used as the importance index of the cluster. The specific formula model is as follows:

[0096] ;

[0097] in, Indicated in Among the parameter data sets, The importance index of each cluster; Indicated in Among the parameter data sets, The total number of labels in the clusters; Indicated in Among the parameter data sets, The first The frequency of occurrence of a tag; Represents the normalization function.

[0098] In this formula model, the frequency of occurrence of each label in all labels in the parameter data set is calculated. The greater the frequency of occurrence, the more times the data corresponding to the label appears in the parameter data set, and the more likely the data corresponding to the label is normal data, and the more it can characterize the data characteristics of the process flow. Therefore, if the frequency of occurrence of a label in a cluster is greater, it means that the importance of the cluster is higher.

[0099] Then, in the clustering result corresponding to the parameter data set, all label sequences in each cluster are analyzed based on the FP-Growth algorithm to obtain a frequent pattern tree corresponding to each cluster. It should be noted that the method of constructing a frequent pattern tree is a technical means well known to those skilled in the art and will not be described in detail here.

[0100] Since the larger the importance index is, the more likely the data in the cluster is normal data and the more it can characterize the characteristic data of the process flow, the frequent pattern trees corresponding to all clusters are finally arranged in descending order based on the importance index to obtain an arrangement sequence. In the arrangement sequence, the first frequent pattern tree is used as the target tree; the frequent pattern tree after the target tree is used as the tree to be analyzed, and the part where the tree to be analyzed differs from the target tree is added to the target tree to obtain a new target tree; the frequent pattern tree after the tree to be analyzed is used as the new tree to be analyzed, and the part where the new tree to be analyzed differs from the new target tree is added to the new target tree, and the new target tree is continuously determined until the frequent pattern trees in the arrangement sequence are traversed and stopped, and the new target tree at this time is used as the frequent pattern model corresponding to the parameter data set.

[0101] An example is given to illustrate the process of merging and constructing a frequent pattern model: if there are 4 frequent pattern trees in the permutation sequence, recorded as 1, 2, 3, and 4. Then frequent pattern tree 1 is used as the target tree, frequent pattern tree 2 is compared with frequent pattern tree 1, and the parts that exist in frequent pattern tree 2 but not in frequent pattern tree 1 are added to frequent pattern tree 1 to obtain a frequent pattern tree ; Then the frequent pattern tree 3 and the frequent pattern tree Compare and compare the frequent pattern tree 3, the frequent pattern tree The parts that do not exist in the tree are added to the frequent pattern tree In the above example, we get the frequent pattern tree ; Next, the frequent pattern tree 4 and the frequent pattern tree For comparison, the frequent pattern tree 4 exists in the frequent pattern tree The parts that do not exist in the tree are added to the frequent pattern tree In the above example, we get the frequent pattern tree , the frequent pattern tree at this time This is the final frequent pattern model. Figure 3 , which shows an example schematic diagram of a merging sub-step in the frequent pattern model construction process.

[0102] At this point, the frequent pattern model can be obtained, and because the frequent pattern tree is constructed for each cluster first, each tree can well retain the data characteristics and association patterns within its corresponding cluster. At the same time, since noise data is often assigned to smaller clusters during the clustering process, the frequent pattern trees are merged based on the importance values ​​of each cluster. The obtained frequent pattern model can better retain the normal data in the current parameter data set, which is also the feature data, thereby distinguishing it from the noise data. Among them, normal data will have a more significant representation in the frequent pattern model due to its frequency and association, while noise data may appear as isolated or low-frequency items.

[0103] Step S3: In the frequent pattern model, based on the distribution of each item in all parameter time series data, the similarity between frequent item sets, and the position of frequent item sets in the frequent pattern model, determine the probability of noise existence in each parameter data set; according to the probability of noise existence in each parameter data set, the change trend of the numerical value in each parameter time series data in the parameter data set, and the numerical fluctuation, obtain the data quality of each parameter data set.

[0104] The above steps can complete the construction of the frequent pattern model corresponding to the parameter data set. This is because normal data is a high-frequency item compared to noise data and tends to constitute the backbone of the frequent pattern model. Noise data is a low-frequency item and will form an atypical path, which is significantly different from the path formed by normal data. Therefore, the probability of noise in the parameter data set can be determined based on the distribution of each item, the similarity between frequent item sets, and the position of frequent item sets in the frequent pattern model. At the same time, compared with normal data, noise data has no obvious trend distribution and the trend changes more frequently, so the changing trend of the numerical values ​​in the parameter time series data can be used as a reference for evaluating data quality; since both noise data and abnormal data appear as outlier data points, they will have relatively consistent characteristics in the frequent pattern model, and abnormal data can be used to simulate the fault conditions that may be encountered in the electric control cabinet. Conversely, noise data is not helpful for the simulation detection of the electric control cabinet. Therefore, it is also necessary to distinguish between noise data and abnormal data. Compared with noise data, the degree of outlier of abnormal data will be more obvious, and the impact on the fluctuation of the overall data will also be greater. Therefore, the data fluctuation in the parameter time series data is also used as a reference for evaluating data quality; at the same time, combined with the probability of noise existence in the parameter data set, the data quality of the parameter data set is analyzed more comprehensively.

[0105] Because noise data usually appear as infrequent items, items that are dissimilar to other item sets, or items with abnormal positions in the tree, by comprehensively considering the frequency of occurrence of items, item set similarity, and position information, we can have a more comprehensive understanding of the characteristics and structure of the data, thereby calculating the probability of noise existence in the parameter data set for identifying noise data.

[0106] Preferably, in one embodiment of the present invention, the method for obtaining the noise existence probability includes:

[0107] See also Figure 4 , which shows a method flow chart of a method for obtaining noise existence probability provided by an embodiment of the present invention, the method comprising the following steps:

[0108] Step S401: Determine a first noise index of a parameter data set based on the distribution of each item in all parameter time series data and the position of a frequent item set in a frequent pattern model.

[0109] For any parameter data set, obtain all frequent item sets in the corresponding frequent pattern model.

[0110] The frequency of occurrence of each item in a frequent item set is the frequency of occurrence of the label corresponding to the item in each parameter data set. By considering the frequency of occurrence of the item, it is possible to more effectively identify items with low frequency of occurrence, which are likely to correspond to noise data. For any frequent item set, based on the frequency of occurrence of each item in the frequent item set, that is, the frequency of occurrence of each label, the first overall noise factor of the frequent item set is determined, and the first overall noise factor is negatively correlated with the frequency of occurrence.

[0111] The frequent pattern model reflects the association between items through its tree structure. The items corresponding to normal data tend to form the trunk of the tree model, while the items corresponding to noise data tend to form the branches of the tree model. Therefore, the location information of frequent item sets in the frequent pattern model, especially the number of bifurcation points, can be used to further capture the intrinsic structural characteristics of the data. Considering the number of bifurcation points can help understand the complexity and diversity of data, so as to more accurately evaluate the existence of noise. Based on the number of bifurcation points of the path corresponding to the frequent item set in the frequent pattern model, the second overall noise factor of the frequent item set is determined, and the second overall noise factor is negatively correlated with the number of bifurcation points.

[0112] Finally, the first overall noise factor and the second overall noise factor of all frequent item sets are fused and averaged to obtain the first noise index of the parameter data set, and both the first overall noise factor and the second overall noise factor are positively correlated with the first noise index. The formula model of the first noise index can be, for example, as follows:

[0113] ;

[0114] in, Indicates The first noisy indicator of the parameter data set; Indicates The total number of frequent item sets in the frequent pattern model corresponding to the parameter data set; Indicates The parameter data set corresponds to the The number of bifurcation points of the paths of frequent item sets in the frequent pattern model; Indicates The parameter data set corresponds to the The number of items in a frequent itemset; Indicates The parameter data set corresponds to the Frequent item set The frequency of occurrence of each item; Indicates the preset first parameter.

[0115] In the formula model of the first noisy index, since the items corresponding to normal data will appear more frequently in the parameter data set, for each frequent item set, if the frequency of occurrence of the items in the frequent item set is greater, it means that the possibility that the frequent item set contains noise data is lower, so based on the frequency of occurrence of each item in the frequent item set , determine the first overall noise factor of the frequent itemset , and in order to achieve logical relationship correction, the first overall noise factor is negatively correlated with the frequency of occurrence. At the same time, because the items corresponding to normal data are more distributed in the main part of the frequent pattern model, for a frequent item set, if the number of bifurcations in its path in the frequent pattern model is more, it means that the path of the frequent item set is in the main part, that is, in the core position, then the data corresponding to the items in the frequent item set is more likely to be normal data, and the frequent item set is less likely to contain noise data. Therefore, the number of bifurcations is negatively correlated to achieve logical relationship correction, and the second overall noise factor of the frequent item set is obtained. Finally, the first overall noise factor of the frequent item set is combined with the second overall noise factor, and the multiplied value is used as the comprehensive noise factor of the frequent item set. , and average the comprehensive noise factors of all frequent itemsets to obtain the first noise index of the parameter data set.

[0116] It should be noted that the first parameter is preset The function of is to prevent the denominator from being 0. Here, the value can be 0.001. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0117] Step S402: Determine a second noise index of the parameter data set based on similarities between frequent item sets.

[0118] Among all frequent item sets, those with the same number of items are regarded as a class of frequent item sets.

[0119] In any type of frequent item set, any two frequent item sets are combined to obtain all non-repeating combinations; for any combination, the differences of the items in the same position in the two frequent item sets are analyzed. When the items in the same position are the same, the difference judgment index of the position is set to 0, otherwise it is set to 1. Then, based on the difference judgment index of all positions and the ordinal number of the position, the independence value of the frequent item set in the combination is determined. The independence value can be used to reflect the internal relationship between two frequent item sets, because similar frequent item sets should have the same items in the same position. The formula model of the independence value includes:

[0120] ;

[0121] in, Indicated in In the frequent item set of this class, The independence value between the frequent item sets of the combination; Indicates The number of items in each frequent itemset among the class frequent itemsets; Represents the index of the number of items in the frequent itemset; Indicated in In the frequent item set of this class, The first The difference judgment index of each item.

[0122] In the formula model of the independence value, since normal data is the main part compared to noise data, the frequent item sets corresponding to normal data should have higher similarity. Specifically, the items in the same position should be the same, and the positions of different items should be further back in the frequent item sets. Therefore, the differences of items in the same position are compared first. If they are different, the difference judgment index at that position is set to 1. If they are the same, it is set to 0. At this time, if the items in the same position in two frequent item sets are compared, the more the difference judgment index is 1, the greater the difference between the two frequent item sets, the lower the similarity, and the more likely it is to contain noise. At the same time, when comparing, the position of the item, that is, the index of the number of items, is used as a weight. If the position where the difference appears is closer to the front, it means that it may contain more noise components. Therefore, the number of items is negatively correlated and mapped to achieve logical relationship correction, and then used as an adjustment weight. Finally, the difference judgment index is weighted by adjusting the weight , and conduct a comprehensive analysis of all items to obtain the independence degree value. At this time, the larger the independence degree value, the stronger the difference between the two frequent item sets in the combination, and the higher the possibility of noise.

[0123] It should be noted that the method of determining the adjustment weight and difference judgment index is illustrated here. For example, two binomial frequent item sets are {A, F} and {C, F}, then A and C are the first items, the index of the number of items is 1, and the adjustment weight value of the first item is 1, and the difference judgment index is 1; F and F are the second items, the index of the number of items is 2, and the adjustment weight value of the second item is , the difference judgment index is 0.

[0124] So far, each combination in each type of frequent item set corresponds to an independence degree value, so the mean of the independence degree values ​​corresponding to all combinations is taken as the second noise index of the parameter data set, recorded as , and the larger the value is, the higher the possibility of noise in the parameter data set is.

[0125] Step S403: Obtain the noise existence probability of the parameter data set according to the first noise index and the second noise index of the parameter data set.

[0126] The sum of the first noise index and the second noise index of the parameter data set is normalized and processed as the noise existence probability of the parameter data set. The formula model of the noise existence probability includes:

[0127] ;

[0128] in, Indicates The probability of noise in the parameter data set; Indicates The first noisy indicator of the parameter data set; Indicates The second noise index of the parameter data set; Represents the normalization function.

[0129] In the formula model of the probability of noise existence, based on the analysis in step S401 and step S402, it can be known that the larger the first noise index is, the greater the possibility that noise exists in the parameter data set is, and the larger the second noise index is, the greater the possibility that noise exists in the parameter data set is. Therefore, here, the sum of the first noise index and the second noise index is normalized to obtain the probability of noise existence in the parameter data set.

[0130] It should be noted that, since both the first noise index and the second noise index are positively correlated with the probability of noise existence, in other embodiments of the present invention, the product of the first noise index and the second noise index can also be normalized and used as the probability of noise existence in the parameter data set.

[0131] At this point, based on the above method, the noise existence probability of the parameter data set in each test period can be obtained. In the subsequent process, the data quality of each parameter data set can be further analyzed in combination with the noise existence probability.

[0132] Since abnormal data can be used to simulate the fault conditions that the electric control cabinet may encounter, thus helping to more effectively discover faults during the simulation process, on the contrary, noise data is not helpful for the fully automated simulation test of the electric control cabinet, but may affect the final simulation results. Therefore, the more noise data there is, the worse the data quality. At the same time, since both noise data and abnormal data represent outlier characteristics, and the outlier characteristics of abnormal data points are more obvious, it is necessary to first filter out the normal data set and set the data quality of the normal data set to a fixed value, so as to further distinguish the abnormal data set and the noise data set in the remaining data set according to the characteristics such as the fluctuation of the numerical value, so as to comprehensively measure the data quality of the parameter data set, so as to facilitate the accurate screening of the noise data set in the subsequent process.

[0133] Preferably, in one embodiment of the present invention, the method for obtaining data quality includes:

[0134] See also Figure 5 , which shows a method flow chart of a data quality acquisition method provided by an embodiment of the present invention, the method comprising the following steps:

[0135] Step S501: distinguishing a normal data set from a target data set based on the noise existence probability of the parameter data set, and obtaining the data set quality of the normal data set.

[0136] The parameter data set with a noise existence probability greater than or equal to the preset noise threshold is used as the target data set, the parameter data set with a noise existence probability less than the preset noise threshold is used as the normal data set, and the data quality of the normal data set is set to a fixed value not less than 1. It should be noted that the preset noise threshold is set to 0.6, and in this embodiment of the present invention, the data quality of the normal data set is set to 1, and the specific values ​​can be adjusted according to the implementation scenario and are not limited here.

[0137] Step S502: for any target data set, based on the change trend of each parameter time series data in the target data set, determine the first quality parameter of the target data set.

[0138] Since noise data appears as irregularly distributed data points in the data set, that is, there is no obvious trend distribution, the more frequently the data trend of the parameter time series data changes, the more noise components may be in the target data set.

[0139] Because the extreme points in a set of data include maximum points and minimum points, which are the local highest or local lowest points in the data sequence, and are the nodes marking the transition of data from one trend to another, for any target data set, the number of extreme points in each parameter time series data in the target data set is obtained, and the first quality parameter of the target data set is determined based on the proportion of the sum of the number of extreme points of all parameter time series data in the total number of values ​​of all parameter time series data, which is recorded as The first quality parameter quantifies the frequency of data change trends in the parameter time series data in the target data set, and the larger the first quality parameter is, the more noise data is contained in the target data set, and the worse the data quality is.

[0140] Step S503: obtaining a second quality parameter of the target data set according to fluctuations in the values ​​of the time series data of each parameter in the target data set.

[0141] The target data set may contain noise data and abnormal data. Since abnormal data is more valuable than noise data, in order to more accurately measure the data quality of the target data set, abnormal data and noise data should be distinguished. Although both noise data and abnormal data show outlier characteristics, since abnormal data usually deviates more from normal data, the difference between the data can be analyzed to distinguish noise data from abnormal data.

[0142] In each parameter time series data corresponding to the target data set, the deviation of each value from the mean of all values ​​is taken as the deviation factor of each value, and then the sum of the deviation factors of all values ​​is negatively correlated and mapped as the overall quality factor of each parameter time series data, wherein the smaller the deviation factor, the larger the overall quality factor, indicating that the parameter time series data contains more noise data.

[0143] The overall quality factor of all parameter time series data is averaged to obtain the second quality parameter of the target data set. The larger the second quality parameter, the more noise data is contained in the target data set, and the worse the data quality is. The formula model of the second quality parameter includes:

[0144] ;

[0145] in, Indicates a second quality parameter of the target data set; Indicates The number of parameter time series data in the target dataset; Indicates The target dataset The total number of values ​​in the parameter time series data; Indicates The target dataset The first parameter in the time series data numerical values; Indicates The target dataset The numerical mean of the parameter time series data; Indicates the preset second parameter.

[0146] In the formula model of the second quality parameter, for each parameter time series data in any target data set, calculate the mean of all values , the mean represents the average level of the values ​​in a parameter time series data, and then the difference between each value and the mean is calculated to obtain the deviation factor , the deviation factor of noise data is smaller than that of abnormal data, so the sum of the deviation factors of all values ​​is negatively correlated to obtain the overall quality factor of each parameter time series data The larger the value is, the more noise data is contained in the parameter time series data. Finally, the overall quality factor of all parameter time series data in the target data set is averaged to obtain the second quality parameter of the target data set.

[0147] It should be noted that the second parameter is preset The function of is to prevent the denominator from being 0. Here, the value can be 0.001. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0148] Step S504: The noise existence probability, the first quality parameter and the second quality parameter of the target data set are integrated to obtain the data quality of the target data set.

[0149] The product of the noise existence probability of the target data set, the first quality parameter and the second quality parameter is negatively correlated and normalized to obtain the value as the data quality of the target data set. The formula model of data quality can be, for example:

[0150] ;

[0151] in, Indicates The data quality of the target dataset; Indicates The probability of noise existing in the target data set; Indicates A first quality parameter of a target data set; Indicates a second quality parameter of the target data set; Indicated by natural constant An exponential function with base .

[0152] In the formula model of data quality, the greater the probability of noise existence, the higher the possibility that the target data set contains noise, and the worse the data quality; the larger the first quality parameter, the more noise data contained in the target data set, and the worse the data quality; the larger the second quality parameter, the more noise data contained in the target data set, and the worse the data quality. Therefore, here, the product of the probability of noise existence, the first quality parameter and the second quality parameter is negatively correlated and normalized to obtain the data quality of the target parameter data.

[0153] At this point, the data quality of all parameter data sets can be obtained, and the noise data sets can be screened out and denoised based on the data quality in the subsequent process.

[0154] Step S4: According to the data quality corresponding to the parameter data sets of all test time periods, the noise data sets are screened from all parameter data sets and the parameter timing data in the noise data sets are denoised to obtain high-quality data sets corresponding to all test time periods; and the performance test of the electric control cabinet is performed based on all high-quality data sets corresponding to all performance parameters.

[0155] Since the worse the data quality is, the more noise data it contains, in order to ensure the accuracy of subsequent simulation test results, the noise data set that needs to be denoised should be screened out and denoised, so as to form a high-quality data set together with the parameter data set that does not need to be denoised.

[0156] Preferably, in one embodiment of the present invention, the method for acquiring a high-quality data set includes:

[0157] Among all parameter data sets, the parameter data sets whose data quality is less than or equal to a preset quality threshold are regarded as noise data sets.

[0158] Then, all parameter time series data in each noise data set are denoised based on the Kalman filtering method to obtain a denoised data set corresponding to each noise data set.

[0159] Finally, the parameter data sets with data quality greater than the preset quality threshold and all denoised data sets are used as high-quality data sets corresponding to all test periods.

[0160] It should be noted that the preset quality threshold is set to 0.6, and the specific value can be adjusted according to the implementation scenario and is not limited here; the Kalman filtering method is a technical means well known to those skilled in the art, and the specific process will not be repeated here.

[0161] Based on the above process, all performance parameters under a certain process flow can be analyzed to obtain a high-quality data set of all performance parameters, and then the performance test of the electric control cabinet can be carried out based on the high-quality data set of all performance parameters.

[0162] Preferably, in one embodiment of the present invention, the performance test of the electric control cabinet is performed based on all high-quality data sets corresponding to all performance parameters, including:

[0163] According to the design drawings and specifications of the electric control cabinet, a simulation test model corresponding to the electric control cabinet is established, and all high-quality data sets corresponding to all performance parameters are used as the database of the simulation test model and input into the simulation test model. Then, the simulation can be run in the software environment and the behavior and output of the system can be observed to perform performance tests on the electric control cabinet, that is, to simulate various fault conditions, such as component damage, power supply fluctuations, load mutations, etc., and obtain test results.

[0164] Finally, the control strategy of the electric control cabinet is adjusted based on the deviation between the test results and the expected results.

[0165] In summary, the embodiment of the present invention firstly needs to obtain the parameter data set of the performance parameter in multiple test periods for any performance parameter, and the parameter data set contains multiple parameter time series data. Since noise data is an accidental phenomenon compared to normal data, the parameter time series data in each parameter data set can be clustered based on the similarity between the data to obtain the clustering results. At this time, the parameter time series data in each cluster cluster in each cluster result can distinguish the normal data and noise data in the parameter data set to a certain extent. Further, based on the numerical characteristics of all parameter time series data in the parameter data set and the FP-Growth algorithm, each cluster cluster in the cluster result is fused and analyzed to obtain a frequent pattern model. At this time, the frequent pattern model can more significantly characterize the difference between normal data and noise data. Further, the frequent pattern model can be analyzed to analyze the correlation between data and the frequent patterns and the position distribution of frequent itemsets in the frequent pattern model, so as to quantify the possibility of the existence of noise data and obtain the probability of the existence of noise in the parameter data set, which provides a reference for the subsequent evaluation of the quality of the parameter data set. Furthermore, since noise data has characteristics in data fluctuation and trend change, the data quality of the parameter data set can be more accurately evaluated by comprehensively considering the change trend and data fluctuation of the parameter time series data and combining the probability of noise existence. This helps to accurately identify and filter out the noise data set, and then denoise it to obtain a high-quality data set. Finally, the performance test of the electric control cabinet is performed based on the high-quality data set corresponding to all performance parameters to obtain the test results. Since the embodiment of the present invention can accurately filter out the noise data set, only the noise data set is denoised in a large number of data sets, which can improve the efficiency of data denoising processing and ensure the accuracy and reliability of the test results.

[0166] This embodiment also provides a comprehensive automated testing system for an electric control cabinet, the system comprising a processor, a memory and a computer program, wherein the memory is used to store a corresponding computer program, the processor is used to run the corresponding computer program, and when the computer program runs on the processor, it can implement any one of the steps of a comprehensive automated testing method for an electric control cabinet.

[0167] See also Figure 6 , which shows a system block diagram of a comprehensive automated testing system for an electric control cabinet provided by an embodiment of the present invention, including: a data acquisition module 601: used to implement step S1 in the above method, a frequent pattern model construction module 602: used to implement step S2 in the above method, a data quality analysis module 603: used to implement step S3 in the above method, and a performance testing module 604: used to implement step S4 in the above method.

[0168] See also Figure 7, which shows a system structure diagram of a comprehensive automated testing system for an electric control cabinet provided by an embodiment of the present invention, including a processor 700, a memory 701, a bus 702 and a communication interface 703, wherein the processor 700, the communication interface 703 and the memory 701 are connected via the bus 702; wherein the memory 701 may include a high-speed random access memory, the bus 702 may be an ISA bus, a PCI bus or an EISA bus, etc., and the processor 700 may be an integrated circuit chip having signal processing capabilities.

[0169] The present invention also provides a computer-readable storage medium corresponding to the method provided in the above-mentioned embodiment. Figure 8 , the storage medium shown is a CD, on which a computer program (ie, a program product) is stored. When the computer program is run by a processor, it will execute the method provided by any of the aforementioned embodiments.

[0170] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), read-only memory (ROM), and other optical and magnetic storage media, which are not listed here one by one.

[0171] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0172] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A comprehensive automated testing method for an electric control cabinet, characterized in that: The method comprises: For any performance parameter of the electric control cabinet, a parameter data set of the performance parameter in each test period is obtained, each of the parameter data sets including parameter time series data of multiple data cycles; According to the similarity between each parameter time series data in each parameter data set, all parameter time series data are clustered to obtain clustering results; according to the numerical characteristics of all parameter time series data in each parameter data set and the frequent pattern mining algorithm, the parameter time series data in each cluster cluster in the clustering result are fused and analyzed to obtain the frequent pattern model corresponding to each parameter data set; In the frequent pattern model, based on the distribution of each item in all parameter time series data, the similarity between frequent item sets, and the position of frequent item sets in the frequent pattern model, the probability of noise existence in each parameter data set is determined; according to the probability of noise existence in each parameter data set, the change trend of the value in each parameter time series data in the parameter data set, and the value fluctuation, the data quality of each parameter data set is obtained; According to the data quality corresponding to the parameter data sets of all test periods, the noise data sets are screened from all parameter data sets and the parameter time series data in the noise data sets are denoised to obtain high-quality data sets corresponding to all test periods; the performance test of the electric control cabinet is performed based on all high-quality data sets corresponding to all performance parameters; The method for acquiring the frequent pattern model includes: For any parameter data set, a numerical range is determined based on the maximum and minimum values ​​in the parameter data set, the numerical range is evenly divided to obtain a preset number of interval ranges, and different interval ranges are marked with different labels; Using the label to replace each value in each parameter time series data, to obtain a label sequence corresponding to each parameter time series data; Obtain the frequency of occurrence of each tag in the parameter data set; in each cluster, according to the frequency of occurrence of all tags and the number of tag sequences in each cluster, obtain the importance index of each cluster, and the number of tag sequences in each cluster and the frequency of occurrence of tags are positively correlated with the importance index; In the clustering result corresponding to the parameter data set, all label sequences in each cluster are analyzed based on the FP-Growth algorithm to obtain the frequent pattern tree corresponding to each cluster; the frequent pattern trees corresponding to all clusters are arranged in descending order based on the importance index to obtain an arrangement sequence, in which the first frequent pattern tree is used as the target tree, the frequent pattern tree after the target tree is used as the tree to be analyzed, and the part where the tree to be analyzed and the target tree have differences is added to the target tree to obtain a new target tree; the frequent pattern tree after the tree to be analyzed is used as the new tree to be analyzed, and the part where the new tree to be analyzed and the new target tree have differences is added to the new target tree, and the new target tree is continuously determined until the frequent pattern trees in the arrangement sequence are traversed and stopped, and the new target tree at this time is used as the frequent pattern model corresponding to the parameter data set; The method for obtaining the noise existence probability includes: For any parameter data set, obtain all frequent item sets in the corresponding frequent pattern model; the frequency of occurrence of each item in the frequent item set is the frequency of occurrence of the label corresponding to the item in the parameter data set; For any frequent item set, based on the occurrence frequency of each item in the frequent item set, determine the first overall noise factor of the frequent item set, and the first overall noise factor is negatively correlated with the occurrence frequency; based on the number of bifurcation points of the path corresponding to the frequent item set in the frequent pattern model, determine the second overall noise factor of the frequent item set, and the second overall noise factor is negatively correlated with the number of bifurcation points; The first overall noise factor and the second overall noise factor of all frequent item sets are merged and averaged to obtain a first noise index of the parameter data set, wherein the first overall noise factor and the second overall noise factor are both positively correlated with the first noise index; Based on the similarities between all frequent item sets, determine the second noise index of the parameter data set; According to a first noise index and a second noise index of the parameter data set, a noise existence probability of the parameter data set is obtained, wherein the first noise index and the second noise index are both positively correlated with the noise existence probability; The method for obtaining the data quality includes: The parameter data set with a noise probability greater than or equal to a preset noise threshold is taken as the target data set, and the parameter data set with a noise probability less than the preset noise threshold is taken as the normal data set, and the data quality of the normal data set is set to a fixed value not less than 1; For any target data set, based on the change trend of each parameter time series data in the target data set, determine the first quality parameter of the target data set; In each parameter time series data corresponding to the target data set, based on the deviation between each value and the mean of all values, a deviation factor of each value is obtained, and based on the deviation factors of all values, an overall quality factor of each parameter time series data is determined, and the overall quality factor is negatively correlated with the deviation factor; the overall quality factors of all parameter time series data are averaged to obtain a second quality parameter of the target data set; The data quality of the target data set is obtained according to the noise existence probability, the first quality parameter and the second quality parameter of the target data set. The noise existence probability, the first quality parameter and the second quality parameter are all negatively correlated with the data quality, and the value of the data quality is a normalized value.

2. A comprehensive automated testing method for an electric control cabinet according to claim 1, characterized in that: The method for obtaining the clustering result includes: In each parameter data set, for any two parameter time series data in different data periods, based on the difference between the numerical values ​​and the difference between the slope values ​​of the numerical values, determine the difference characteristic value between the two parameter time series data; The difference characteristic values ​​between the parameter time series data are used as the distance metric, and K-means clustering analysis is performed on all the parameter time series data in the parameter data set based on a preset K value to obtain the clustering result.

3. A comprehensive automated testing method for an electric control cabinet according to claim 1, characterized in that: The method for obtaining the second noise index includes: Among all frequent item sets, the frequent item sets with the same number of items are regarded as a class of frequent item sets; In any type of frequent item sets, any two frequent item sets are combined to obtain all non-repeating combinations; for any combination, the independence value of the frequent item sets in the combination is determined based on the differences in the items at the same position in the two frequent item sets; The mean of the independence values ​​corresponding to all combinations of frequent item sets of all classes is taken as the second noise index of the parameter data set.

4. A comprehensive automated testing method for an electric control cabinet according to claim 1, characterized in that: The method for obtaining the first quality parameter includes: The number of extreme value points in each parameter time series data in the target data set is obtained, and the first quality parameter of the target data set is determined based on the proportion of the sum of the number of extreme value points of all parameter time series data in the total number of values ​​of all parameter time series data.

5. A comprehensive automated testing method for an electric control cabinet according to claim 1, characterized in that: The method for obtaining the high-quality data set includes: Among all parameter data sets, the parameter data sets whose data quality is less than or equal to the preset quality threshold are regarded as noise data sets; Based on the Kalman filtering method, all parameter time series data in each noise data set are denoised to obtain a denoised data set corresponding to each noise data set; The parameter data sets with data quality greater than the preset quality threshold and all denoised data sets are regarded as high-quality data sets corresponding to all test periods.

6. A comprehensive automated testing method for an electric control cabinet according to claim 1, characterized in that: The performance test of the electric control cabinet based on all high-quality data sets corresponding to all performance parameters includes: A simulation test model corresponding to the electric control cabinet is established, all high-quality data sets corresponding to all performance parameters are used as the database of the simulation test model, and the performance test of the electric control cabinet is performed to obtain the test results.

7. A comprehensive automated testing system for an electric control cabinet, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

Images