A process state labeling method and device based on an adaptive density clustering algorithm

Abstract

The application discloses a process state labeling method and device based on adaptive density clustering, comprising an adaptive density calculation module and an adaptive neighborhood search module. Firstly, a non-parametric density calculation method is used to realize automatic calculation of density; secondly, the adaptive neighborhood of the data set is calculated; finally, the density center is selected in a loop according to the density size from high to low, and the samples meeting the merging condition are classified into a unified cluster. A prominent feature of the application is that the parameters can be automatically set, which is not possessed by similar methods. The application can automatically label the class of real-time generated unlabeled process data, and solves the problems of high cost, high time consumption, non-uniform standard and low accuracy rate of current manual labeling.

Description

Technical Field

[0001] This invention belongs to the field of data mining and analysis technology, and specifically relates to a process state labeling method and apparatus based on an adaptive density clustering algorithm. Background Technology

[0002] With the rapid development of the Internet of Things (IoT), the improvement of information processing capabilities, and the diminishing demographic dividend, improving quality and efficiency has become an urgent need for my country's industrial development. Achieving digitalization, automation, and informatization enables comprehensive perception, automatic control, and centralized integration of industrial production sites, effectively reducing the intensity of manual operations in harsh natural environments and simplifying production management. Through intelligent IoT, problems can be diagnosed promptly, management processes optimized, work efficiency improved, refined management achieved, costs reduced, and efficiency increased. By integrating IoT technology with industrial production processes, a real-time information collection, transmission, and analysis application system for industrial production processes can be established, enabling process monitoring and optimized control, thereby significantly reducing energy consumption and effectively improving production efficiency.

[0003] Since most datasets in the Internet of Things (IoT) are collected by sensors, their samples are few or even unlabeled. This limits their effective use and positive feedback. Data annotation is time-consuming and laborious, especially when the data volume is large and updates rapidly. Therefore, manual annotation of large-scale IoT datasets is impossible. Furthermore, the strong correlations between samples necessitate research into sample decoupling. Therefore, clustering, as an unsupervised machine learning algorithm, has high application value in quickly and accurately labeling process states and is a prerequisite for maximizing the mining of process data, thereby achieving comprehensive perception, automatic control, and centralized integration of industrial production sites. Summary of the Invention

[0004] To address the aforementioned needs and problems, this invention provides a process state labeling method and apparatus based on an adaptive density clustering algorithm, which automatically labels unlabeled process data generated in real time, solving the problems of high cost, high time consumption, inconsistent standards, and low accuracy of current manual labeling.

[0005] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0006] A process state labeling method based on an adaptive density clustering algorithm includes the following steps:

[0007] 1) Obtain unlabeled sensor real-time process data as samples to construct an unlabeled dataset;

[0008] 2) Calculate the Euclidean distance matrix and adaptive neighborhood for all samples in the unlabeled dataset;

[0009] 3) Calculate the adaptive density of the unlabeled dataset;

[0010] 4) Draw a decision graph based on the adaptive density of the samples and the Euclidean distance between the samples, and select the density center as the starting set for the search;

[0011] 5) All starting points are searched for neighborhoods in descending order of adaptive density. All samples in their neighborhood that meet the merging conditions are labeled with the same cluster as the starting point. The labeled samples are then used as the starting point for the next round of neighborhood search until there are no more samples in the neighborhood that meet the conditions, and the search ends.

[0012] 6) Return to step 5) and select the next unlabeled starting point from the starting point set to perform a neighborhood search until the starting point set is cleared, thus completing the state labeling of the sensor process data.

[0013] The adaptive density for:

[0014]

[0015] in, n Indicates the number of samples. d ij Indicates sample i and samples j Euclidean distance, d ave The adaptive neighborhood represents the average distance from all points to their nearest points.

[0016] .

[0017] Step 4) specifically refers to:

[0018] In the unlabeled dataset, for each sample point i Find all points that are larger than the sample point. i High density sample points j Select the smallest one. d ij , recorded as δ i ,Will δ i and i Plotting these axes as the vertical and horizontal axes respectively yields a decision graph, with the density center being the... δ Samples with values ​​greater than the threshold.

[0019] Step 5) includes the following steps:

[0020] Arrange the selected density centers in descending order of density as the starting set for the search, and use one of the search starting points as the center of a circle.d ave As the radius of the adaptive neighborhood, samples with a density greater than a density threshold within the neighborhood are all labeled as cluster C, which is the same as the search starting point. i That is, satisfying the following expression:

[0021] C[ i ] d ave

[0022] in, D ' It is a sample set that satisfies the condition that the density is greater than the density threshold;

[0023] The labeled samples are used as the starting point for the next round of neighborhood search, and the next round of neighborhood search is carried out until there are no more samples that meet the density condition in the neighborhood of all samples.

[0024] Select the next search starting point from the set of search starting points. If the starting point has already been labeled, discard it and select the next search starting point for neighborhood search. Continue in this manner until there are no more unlabeled starting points in the set of search starting points. The neighborhood search process ends, and the remaining unlabeled points are labeled with the same label as the nearest point.

[0025] A process state labeling device based on an adaptive density clustering algorithm, comprising:

[0026] The data acquisition module is used to acquire unlabeled sensor real-time process data as samples to construct an unlabeled dataset;

[0027] The adaptive density calculation module is used to calculate the Euclidean distance matrix and adaptive density of all samples in the unlabeled dataset; and to calculate the adaptive neighborhood of the unlabeled dataset.

[0028] The adaptive neighborhood search module is used to draw a decision graph based on the adaptive density of samples and the Euclidean distance between samples, and select density centers as the starting point set for the search. All starting points are searched for neighborhoods in descending order of adaptive density. All samples in their neighborhood that meet the merging conditions are labeled with the same cluster as the starting point. The labeled samples are then used as the starting points for the next round of neighborhood search, until there are no more samples in the neighborhood that meet the conditions, at which point the search ends. The next unlabeled starting point is selected from the starting point set for neighborhood search, until the starting point set is cleared, thus completing the state labeling of the sensor process data.

[0029] A process state labeling device based on an adaptive density clustering algorithm includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement the process state labeling method based on the adaptive density clustering algorithm when the computer program is executed.

[0030] A computer-readable storage medium storing a computer program that, when executed by a processor, implements a process state labeling method based on an adaptive density clustering algorithm.

[0031] The present invention has the following beneficial effects and advantages:

[0032] 1. This invention proposes a non-parametric density calculation method. Based on the size and characteristics of the dataset, the average distance from a sample to its nearest sample is calculated, and this distance is used as a parameter to calculate the density of each point.

[0033] 2. This invention proposes the concept of density centers and its association selection method based on decision graphs. Density centers are the starting points for clustering and directly affect the clustering results. A density center selection strategy based on decision graphs is used to adjust the clustering results.

[0034] 3. This invention proposes a neighborhood-adaptive neighborhood search clustering method. The neighborhood radius of each dataset is only related to the density of data points and can be calculated automatically, thus avoiding manual setting.

[0035] 4. This invention is standardized, has a simple structure, and is easy to standardize and promote on a large scale. The device has a wide range of applications, strong versatility, low manufacturing cost, is easy to promote and use, and has a wide range of applications, resulting in good economic and social benefits. Attached Figure Description

[0036] Figure 1 A schematic diagram of the method flow of this invention;

[0037] Figure 2 Decision diagram illustration;

[0038] Figure 3 Neighborhood search diagram;

[0039] Figure 4 A two-dimensional display of typical sample points;

[0040] Figure 5 A diagram showing the first round of neighborhood search for density center 1;

[0041] Figure 6 A diagram illustrating the first round of cluster merging at density center 1;

[0042] Figure 7The second round of neighborhood search for density center 1 is shown in the diagram.

[0043] Figure 8 A diagram illustrating the second round of cluster merging at density center 1;

[0044] Figure 9 The third round of neighborhood search for density center 1 is shown in the diagram.

[0045] Figure 10 The third round of cluster merging at density center 1 is illustrated in the diagram.

[0046] Figure 11 The first round of neighborhood search for density center 2 is shown in the diagram.

[0047] Figure 12 The first round of cluster merging at density center 2 is illustrated in the diagram.

[0048] Figure 13 The second round of neighborhood search for density center 2 is shown in the diagram.

[0049] Figure 14 The second round of cluster merging at density center 2 is illustrated in the diagram.

[0050] Figure 15 The final cluster diagram shows the two clusters. Detailed Implementation

[0051] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0052] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the steps as sequential processes, many of these steps can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the steps can be rearranged. The process can be terminated when its operation is complete, but may also have additional steps not included in the figures. The process can correspond to a method, function, procedure, subroutine, subroutine, etc.

[0053] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0054] like Figure 1 As shown, in a first aspect, the present invention provides a process state labeling method based on an adaptive density clustering algorithm, the method comprising:

[0055] Acquire unlabeled real-time process data;

[0056] Calculate the Euclidean distance matrix for all samples in the data, and calculate the adaptive neighborhood for all samples;

[0057] Calculate the adaptive density of the dataset, draw a decision graph based on the density of the samples and their distances to each other, and then manually select the density centers as the starting set for the search;

[0058] All starting points are searched for neighborhoods in descending order of density. All samples in their neighborhood that meet the merging criteria are labeled with the same cluster as the starting point. The labeled samples then serve as the starting point for the next round of neighborhood search, until there are no more samples in the neighborhood that meet the criteria, at which point the next round of search ends.

[0059] Then, select the next unlabeled starting point from the sample set and perform a neighborhood search, repeating this process until the starting point set is cleared.

[0060] The method for calculating the adaptive density:

[0061] For sample points from the process dataset i , j ,sample i The adaptive density satisfies the following expression:

[0062]

[0063] Special:

[0064]

[0065] The fundamental equation for adaptive density is a typical Gaussian kernel function. .sample j Located in the sample i Within its neighborhood, yes j right i The contribution of density, its contribution is related to the surrounding points and i The distance is inversely proportional to the sample, that is, the sample i The density is only sensitive to points in its vicinity. d ij Indicates sample i and j Euclidean distance, d ave It is the average distance from all points to their nearest points, i.e., the adaptive neighborhood. This invention presents a new Gaussian kernel-based definition of distance without manually setting the parameters of the density formula; all parameters are determined based on the selected dataset.

[0066] The method for drawing a decision map and selecting a density center specifically includes:

[0067] In the process dataset, for each sample point i Find all points that are larger than the sample point. i High density sample points j Select the smallest one. d ij , recorded as δ i .Will δ i and i Plotting these axes as the vertical and horizontal axes creates a decision graph. The density center represents the area where choices are relatively more frequent. δ A portion of the value (the specific quantity is determined by the operator based on the actual situation).

[0068] The neighborhood search method specifically includes:

[0069] Arrange the selected density centers in descending order of density as the search starting set, and use the search starting point as the center of the circle. d ave As the radius of the adaptive neighborhood, samples with a density greater than a density threshold within the neighborhood are all labeled as cluster C, which is the same as the search starting point. i That is, satisfying the following expression:

[0070] C[ i ] d ave

[0071] in D ' It is a sample set that satisfies the condition that the density is greater than the density threshold.

[0072] like Figure 3 As shown, the labeled samples in this process serve as the starting point for the next round of neighborhood search, continuing until no samples satisfy the density condition remain in their neighborhoods. At this point, the next search starting point is selected from the search starting point set. If this starting point has already been labeled in the previous neighborhood search process, it is discarded, and the next search starting point is selected for the neighborhood search step. This process continues until there are no more unlabeled starting points in the search starting point set, at which point the neighborhood search process ends.

[0073] The remaining unlabeled points are labeled with the same tag as the nearest point.

[0074] A second aspect of the present invention provides an automatic image category labeling device based on a density peak clustering algorithm, comprising: a processor; and a memory, wherein the memory stores a computer-executable program, and when the processor executes the computer-executable program, the above-described automatic image category labeling method based on the density peak clustering algorithm is executed.

[0075] In a third aspect, the present invention provides a computer-readable storage medium having instructions stored thereon, which, when executed by a processor, cause the processor to perform the above-described automatic image category labeling method based on density peak clustering algorithm.

[0076] This invention discloses an automatic image category labeling method, apparatus, device, and computer storage medium based on a density peak clustering algorithm. It employs a non-parametric density calculation method to automatically calculate density; calculates the adaptive neighborhood of the dataset; and iteratively selects density centers from high to low density, grouping samples that meet the merging criteria into a unified cluster. A key feature is its ability to automatically set parameters, a capability not found in similar methods. This invention can automatically label unlabeled process data generated in real time, solving the problems of high cost, high time consumption, inconsistent standards, and low accuracy associated with current manual labeling methods.

[0077] This invention provides the following embodiments for an automatic image category labeling method, apparatus, and storage medium based on a density peak clustering algorithm:

[0078] Based on Embodiment 1 of the present invention

[0079] Obtain oil drilling process data for a preset time period.

[0080] The data is in the form of multi-sensor time-series data, taken from a digitization project of 300 oil wells in the oilfield, utilizing a wireless network of 1,500 monitoring sensors to generate process data. Key production parameters such as temperature, oil pressure, casing pressure, and electrical signals are collected through sensing technology. Liquid volume, dynamic liquid level, and pump efficiency are calculated from this IoT data. Through real-time monitoring of the acquired data, the oil well data is aggregated and labeled with operating conditions, providing data support for subsequent functions such as single-well operating condition diagnosis, single-well measurement, intelligent frequency conversion control, and balance adjustment.

[0081] First, data preprocessing was performed based on the displacement data; after removing missing and duplicate samples, approximately 99.8% of the data was retained. Second, 56,500 samples with a displacement stroke of 3000 were selected, which is less than 68.9%; the displacement curves of these samples, after smoothing, are all sinusoidal curves, consistent with the physical model of the pumping unit. Displacement and pressure per stroke are the most critical parameters for oil meter readings. Load pressure data contains a large amount of information; therefore, this embodiment focuses on load analysis.

[0082] For example, as shown in Table 1, the oil pressure values ​​obtained by the oil well drill bit at 200 sampling points in each cycle.

[0083] Table 1

[0084]

[0085] Based on Embodiment 2 of the present invention

[0086] Calculate the adaptive density value of the oil well drill bit pressure sample.

[0087] The distance matrix between samples in Table 1 is calculated. In this embodiment, the distance calculation method can be expressed as follows:

[0088]

[0089] i, j =1,2,…200

[0090] To calculate the adaptive neighborhood value of this segment of oil drilling data, for each sample, select the nearest distance to that sample from all samples in the dataset, and then calculate the average of these distances as the adaptive neighborhood of that segment of data. The calculation method can be expressed as:

[0091]

[0092] i, j =1,2,…200

[0093] This value is used to calculate the adaptive neighborhood of each sample in this data segment. iBy using the Gaussian kernel function, the contribution of all surrounding samples to the density is estimated. Samples that are closer to the density contribute more to the density, and vice versa. The calculation method can be expressed as:

[0094]

[0095] Based on Embodiment 3 of the present invention

[0096] Methods for selecting density centers

[0097] Through Example 2, we obtained the distance matrix and adaptive density of this data segment. For each sample point... i We found all points that were larger than that sample point. i High density sample points j Select the smallest one. d ij , recorded as δi The calculation method can be expressed as:

[0098]

[0099] All samples in this data segment δ and ρ Plot the graphs using the vertical and horizontal axes respectively, as follows: Figure 2 As shown:

[0100] according to Figure 2 Select the density center of this data segment, and then set... δ The threshold size is 6, and the filter is... δ The six samples exceeding the threshold are designated as density centers. Figure 2 Part A of the equation. At this point, we have obtained the starting point set for the next neighborhood search: { p 1 ,p 2 ,…,p 6 Then, the set is sorted in descending order according to the density of each point, resulting in: { p 1 ,p 2 ,p 4 ,p 5 ,p 3 , p 6}

[0101] Based on Embodiment 4 of the present invention

[0102] Adaptive Neighborhood Search Method

[0103] To facilitate a more intuitive demonstration of this embodiment, we have reduced the 200 dimensions of this data segment and extracted 21 typical samples for neighborhood search. The following operations are all 2D displays of typical sample points of this oil well process data segment.

[0104] There are 21 sample points participating in the adaptive neighborhood search in this embodiment. The sample numbers are shown below. Figure 4 Samples 1 and 15 are density centers selected according to Example 3, which are the search starting points of this example.

[0105] Because the density of sample 1 is greater than that of sample 15, the neighborhood search is performed starting from sample 1, with a neighborhood radius of . d ave ,like Figure 5 Samples 2, 3, 4, 5, and 6 shown are all within this neighborhood; however, none of these five samples fall within the specified neighborhood. Figure 2 Within region B, the conditions for cluster merging are met. Therefore, these 5 samples are labeled as belonging to the same cluster as point 1. Figure 6 .

[0106] At this point, samples 2, 3, 4, 5, and 6 become the origin for the next round of neighborhood search, with a neighborhood radius of... d ave Sample points 7, 8, 9, 10, 11, etc., within the neighborhood were obtained through searching. Figure 7 , Figure 8 These 5 samples are then labeled as belonging to the same cluster as sample 1, and also serve as the origin for the next round of neighborhood search, with a neighborhood radius of [missing value]. d ave Sample 12 was obtained from the neighborhood through the search, such as Figure 9 , Figure 10 .

[0107] At this point, there is no starting point for the next round of neighborhood search within the neighborhood, so we search for the next unlabeled starting point from the initial starting point set, which in this example is sample 15. We take this sample as the origin. d ave A neighborhood search was performed on the neighborhood radius, yielding samples 16, 17, 18, and 19, which were labeled as belonging to the same cluster as sample 15. Figure 11 , Figure 12 .

[0108] Using these four samples as the origin, d ave Perform another neighborhood search based on the neighborhood radius, obtaining samples 20 and 22, which are then labeled as belonging to the same cluster as sample 15, such as... Figure 13 .

[0109] At this point, the neighborhood search for this round ends, and there are no origin samples for the next round. Note that the initial starting point set no longer contains labeled starting points; thus, the neighborhood search process concludes.

[0110] The remaining samples are 13, 14, 21, and 23. Samples 13 and 21 have a density greater than the set threshold of 0.0045, so these two samples are labeled with the same label as their nearest neighbor. Sample 13 is labeled with the same class label as sample 8, i.e., label 1. Sample 21 is the closest to sample 21, so it is labeled with label 15. The remaining samples 14 and 23, since they are located in region B of the decision graph (i.e., their density value is less than the threshold of 0.0045), are considered noise points and are not labeled. The final result is as follows. Figure 15 As shown, samples 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, and 13 are labeled as cluster 1, and samples 15, 16, 17, 18, 19, 20, 21, and 22 are labeled as cluster 15. Samples 14 and 23 are labeled as noise points, as shown below. Figure 14 .

[0111] Through this embodiment, the sample in this segment can be clustered into two categories, and these two clusters can be labeled as two different oil well process states, thus completing the process state labeling based on the adaptive density clustering algorithm.

[0112] Furthermore, in conjunction with the data indicator determination methods in the above embodiments, this application embodiment can provide a computer storage medium for implementation. This computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the data indicator determination methods in the above embodiments.

[0113] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0114] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A process state labeling method based on an adaptive density clustering algorithm, characterized in that, Includes the following steps: 1) Obtain unlabeled sensor real-time process data as samples to construct an unlabeled dataset; 2) Calculate the Euclidean distance matrix and adaptive neighborhood for all samples in the unlabeled dataset; 3) Calculate the adaptive density of the unlabeled dataset; 4) Draw a decision graph based on the adaptive density of the samples and the Euclidean distance between the samples, and select the density center as the starting set for the search; 5) All starting points are searched for neighborhoods in descending order of adaptive density. All samples in their neighborhood that meet the merging conditions are labeled with the same cluster as the starting point. The labeled samples are then used as the starting point for the next round of neighborhood search until there are no more samples in the neighborhood that meet the conditions, and the search ends. 6) Return to step 5) and select the next unlabeled starting point from the starting point set to perform a neighborhood search until the starting point set is cleared, thus completing the state labeling of the sensor process data; The adaptive density for: ； in, n Indicates the number of samples. d ij Indicates sample i and samples j Euclidean distance, d ave The adaptive neighborhood represents the average distance from all points to their nearest points. 。 2. The process state labeling method based on adaptive density clustering algorithm according to claim 1, characterized in that, Step 4) specifically refers to: In the unlabeled dataset, for each sample point i Find all points that are larger than the sample point. i High density sample points j Select the smallest one. d ij , recorded as δ i ,Will δ i and i Plotting these axes as the vertical and horizontal axes respectively yields a decision graph, with the density center being the... δ Samples with values ​​greater than the threshold.

3. The process state labeling method based on adaptive density clustering algorithm according to claim 1, characterized in that, Step 5) includes the following steps: Arrange the selected density centers in descending order of density as the starting set for the search, and use one of the search starting points as the center of a circle. d ave As the radius of the adaptive neighborhood, samples with a density greater than a density threshold within the neighborhood are all labeled as cluster C, which is the same as the search starting point. i That is, satisfying the following expression: C[ i ] d ave ； in, D ' It is a sample set whose density is greater than the density threshold; The labeled samples are used as the starting point for the next round of neighborhood search, and the next round of neighborhood search is carried out until there are no more samples that meet the density condition in the neighborhood of all samples. Select the next search starting point from the set of search starting points. If the starting point has already been labeled, discard it and select the next search starting point for neighborhood search. Continue in this manner until there are no more unlabeled starting points in the set of search starting points. The neighborhood search process ends. The remaining unlabeled points are labeled with the same label as the nearest point.

4. A process state labeling device based on an adaptive density clustering algorithm, the device being used to implement the process state labeling method based on an adaptive density clustering algorithm as described in any one of claims 1-3, characterized in that, include: The data acquisition module is used to acquire unlabeled sensor real-time process data as samples to construct an unlabeled dataset; The adaptive density calculation module is used to calculate the Euclidean distance matrix and adaptive density of all samples in the unlabeled dataset. Calculate the adaptive neighborhood of an unlabeled dataset; The adaptive neighborhood search module is used to draw a decision graph based on the adaptive density of samples and the Euclidean distance between samples, and select density centers as the starting point set for the search. All starting points are searched for neighborhoods in descending order of adaptive density. All samples in their neighborhood that meet the merging conditions are labeled with the same cluster as the starting point. The labeled samples are then used as the starting point for the next round of neighborhood search, until there are no more samples in the neighborhood that meet the conditions, at which point the search ends. The next unlabeled starting point is selected from the starting point set for neighborhood search, until the starting point set is cleared, thus completing the state labeling of the sensor process data.

5. A process state labeling device based on an adaptive density clustering algorithm, characterized in that, It includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement, when executing the computer program, a process state labeling method based on an adaptive density clustering algorithm as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements a process state labeling method based on an adaptive density clustering algorithm as described in any one of claims 1-3.

Images