A digital twin system for sensing and predicting the state of a machine room device

Abstract

The present application relates to the field of device twin monitoring, and more particularly to a digital twin system for sensing and predicting the state of equipment in a computer room, which converts server historical operation logs into log images through an image analysis module, generates binary pixel images through pixelization processing, and extracts image parameter change curves using a sliding window; a regularity analysis module analyzes periodicity based on the curves and verifies text coincidence, divides server nodes into visual regularity categories; a twin detection module performs differential monitoring based on the visual regularity categories, identifies non-periodic curve segments in strong visual regularity nodes for text traversal analysis, and performs full-text traversal analysis on weak visual regularity nodes to determine whether server nodes have abnormalities. The present application is based on image modal analysis and log periodicity identification, realizes feature extraction and classification monitoring of massive log data, reduces the amount of calculation and data transmission, and improves the overall performance of the system in processing large-scale log data.

Description

Technical Field

[0001] This invention relates to the field of equipment twin monitoring, and more particularly to a digital twin system for sensing and predicting the status of equipment in a computer room. Background Technology

[0002] As the core physical carrier of enterprise information systems, the continuous and stable operation of internal servers, network equipment, and environmental systems in data centers is crucial for ensuring business continuity. Traditional data center monitoring systems mainly rely on fixed thresholds for anomaly alerts, which are insufficient to cope with the combined effects of equipment performance degradation, complex environmental changes, and sudden load fluctuations. In recent years, digital twin technology has been introduced into the field of industrial operations and maintenance. By constructing virtual mappings of physical entities, it enables real-time status perception and operational prediction, thereby improving the level of intelligent operations and maintenance.

[0003] For example, Chinese Patent Publication No. CN115941453A discloses a data center alarm processing method, device, data center, and medium based on digital twins. The data center alarm processing method includes: acquiring a digital twin virtual model corresponding to the generated data center and constructing a digital twin data center; simulating the operation of the digital twin data center in multiple scenarios; real-time monitoring of a first state parameter of the digital twin data center; determining whether the first state parameter exceeds a corresponding first threshold range parameter; if yes, storing the first state parameter in a first historical alarm information database; if no, storing the first state parameter in a second historical alarm information database; real-time monitoring of a second state parameter during actual operation of the data center; determining whether the second state parameter exceeds a corresponding second threshold range parameter; if no, the data center operates normally; if yes, storing the second state parameter in an actual alarm information database and issuing a total alarm message.

[0004] However, the following problems still exist in the existing technology:

[0005] 1. Existing technologies have not achieved lightweight and feature compression processing of massive log data, resulting in high computational pressure and high data transmission load on central nodes;

[0006] 2. Existing technologies do not take into account the inherent periodic patterns in equipment operation logs, and cannot differentiate monitoring strategies based on these patterns, resulting in redundant consumption of computing and communication resources. Summary of the Invention

[0007] To address this, the present invention provides a digital twin system for data center equipment status perception and prediction, which overcomes the problems of not achieving lightweight and feature compression processing of massive log data, high computational pressure, high data transmission load of central nodes, failure to consider the inherent periodicity in equipment operation logs, inability to differentiate monitoring strategies based on regularity, and redundant consumption of computing and communication resources.

[0008] To achieve the above objectives, the present invention provides a digital twin system for data center equipment status perception and prediction, comprising:

[0009] The image parsing module is used to obtain historical operation logs of several server nodes in the computer room, convert the historical operation logs into log images, and perform image modality parsing, including pixelating the log images to obtain binary pixel images, constructing a sliding window in the binary pixel images, extracting image parameters during the sliding process of the sliding window, and obtaining historical image parameter change curves.

[0010] The pattern analysis module, which is connected to the image analysis module, is used to analyze the periodicity of the curve based on the parameter change curve of the historical image, determine the proportion of periodic curve segments, extract the running log text fragments corresponding to each periodic curve segment, and verify the text overlap within the running log text fragments in order to classify the visual pattern categories corresponding to the server nodes.

[0011] A twin detection module, which is connected to the pattern parsing module, responds to the server node generating new operation logs and performs monitoring based on the visual pattern category corresponding to the server node, including:

[0012] The operation log is converted into a log image, an image parameter change curve is constructed, non-periodic curve segments are identified, and the text traversal and parsing of the operation log segments corresponding to the non-periodic curve segments are performed to determine whether there are any abnormalities in the server nodes.

[0013] Alternatively, the runtime logs can be parsed to determine if any server nodes are abnormal.

[0014] Furthermore, the image parsing module performs pixelation on the log image to obtain a binary pixel image, including:

[0015] This is used to identify text and non-text regions in a log image, covering each text region with a black pixel block and the non-text region with a white pixel block to obtain a binary pixel image.

[0016] Furthermore, the image parsing module constructs a sliding window in the binary pixel image, wherein,

[0017] The sliding window is a rectangular window with a preset height and a width that is the same as the total width of the binary pixel image.

[0018] Furthermore, the sliding window constructed by the image parsing module slides from top to bottom along the vertical direction of the binary pixel image with a preset step size.

[0019] Furthermore, the image analysis module extracts image parameters during the sliding window's movement and obtains historical image parameter change curves, including...

[0020] After each movement of the sliding window, the proportion of black pixels in the current sliding window area is calculated, and the proportion is used as the image parameter of the corresponding sliding window.

[0021] Record the time and image parameters corresponding to the sliding window at different positions, and construct the historical image parameter change curve with time as the horizontal axis and image parameters as the vertical axis.

[0022] Furthermore, the pattern analysis module analyzes the periodicity of the curve based on the historical image parameter change curve, and determines the proportion of periodic curve segments, including:

[0023] Used to perform periodic detection on the parameter change curves of the historical images, and to identify periodic curve segments in the parameter change curves of the historical images that conform to periodic patterns;

[0024] The ratio of the cumulative time length of the periodic curve segment to the total time length of the historical image parameter change curve is used as the periodic curve segment ratio.

[0025] Furthermore, the pattern analysis module extracts the corresponding runtime log text segments for each periodic curve segment and verifies the text overlap within the runtime log text segments, including...

[0026] Based on the time period corresponding to the periodic curve segment on the historical image parameter change curve, the text content of the corresponding time period in the historical operation log is mapped and extracted as the operation log text fragment.

[0027] Calculate the similarity between runtime log text segments corresponding to different periodic curve segments;

[0028] When the average similarity between the various log text segments is higher than a preset threshold, the text overlap verification is considered successful.

[0029] Furthermore, the pattern parsing module categorizes the visual patterns corresponding to the server nodes, including:

[0030] If the proportion of periodic curve segments is greater than the proportion threshold and the text overlap verification passes, the server node is classified as a node with strong visual regularity; otherwise, the server node is classified as a node with weak visual regularity.

[0031] Furthermore, the twin detection module responds to new operation logs generated by the server node and performs monitoring based on the visual pattern category corresponding to the server node, including:

[0032] If the server node is a node with strong visual regularity, the new operation log is converted into a log image, a real-time image parameter change curve is constructed, and the non-periodic curve segment is identified and locked.

[0033] Extract the running log text fragments corresponding to the non-periodic curve segments, and perform text traversal parsing on the running log text fragments to determine whether there are any abnormalities in the server nodes.

[0034] Furthermore, the twin detection module, in response to new operation logs generated by the server node, monitors based on the visual pattern categories corresponding to the server node, and also includes...

[0035] If the server node is a weakly visually predictable node, then the new running log is directly traversed and parsed in its entirety to determine whether the server node is abnormal.

[0036] Compared with existing technologies, this invention uses an image analysis module to convert historical server operation logs into log images, generates binary pixel images through pixelation, and extracts image parameter change curves using a sliding window. A pattern analysis module analyzes the periodicity of these curves and verifies text overlap, classifying server nodes into visual pattern categories. A twin detection module performs differentiated monitoring based on these visual pattern categories, identifying non-periodic curve segments within nodes with strong visual patterns and performing full-text traversal analysis on nodes with weak visual patterns to determine if any server nodes are abnormal. This invention, based on image modality analysis and log periodicity pattern identification, achieves feature extraction and classification monitoring of massive log data, reducing computational and data transmission loads and improving the overall performance of the system in processing large-scale log data.

[0037] In particular, this invention considers the visualization and pixel compression mechanisms of log data. In practice, when faced with massive amounts of text logs, direct processing presents challenges such as large data transmission volumes, abundant redundant information, and heavy real-time parsing computational load. This invention uses an image parsing module to convert log text into an image modality, and further performs pixelation processing to generate binary pixel images, compressing complex text information into black-and-white pixelated images. By transforming unstructured log content into structured images, subsequent processing into image dimensions is facilitated, reducing the complexity of data processing. This transforms the dynamic fluctuations of log content into visual and quantifiable time-series image signals, laying the foundation for subsequent periodic analysis.

[0038] In particular, this invention considers the construction and sliding mechanism of a sliding window. In actual pixel image processing, it is necessary to effectively convert the pixel information distributed in two-dimensional space into analyzable temporal signals. This invention constructs a rectangular sliding window, whose width is consistent with the total width of the binary pixel image, while the window height is preset according to the analysis requirements. This sliding window slides systematically from top to bottom along the vertical direction of the binary pixel image with a preset fixed step size. By continuously sampling the image in segments, it achieves the information conversion from the spatial dimension to the temporal dimension. This sliding mechanism provides an effective data organization form for subsequent feature extraction, ensuring that all feature information in the image can be captured.

[0039] In particular, this invention considers the extraction and curve construction mechanism of image parameters. During the sliding window traversal, it is necessary to convert the spatial information of the image into analyzable temporal data. This invention calculates the proportion of black pixels within the current window area in real time after each movement of the sliding window and defines this proportion as an image parameter. By recording the image parameter values ​​corresponding to different positions of the sliding window and arranging them in chronological order to form a sequence, a historical image parameter change curve is constructed. The historical image parameter change curve transforms the spatial distribution characteristics of a two-dimensional image into a one-dimensional temporal signal. Furthermore, since the image parameters are extracted based on black pixels, which are an inherent attribute of the image, the extraction is rapid. Based on this, a quantitative representation of the density changes in log content is achieved, providing an accurate data foundation for subsequent periodic analysis.

[0040] In particular, this invention considers a differentiated monitoring triggering mechanism based on pattern categories. In actual operation and maintenance environments, the operating status and log generation patterns of different server nodes differ significantly. Most devices in the data center perform a large number of repetitive or periodic tasks under normal operating conditions, resulting in stable patterns in the generation sequence and content structure of their operation logs. However, under abnormal operating conditions, these inherent patterns are easily disrupted by sudden errors, performance fluctuations, or unexpected interruptions. Traditional unified monitoring methods waste computing and communication resources. This invention identifies the visual pattern categories of nodes through a pattern parsing module, and the twin detection module adaptively activates corresponding monitoring strategies based on these categories. The core of this mechanism is that it does not require real-time reading and parsing of the entire log text content itself, but instead prioritizes monitoring the stability of macroscopic patterns through image parameter change curves, thereby quickly capturing abnormal device signs. For nodes with strong visual patterns, text parsing is only performed on the corresponding log text fragments when their operation logs deviate from the identified periodic patterns; for nodes with weak visual patterns, full log text analysis is performed. This mechanism achieves dynamic allocation of computing resources by differentiating node characteristics, avoids redundant processing of periodic normal logs, effectively reduces computing load and network transmission pressure, and achieves the goal of accurate and efficient anomaly detection. Attached Figure Description

[0041] Figure 1 A schematic diagram of the structural connection of a digital twin system for sensing and predicting the status of computer room equipment, as shown in an embodiment of the invention;

[0042] Figure 2 This is a schematic diagram of pixelation processing in an embodiment of the invention;

[0043] Figure 3 This is a logic block diagram for determining whether the text overlap verification is passed in an embodiment of the invention.

[0044] Figure 4 This is a logic block diagram illustrating the classification of visual patterns corresponding to server nodes in an embodiment of the invention. Detailed Implementation

[0045] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0046] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0047] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0048] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0049] Please see Figure 1 As shown, Figure 1 This is a schematic diagram of the structural connection of a digital twin system for data center equipment status perception and prediction. The digital twin system for data center equipment status perception and prediction according to an embodiment of the present invention includes:

[0050] The image parsing module is used to obtain historical operation logs of several server nodes in the computer room, convert the historical operation logs into log images, and perform image modality parsing, including pixelating the log images to obtain binary pixel images, constructing a sliding window in the binary pixel images, extracting image parameters during the sliding process of the sliding window, and obtaining historical image parameter change curves.

[0051] The pattern analysis module, which is connected to the image analysis module, is used to analyze the periodicity of the curve based on the parameter change curve of the historical image, determine the proportion of periodic curve segments, extract the running log text fragments corresponding to each periodic curve segment, and verify the text overlap within the running log text fragments in order to classify the visual pattern categories corresponding to the server nodes.

[0052] A twin detection module, which is connected to the pattern parsing module, responds to the server node generating new operation logs and performs monitoring based on the visual pattern category corresponding to the server node, including:

[0053] The operation log is converted into a log image, an image parameter change curve is constructed, non-periodic curve segments are identified, and the text traversal and parsing of the operation log segments corresponding to the non-periodic curve segments are performed to determine whether there are any abnormalities in the server nodes.

[0054] Alternatively, the runtime logs can be parsed to determine if any server nodes are abnormal.

[0055] Please see Figure 2 As shown, Figure 2 This is a schematic diagram of pixelation processing according to an embodiment of the invention. The image parsing module performs pixelation processing on the log image to obtain a binary pixel image, including:

[0056] This is used to identify text and non-text regions in a log image, covering each text region with a black pixel block and the non-text region with a white pixel block to obtain a binary pixel image.

[0057] There are no specific limitations on the method of identifying text regions and non-text regions in log images. When identifying text regions, it is not necessary to identify the content of the text. It is only necessary to locate the text position to save computing power. Subsequently, black pixel blocks can be used to cover the text. For example, an image segmentation algorithm can be used to identify the text outline and then cover it with black pixel blocks.

[0058] This invention considers the visualization and pixel compression mechanisms of log data. In practice, when faced with massive amounts of text logs, direct processing presents challenges such as large data transmission volumes, abundant redundant information, and heavy real-time parsing computational load. This invention utilizes an image parsing module to convert log text into an image modality, and further performs pixelation processing to generate binary pixel images, compressing complex text information into black-and-white pixelated images. By transforming unstructured log content into structured images, subsequent image-level processing is facilitated, reducing data processing complexity. This transforms the dynamic fluctuations of log content into visual and quantifiable time-series image signals, laying the foundation for subsequent periodic analysis.

[0059] Specifically, the image parsing module constructs a sliding window in the binary pixel image, wherein,

[0060] The sliding window is a rectangular window with a preset height and a width that is the same as the total width of the binary pixel image.

[0061] Specifically, the window height setting needs to strike a balance between the coverage of log content and the accuracy of periodic identification. Theoretically, the window height should not be too large to avoid a single window containing too much content, thereby blurring or obscuring potential short-term periodic patterns. The window height should be set between 2 and 4 times the standard line height of the log text, preferably 2 times.

[0062] Specifically, the sliding window constructed by the image parsing module slides from top to bottom along the vertical direction of the binary pixel image with a preset step size.

[0063] Understandably, the preset step size is consistent with the window height to achieve complete sampling of the binary pixel image without repetition. When the sliding step size equals the window height, the distance the window moves vertically downwards each time is exactly equal to its own height. This ensures that two adjacent sampling windows are connected end-to-end in vertical space, with no overlapping areas or gaps.

[0064] This invention considers the construction and sliding mechanism of a sliding window. In actual pixel image processing, it is necessary to effectively convert the two-dimensional spatially distributed pixel information into analyzable temporal signals. This invention constructs a rectangular sliding window, whose width is consistent with the total width of the binary pixel image, while the window height is preset according to the analysis requirements. This sliding window slides systematically from top to bottom along the vertical direction of the binary pixel image with a preset fixed step size. By continuously sampling the image in segments, it achieves the information conversion from the spatial dimension to the temporal dimension. This sliding mechanism provides an effective data organization form for subsequent feature extraction, ensuring that all feature information in the image can be captured.

[0065] Specifically, the image parsing module extracts image parameters during the sliding window's movement and obtains historical image parameter change curves, including...

[0066] After each movement of the sliding window, the proportion of black pixels in the current sliding window area is calculated, and the proportion is used as the image parameter of the corresponding sliding window.

[0067] Record the time and image parameters corresponding to the sliding window at different positions, and construct the historical image parameter change curve with time as the horizontal axis and image parameters as the vertical axis.

[0068] Specifically, the pattern analysis module analyzes the periodicity of curves based on the parameter change curves of historical images, determines the proportion of periodic curve segments, including,

[0069] Used to perform periodic detection on the parameter change curves of the historical images, and to identify periodic curve segments in the parameter change curves of the historical images that conform to periodic patterns;

[0070] The ratio of the cumulative time length of the periodic curve segment to the total time length of the historical image parameter change curve is used as the periodic curve segment ratio.

[0071] Understandably, the historical image parameter change curves are constructed based on historical operation logs, and the acquisition period for historical operation logs can be set to 5 days to ensure sufficient data volume.

[0072] The method for periodic detection of the historical image parameter change curves is not limited, as long as curve segments exhibiting regular repetition characteristics in waveform shape, amplitude changes, and time intervals can be identified. For example, the potential period length can be determined by calculating the autocorrelation function of the sequence and detecting its significant peak, and then the recurring curve segments can be divided according to the period. Of course, other methods for analyzing the periodicity of time-domain curve segments can also be used, which are existing technologies and will not be elaborated further.

[0073] Please see Figure 3 As shown, Figure 3 This is a logic diagram illustrating how the text overlap verification is passed in an embodiment of the invention. The pattern analysis module extracts the running log text segments corresponding to each periodic curve segment and verifies the text overlap within the running log text segments, including:

[0074] Based on the time period corresponding to the periodic curve segment on the historical image parameter change curve, the text content of the corresponding time period in the historical operation log is mapped and extracted as the operation log text fragment.

[0075] Calculate the similarity between runtime log text segments corresponding to different periodic curve segments;

[0076] When the average similarity between the various log text segments is higher than a preset threshold, the text overlap verification is considered successful.

[0077] Specifically, the preset threshold is a standard used to determine whether periodic log fragments have sufficient consistency at the textual semantic level. Ideally, the similarity between log text fragments should be 1. However, in reality, even if a server node runs a program periodically, its corresponding running logs do not exhibit perfect periodicity because some logs may display performance indices, and these performance indices fluctuate, resulting in the logs not exhibiting perfect periodicity. Based on this, the preset threshold is set as the product of the ideal similarity and the error coefficient. The error coefficient should be less than 1. This error coefficient aims to balance the strictness and fault tolerance of the judgment, and by appropriately reducing the ideal similarity, it adapts to the reasonable textual fluctuations that exist in actual application scenarios. In this embodiment of the invention, the error coefficient ranges from [0.85, 0.95], preferably 0.9.

[0078] This invention considers the extraction and curve construction mechanism of image parameters. During the sliding window traversal, it is necessary to convert the spatial information of the image into analyzable temporal data. This invention calculates the proportion of black pixels within the current window area in real time after each movement of the sliding window and defines this proportion as an image parameter. By recording the image parameter values ​​corresponding to different positions of the sliding window and arranging them in chronological order to form a sequence, a historical image parameter change curve is constructed. The historical image parameter change curve transforms the spatial distribution characteristics of a two-dimensional image into a one-dimensional temporal signal. Furthermore, since the image parameters are extracted based on black pixels, which are an inherent attribute of the image, the extraction is rapid. Based on this, a quantitative representation of the density changes in log content is achieved, providing an accurate data foundation for subsequent periodic analysis.

[0079] Please see Figure 4 As shown, Figure 4 This is a logic block diagram illustrating the process of classifying visual pattern categories corresponding to server nodes according to an embodiment of the invention. The pattern parsing module classifies visual pattern categories corresponding to server nodes, including:

[0080] If the proportion of periodic curve segments is greater than the proportion threshold and the text overlap verification passes, the server node is classified as a node with strong visual regularity; otherwise, the server node is classified as a node with weak visual regularity.

[0081] Specifically, the purpose of setting a percentage threshold is to take into account the sporadic nature of periodic curve segments. Therefore, a lower limit is set to indicate that the identified periodic curve segments are not sporadic. In practice, the lower limit is selected between the range [10%, 20%], preferably 10%.

[0082] Specifically, the twin detection module responds to new operation logs generated by the server node and monitors based on the visual pattern categories corresponding to the server node, including:

[0083] If the server node is a node with strong visual regularity, the new operation log is converted into a log image, a real-time image parameter change curve is constructed, and the non-periodic curve segment is identified and locked.

[0084] Extract the running log text fragments corresponding to the non-periodic curve segments, and perform text traversal parsing on the running log text fragments to determine whether there are any abnormalities in the server nodes.

[0085] Specifically, the text traversal and parsing of the operation log text fragments to determine whether the server node has an anomaly is not limited in the specific form. In implementation, an anomaly sample database can be built to store sample anomaly texts. By traversing the log text, it can be determined whether there are sample anomaly texts in the anomaly sample database in the daily fragments. If sample anomaly texts are found, it is determined that the server node has an anomaly. Of course, other methods can also be used, which will not be elaborated here.

[0086] Specifically, the twin detection module responds to new operation logs generated by the server node, monitors based on the visual pattern categories corresponding to the server node, and also includes,

[0087] If the server node is a weakly visually predictable node, then the new running log is directly traversed and parsed in its entirety to determine whether the server node is abnormal.

[0088] This invention considers a differentiated monitoring triggering mechanism based on pattern categories. In actual operation and maintenance environments, the operating status and log generation patterns of different server nodes differ significantly. Most devices in the data center perform a large number of repetitive or periodic tasks under normal operating conditions, resulting in stable patterns in the generation sequence and content structure of their operation logs. However, under abnormal operating conditions, these inherent patterns are easily disrupted by sudden errors, performance fluctuations, or unexpected interruptions. Traditional unified monitoring methods waste computing and communication resources. This invention identifies the visual pattern categories of nodes through a pattern parsing module, and the twin detection module adaptively activates corresponding monitoring strategies based on these categories. The core of this mechanism is that it does not require real-time reading and parsing of the entire log text content itself, but instead prioritizes monitoring the stability of macroscopic patterns through image parameter change curves, thereby quickly capturing abnormal device signs. For nodes with strong visual patterns, text parsing is only performed on the corresponding log text fragments when their operation logs deviate from the identified periodic patterns; for nodes with weak visual patterns, full log text analysis is performed. This mechanism achieves dynamic allocation of computing resources by differentiating node characteristics, avoids redundant processing of periodic normal logs, effectively reduces computing load and network transmission pressure, and achieves the goal of accurate and efficient anomaly detection.

[0089] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A digital twin system for sensing and predicting the status of data center equipment, characterized in that, Comprise: An image analysis module is used to obtain the historical running log of a plurality of server nodes in a machine room, convert the historical running log into a log image, and perform image modal analysis, including pixelizing the log image to obtain a binary pixel image, constructing a sliding window in the binary pixel image, extracting image parameters in the sliding process of the sliding window, and obtaining a historical image parameter change curve; A rule analysis module connected with the image analysis module is used to analyze the periodicity of the historical image parameter change curve, determine the proportion of the periodic curve segment, extract the running log text segment corresponding to each periodic curve segment, and verify the text coincidence degree in the running log text segment to divide the visual rule category corresponding to the server node; A twin detection module connected with the rule analysis module is used to monitor based on the visual rule category corresponding to the server node in response to the generation of new running log of the server node, including: Converting the running log into a log image, constructing an image parameter change curve, locking a non-periodic curve segment, performing text traversal analysis on the running log segment corresponding to the non-periodic curve segment, and determining whether the server node is abnormal; Or, performing text traversal analysis on the running log to determine whether the server node is abnormal.

2. The digital twin system for machine room equipment condition awareness and prediction as claimed in claim 1 wherein, The image analysis module pixelizes the log image to obtain a binary pixel image, including: To identify each text area and non-text area in the log image, cover each text area as a black pixel block, and cover the non-text area as a white pixel block to obtain a binary pixel image.

3. The digital twin system for machine room equipment condition awareness and prediction as claimed in claim 1 wherein, The image analysis module constructs a sliding window in the binary pixel image, wherein The sliding window is a rectangular window, the window height is pre-set, and the window width is consistent with the total width of the binary pixel image.

4. The digital twin system for machine room equipment condition awareness and prediction as claimed in claim 3 wherein, The sliding window constructed by the image analysis module slides from top to bottom along the vertical direction of the binary pixel image at a preset step.

5. The digital twin system for machine room equipment condition awareness and prediction as claimed in claim 1 wherein, The image analysis module extracts the image parameters in the sliding process of the sliding window, and obtains a historical image parameter change curve, including: After each movement of the sliding window, calculate the proportion of black pixels in the current sliding window area, and take the proportion as the image parameter of the corresponding sliding window; Record the corresponding time and image parameter of the sliding window at different positions, and construct the historical image parameter change curve with time as the horizontal axis and image parameter as the vertical axis.

6. The digital twin system for machine room equipment condition awareness and prediction as claimed in claim 1 wherein, The rule analysis module analyzes the periodicity of the historical image parameter change curve, determines the proportion of the periodic curve segment, including: To detect the periodicity of the historical image parameter change curve, identify the periodic curve segment in the historical image parameter change curve that meets the periodicity rule; Calculate the ratio of the cumulative time length of the periodic curve segment to the total time length of the historical image parameter change curve as the proportion of the periodic curve segment.

7. The digital twin system for machine room equipment condition awareness and prediction as claimed in claim 1 wherein, The rule analysis module extracts the running log text segment corresponding to each periodic curve segment, and verifies the text coincidence degree in the running log text segment, including: According to a time period corresponding to the periodic curve segment on the historical image parameter change curve, text content of a corresponding time period in a historical operation log is mapped and extracted as an operation log text segment; Similarities between operation log text segments corresponding to different periodic curve segments are calculated; When an average of the similarities between the operation log text segments is higher than a preset threshold, it is determined that the text coincidence degree verification is passed.

8. The digital twin system for machine room equipment condition awareness and prediction of claim 1, wherein, The rule analysis module divides a visual rule category corresponding to a server node, including: If the proportion of the periodic curve segment is greater than a proportion threshold and the text coincidence degree verification is passed, the server node is divided into a strong visual rule node, otherwise the server node is divided into a weak visual rule node.

9. The digital twin system for machine room equipment condition awareness and prediction of claim 1, wherein, The twin detection module, in response to a new operation log generated by a server node, monitors based on a visual rule category corresponding to the server node, including: If the server node is a strong visual rule node, the new operation log is converted into a log image, a real-time image parameter change curve is constructed, a non-periodic curve segment is identified and locked; An operation log text segment corresponding to the non-periodic curve segment is extracted, and text traversal analysis is performed on the operation log text segment to determine whether the server node is abnormal.

10. The digital twin system for data center equipment state awareness and prediction of claim 1, wherein, The twin detection module, in response to a new operation log generated by a server node, monitors based on a visual rule category corresponding to the server node, and further includes: If the server node is a weak visual rule node, full-text traversal analysis is directly performed on the new operation log to determine whether the server node is abnormal.

Images