A data classification and storage method for compressor valve plate fault diagnosis

Abstract

The application discloses a data classification and storage method for compressor valve plate fault diagnosis. The method acquires compressor operation state signals to form first fine-grained data, extracts features and aggregate statistics in sequence to generate second feature data and third long-term trend indicators; a multi-level storage index structure is constructed to associate and map the long-term trend indicators, feature data identifiers and physical addresses of the fine-grained data; when the long-term trend indicators meet abnormal conditions, reverse tracing is triggered, the identifiers of feature mutations are located according to the index structure, and the physical addresses are acquired to read underlying fine-grained data. The application takes into account both diagnostic accuracy and storage resources, solves the problem of difficulty in micro original data backtracking in abnormal conditions, and realizes efficient root cause positioning.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment condition monitoring and data processing technology, and more specifically, to a data classification and storage method for compressor valve plate fault diagnosis. Background Technology

[0002] As the core equipment for gas pressurization and transportation in industrial production, the valve plate component of the compressor directly determines the reliability and efficiency of the entire machine. Once a fault occurs, it often leads to shutdown, surge in energy consumption, or even safety accidents. Therefore, accurate and timely fault diagnosis of the valve plate status has extremely high engineering value and economic significance.

[0003] Most current compressor fault diagnosis systems face a significant challenge in managing the imbalance of data granularity when processing massive amounts of sensor data. These systems typically handle a range of data simultaneously, from raw microscopic signals acquired at high frequencies by sensors to pre-processed feature data and statistical indicators reflecting long-term trends. However, these different levels of data vary greatly in acquisition frequency, storage capacity, and information density, making it difficult to maintain efficient coordination throughout the data flow chain. An excessive pursuit of complete retention of fine-grained data leads to rapid depletion of storage resources, while over-compressing or discarding early data results in a lack of crucial original evidence for subsequent in-depth analysis. Both extremes severely limit the practicality of the diagnostic system.

[0004] At a deeper level, this granularity imbalance directly leads to the contradiction between the unidirectional flow of data and the difficulty of anomaly backtracking. During normal operation, data flows unidirectionally from the collection end to the analysis and storage ends, and the system can perform routine classification tasks well. However, when coarse-grained trend indicators show abnormal signs, and it is necessary to trace the detailed process of the specific time of the fault, it is difficult to quickly locate and extract the original fine-grained data within the corresponding time window. This phenomenon of smooth forward flow but obstructed reverse backtracking often leaves maintenance personnel in a dilemma of "knowing there is a high probability of a problem but not being able to find concrete evidence" when facing sudden fault upgrades or difficult cases, severely restricting the timeliness and accuracy of diagnosis.

[0005] Therefore, establishing a data classification and storage mechanism in the compressor valve plate fault diagnosis system that supports reasonable conversion of multi-granular data, smooth forward flow, and efficient reverse backtracking capability has become a key issue in achieving a closed-loop system from real-time monitoring to long-term trend analysis, while also taking into account diagnostic accuracy and resource controllability.

[0006] Currently, there is a lack of a good technology that can effectively solve the above problems.

[0007] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0008] To address the contradictions in existing compressor valve plate fault diagnosis systems when processing massive amounts of sensor data, such as imbalances in data granularity management, excessive storage resources, and lack of original micro-diagnostic evidence, as well as the difficulties in reverse backtracking and root cause localization when coarse-grained trend indicators are abnormal, this invention provides a data classification and storage method for compressor valve plate fault diagnosis.

[0009] The technical solution of the present invention is as follows:

[0010] A data classification and storage method for compressor valve plate fault diagnosis includes the following steps: acquiring compressor operating status signals to form first fine-grained data, and sequentially performing feature extraction and statistical aggregation on the first fine-grained data to generate second feature data and a third long-term trend indicator; constructing a multi-level storage index structure to associate and map the third long-term trend indicator, the identifier of the second feature data, and the physical address of the first fine-grained data in persistent storage; triggering a reverse tracing process when the third long-term trend indicator meets a preset anomaly judgment condition, and locating the second feature data associated with the abnormal third long-term trend indicator according to the multi-level storage index structure; and obtaining the physical address of the corresponding first fine-grained data in persistent storage from the multi-level storage index structure based on the target data identifier of the feature mutation in the located second feature data, so as to read the corresponding first fine-grained data.

[0011] Optionally, the step of acquiring the compressor's operating status signal to form first fine-grained data, and sequentially performing feature extraction and statistical aggregation on the first fine-grained data to generate second feature data and a third long-term trend indicator, includes: acquiring raw vibration and pressure signals collected at high frequency by sensors during compressor operation to form the first fine-grained data; segmenting the first fine-grained data using a preset sliding time window, extracting time-domain and frequency-domain features for the data within each time window to generate the second feature data; calculating the mean and variance of statistics for each time window based on the second feature data, and aggregating the mean and variance of statistics according to a preset aggregation period to generate the third long-term trend indicator.

[0012] Optionally, the step of obtaining the physical address of the corresponding first fine-grained data in persistent storage from the multi-level storage index structure based on the target data identifier of the feature mutation in the located second feature data, in order to read the corresponding first fine-grained data, includes: identifying a sub-window of feature mutation from the located second feature data; obtaining the physical address of the corresponding first fine-grained data in persistent storage from the multi-level storage index structure based on the identifier of the sub-window as the target data identifier; and reading the first fine-grained data of the corresponding time segment from the persistent storage according to the obtained physical address.

[0013] Optionally, the method further includes: dynamically adjusting the length of the sliding time window, the dimension of feature extraction, and the period of statistical aggregation based on the operating conditions of the compressor and the health status of the third long-term trend indicator; and automatically reducing the sliding time window, shortening the aggregation period, and increasing the proportion of high-frequency dimensions in feature extraction when the third long-term trend indicator approaches the anomaly judgment threshold.

[0014] Optionally, after reading the first fine-grained data for the corresponding time segment from the persistent storage based on the acquired physical address, the method further includes: after completing refined fault diagnosis and determining the root cause of the fault based on the read first fine-grained data, synchronously updating the anomaly label corresponding to the third long-term trend indicator and the fault feature weight of the second feature data; dynamically adjusting the subsequent granular flow rules, wherein the dynamic adjustment includes increasing the acquisition frequency of the corresponding valve plate part, reducing the sliding window, or extending the storage period of the fine-grained data.

[0015] Optionally, each level of the multi-level storage index structure embeds a fault classification label for the compressor valve plate, specifically including: a first-level index embedding a fault category label, which includes mechanical damage and performance degradation; a second-level index embedding a fault sub-category label, which includes fatigue cracks, wear, corrosion, and valve plate stiffness degradation; and a third-level index embedding a valve plate part code, which includes valve plate, valve seat, spring, and lift limiter. When the second feature data associated with the third long-term trend indicator that is found to be abnormal is located, the fault category of the first-level index is used for quick location, and then the second- and third-level labels are used for precise matching.

[0016] Optionally, the method further includes a dynamic maintenance mechanism for the multi-level storage index structure, specifically including: when the first fine-grained data and the second feature data are cleaned or merged after a preset storage period, the entries to be updated in the corresponding third-level index and the second-level index are locked to block the temporary access of the entry to be updated by the reverse tracing process; if the corresponding data is merged into the upper-level granular data, the physical address mapping relationship of the entry to be updated and the fault classification label are synchronously updated to the index entry corresponding to the merged upper-level granular data to complete the index merging and reconstruction; if the corresponding data is cleaned and there is no upper-level merged data, the corresponding index entry is marked as invalid and the association link between the index entry and the upper and lower level indexes is disconnected, and the cleanup time and reason notes of the invalid entry are added to the multi-level storage index structure to complete the link break repair; and after the index update is completed, the locking is released and the version number of the multi-level storage index structure is synchronously updated.

[0017] Optionally, the method further includes a conflict handling mechanism for concurrent backtracking, specifically including: when two or more of the third long-term trend indicators simultaneously meet the preset anomaly judgment conditions, thereby triggering a multi-path reverse tracing process, collecting the fault priority, data reading volume, and timeliness requirements corresponding to each reverse tracing process; allocating corresponding persistent storage I / O resource quotas to each reverse tracing process in order of fault priority from high to low and data reading volume from small to large, wherein the fault priority is preset based on the degree of abnormal deviation of the third long-term trend indicator and the valve plate fault risk level; and when the I / O load of the underlying persistent storage exceeds a preset load threshold, suspending the I / O resource allocation of low-priority and non-urgent reverse tracing processes, prioritizing data reading of high-priority processes, and resuming resource allocation and data retrieval of the suspended processes after the I / O load falls back to within the load threshold.

[0018] Optionally, the persistent storage includes three levels of storage media: memory, solid-state drive (SSD), and cold backup storage. The method further includes a triggering and execution mechanism for data migration across media, specifically including: establishing preset triggering conditions for silent migration, the triggering conditions including time-window-based triggering conditions and access frequency-based triggering conditions; wherein, the time-window-based triggering conditions include: the first fine-grained data being stored on the SSD for more than a first preset time, the second feature data being stored on the SSD for more than a second preset time, and the third long-term trend indicator being stored on the SSD for more than a third preset time; the access frequency-based triggering conditions... The triggering conditions include: the second characteristic data with no access records for a first consecutive preset period, and the third long-term trend indicator with no access records for a second consecutive preset period; when any of the triggering conditions is met, a silent migration process is initiated during a preset low-peak business time window to migrate the corresponding data from the current storage medium to the next-level cold backup storage medium, while simultaneously migrating the frequently accessed hot data from the solid-state drive to the memory for caching; and after the data migration is completed, the storage medium identifier and physical address mapping relationship of the corresponding data in the multi-level storage index structure are updated synchronously, and after verifying that the data integrity is correct, the source data in the original storage medium is deleted.

[0019] Optionally, the method further includes: summarizing and archiving the index entries marked as invalid after the chain break repair is completed, according to a preset archiving period, and storing them in a preset archiving medium; and when the reverse tracing process needs to retrieve historical data for the corresponding time period of the invalid index entries, first retrieving the archived invalid index entries from the archiving medium, and locating the upper-level granularity merged data for the corresponding time period through the cleanup time, the reason notes, and the original associated link information recorded in the invalid index entries, so as to complete the fallback retrieval of historical data.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] This invention overcomes the limitations of unidirectional data flow in traditional diagnostic systems by establishing a multi-granularity data dimensionality reduction and transfer mechanism and a multi-level storage index structure. Under normal conditions, the system progressively reduces the dimensionality of massive high-frequency raw signals, significantly reducing the storage pressure of invalid, massive amounts of data and greatly saving storage resources. When anomalies are triggered, the strong binding relationship between long-term indicators, mid-level feature identifiers, and low-level physical addresses in the index structure enables coarse-to-medium-to-fine reverse tracing, allowing the system to directly locate and extract the microscopic raw fragments that trigger mutations, thus resolving the contradiction between limited storage space and the need for precise microscopic diagnosis.

[0022] Meanwhile, by introducing a dynamic adjustment mechanism and a fully closed-loop logic of reverse tracing, this invention deeply couples the data flow parameters of pure software with the physical operating conditions and health status of the compressor, automatically switching the high-frequency acquisition strategy when there is a critical anomaly; and after completing the reading of the underlying data and root cause location, it can dynamically correct the fault weight of the upper-level index and the subsequent flow rules, so that the entire classification and storage architecture has the technical effect of self-evolution and optimization.

[0023] Furthermore, this invention forcibly binds the data storage index architecture to the compressor valve plate and its specific physical fault modes in three levels. It also incorporates a multi-level index chain break repair and merging reconstruction mechanism, a fault-priority-based underlying persistent storage concurrent I / O control mechanism, and silent migration scheduling of hot and cold data across different media layers. This transforms the abstract database structure into a dedicated high-availability storage architecture with strong valve plate physical attributes, significantly improving system stability and response efficiency during high-concurrency read / write operations and sudden anomaly backtracking in complex industrial scenarios. Attached Figure Description

[0024] Figure 1 A flowchart illustrating a data classification and storage method for compressor valve plate fault diagnosis provided in an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of the multi-level storage index structure and data flow mechanism provided in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described embodiments are merely some embodiments of the invention, and not all embodiments. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0027] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0028] It should be noted that 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 one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0029] As mentioned earlier, most current compressor fault diagnosis systems face a prominent problem of unbalanced data granularity management when processing massive amounts of sensor data. Excessively pursuing the complete retention of fine-grained data will cause storage resources to be exhausted quickly, while over-compressing or discarding early data makes it difficult to trace back anomalies. Maintenance personnel often find themselves in the dilemma of "knowing that there is a high probability of a problem but not being able to find concrete evidence," which severely restricts the timeliness and accuracy of diagnosis.

[0030] To address this issue, this invention provides a data classification and storage method for compressor valve plate fault diagnosis. By establishing a multi-granularity data dimensionality reduction flow mechanism and a multi-level storage index structure, it saves storage resources under normal conditions and achieves accurate reverse tracing from coarse to medium to fine during anomalies, thus balancing diagnostic accuracy and resource controllability. This invention solves the problem in the following way.

[0031] Example 1:

[0032] Please refer to the instruction manual appendix. Figure 1 , Figure 1 This is a flowchart illustrating a data classification and storage method for compressor valve plate fault diagnosis provided in Embodiment 1 of the present invention. Figure 1 As shown, in one embodiment of the present invention, the executing entity can be an edge computing node (such as a PLC or industrial computer) deployed locally on the compressor, employing a common hardware architecture environment in the art. The method mainly includes the following steps:

[0033] S101. Acquire the compressor's operating status signal to form the first fine-grained data. For example, the first fine-grained data is formed by acquiring the raw vibration and pressure signals collected at high frequency by sensors through an edge computing node deployed locally on the compressor.

[0034] S102. Features are extracted sequentially from the first fine-grained data to generate second feature data. A preset sliding time window is used to segment the first fine-grained data. For example, the length of the sliding time window is set to 1000 sampling points, and the step size is 500 sampling points. For the data within each time window, a common algorithm in the field is used to extract time-domain and frequency-domain features to generate the second feature data. When extracting time-domain features, the root mean square (RMS) algorithm is used, and its calculation formula is as follows:

[0035]

[0036] in, This represents the root mean square value of the signal within the time window. This indicates the total number of data points contained within the time window; Indicates the first time within the time window The original signal amplitude of each data point.

[0037] For example, suppose that within a sub-time window, data is collected... The vibration acceleration data points have the following amplitudes: Substitute the example data above into the formula to calculate:

[0038]

[0039] When extracting frequency domain features, a Fast Fourier Transform (FFT) is performed on the data within the window to extract features such as the main frequency energy, and these features are combined with time domain features to generate the second feature data.

[0040] S103. Perform statistical aggregation on the second feature data to generate a third long-term trend indicator.

[0041] Based on the second feature data, the mean and variance of the statistics for each time window are calculated, and the mean and variance of the statistics are aggregated according to a preset aggregation period. Mean ( ) and variance The calculation formulas for ) are as follows:

[0042]

[0043]

[0044] in, This represents the characteristic mean within the aggregation period; $ represents the characteristic variance within the aggregation period; This indicates the total number of second feature data items included within the preset aggregation period; Indicates the first The value of the second feature data corresponding to each time window.

[0045] For example, suppose that within an aggregation cycle, there are... The root mean square value characteristics of each time window are as follows: , , Substitute the example data above into the formula to calculate:

[0046] mean

[0047] variance

[0048] Finally, the aggregated mean and variance are used as the third long-term trend indicator.

[0049] S104. Construct a multi-level storage index structure to associate and map the third long-term trend indicator, the identifier of the second feature data, and the physical address of the first fine-grained data in persistent storage.

[0050] S105. When the third long-term trend indicator meets the preset anomaly judgment conditions, a reverse tracing process is triggered, and the second feature data associated with the abnormal third long-term trend indicator is located according to the multi-level storage index structure.

[0051] S106. Based on the target data identifier of the feature mutation in the located second feature data, obtain the physical address of the corresponding first fine-grained data in the persistent storage from the multi-level storage index structure, so as to read the corresponding first fine-grained data.

[0052] Specifically, step S106 includes: identifying a sub-window with a feature mutation from the located second feature data; using the identifier of the sub-window as the target data identifier, obtaining the physical address of the corresponding first fine-grained data in the persistent storage from the multi-level storage index structure; and reading the first fine-grained data of the corresponding time segment from the persistent storage according to the obtained physical address.

[0053] Example 2:

[0054] Based on the above embodiments, in order to provide a clearer and more complete explanation of the technical solution provided by the present invention, the present invention also provides a second embodiment, which provides a dynamic control and closed-loop optimization mechanism to adapt to complex industrial field conditions.

[0055] Specifically, the method further includes: dynamically adjusting the length of the sliding time window, the dimension of feature extraction, and the period of statistical aggregation based on the compressor's operating conditions and the health status of the third long-term trend indicator. When the third long-term trend indicator approaches the anomaly detection threshold, the sliding time window is automatically reduced, the aggregation period is shortened, and the proportion of high-frequency dimensions in feature extraction is increased.

[0056] In this second embodiment, the third long-term trend indicator (such as variance) is set. The anomaly detection threshold is: Define the deviation ratio. .when When the time window is approaching the anomaly threshold, dynamic adjustment is triggered. The formula for the dynamic adjustment of the sliding time window length is as follows:

[0057]

[0058] in, This indicates the adjusted sliding time window length; This indicates the initially set length of the sliding time window; This is a preset adjustment factor, typically set to a value of [value missing]. ; This represents the deviation ratio.

[0059] For example, suppose the initial sliding time window length is... Each sampling point, anomaly detection threshold The variance calculated so far. Then the deviation ratio Substitute the example data above into the formula to calculate:

[0060]

[0061] Based on this, the system automatically reduces the sliding window length from 1000 to 950, thereby improving the temporal resolution of feature extraction.

[0062] Furthermore, after reading the first fine-grained data corresponding to the time segment from the persistent storage based on the acquired physical address, the method further includes: after completing refined fault diagnosis and determining the root cause of the fault based on the read first fine-grained data, synchronously updating the anomaly label corresponding to the third long-term trend indicator and the fault feature weight of the second feature data; and dynamically adjusting the subsequent granular flow rules, wherein the dynamic adjustment includes increasing the acquisition frequency of the corresponding valve plate part, reducing the sliding window, or extending the storage period of the fine-grained data.

[0063] When updating the fault feature weights of the second feature data, a moving average iterative formula based on diagnostic contribution is used:

[0064]

[0065] in, Indicates the updated number The weights of each feature; Indicates the historical weights before the update; Forgetting factor, usually taken as ; This is the actual contribution score of this feature in this refined diagnosis (the value range is...). to ).

[0066] For example, suppose that in the time-frequency domain feature library, the historical weight of the feature of high-frequency band energy... In a reverse tracing diagnosis, early fatigue cracks were confirmed in the valve plate, and the contribution score of this feature to the locking crack was [not specified]. Substitute the example data into the formula to calculate:

[0067]

[0068] After the update, the weight of this feature was increased to 0.58, making the system more sensitive to this type of fault in future monitoring.

[0069] To achieve more efficient reverse tracing and location, each level of the multi-level storage index structure embeds a fault classification label for the compressor valve plate.

[0070] Specifically, in the underlying data structure, a hierarchical string encoding concatenation is used as the index key. The first-level index embeds fault category labels, for example, using two-digit character encoding, where "A1" represents mechanical damage and "A2" represents performance degradation; the second-level index embeds fault sub-category labels, for example, "B1" represents fatigue cracks, "B2" represents wear, "B3" represents corrosion, and "B4" represents valve plate stiffness degradation; the third-level index embeds valve plate part codes, for example, "C1" represents valve plate, "C2" represents valve seat, "C3" represents spring, and "C4" represents lift limiter.

[0071] In actual storage mapping, the aforementioned tags are concatenated with timestamps or unique identifiers (UUIDs). For example, when the system detects signs of suspected valve plate fatigue cracks, the associated index key value format is encoded as "A1-B1-C1-UUID". When locating the second feature data associated with the abnormal third long-term trend indicator, the system first performs hash prefix matching through the fault category of the first-level index (such as "A1") to achieve rapid filtering of a large range of data, and then performs precise matching through second-level ("B1") and third-level ("C1") tags, thereby significantly reducing the system I / O overhead when retrieving massive amounts of data.

[0072] Example 3:

[0073] Based on the above embodiments, in order to further clarify and completely explain the technical solution provided by the present invention, the present invention also provides Embodiment Three. Building upon the foregoing embodiments, and addressing the issues of concurrent access and full lifecycle management of massive data in underlying persistent storage, Embodiment Three further provides a dynamic maintenance mechanism for a multi-level storage index structure, a conflict handling mechanism for concurrent backtracking, and a data cross-media migration mechanism, specifically combined with… Figure 2 Please provide an explanation.

[0074] The method also includes a dynamic maintenance mechanism for the multi-level storage index structure. Specifically, when the first fine-grained data and the second feature data are cleaned up or merged after a preset storage period, the system first locks the entries to be updated in the corresponding third-level index and the second-level index (for example, using an exclusive lock mechanism in a relational database) to block the temporary access of the reverse tracing process to the entry to be updated, thus avoiding read-write conflicts.

[0075] If the corresponding data is merged into the higher-level granular data, the physical address mapping relationship of the entry to be updated and the fault classification label are synchronously updated to the index entry corresponding to the merged higher-level granular data to complete the index merging and reconstruction.

[0076] If the corresponding data is cleaned up and there is no upper-level merged data, the corresponding index entry is marked as invalid (e.g., the status field is set to 0), and the association link between the index entry and the upper and lower level indexes is disconnected. Subsequently, the cleanup time (e.g., the current system timestamp) and reason remarks (e.g., marked as "Time_Expired") of the invalid entry are added to the multi-level storage index structure to complete the link break repair;

[0077] After the index update is complete, the lock is released, and the version number of the multi-level storage index structure is updated synchronously (e.g., the version number is incremented). This ensures eventual consistency of index readings in a distributed architecture.

[0078] Meanwhile, in response to potential underlying disk read congestion under extreme operating conditions, the method also includes a concurrent backtracking conflict handling mechanism.

[0079] Specifically, this includes: when two or more of the third long-term trend indicators simultaneously meet the preset anomaly judgment conditions, thereby triggering a multi-path reverse tracing process, collecting the fault priority, data reading volume, and timeliness requirements corresponding to each reverse tracing process.

[0080] For each reverse tracing process, the fault priority is calculated using a linear weighted evaluation model, and the calculation formula is as follows:

[0081]

[0082] in, This represents the calculated overall fault priority score; Indicates the degree of outlier deviation after normalization (range of values) ); This indicates the quantified valve plate failure risk level (e.g., Level 1 severe risk). Level 2, Medium Risk ); These are preset weight coefficients, and satisfy... .

[0083] For example, setting weighting coefficients , Suppose the current system simultaneously triggers both tracing link A and tracing link B. What is the degree of abnormal deviation in tracing link A? This corresponds to a serious risk of fatigue cracks in the valve plate. ; Trace the degree of abnormal deviation in link B Medium risk corresponding to slight wear of the valve seat Substitute the example data above into the formula to calculate:

[0084] Priority for tracing link A:

[0085]

[0086] Priority for tracing link B:

[0087]

[0088] The system prioritizes faults based on their calculated overall fault priority scores, from highest to lowest (i.e., handles them first). Link A, with a data read volume as high as 0.94, allocates corresponding persistent storage I / O resource quotas to each reverse tracing process in ascending order of data read volume.

[0089] When the I / O load of the underlying persistent storage exceeds a preset load threshold (e.g., I / O utilization reaches 80%), the system automatically suspends I / O resource allocation for low-priority and non-urgent reverse tracing processes (such as link B), prioritizing bandwidth for data reading of high-priority processes (such as link A). Once the I / O load falls back below the load threshold, resource allocation and data retrieval for the suspended link B are resumed.

[0090] Additionally, please see the appendix. Figure 2 The persistent storage, in terms of physical architecture, includes three levels of storage media: memory, solid-state drives (SSDs), and cold backup storage (such as high-capacity hard disk drives or tape libraries). The method also includes a triggering and execution mechanism for data migration across media.

[0091] The system establishes preset trigger conditions for silent migration, including time-window-based trigger conditions and access frequency-based trigger conditions. As an engineering example: the time-window-based trigger conditions include: the first fine-grained data being stored on the solid-state drive for more than a first preset time (e.g., 72 hours); the second feature data being stored on the solid-state drive for more than a second preset time (e.g., 30 days); and the third long-term trend indicator being stored on the solid-state drive for more than a third preset time (e.g., 180 days). The access frequency-based trigger conditions include: the second feature data having no access records for a consecutive first preset period (e.g., 30 days), and the third long-term trend indicator having no access records for a consecutive second preset period (e.g., 90 days).

[0092] When any of the aforementioned triggering conditions are met, the system initiates a silent migration process during a preset off-peak business time window (such as 02:00 to 04:00 AM), migrating the corresponding data from the current storage medium to the next-level cold backup storage medium. Simultaneously, frequently accessed hot data is migrated from the solid-state drive to the memory for caching to improve the hit rate of subsequent tracing. After the data migration is completed, the storage medium identifier and physical address mapping relationship of the corresponding data in the multi-level storage index structure are updated synchronously. After verifying the data integrity through a hash algorithm (such as MD5), the source data in the original storage medium is securely deleted.

[0093] Finally, to ensure an absolute closed loop for data backtracking, the method further includes: summarizing and archiving index entries marked as invalid after the link break repair is completed, according to a preset archiving cycle (such as weekly), and storing them in a preset archiving medium; when the reverse tracing process needs to retrieve historical data for the corresponding time period of the invalid index entry, the system first retrieves the archived invalid index entry from the archiving medium, and locates the upper-level granular merged data for the corresponding time period through the cleanup time, the reason notes, and the original associated link information recorded in the invalid index entry, so as to complete the fallback retrieval of historical data.

[0094] It should be understood that, in the embodiments of the present invention, "B corresponding to A" means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean that B is determined solely based on A; B can also be determined based on A and / or other information.

[0095] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0096] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0097] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, apparatuses, or units, or they may be electrical, mechanical, or other forms of connection.

[0098] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0099] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0100] From the above description of the embodiments, those skilled in the art will clearly understand that the present invention can be implemented in hardware, firmware, or a combination thereof. When implemented in software, the above-described functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media can be any available medium accessible to a computer. For example, but not limited to, computer-readable media can include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer. Furthermore, any connection can suitably be a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of the medium. As used in this invention, disk and disc include compressed optical discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks, and Blu-ray discs, wherein disks typically magnetically copy data, while discs optically copy data using lasers. The combinations described above should also be included within the scope of protection for computer-readable media.

[0101] In summary, the above description is merely a preferred embodiment of the technical solution of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A data classification and storage method for compressor valve plate fault diagnosis, characterized in that, Includes the following steps: The compressor's operating status signal is acquired to form the first fine-grained data, and the first fine-grained data is sequentially subjected to feature extraction and statistical aggregation to generate the second feature data and the third long-term trend indicator. A multi-level storage index structure is constructed to associate and map the third long-term trend indicator, the identifier of the second feature data, and the physical address of the first fine-grained data in persistent storage. When the third long-term trend indicator meets the preset anomaly judgment conditions, the reverse tracing process is triggered, and the second feature data associated with the abnormal third long-term trend indicator is located according to the multi-level storage index structure. as well as Based on the target data identifier of the feature mutation in the located second feature data, the physical address of the corresponding first fine-grained data in the persistent storage is obtained from the multi-level storage index structure to read the corresponding first fine-grained data. In this multi-level storage index structure, each level of the index embeds a fault classification label for the compressor valve plate, including: a first-level index embedding a fault category label, which includes mechanical damage and performance degradation; a second-level index embedding a fault sub-category label, which includes fatigue cracks, wear, corrosion, and valve plate stiffness degradation; and a third-level index embedding a valve plate part code, which includes valve plate, valve seat, spring, and lift limiter. When the second feature data associated with the third long-term trend indicator that is found to be abnormal is located, the fault category of the first-level index is used for quick location, and then the second- and third-level labels are used for precise matching. The method also includes a dynamic maintenance mechanism for the multi-level storage index structure, specifically including: when the first fine-grained data and the second feature data are cleaned or merged after a preset storage period, the entries to be updated in the corresponding third-level index and the second-level index are locked to block the temporary access of the entry to be updated by the reverse tracing process; if the corresponding data is merged into the upper-level granular data, the physical address mapping relationship of the entry to be updated and the fault classification label are synchronously updated to the index entry corresponding to the merged upper-level granular data to complete the index merging and reconstruction; if the corresponding data is cleaned and there is no upper-level merged data, the corresponding index entry is marked as invalid and the association link between the index entry and the upper and lower level indexes is disconnected, and the cleanup time and reason notes of the invalid entry are added to the multi-level storage index structure to complete the link break repair; and after the index update is completed, the locking is released and the version number of the multi-level storage index structure is synchronously updated.

2. The data classification and storage method according to claim 1, characterized in that, The process of acquiring the compressor's operating status signal to form first fine-grained data, and then sequentially performing feature extraction and statistical aggregation on the first fine-grained data to generate second feature data and a third long-term trend indicator, includes: The original vibration and pressure signals collected at high frequency by the sensors during the operation of the compressor are acquired to form the first fine-grained data; The first fine-grained data is segmented using a preset sliding time window, and time-domain and frequency-domain features are extracted for the data within each time window to generate the second feature data; Based on the second feature data, the mean and variance of the statistics for each time window are calculated, and the mean and variance of the statistics are aggregated according to a preset aggregation period to generate the third long-term trend indicator.

3. The data classification and storage method according to claim 1, characterized in that, The step of retrieving the physical address of the corresponding first fine-grained data in persistent storage from the multi-level storage index structure based on the target data identifier of the feature mutation in the located second feature data, in order to read the corresponding first fine-grained data, includes: From the located second feature data, identify the sub-windows with feature mutations; Based on the identifier of the sub-window as the target data identifier, the physical address of the corresponding first fine-grained data in the persistent storage is obtained from the multi-level storage index structure; Based on the obtained physical address, the first fine-grained data corresponding to the time segment is read from the persistent storage.

4. The data classification and storage method according to claim 2, characterized in that, The method further includes: Based on the compressor's operating conditions and the health status of the third long-term trend indicator, the length of the sliding time window, the dimension of feature extraction, and the period of statistical aggregation are dynamically adjusted. When the third long-term trend indicator approaches the anomaly detection threshold, the sliding time window is automatically reduced, the aggregation cycle is shortened, and the proportion of high-frequency dimensions in feature extraction is increased.

5. The data classification and storage method according to claim 1, characterized in that, After reading the first fine-grained data corresponding to the time segment from the persistent storage based on the acquired physical address, the method further includes: After completing the refined fault diagnosis and determining the root cause of the fault based on the first fine-grained data read, the abnormal label corresponding to the third long-term trend indicator and the fault feature weight of the second feature data are updated synchronously. The subsequent granularity transfer rules are dynamically adjusted, including increasing the acquisition frequency of the corresponding valve plate part, reducing the sliding window, or extending the storage period of fine-grained data.

6. The data classification and storage method according to claim 1, characterized in that, The method also includes a conflict handling mechanism for concurrent backtracking, specifically including: When two or more of the third long-term trend indicators simultaneously meet the preset anomaly judgment conditions, thereby triggering a multi-path reverse tracing process, the fault priority, data reading volume, and timeliness requirements corresponding to each reverse tracing process are collected. Persistent storage I / O resource quotas are allocated to each reverse tracing process according to the order of fault priority from high to low and the order of data read volume from small to large. The fault priority is preset based on the degree of abnormal deviation of the third long-term trend indicator and the valve plate fault risk level. When the I / O load of the underlying persistent storage exceeds the preset load threshold, the I / O resource allocation of low-priority and non-urgent reverse trace processes is suspended, and data reading of high-priority processes is prioritized. Once the I / O load falls back to within the load threshold, the resource allocation and data retrieval of the suspended processes are resumed.

7. The data classification and storage method according to claim 1, characterized in that, The persistent storage includes three levels of storage media: memory, solid-state drive, and cold backup storage. The method also includes a triggering and execution mechanism for cross-media data migration, specifically including: Establish preset trigger conditions for silent migration, including time-window-based trigger conditions and access frequency-based trigger conditions; wherein, the time-window-based trigger conditions include: the first fine-grained data being stored in the solid-state drive for more than a first preset time, the second feature data being stored in the solid-state drive for more than a second preset time, and the third long-term trend indicator being stored in the solid-state drive for more than a third preset time; the access frequency-based trigger conditions include: the second feature data having no access records for a consecutive first preset period, and the third long-term trend indicator having no access records for a consecutive second preset period; When any of the aforementioned triggering conditions is met, a silent migration process is initiated during a preset off-peak business time window, migrating the corresponding data from the current storage medium to a lower-level cold backup storage medium, while simultaneously migrating frequently accessed hot data from the solid-state drive to the memory for caching; and After the data migration is completed, the storage medium identifier and physical address mapping relationship of the corresponding data in the multi-level storage index structure are updated synchronously, and the source data in the original storage medium is deleted after verifying that the data integrity is correct.

8. The data classification and storage method according to claim 1, characterized in that, The method further includes: Index entries marked as invalid after link break repair are summarized and archived according to a preset archiving cycle and stored in a preset archiving medium; and When the reverse tracing process needs to retrieve historical data for the corresponding time period of an expired index entry, it first retrieves the archived expired index entry from the archive medium, and then locates the upper-level granular merged data for the corresponding time period through the cleanup time, the reason notes, and the original associated link information recorded in the expired index entry, so as to complete the fallback retrieval of historical data.

Images