An abnormality monitoring method and system for a shared device based on big data

Abstract

The application belongs to the technical field of shared device monitoring, and provides an abnormal monitoring method and system for shared devices based on big data. The method comprises the following steps: constructing a performance degradation knowledge graph of the same type of shared devices based on the operation history big data of the same type of shared devices of a target shared device; obtaining a bad use data set from the use big data of the target shared device by users, performing data enhancement processing on the bad use data set based on an attention mechanism, and analyzing the performance degradation knowledge graph to obtain a potential loss degree set; optimizing and adjusting corresponding monitoring indicators in a monitoring indicator set based on the historical operation data of the target shared device and the potential loss degree set to obtain a target monitoring indicator set; and comparing real-time operation data of the target shared device with the target monitoring indicator set to realize abnormal monitoring. The application can early warn potential faults, reduce maintenance costs, and improve the operation stability and user experience of shared devices.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of shared device monitoring, in particular to an abnormal monitoring method and system for shared devices based on big data. BACKGROUND

[0002] With the deep integration of Internet of Things technology and sharing economy model, shared devices have been widely penetrated into multiple scenarios such as residential life, commercial office, and campus dormitory. Among them, shared shower products, shared water dispensers, shared washing machines, shared hair dryers and other devices have become important carriers for improving public service efficiency due to their convenience and economy. Such devices usually have characteristics such as high frequency of use, decentralized deployment, and strong user mobility, and their stable operation is directly related to user experience and operational safety.

[0003] The prior art lacks deep consideration of user usage big data in shared device abnormal monitoring. The abnormal condition of a shared device is often closely related to the user's usage behavior. For example, the user putting too much laundry in a shared washing machine can cause motor overload, frequently turning on and off a shared hair dryer can accelerate internal component wear, and the user's unreasonable adjustment of water temperature and water pressure when using a shared shower product can affect the device pipeline life. However, current monitoring technology mostly focuses on the running parameters of the device itself (such as determining whether the shower device is overheating through a water temperature sensor, detecting whether the water dispenser is leaking through a flow sensor, etc.), ignoring the analysis of key data such as user usage duration, usage frequency, operation habits, and usage time period.

[0004] Therefore, how to use big data technology to deeply mine the full amount of running data of shared devices and achieve early and accurate warning of device failures is a key problem to be solved in the current shared device operation and management field. SUMMARY

[0005] To this end, the present application provides an abnormal monitoring method and system for shared devices based on big data, an electronic device, a computer storage medium, and a computer program product to solve at least one of the above technical problems.

[0006] In a first aspect, the present application provides a big data-based shared device anomaly monitoring method, comprising the following method steps: obtaining operation history big data of a same type of shared device of a target shared device, and constructing a performance degradation knowledge graph of the same type of shared device based on the operation history big data; obtaining usage big data of the target shared device by a user, obtaining a bad usage data set from the usage big data, performing data enhancement processing on the bad usage data set based on an attention mechanism, and analyzing a potential loss degree set of the target shared device based on the performance degradation knowledge graph and the bad usage data set after the enhancement processing; optimizing and adjusting corresponding monitoring indicators in a monitoring indicator set based on historical operation data of the target shared device and the potential loss degree set to obtain a target monitoring indicator set; and comparing real-time operation data of the target shared device with the target monitoring indicator set to realize anomaly monitoring.

[0007] In a second aspect, the present application provides a big data-based shared device anomaly monitoring system, comprising a knowledge graph construction unit, a potential loss analysis unit, an optimization adjustment unit, and an anomaly monitoring unit; the knowledge graph construction unit obtains operation history big data of a same type of shared device of a target shared device, and constructs a performance degradation knowledge graph of the same type of shared device based on the operation history big data; the potential loss analysis unit obtains usage big data of the target shared device by a user, obtains a bad usage data set from the usage big data, performs data enhancement processing on the bad usage data set based on an attention mechanism, and analyzes a potential loss degree set of the target shared device based on the performance degradation knowledge graph and the bad usage data set after the enhancement processing; the optimization adjustment unit optimizes and adjusts corresponding monitoring indicators in a monitoring indicator set based on historical operation data of the target shared device and the potential loss degree set to obtain a target monitoring indicator set; and the anomaly monitoring unit compares real-time operation data of the target shared device with the target monitoring indicator set to realize anomaly monitoring.

[0008] In a third aspect, the present application provides an electronic device, comprising at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the computer program is executed by the processor to implement the method according to any one of the preceding aspects.

[0009] In a fourth aspect, the present application provides a computer storage medium storing a computer program executable by a processor to implement the method according to any one of the preceding aspects.

[0010] In a fifth aspect, the present application provides a computer program product comprising a computer program executable by a processor to implement the method according to any one of the preceding aspects.

[0011] The application realizes accurate monitoring of shared equipment abnormalities by constructing a performance degradation knowledge graph, mining hidden bad use behaviors and optimizing monitoring indexes. The traditional method can solve the problems of missed and false reports, early warning of potential faults, reduction of maintenance costs and improvement of the operation stability and user experience of shared equipment. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0013] Figure 1 is a flowchart of a shared equipment abnormality monitoring method based on big data disclosed by the embodiments of the application.

[0014] Figure 2 is a system architecture diagram disclosed by the embodiments of the application.

[0015] Figure 3 is a structure diagram of a shared equipment abnormality monitoring system based on big data disclosed by the embodiments of the application. DETAILED DESCRIPTION

[0016] The following describes the embodiments of the application by specific and concrete embodiments. Those skilled in the art can easily understand other advantages and effects of the application from the content disclosed in the specification. Obviously, the described embodiments are part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0017] In addition, the technical features involved in different embodiments of the application described below can be combined with each other as long as there is no conflict.

[0018] As shown in Figure 1 The embodiments of the application disclose a shared equipment abnormality monitoring method based on big data, which comprises the following method steps: S10, obtaining the operation history big data of the same type of shared equipment of a target shared equipment, and constructing a performance degradation knowledge graph of the same type of shared equipment based on the operation history big data.

[0019] As shown in Figure 2As shown, the present application constructs a management platform for shared devices, which accesses multiple types of shared devices, including but not limited to shared shower products, shared water dispensers, shared washing machines, shared hair dryers, etc. The management platform can receive the operation data of these shared devices and store them in the cloud, and then classify and organize the operation history big data of the same type of shared devices and the use big data of users on the target shared device. The management platform can predict whether the target shared device has operation abnormalities based on these data.

[0020] First, collect the operation history big data of a large number of same type of devices of the target shared device, covering the whole life cycle information of the shared device from the time it is put into use to the time it is scrapped, including the running parameters of the shared device at different time periods (such as the washing speed and water temperature of the shared washing machine, the water flow and water temperature of the shared water dispenser), the occurrence time of previous failures, the failure type (such as motor failure, pipeline blockage), maintenance records (such as replaced parts, performance recovery after maintenance), and environmental data (such as temperature, humidity, voltage stability of the installation location, etc.) of the device.

[0021] Then, use big data processing technology to deeply analyze these operation history big data. Through data cleaning to remove invalid and erroneous data, and then feature extraction, the key features related to device performance degradation are extracted, such as the change rule of different part performance parameters with use time. Then, through correlation analysis, the relationship between device performance degradation and various factors (such as running parameters, use frequency, environmental conditions) is found out, for example, the correlation degree between high use frequency of shared hair dryer and motor performance degradation.

[0022] Finally, the rules and relationships obtained through analysis are constructed in the form of knowledge graph, and the nodes in the knowledge graph represent device parts, performance parameters, failure types, etc., and the edges represent the causal relationship or correlation strength between them.

[0023] S20, obtain the use big data of the target shared device by the user, extract the bad use data set from the use big data, perform data augmentation processing on the bad use data set based on the attention mechanism, and analyze the potential damage degree set of the target shared device based on the performance degradation knowledge graph and the bad use data set after the augmentation processing.

[0024] Obtain user usage big data of the target shared device to be monitored, including the time (such as the specific time period) of the user using the target shared device, the single use duration, the usage frequency (such as the number of daily uses), the operation mode (such as the water temperature adjustment range of the shared shower product, the mode selection of the shared washing machine), and the like. From these usage big data, according to the preset bad usage judgment standard (such as the shared washing machine load exceeding the rated value, the shared hair dryer continuous use duration exceeding the safety threshold, and the like), bad usage data is screened out to form a bad usage data set.

[0025] At the same time, since part of the bad usage cases have concealment, it is difficult to identify only by the preset standard, and therefore data enhancement processing based on the attention mechanism is needed. The attention mechanism can focus on those usage features that have potential impact on device wear and tear but are not obviously marked as "bad", such as the shared washing machine single load slightly lower than the rated value but frequently running in high speed mode, the shared shower product adjusting the water temperature multiple times in a short time, and the like. By mining these not obvious "bad" cases, they are supplemented to the bad usage data set, thereby increasing the data volume and making the data set more comprehensively cover various usage behaviors that may cause device wear and tear.

[0026] Subsequently, the enhanced bad usage data set is matched and analyzed with the rules and relationships of device performance degradation in the performance degradation knowledge graph constructed in step S10. Through the associated information of bad usage behaviors and device component wear and tear, fault risk in the knowledge graph, the degree of wear and tear that the bad usage behaviors may cause to each component of the target shared device is quantitatively evaluated, and finally a potential wear and tear degree set containing information such as the potential wear and tear level of each component, the fault type that may be triggered, and the occurrence probability is formed.

[0027] S30, based on the historical running data of the target shared device and the potential wear and tear degree set, the corresponding monitoring indicators in the monitoring indicator set are optimized and adjusted to obtain a target monitoring indicator set.

[0028] Collect the historical running data of the target shared device itself, including the past running parameter change range of the target shared device, the faults that have occurred and the handling situation, and the like. The initial monitoring indicator set pre-configured for the target shared device is optimized and adjusted in combination with the potential wear and tear degree set obtained in the above step S20. The initial monitoring indicator set contains device regular monitoring parameters and threshold values, and according to the high-risk components and fault types in the potential wear and tear degree set, the corresponding monitoring indicators are adjusted, for example, the monitoring frequency of the high-risk component related parameters is increased, and the warning threshold is tightened, so as to obtain the target monitoring indicator set suitable for the target device.

[0029] S40, by comparing the real-time running data of the target shared device with the target monitoring indicator set to realize abnormal monitoring.

[0030] In the actual monitoring process, the running data (such as current running parameters, working state, etc.) of the target shared device is collected in real time, and the real-time running data is compared with the target monitoring index set in real time. When the real-time running data exceeds the range set by the target monitoring index set, it is determined that the device has an abnormality, and a warning is sent in time, so as to realize accurate and efficient abnormal monitoring of the target shared device.

[0031] The present application realizes accurate monitoring of shared device abnormalities by constructing a performance degradation knowledge graph, mining hidden bad use behaviors, and optimizing monitoring indicators. It can solve the problem of false positives and false negatives in traditional methods, provide early warning of potential faults, reduce maintenance costs, and improve the operational stability and user experience of shared devices.

[0032] As an example, the running history big data includes real-time state parameter sequences, fault repair records, component replacement cycle data, environmental interference parameters, and corresponding performance degradation coefficients of the same type of shared device; and the performance degradation knowledge graph of the same type of shared device is constructed based on the running history big data, including: using a time series difference algorithm to extract the trend of the real-time state parameter sequence to obtain a performance parameter decay curve with time; extracting fault type entities, component entities and associated time nodes from the fault repair records and component replacement cycle data through entity recognition technology; combining the mapping relationship between the environmental interference parameters and the performance degradation coefficients, constructing a multi-level association network containing components, parameter decay curves, fault types, and environmental influence weights, and forming a performance degradation knowledge graph.

[0033] The real-time state parameter sequence in the running history big data is the record of various performance indicators of the same type of shared device at different times, such as the continuous change data of the speed and water temperature of a shared washing machine. By using a time series difference algorithm to extract the trend, the difference between the parameters at different time points is calculated, which can accurately capture the decay trend of the performance parameters over time, and finally obtain a performance parameter decay curve with time. The curve reflects the degradation of device performance.

[0034] The fault repair record contains the description of the fault problem of the device, and the component replacement cycle data records the replacement time of each component. By using entity recognition technology to process these data, the fault type entity (such as motor failure), component entity (such as washing machine motor) and their associated time nodes (such as motor failure time and replacement time) can be accurately extracted from the text information, and the correspondence between the fault and the component and the time association are clear. Entity recognition technology is realized by, for example, BERT, DeepSeek, etc. large language model, and specific details are not repeated.

[0035] Environmental interference parameters (such as temperature, voltage) will affect the performance of the device, and there is a specific mapping relationship between it and the performance decay coefficient, that is, different environmental parameters correspond to different degrees of performance decay. The following Table 1 shows an example:

[0036] Table 1: Mapping relationship between environmental interference parameters and performance decay coefficient:

[0037]

[0038] In combination with the above mapping relationship, the previously obtained component, parameter decay curve, fault type, and environmental influence weight (the influence degree of environmental parameters on performance decay) are associated to construct a multi-level association network. Through this multi-level association network, the association between component performance decay curve and fault type is clearly displayed, and the influence weight of environmental factors on this process is also displayed, ultimately forming a complete performance decay knowledge graph.

[0039] As an example, the data enhancement processing of the bad use data set based on the attention mechanism includes: mapping the bad use data set with the wear and tear associated features in the performance decay knowledge graph to construct a usage behavior-component wear attention weight matrix; calculating the contribution degree of each usage behavior feature to the device wear and tear through the attention weight matrix, and screening out potential bad features with a contribution degree higher than a preset threshold; interpolating and expanding the usage data corresponding to the potential bad features and combining the features to generate new bad usage samples, which are added to the bad use data set to form the enhanced bad use data set.

[0040] The bad use data set contains records of various bad use behaviors of users (such as overloading a shared washing machine, high-frequency switching of a hair dryer, etc.), and the wear and tear associated features in the performance decay knowledge graph record the association between different usage behaviors and device component wear and tear (such as the association between overloading and motor wear and tear). Mapping the two to establish an association bridge between usage behavior and component wear and tear.

[0041] The influence degree of different usage behaviors on the wear and tear of each component is calculated through the attention mechanism to form an attention weight matrix, and the elements in the attention weight matrix represent the influence weight of a usage behavior feature on a specific component wear and tear. The higher the weight, the stronger the association between the behavior and the component wear and tear. Then, based on the above attention weight matrix, the contribution degree of each usage behavior feature (such as usage time, operation frequency, etc.) to the overall wear and tear of the device (i.e. the total weight of the wear and tear of various components caused by this feature) is calculated.

[0042] Take shared washing machines and shared shower products as examples to illustrate the construction process of attention weight matrix and the calculation process of contribution degree: (1) For shared washing machines, the use behavior feature is "single overload of 5 kg or more", and the associated components are motor, clutch, drain pipe. Through the attention mechanism, combined with the historical association data of the performance degradation knowledge graph of the behavior and the loss of each component, the attention weight matrix is calculated as "single overload of 5 kg or more, motor" = 0.85, "single overload of 5 kg or more, clutch" = 0.78, "single overload of 5 kg or more, drain pipe" = 0.32. This matrix reflects the influence weight of the use behavior on the loss of each component.

[0043] Then calculate the contribution degree of the use behavior to the overall loss of the equipment, add the above weights, 0.85+0.78+0.32=1.95, that is, the contribution degree of "single overload of 5 kg or more" to the overall loss of the shared washing machine is 1.95.

[0044] (2) For shared shower products, the use behavior feature is "water temperature adjustment more than 3 times per minute (high frequency temperature adjustment)", and the associated components are valve core, heating rod, water pipe interface. Through the attention mechanism, the attention weight matrix is calculated as "high frequency temperature adjustment, valve core" = 0.91, "high frequency temperature adjustment, heating rod" = 0.45, "high frequency temperature adjustment, water pipe interface" = 0.28. Its contribution degree to the overall loss of the equipment is 0.91+0.45+0.28=1.64.

[0045] Set a preset threshold (such as contribution degree ≥ 0.6), and select the use behavior features with contribution degree exceeding the threshold as potential bad features. These features may not be identified by the initial bad judgment standard (such as the combination behavior of "close to rated load + high speed" of shared washing machines), but through the contribution degree calculation, the significant impact of the behavior on the equipment loss can be found.

[0046] For the selected potential bad features, the corresponding original use data is processed: on the one hand, interpolation expansion is carried out (such as generating intermediate value samples within the time interval of the existing use time data to simulate more similar potential bad behaviors); on the other hand, feature combination is carried out (such as combining "high water temperature adjustment" and "short time frequent operation" features to generate more complex bad use scenario samples). These newly generated samples supplement the coverage of hidden bad behaviors in the original data. After adding them to the original bad use data set, the enhanced data set formed can more comprehensively reflect various use behaviors that may cause equipment loss.

[0047] As an example, the interpolation expansion of the use data corresponding to the potential adverse features and the feature combination generate new adverse use samples, including: obtaining the alarm records of the target shared device in the recent preset number of days, and obtaining the alarm type, triggering time and associated use behavior features therefrom; if the alarm type is a continuous parameter overrun, and the proportion exceeds a first threshold, then the linear interpolation method is adjusted to a Gaussian interpolation based on the data distribution before and after the triggering time, and the sample density of the abnormal fluctuation interval is increased; if the alarm type is a composite alarm, and the proportion exceeds a second threshold, then in the feature combination, the discrete features that commonly appear in the alarm records are preferentially associated to improve the matching degree of the composite sample and the actual fault scene; the new adverse use samples generated after adjustment are subjected to effectiveness verification, and the adverse use samples with an alarm feature association degree higher than a third threshold are retained.

[0048] Firstly, all alarm records of the target shared device in the recent preset number of days (such as 30 days) are collected, and these alarm records contain the specific reasons for triggering the alarm of the device (such as the shared washing machine "rotation speed overrun", the shared shower product "water temperature fluctuation too large"), the accurate time of the alarm, and the use behavior features associated before the alarm (such as "high rotation speed mode selection" and "water temperature adjustment frequency" within 10 minutes before the alarm). If the proportion of continuous parameter overrun (such as shared hair dryer "continuous running time exceeds threshold", shared water dispenser "water temperature fluctuation amplitude exceeds upper limit") in the alarm record exceeds a first threshold (such as 60%), it indicates that the target shared device is more likely to be triggered by the fluctuation of continuous use data.

[0049] Since linear interpolation is suitable for smooth data, and Gaussian interpolation can simulate the random fluctuation characteristics of data. Therefore, the original linear interpolation method (uniformly generating intermediate values) is adjusted to a Gaussian interpolation method. Specifically: based on the continuous data before and after the alarm time (such as "use time" and "temperature adjustment amplitude" within 5 minutes before and after the alarm), the probability distribution of the data in the abnormal fluctuation interval (such as the critical value range close to the threshold) is simulated, and more intermediate value samples concentrated in the high risk interval are generated. For example, for the "continuous running time exceeds threshold" alarm, samples of 13-15 minutes are generated in the 10-15 minute time interval to improve the coverage of critical abnormal behavior. In this way, the sample density of the abnormal fluctuation interval can be increased, and the potential adverse use behavior of the continuous parameter anomaly can be more accurately captured.

[0050] When the proportion of composite alarms (such as alarms triggered simultaneously by the shared washing machine "high speed + overload", and alarms triggered by the shared shower product "high frequency temperature adjustment + high water temperature") exceeds the second threshold (such as 40%), it indicates that the synergistic effect of multiple behavior characteristics is the main cause of the fault. At this time, when combining features, no longer randomly associate discrete features (such as "operation frequency" and "mode selection"), but preferentially select feature combinations that commonly appear in alarm records. For example, if "high speed mode" and "overload 80%-90%" frequently appear together in historical alarms, then focus on combining these two discrete features to generate "high speed + near rated load" composite samples, so that the new samples are closer to the actual behavior combination pattern that triggers the fault, improve the effectiveness of the samples, and avoid generating irrelevant feature combinations.

[0051] After generating new bad use samples, the association degree between the samples and the alarm features (such as the similarity between the use behavior features in the samples and the behavior features before historical alarms) is calculated for verification. If the association degree is higher than the third threshold (such as 0.7), it means that the new bad use sample can effectively simulate the bad use behavior in the actual fault scenario, and is retained and added to the original dataset; otherwise, it is rejected. For example, if the similarity between the newly generated "high frequency temperature adjustment + high water temperature" sample and the behavior features before the historical similar alarms is 0.85 (higher than 0.7), it is determined to be an effective sample, ensuring that the enhanced dataset is both rich and consistent with the actual operation risks of the device.

[0052] As an example, the analysis of the potential damage degree set of the target shared device based on the performance degradation knowledge graph and the enhanced bad use dataset includes: classifying the enhanced bad use dataset by use behavior type, and matching with the corresponding component damage rules in the performance degradation knowledge graph; according to the matching result, combining the occurrence frequency and duration of each use behavior, calculating the cumulative damage value of each component, based on the cumulative damage value and the damage level division standard in the performance degradation knowledge graph, determining the potential damage level of each component; associating the potential damage level with the fault type and occurrence probability data in the performance degradation knowledge graph, generating a potential damage degree set containing component identifier, potential damage level, associated fault type and occurrence probability.

[0053] The enhanced post-processing bad use dataset contains various bad use behaviors, such as "overload running" of shared washing machines, "high-speed frequent start-stop", "high-frequency temperature adjustment" of shared shower products, and "long-time high-temperature running". After classifying these bad use behaviors by type, they are matched with the preset component wear-out rules in the performance degradation knowledge graph. For example, "overload running" corresponds to the rule "washing machine motor load exceeds standard → winding overheating wear-out" in the performance degradation knowledge graph, and "high-frequency temperature adjustment" corresponds to the rule "frequent friction of shower valve core → sealing performance degradation". Through the above matching, the specific components and wear-out mechanisms that may be affected by each bad behavior are determined.

[0054] According to the matching results, the occurrence frequency (such as "overload running" occurs 5 times a week) and duration (such as each lasts for 20 minutes) of each bad use behavior are combined, and the quantified relationship between "behavior intensity - wear-out amount" in the performance degradation knowledge graph (such as overload 1 kg / hour corresponds to motor wear-out value 0.02) is referenced to calculate the cumulative wear-out value of each component. For example, the cumulative wear-out value of the washing machine motor due to "overload running" = 5 times / week × 20 minutes / time × corresponding wear-out coefficient. Then, the cumulative wear-out value is compared with the wear-out level division standard in the performance degradation knowledge graph (such as cumulative wear-out value 0-0.3 is "mild", 0.3-0.7 is "moderate", and >0.7 is "severe"), and the potential wear-out level of the motor is determined to be "moderate".

[0055] The potential wear-out level of each component is associated with the fault type and occurrence probability corresponding to that level in the performance degradation knowledge graph, for example, "motor moderate wear-out" is associated with "bearing wear" (occurrence probability 60%) and "winding short circuit" (occurrence probability 30%). The final generated potential wear-out level set is presented in a structured form, such as "component: washing machine motor; wear-out level: moderate; associated faults: bearing wear (60%), winding short circuit (30%)", which fully covers the wear-out risk information of each component.

[0056] As an example, the target monitoring indicator set is obtained by optimizing and adjusting the corresponding monitoring indicator in the monitoring indicator set based on the historical running data of the target shared device and the potential wear-out level set, including: extracting the abnormal fluctuation threshold and fault warning accuracy correlation curve of each monitoring indicator from the historical running data; determining the components that need to be monitored and the corresponding core indicators according to the potential wear-out level and the occurrence probability of each component in the potential wear-out level set; dynamically adjusting the abnormal fluctuation threshold of the core indicators based on the potential wear-out level, retaining the indicators in the non-core indicators that have a correlation degree higher than a preset value with the historical fault records, and eliminating redundant indicators to form the target monitoring indicator set.

[0057] The historical operation data contains fault warning records of each monitoring indicator under different abnormal fluctuation thresholds. For example, the "motor current" indicator of the shared washing machine, when the threshold is set to 10A, the fault warning accuracy is 70%; when the threshold is set to 9A, the accuracy is increased to 85%. By extracting such data, a correlation curve of motor current abnormal fluctuation threshold-fault warning accuracy can be drawn to present the influence of threshold adjustment on the warning effect.

[0058] According to the set of potential loss degrees, if the potential loss level of the "motor" of the shared washing machine is "severe" and the occurrence probability of "bearing wear" is 60%, and the loss level of "drain pipe" is "mild" and the occurrence probability of "blockage" is 20%, the "motor" is preferentially listed as the key monitoring component, and its corresponding indicators such as "motor current", "rotational speed stability" are determined as core indicators. This is because components with high loss level and high occurrence probability are more likely to cause faults and need to be monitored through core indicators.

[0059] For core indicators, adjust the threshold according to the potential loss level: if the motor loss level is "severe", the original threshold of "motor current" is tightened to 6-8A (60%-80% of the original threshold); if the loss level is "moderate", it is adjusted to 8-9A (80%-90% of the original threshold) to improve the sensitivity to potential faults.

[0060] For non-core indicators (such as "washing machine shell temperature"), calculate its correlation degree with historical fault records, if the correlation degree is higher than the preset value (such as 50%), it means that the indicator has auxiliary effect on fault warning, and it is retained; if the correlation degree is lower than the preset value, it is determined as a redundant indicator (such as "control panel key response speed" has low correlation with motor fault), and it is removed, finally forming a simplified and effective target monitoring indicator set.

[0061] For example, Figure 3As shown, the embodiment of the present application also provides an abnormality monitoring system of a shared device based on big data, comprising a knowledge graph construction unit 1001, a potential loss analysis unit 1002, an optimization adjustment unit 1003, and an abnormality monitoring unit 1004. The knowledge graph construction unit 1001 acquires running history big data of a same type of shared device of a target shared device, and constructs a performance attenuation knowledge graph of the same type of shared device based on the running history big data. The potential loss analysis unit 1002 acquires usage big data of the target shared device from a user, extracts a bad usage data set from the usage big data, performs data enhancement processing on the bad usage data set based on an attention mechanism, and analyzes a potential loss degree set of the target shared device based on the performance attenuation knowledge graph and the bad usage data set after the enhancement processing. The optimization adjustment unit 1003 optimizes and adjusts corresponding monitoring indicators in a monitoring indicator set based on historical running data of the target shared device and the potential loss degree set to obtain a target monitoring indicator set. The abnormality monitoring unit 1004 compares real-time running data of the target shared device with the target monitoring indicator set to realize abnormality monitoring.

[0062] As an example, the running history big data comprises real-time state parameter sequences, fault repair records, component replacement period data, environmental interference parameters, and corresponding performance attenuation coefficients of the same type of shared device. The knowledge graph construction unit 1001 specifically: extracts a trend of the real-time state parameter sequences by using a time series difference algorithm to obtain a performance parameter attenuation curve with time; extracts fault type entities, component entities, and associated time nodes from the fault repair records and the component replacement period data by using an entity recognition technology; and constructs a multi-level association network comprising components, parameter attenuation curves, fault types, and environmental influence weights based on a mapping relationship between the environmental interference parameters and the performance attenuation coefficients to form a performance attenuation knowledge graph.

[0063] As an example, the potential loss analysis unit 1002 specifically: maps the bad usage data set and loss association features in the performance attenuation knowledge graph to construct an attention weight matrix of usage behavior-component loss; calculates a contribution degree of each usage behavior feature to device loss by using the attention weight matrix, and screens out potential bad features with a contribution degree higher than a preset threshold; interpolates and expands usage data corresponding to the potential bad features and combines features to generate new bad usage samples, adds the new bad usage samples to the bad usage data set, and forms the bad usage data set after the enhancement.

[0064] As an example, the potential loss analysis unit 1002 specifically: obtains alarm records of the target shared device in a preset number of days in the near future, and extracts alarm types, triggering time and associated use behavior characteristics therefrom; if the proportion of alarm types that are continuous parameter over-limit exceeds a first threshold value, adjusts the linear interpolation method to a Gaussian interpolation based on data distribution before and after the triggering time, and increases the sample density of the abnormal fluctuation interval; if the proportion of alarm types that are composite alarms exceeds a second threshold value, preferentially associates discrete features that commonly appear in alarm records when combining features, to improve the matching degree of composite samples and actual fault scenarios; and performs effectiveness verification on the new bad use samples generated after adjustment, and retains bad use samples with an alarm feature association degree higher than a third threshold value.

[0065] As an example, the potential loss analysis unit 1002 specifically: classifies the bad use data set after enhancement processing according to use behavior types, and matches corresponding component loss rules in the performance degradation knowledge graph; according to the matching result, combines the occurrence frequency and duration of each type of use behavior, calculates the cumulative loss value of each component, determines the potential loss level of each component based on the cumulative loss value and the loss level division standard in the performance degradation knowledge graph; and associates the potential loss level with the fault type and occurrence probability data in the performance degradation knowledge graph, to generate a potential loss degree set containing component identifiers, potential loss levels, associated fault types and occurrence probabilities.

[0066] As an example, the optimization adjustment unit 1003 specifically: extracts an association curve of abnormal fluctuation threshold values of each monitoring index and fault early warning accuracy from the historical running data; determines components and corresponding core indexes that need to be monitored in priority according to the potential loss level and the occurrence probability of each component in the potential loss degree set; dynamically adjusts the abnormal fluctuation threshold values of core indexes based on the potential loss level, retains indexes in non-core indexes with an association degree higher than a preset value with historical fault records, eliminates redundant indexes, and forms a target monitoring index set.

[0067] The embodiment of the application further provides an electronic device, which comprises at least one processor, a memory and a computer program stored in the memory and executable on the at least one processor, and the computer program is executed by the processor to implement the method according to any one of the preceding embodiments.

[0068] The embodiment of the application further provides a computer storage medium, which stores a computer program executable by a processor to implement the method according to any one of the preceding embodiments.

[0069] The embodiments of the present application also provide a computer program product, which comprises a computer program executable by a processor to implement the method according to any one of the preceding embodiments.

[0070] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in the above description in a general manner. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0071] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for monitoring abnormality of a shared device based on big data, the method comprising: The method comprises the following steps: obtaining operation history big data of a same type of shared device of a target shared device, and constructing a performance degradation knowledge graph of the same type of shared device based on the operation history big data; obtaining usage big data of the target shared device, and obtaining a bad usage data set from the usage big data; performing data enhancement processing on the bad usage data set based on an attention mechanism; and analyzing a potential damage degree set of the target shared device based on the performance degradation knowledge graph and the bad usage data set after the enhancement processing. Based on the historical operation data of the target shared device and the potential damage degree set, corresponding monitoring indicators in a monitoring indicator set are optimized and adjusted to obtain a target monitoring indicator set; and real-time operation data of the target shared device is compared with the target monitoring indicator set to realize abnormal monitoring. The potential damage degree set of the target shared device is analyzed based on the performance degradation knowledge graph and the bad usage data set after the enhancement processing, which comprises: the bad usage data set after the enhancement processing is classified according to usage behavior types, and matched with corresponding component damage rules in the performance degradation knowledge graph; according to the matching results, the cumulative damage values of each component are calculated by combining the occurrence frequencies and the duration of each type of usage behavior; based on the cumulative damage values and damage grade division standards in the performance degradation knowledge graph, the potential damage grades of each component are determined; the potential damage grades are associated with fault types and occurrence probability data in the performance degradation knowledge graph to generate a potential damage degree set containing component identifiers, potential damage grades, associated fault types and occurrence probabilities.

2. The method of claim 1, wherein: The operation history big data comprises real-time state parameter sequences, fault repair records, component replacement cycle data, environmental interference parameters and corresponding performance degradation coefficients of the same type of shared device. The performance degradation knowledge graph of the same type of shared device is constructed based on the operation history big data, which comprises: a time series difference algorithm is used to extract trends of the real-time state parameter sequences to obtain parameter degradation curves of performance parameters over time. Fault type entities, component entities and associated time nodes are extracted from the fault repair records and the component replacement cycle data by entity recognition technology; a multi-level association network containing components, parameter degradation curves, fault types and environmental influence weights is constructed based on the mapping relationship between the environmental interference parameters and the performance degradation coefficients to form the performance degradation knowledge graph. 3.The method of claim 1, wherein: The data enhancement processing on the bad usage data set based on the attention mechanism comprises: the bad usage data set is mapped with damage association features in the performance degradation knowledge graph to construct an attention weight matrix of usage behavior-component damage; the contribution degrees of each usage behavior feature to device damage are calculated through the attention weight matrix, and potential bad features with a contribution degree higher than a preset threshold are screened out; the usage data corresponding to the potential bad features are interpolated and expanded, and new bad usage samples are generated, which are added to the bad usage data set to form the bad usage data set after the enhancement processing. 4.The method of claim 3, wherein: The potential adverse characteristics corresponding to the use data are interpolated and expanded and combined with characteristics to generate new adverse use samples, including: obtaining alarm records of the target shared device in the recent preset number of days, obtaining alarm types, triggering time and associated use behavior characteristics therefrom; if the alarm type is a continuous parameter overrun ratio exceeding a first threshold, adjusting the linear interpolation method to a Gaussian interpolation based on data distribution before and after the triggering time to increase the sample density of the abnormal fluctuation interval; if the alarm type is a composite alarm ratio exceeding a second threshold, preferentially associating discrete characteristics that commonly appear in the alarm records during feature combination to improve the matching degree of the composite sample and the actual fault scenario; performing effectiveness verification on the new adverse use samples generated after adjustment, and retaining adverse use samples with an alarm characteristic association degree higher than a third threshold.

5. The method of claim 1, wherein: Based on the historical running data of the target shared device and the set of potential damage degrees, corresponding monitoring indicators in the monitoring indicator set are optimized and adjusted to obtain a target monitoring indicator set, including: extracting the correlation curve of the abnormal fluctuation threshold and the fault early warning accuracy of each monitoring indicator from the historical running data; determining the components that need to be monitored and the corresponding core indicators according to the potential damage levels and the occurrence probabilities of the components in the set of potential damage degrees; dynamically adjusting the abnormal fluctuation threshold of the core indicators based on the potential damage levels, retaining indicators in non-core indicators with an association degree higher than a preset value with historical fault records, and eliminating redundant indicators to form the target monitoring indicator set. 6.A big data based shared device abnormality monitoring system, characterized in that, It includes a knowledge graph construction unit, a potential damage analysis unit, an optimization adjustment unit, and an abnormal monitoring unit. The knowledge graph construction unit obtains running historical big data of similar shared devices of the target shared device, and constructs a performance degradation knowledge graph of the shared devices based on the running historical big data. The potential damage analysis unit obtains use big data of the target shared device from the user, extracts a set of adverse use data from the use big data, performs data enhancement processing on the set of adverse use data based on an attention mechanism, and analyzes the potential damage degree set of the target shared device based on the performance degradation knowledge graph and the set of adverse use data after enhancement processing. The optimization adjustment unit optimizes and adjusts corresponding monitoring indicators in the monitoring indicator set based on the historical running data of the target shared device and the set of potential damage degrees to obtain a target monitoring indicator set. The abnormal monitoring unit compares the real-time running data of the target shared device with the target monitoring indicator set to realize abnormal monitoring. The potential damage analysis unit specifically: classifies the set of adverse use data after enhancement processing according to use behavior types, and matches with the corresponding component damage rules in the performance degradation knowledge graph; according to the matching result, combines the occurrence frequency and duration of each type of use behavior to calculate the cumulative damage value of each component, and determines the potential damage level of each component based on the cumulative damage value and the damage level division standard in the performance degradation knowledge graph. Correlate the potential loss level with the fault type and occurrence probability data in the performance degradation knowledge graph to generate a potential loss degree set containing component identification, potential loss level, associated fault type and occurrence probability.

7. An electronic device, the electronic device comprising: The computer program is executed by the processor to implement the method of any one of claims 1-5.

8. A computer storage medium, characterized in that: The computer storage medium stores a computer program executable by the processor to implement the method of any one of claims 1-5.

9. A computer program product, characterised in that: The computer program product contains a computer program executable by the processor to implement the method of any one of claims 1-5.

Images