Search

CN-122026749-A - Method for monitoring performance of motor equipment

CN122026749ACN 122026749 ACN122026749 ACN 122026749ACN-122026749-A

Abstract

The invention discloses a method for monitoring the performance of motor equipment, which relates to the technical field of motor equipment monitoring, wherein the accuracy and the real-time performance of data processing are remarkably improved through advanced technologies such as multithread management, self-calibration learning, topology virtualization, load balancing and the like, the multithread management module ensures the high efficiency and the high accuracy of data analysis through dynamically adjusting the analysis range and the precision, the self-calibration learning technology realizes the accurate evaluation of the performance of the equipment by analyzing fine waveform information and forming sample data, the topology virtualization technology constructs a virtual entity consistent with the performance of the actual equipment through recombining analysis data, the user can intuitively know the state of the equipment conveniently, the load balancing technology optimizes the overall performance of a system through dynamically adjusting the loads of an edge end and a central server, the real-time performance of data processing is ensured, and the comprehensive application of the technologies greatly improves the efficiency and the reliability of the performance monitoring of the motor equipment.

Inventors

  • CHEN DEKE

Assignees

  • 先端领航(广州)技术有限公司

Dates

Publication Date
20260512
Application Date
20260127

Claims (10)

  1. 1. A method for monitoring the performance of motor equipment is characterized by comprising the following specific steps: Constructing a star-shaped framework of an edge end and a central server, wherein the star-shaped framework is compatible with high-availability cloud deployment and high-safety local deployment, integrates a resource management technology and a remote upgrading technology, and establishes an independent, universal and controllable monitoring network on the premise of not modifying, influencing and interacting field motor equipment; the multithread management comprises a multithread management module, a self-adaptive unit, a self-learning unit and a self-evaluation unit, wherein the multithread management module comprises a resource establishment unit, a self-adaptive unit, a self-learning unit and a self-evaluation unit, the resource establishment unit completes hardware setting and software setting based on a setting document and starts data acquisition, the self-adaptive unit triggers when acquired data reach a preset level, the self-learning unit selects the analysis range and the precision of the data, the self-learning unit decomposes and reorganizes the data in the analysis range and the precision, historical data of embedded load conditions form footprint information, the self-evaluation unit activates when the footprint information reaches a triggering threshold value, carries out high-precision evaluation on equipment performance and outputs an alarm according to probability screening rules; Self-calibration learning, namely adopting a self-calibration learning method, collecting fine waveform information of equipment through an edge end, analyzing the fine waveform information into a plurality of characteristic values, mapping part of characteristic values to a mark group, and mapping part of characteristic values to the characteristic group, wherein the mark group and the characteristic group are combined to form sample data; Topology virtualization, namely, applying a topology virtualization technology, receiving analysis data uploaded by an edge, wherein the analysis data comprises specific numerical values and alarm information of all analysis channels, recombining the analysis data according to a user-specified relation, and mapping to form a virtual entity consistent with the performance of objective equipment, or constructing objective non-existent virtual equipment meeting the attention requirement of a user; The load balancing method comprises the steps of deploying a load balancing technology, setting an edge load pool at an edge, setting a center service load pool and a load balancing module at a center server, wherein the edge load pool is responsible for data acquisition, model learning and information output, and the center service load pool is responsible for data stacking, virtual mapping, entity recombination, data query and pushing; The data safety backup comprises the steps of implementing a data safety backup technology, arranging a high-speed data pool in an edge end for storing long-time information for rapid model scheduling, arranging a large-capacity data lake in a central server, collecting key data of each edge end after model learning evaluation, and simultaneously collecting original data of the edge end by the central server and storing the original data in an external data warehouse in a bypass mode for subsequent review comparison.
  2. 2. A method for monitoring the performance of motor-class equipment according to claim 1, characterized in that the high robustness of the contact assembly in the architecture construction step is realized by the following technical steps that firstly, when the contact assembly of an edge end and a center service end is initialized, an independent process daemon is configured for each sub-assembly, the daemon circularly detects the running state of the corresponding sub-assembly according to a period of 50ms, the detection content comprises a CPU occupied by the sub-assembly process, whether the memory resource is within a preset threshold value, whether the data processing response time exceeds 100ms and the data transmission packet loss rate of other assemblies exceeds 0.5%, secondly, when the daemon detects any abnormal condition of a certain sub-assembly, a pre-deployed standby sub-assembly is started to take over the task currently being executed, the standby sub-assembly and a main assembly share real-time data buffer storage, the buffer storage data adopts a ring queue for storage, and then, in the process state parameters, occupation and data transmission log information are uploaded to the center service end in real time according to the period of 50ms, the key information of the abnormal occurrence time, if the error log is detected, the error type of the error log is detected to be automatically, the error is detected, the error is corrected by the error-correction module is automatically, the error-corrected by the error-correction module is automatically, and the error-corrected by the error-correction module is calibrated, and the error-correction of the error-correction module is automatically, if the error-corrected, and the error-corrected by the error-correction of the error-correction module is automatically, and the error-corrected by the error-correction of the error-correction module, and the error-corrected by the error-correction module, the standby subassembly enters a standby state.
  3. 3. The method for monitoring the performance of a motor device according to claim 1, wherein the dynamic adjustment of the analysis range and the precision of the adaptive unit in the multithreading management step is realized by the following steps: Firstly, after a resource establishment unit starts data acquisition, an adaptive unit counts the accumulated quantity of acquired data in real time, and when the data quantity reaches 80% of a preset magnitude, a data preprocessing flow is started in advance to perform noise filtering and data format standardization processing on the acquired data; secondly, when the data quantity reaches a preset magnitude, the self-adaptive unit firstly performs double cluster analysis of time dimension and feature dimension on the whole data, the time dimension is divided into a plurality of time segments according to the running period of the equipment, and the feature dimension is divided into different feature clusters according to the data type; thirdly, calculating the distribution density and the fluctuation coefficient of the data according to each time segment and each characteristic cluster, wherein the distribution density is calculated by adopting a nuclear density estimation method, the fluctuation coefficient is the ratio of the standard deviation to the average value of the data, the time segments and the characteristic clusters with the distribution density higher than a preset density threshold and the fluctuation coefficient lower than the preset fluctuation threshold are determined as key analysis ranges, and the rest parts are taken as conventional analysis ranges; Determining analysis precision according to the type of the analysis range, wherein the key analysis range adopts a high-precision analysis mode, the data sampling interval is reduced to 50% of the original interval, the feature extraction dimension is increased by 30%, and the conventional analysis range adopts a standard-precision analysis mode, so that the original sampling interval and the feature extraction dimension are maintained; and fifthly, in the running process of the self-learning unit and the self-evaluation unit, the self-adaptive unit continuously monitors the validity of the analysis result, if the matching degree of the key analysis range is lower than 85%, the time segment span of the key analysis range is automatically expanded, and if the abnormal data point ratio of the conventional analysis range exceeds 5%, the conventional analysis range is updated to be the key analysis precision.
  4. 4. The method for monitoring the performance of motor equipment according to claim 1, wherein the variability weighted aggregation in the self-calibration learning step adopts an overall performance evaluation algorithm, and the formula is: wherein Represents an overall performance evaluation index of the device, Is the first The variability of the condition particles of each marker set, Is the first Sample embedding depth of individual condition particles, Is the first The continuous operation time of the equipment corresponding to each condition particle, k is the time attenuation coefficient, Is the first Weights of individual marker set conditional particles.
  5. 5. The method for monitoring the performance of motor equipment according to claim 1, wherein the adaptive optimization of the characteristic value mapping relation in the self-calibration learning step is realized by the following technical steps: Firstly, in an initial stage of self-calibration learning, presetting initial characteristic value mapping relations according to equipment types and monitoring requirements, and setting initial confidence coefficient for each mapping relation; secondly, with the continuous accumulation of sample data, carrying out validity evaluation on the current mapping relation once every 100 groups of sample data, wherein evaluation indexes comprise variability differentiation degree after sample combination and accuracy of evaluation results; Thirdly, if the variation degree of a certain mapping relation is lower than a preset degree threshold or the accuracy is lower than 80%, starting a mapping relation adjustment flow, selecting a new mapping scheme from the feature value combinations which are not mapped currently, wherein the new scheme needs to meet the condition that the feature value types of the mark group and the feature group are not repeated and the key operation parameters of the equipment are covered; Fourthly, verifying the new mapping relation by using a small range sample, selecting 20 groups of historical samples and 10 groups of newly added samples for testing, calculating the degree of variation differentiation and the accuracy of the new mapping relation, replacing the current mapping relation if both indexes are higher than 10% of the current mapping relation, and simultaneously updating the confidence coefficient to be the average value of the testing indexes, and continuously testing other mapping schemes if the confidence coefficient is not reached; and fifthly, when the operation condition of the equipment changes, automatically triggering the reevaluation and adjustment of the mapping relation, and ensuring that the mapping relation is always suitable for the actual operation state of the equipment.
  6. 6. The method for monitoring the performance of the motor equipment according to claim 1, wherein the performance parameter synchronous optimization of the virtual entity and the objective equipment in the topology virtual step is realized by the following technical steps that firstly, when an edge terminal uploads analysis data, a time stamp and a data acquisition sequence number are embedded in a data frame, after the center server terminal receives the data, the consistency of the time stamp and the continuity of the sequence number are firstly verified, and if the time stamp deviation exceeds 10ms or the number is missing, a data retransmission request is immediately sent to the corresponding edge terminal; the topology virtual module of the center server adopts a parallel processing architecture, groups the received multi-edge analysis data according to the association relation of the equipment, distributes independent processing threads for each group of data, simultaneously analyzes and extracts the characteristics of each group of data, wherein the extracted characteristics comprise real-time values, change rates, accumulated values and the like of equipment operation parameters, carries out fusion calculation on the grouped characteristic data according to the association relation appointed by a user, processes multi-source data of the same parameter by adopting a weighted average method in the fusion process, determines the weight according to the acquisition precision grade of each edge, establishes the mapping relation between the performance parameters of the virtual entity and the fused characteristic data, adopts a real-time synchronous update mechanism, updates the performance parameters of the virtual entity once every time when one frame of edge data is received, adopts an incremental update algorithm in the update process, only updates the changed parameters, reduces the calculation cost, finally monitors the synchronous delay of the performance parameters of the virtual entity and objective equipment in real time in the synchronization process, calculates the difference between the update time of the parameters of the virtual entity and the data acquisition time of the edge, and if the delay time exceeds 30ms, and automatically adjusting the priority of the parallel processing threads, increasing the priority of the processing threads corresponding to the virtual entity by one level, and simultaneously temporarily reducing the number of the processing threads of the non-key virtual equipment to ensure that the synchronous delay is controlled within a preset threshold.
  7. 7. A method for monitoring the performance of motor-class equipment according to claim 1, characterized by that, the user-defined virtual equipment in said topological virtual step is built up by means of the visual operation interface of local client or remote client, the user initiates the request for building user-defined virtual equipment, the interface provides the equipment name setting, the attention performance parameter selection, the data source edge assignment, parameter weight configuration function module, the user completes the basic configuration according to the self-monitoring requirement and submits the request, the topological virtual module of central server receives the request, then carries out legal check to the user configuration information, the check content includes whether the appointed edge end is on line, the concerned performance parameter is the parameter which is supported and collected by the edge end, the parameter weight configuration meets the normalization requirement, if the check is not passed, the error prompt is returned and the reason is explained, if the check is passed, the virtual equipment instance is created, the third step, the topological virtual module builds the data subscription relation according to the user-specified edge end and performance parameter, receives the corresponding performance parameter data of the appointed edge end in real time, and carries out the weighted fusion calculation to the multi-source data according to the parameter weight of the user configuration, the user-defined data is obtained, the user-defined comprehensive performance parameter is evaluated by the user-defined virtual equipment, the user-defined virtual equipment is evaluated by the virtual equipment, the user-defined virtual equipment is evaluated by the virtual equipment in real-time, the virtual equipment is evaluated by the virtual module, the user-defined virtual equipment is set up by the virtual equipment, the user-defined virtual equipment is evaluated by the virtual equipment, and the user-defined by the user-defined virtual equipment has a virtual equipment, when the user-defined module has a virtual module, and the user-defined module has a real-defined module, according to real-time device, and a user-defined device, the topology virtual module responds to the configuration change in real time, completes adjustment on the premise of not interrupting the operation of the virtual equipment, automatically stores configuration change records and supports the backtracking of historical configuration.
  8. 8. The method for monitoring the performance of a motor device according to claim 1, wherein the edge acquisition performance adjustment in the load balancing step adopts a load response algorithm, and the formula is: wherein In order to adjust the data acquisition frequency of the edge end, In order to adjust the initial acquisition frequency before adjustment, As a factor of the influence of the load, For the actual load of the current central server load pool, For a preset load safety threshold value, For the maximum load carried by the central server, The coefficients are modified for data importance.
  9. 9. The method for monitoring the performance of motor equipment according to claim 1, wherein the load prediction and the advanced scheduling in the load balancing step are implemented by the following technical steps: Firstly, a load balancing module of a central server is internally provided with a load prediction model, the model is constructed based on a long-period memory network, and input data comprise central server load data, edge data transmission rate, equipment monitoring task number and historical synchronous load data in the past 24 hours; secondly, the load prediction model predicts the central server load change trend of each period within 60 minutes in the future in advance according to 10 minutes as a prediction period, outputs a predicted load value and a predicted confidence coefficient, and the prediction result with the confidence coefficient lower than 70% is corrected by combining a sliding average method; Thirdly, judging whether the scheduling is needed in advance according to the predicted load value, if the predicted load value in a certain period exceeds 75% of a preset threshold value, starting an advanced scheduling process, preferentially sending a load adjustment forecast to the edge end corresponding to the non-key equipment, informing the non-key equipment of the about to perform acquisition performance adjustment, and reserving data buffering time; step four, after the advanced scheduling process is started, calculating the number and the adjustment amplitude of edge ends to be adjusted by the central server load balancing module, preferentially selecting the edge ends with larger data transmission quantity but lower key grade of equipment as adjustment objects, and adopting a gradient adjustment mode, wherein the adjustment amplitude does not exceed 20% of the initial acquisition frequency each time; And fifthly, finishing the adjustment of the collection performance of the edge end 10 minutes before the prediction period, simultaneously monitoring the load change of the adjusted center service end in real time, and dynamically correcting the adjustment amplitude if the deviation between the actual load and the predicted load exceeds 10 percent.
  10. 10. The method for monitoring the performance of a motor device according to claim 1, wherein the multi-layer data encryption transmission and storage in the data security backup step is implemented by the following technical steps: Firstly, the original data collected by an edge end are encrypted by adopting a symmetric encryption algorithm before being transmitted to a central server, a dynamic key is generated by combining unique identification of edge end equipment, a current time stamp and a random number, the key is updated every 10 minutes, and the key is transmitted to the central server through an asymmetric encryption algorithm, so that the data is prevented from being stolen or tampered in the transmission process; Secondly, dividing data into core data and common data according to importance by a high-speed data pool built in an edge end in a partition encryption storage mode, storing the core data by an AES-256 encryption algorithm, storing the common data by the AES-128 encryption algorithm, adding a check code during data storage, and verifying the check code when each time of calling the data; thirdly, a large-capacity data lake of the central server adopts a distributed encryption storage architecture, data fragments are stored on a plurality of physical nodes, each fragment adopts different encryption keys, the keys are uniformly generated and managed by a key management module of the central server, and the key management module adopts a hardware encryption module to store master keys, so that key leakage is prevented; then, when the external data warehouse stores the edge-end original data, adopting a remote disaster recovery backup strategy, deploying at least two backup nodes in different regions, carrying out incremental backup every day in a mode of combining incremental backup and full backup, carrying out full backup once a week, and storing the backup data by encryption; and finally, establishing a data access authority management and control mechanism, distributing different access authorities by different user roles, carrying out identity authentication when accessing data, and recording detailed access logs.

Description

Method for monitoring performance of motor equipment Technical Field The invention relates to the technical field of motor equipment monitoring, in particular to a method for monitoring performance of motor equipment. Background With the rapid development of industrial automation and intelligent manufacturing, motor equipment is used as a core power source in industrial production, and the running state and performance of the motor equipment directly influence the efficiency and stability of the whole production system. Conventional motor equipment performance monitoring techniques have various limitations. Firstly, the traditional method often depends on regular inspection and offline testing, and the intermittent monitoring mode cannot reflect the running state of the equipment in real time, so that failure discovery is not timely easily caused, and the shutdown risk is increased. Secondly, the traditional monitoring means have limited data processing and analysis capability, valuable information is difficult to extract from mass data, evaluation on equipment performance often stays on the surface, and potential problems cannot be deeply mined. In addition, the traditional monitoring system is generally tightly coupled with equipment, is high in transformation difficulty and cost, is difficult to adapt to monitoring requirements in different scenes, and lacks versatility and flexibility. Finally, the conventional method has defects in data security and backup recovery, and once the data is lost or damaged, the normal operation of the monitoring system is seriously affected. Aiming at a plurality of defects existing in the traditional motor equipment performance monitoring technology, the invention provides a method for monitoring the motor equipment performance, which is particularly important. Disclosure of Invention The invention aims to make up the defects of the prior art, provides a method for monitoring the performance of motor equipment, which can realize real-time, accurate and reliable monitoring of the performance of the motor equipment through integrating advanced technologies such as multithreading management, self-calibration learning, topology virtual and load balancing, and the like, the architecture is compatible with high-availability cloud deployment and high-safety local deployment modes, the method can establish a high-efficiency and flexible monitoring network without modifying the existing equipment, and meanwhile, the accuracy and the instantaneity of data processing are remarkably improved through the measures of dynamically adjusting the analysis range and the precision, accurately evaluating the equipment performance, constructing a virtual entity, optimizing the system load and the like, so that powerful guarantee is provided for the safe and stable operation of industrial production. The invention provides a method for monitoring the performance of motor equipment, which aims to solve the technical problems and comprises the following specific steps: Constructing a star-shaped framework of an edge end and a central server, wherein the star-shaped framework is compatible with high-availability cloud deployment and high-safety local deployment, integrates a resource management technology and a remote upgrading technology, and establishes an independent, universal and controllable monitoring network on the premise of not modifying, influencing and interacting field motor equipment; the multithread management comprises a multithread management module, a self-adaptive unit, a self-learning unit and a self-evaluation unit, wherein the multithread management module comprises a resource establishment unit, a self-adaptive unit, a self-learning unit and a self-evaluation unit, the resource establishment unit completes hardware setting and software setting based on a setting document and starts data acquisition, the self-adaptive unit triggers when acquired data reach a preset level, the self-learning unit selects the analysis range and the precision of the data, the self-learning unit decomposes and reorganizes the data in the analysis range and the precision, historical data of embedded load conditions form footprint information, the self-evaluation unit activates when the footprint information reaches a triggering threshold value, carries out high-precision evaluation on equipment performance and outputs an alarm according to probability screening rules; Self-calibration learning, namely adopting a self-calibration learning method, collecting fine waveform information of equipment through an edge end, analyzing the fine waveform information into a plurality of characteristic values, mapping part of characteristic values to a mark group, and mapping part of characteristic values to the characteristic group, wherein the mark group and the characteristic group are combined to form sample data; Topology virtualization, namely, applying a topology virtualization technology, receiving analy