CN112947290A

CN112947290A - Edge cloud cooperation-based equipment state monitoring method and system and storage medium

Info

Publication number: CN112947290A
Application number: CN202110531024.XA
Authority: CN
Inventors: 张颖华
Original assignee: Beijing Cyberconsortium Technology Co ltd
Current assignee: Beijing Cyberconsortium Technology Co ltd
Priority date: 2021-05-16
Filing date: 2021-05-16
Publication date: 2021-06-11
Anticipated expiration: 2041-05-16
Also published as: CN112947290B

Abstract

The invention belongs to the technical field of equipment performance prediction, and relates to an equipment state monitoring method, a monitoring system and a storage medium based on edge cloud cooperation. The method comprises the following steps: collecting the operation information of at least one device; processing and caching the collected running information; extracting the characteristics of the cached operation information; generating a data snapshot by using a timestamp at least comprising the characteristic parameters and the data information corresponding to the extracted characteristic parameters; transmitting an index data set at least comprising a data snapshot, and device information and event description corresponding to the data snapshot to a cloud database and storing the index data set; the cloud end organizes data arrangement of multiple levels including equipment, an event map and a characteristic waveform according to the index data set; the data layout corresponding to the indexed data set is displayed at a terminal in communication with the database. The system can realize edge cloud cooperation in the true sense, comprehensively monitor the state of the equipment and improve the accuracy of equipment state evaluation and fault prediction.

Description

Edge cloud cooperation-based equipment state monitoring method and system and storage medium

Technical Field

The invention belongs to the technical field of equipment performance prediction, and particularly relates to a method for monitoring equipment state based on edge cloud cooperation, a system for monitoring equipment performance based on edge cloud cooperation and a distributed storage medium.

Background

Industrial manufacturing facilities face significant operational and maintenance challenges, and unplanned equipment outages not only severely affect manufacturing efficiency and quality, but also bring high maintenance and repair costs to the manufacturing enterprise. How to effectively and accurately maintain equipment becomes a difficult problem which needs to be overcome urgently in industrial manufacturing digital transformation.

The predictive maintenance mode is strongly advocated by the industry and is considered to be an effective means for dealing with unexpected shutdown of equipment. The predictive maintenance is to send out a fault early warning in advance by monitoring and evaluating the state of the equipment in real time so as to avoid the sudden occurrence of the fault.

For the state perception of the equipment, the traditional approach is to start with a mechanical mechanism and diagnose the machine by measuring direct changes (e.g. vibration) or indirect changes (e.g. sound) of the mechanical state. This approach is usually only useful for finding mechanical anomalies such as wear of the bearings, loosening of the screws, etc., but not for electrical faults that may lead to serious equipment failure.

For electrical aspects, the traditional current measurement mode is that the equipment current obtained by a three-phase electric meter or a Programmable Logic Controller (PLC) and the like only has an average current value within a period of time, usually only can reflect the power consumption condition within a period of time, and the information is single, so that more information cannot be obtained.

Compared with single-machine measurement data, the measurement data of a new industrial internet system structure is greatly increased, a Cloud Storage (Cloud Storage) mode has to be adopted, and on-site data needs to be sent to a Cloud end, namely the data is stored on a plurality of virtual servers hosted by a third party. In the past, the data is low-frequency data, and the data volume is not very large. And intellectuality then requires to have more sensitive perception, and the information of perception contains a large amount of detail information simultaneously, and these information mean very big data bulk, if with all these data that relate to health and operating mode etc. of equipment, all send to the high in the clouds, must cause the resource shortage of high in the clouds data storage server, can lead to submerging in magnanimity data to the useful data of health and the operating mode of diagnostic equipment in addition, make the server fall into a speed retardation, the simple storage device that reduces day by day.

Disclosure of Invention

The technical problem to be solved by the present invention is to provide a method for monitoring a state of a device based on edge cloud coordination, a system for monitoring a performance of a device based on edge cloud coordination, and a distributed storage medium, which can achieve edge cloud coordination in a true sense: the edge side continuously sends the abnormal original data back to the cloud end, the cloud end carries out continuous big data autonomous learning to obtain a judgment threshold value for abnormal judgment, the judgment threshold value is sent to the edge side at any time to correct an abnormal recognition strategy, and multiple state abnormity monitoring can be carried out on the same equipment, so that comprehensive state monitoring is carried out on the equipment on the basis of one hardware platform, the accuracy of equipment state evaluation is improved, and equipment fault prediction is carried out.

The technical scheme adopted for solving the technical problem of the invention is to provide the following scheme:

as one aspect of the present invention, the present invention provides a device status monitoring method based on edge cloud coordination, which includes the steps of:

collecting operation information of at least one device, wherein the operation information comprises at least one of current, voltage, vibration, noise, temperature, strain and attitude;

processing and caching the collected running information;

performing feature extraction on the cached operation information, wherein feature parameters obtained by the feature extraction comprise: at least one or a combination of amplitude, peak value, action time and frequency component;

generating a data snapshot at least comprising the characteristic parameters and timestamps corresponding to the data information for extracting the characteristic parameters;

transmitting an index data set at least comprising the data snapshot, and device information and event description corresponding to the data snapshot to a cloud database and storing the index data set;

the cloud end organizes a plurality of levels of data arrangement including equipment, an event map and a characteristic waveform according to the index data set;

displaying, at a terminal in communication with the database, a data arrangement corresponding to the indexed data set;

wherein, at least one device is the same type of device or different type of device in the same working space, or the same type of device or different type of device in different working spaces.

Preferably, before generating the data snapshot, the method further includes:

generating the data snapshot according to the equipment state by a multivariate regression analysis method based on the Mahalanobis distance according to the characteristic parameters, wherein the method comprises the following steps:

merging the characteristic parameters, the cache information of the collected frames corresponding to the characteristic parameters and the abnormal results of the characteristic parameters which are judged to be abnormal to generate a data snapshot, and stamping a time stamp on the data snapshot;

and generating a data snapshot for the characteristic parameters which are judged to be normal, and stamping a time stamp for the data snapshot.

Preferably, the step of generating the data snapshot according to the device status by using a multivariate regression analysis method based on mahalanobis distance according to the characteristic parameters includes:

presetting a judgment threshold, and setting a rated working condition of a certain characteristic parameter or calculating a mean value of the certain characteristic parameter as an expected value of Mahalanobis distance calculation;

calculating a covariance matrix of the characteristic parameter according to the expected value, wherein at least one of covariance elements forming the covariance matrix is a vector of the characteristic parameter;

calculating to obtain the Mahalanobis distance as a working condition deviation value according to the covariance matrix;

judging the equipment state according to the working condition deviation and the judgment threshold value, wherein the judging step comprises the following steps:

if the working condition deviation is larger than the judgment threshold value, judging that the equipment is abnormal;

if the working condition deviation value is less than or equal to the judgment threshold value, judging that the equipment is normal;

the data snapshot comprises the working condition deviation values.

Preferably, the covariance matrix formula is:

wherein:μ _iis the expected value of the ith element;

the (i, j) th element in the matrix isX _iAndX _jthe covariance of (a);

the mahalanobis distance formula is:

wherein:

is an average value;

Sis a covariance matrix of the multivariate.

Preferably, the preset decision threshold is:

matching sample equipment with the same type as the monitored equipment, and acquiring working condition life lines of all finally failed sample equipment;

selecting the working condition life line with the maximum matching degree with the monitored equipment from the working condition life lines of the sample equipment;

and taking the health value from the selected failure point of the working condition life line to the remaining service life point corresponding to the current latest acquired data as a new judgment threshold value.

Preferably, the method further comprises the step of predicting and reporting the performance of the monitored equipment:

fitting a working condition life line according to the working condition deviation value corresponding to the characteristic parameter of the equipment of the same type of the monitored equipment;

selecting the monitored device in the database, extracting the index dataset of the monitored device;

performing M equal segmentation on the time axis of the working condition life line corresponding to the characteristic parameter from normal equipment to complete equipment failure, wherein the state index value of the normal point of the equipment corresponds to a Mahalanobis distance segment value obtained by bringing the expected value into the time axis, and the state index value of the complete equipment failure point corresponds to a Mahalanobis distance segment value obtained by bringing the characteristic parameter into the time axis when the equipment fails;

determining a health value and fault prediction time according to the position of the working condition deviation in the working condition life line;

and performing prediction report on the performance of the monitored equipment according to the health value and the fault prediction time.

Preferably, in the step of determining the health value, the method comprises:

according to the position of the working condition deviation value falling into the time axis of the working condition life line:

calculating the health value as:

wherein: h_xIs the equipment health value;

m is the number of sections of the working condition life line;

m is the number of the falling section of the working condition life line corresponding to the current working condition deviation value;

D_xthe current Mahalanobis distance working condition deviation value is obtained;

D_m-1the deviation value of the initial working condition of the falling section is obtained;

D_mthe deviation value of the end working condition of the falling section is obtained;

in the step of calculating the failure prediction time, the method comprises the following steps:

establishing a polynomial fitting regression formula based on the Mahalanobis distance of the monitored equipment and the time difference between the current monitoring time and the fault occurrence time:

wherein:ythe time difference between the monitored equipment and the fault occurrence time at the monitoring moment;

xmahalanobis distance of the monitored device at the monitoring time;

a _n， a _n-1，…，a ₁，a ₀is a fitting coefficient;

and obtaining the fault prediction time of the monitored equipment according to the Mahalanobis distance.

Preferably, the processing the collected operation information includes:

before caching: preprocessing the collected running information, wherein the preprocessing comprises filtering and correcting;

after caching and before feature extraction is carried out on the running information:

and performing characteristic analysis on the operation information by adopting time domain analysis processing and frequency domain analysis processing.

As another aspect of the present invention, the present invention further provides an equipment status monitoring system based on edge cloud coordination, which includes an edge side, a cloud side, and a terminal, where the edge side includes a data acquisition unit, a data processing unit, a feature extraction unit, a data snapshot unit, and an event reporting unit, where:

the data acquisition unit is configured to acquire operation information of at least one device, wherein the operation information comprises at least one of current, voltage, vibration, noise, temperature, strain and attitude;

the data processing unit is configured to process and cache the acquired running information;

the feature extraction unit is configured to perform feature extraction on the cached operation information, and feature parameters obtained by feature extraction include: at least one or a combination of amplitude, peak value, action time and frequency component;

the data snapshot unit is configured to generate a data snapshot at least including the characteristic parameters and timestamps corresponding to the data information for extracting the characteristic parameters;

the event reporting unit is configured to transmit an index data set at least comprising the data snapshot, and the device information and the event description corresponding to the data snapshot to a cloud database and store the index data set;

the cloud comprises a data arrangement unit configured to organize a plurality of levels of data arrangement including devices, event maps and characteristic waveforms according to the index data set;

a terminal configured to communicate with the database and to display a data arrangement corresponding to the index data set;

Preferably, the data snapshot unit includes a determining module, and the determining module is configured to generate the data snapshot according to the device state by a multivariate regression analysis method based on mahalanobis distance according to the characteristic parameter, and includes:

Preferably, the data snapshot unit further includes a threshold acquisition module, a preset module, and a calculation module, wherein:

the threshold value obtaining module is configured to obtain a preset decision threshold value;

the preset module is configured to set a rated working condition of a certain characteristic parameter or calculate a mean value of the certain characteristic parameter as an expected value of Mahalanobis distance calculation;

the calculation module is configured to calculate a covariance matrix of the feature parameter according to the expected value, and at least one of covariance elements forming the covariance matrix is a vector of the feature parameter;

further, according to the covariance matrix, calculating to obtain a Mahalanobis distance as a working condition deviation value;

the judging module is configured to judge the device state according to the working condition deviation and the judgment threshold, and includes:

the data snapshot comprises the working condition deviation values.

Preferably, in the calculation module, the covariance matrix formula is:

wherein:μ _iis the expected value of the ith element;

the (i, j) th element in the matrix isX _iAndX _jthe covariance of (a);

the mahalanobis distance formula is:

wherein:

is an average value;

Sa covariance matrix that is multivariate.

Preferably, the cloud is further provided with an update decision threshold unit, and the update decision threshold unit includes a sample device matching module, a working condition life line matching module, and a threshold setting module, wherein:

the sample equipment matching module is used for matching sample equipment with the same type as the monitored equipment and acquiring working condition life lines of all finally failed sample equipment;

the working condition life line matching module is used for selecting a working condition life line with the maximum matching degree with the monitored equipment from the working condition life lines of the sample equipment;

and the threshold setting module is used for taking the health value from the selected failure point of the working condition life line to the remaining service life point corresponding to the currently and latest acquired data as a new judgment threshold.

Preferably, the cloud further sets a reporting unit configured to predict and report the performance of the monitored device, where the reporting unit includes a working condition life line fitting module, a device data extraction module, a point correspondence module, a determination module, and a report generation module, where:

the working condition lifeline fitting module is configured to fit a working condition lifeline according to the working condition deviation value corresponding to the characteristic parameter of the equipment of the same type of the monitored equipment;

the device data extraction module is configured to select a monitored device in a database and extract the index data set of the monitored device;

the point correspondence module is configured to perform M-equal segmentation on the time axis of the working condition life line corresponding to the characteristic parameter from normal equipment to complete equipment failure, wherein the state index value of the normal point of the equipment corresponds to a Mahalanobis distance segment value obtained by bringing the expected value into the time axis, and the state index value of the complete equipment failure point corresponds to a Mahalanobis distance segment value obtained by bringing the characteristic parameter into the time axis when the equipment fails;

the determining module is configured to determine a health value and a fault prediction time according to the position of the working condition deviation in the working condition life line;

and the report generation module is configured to carry out prediction report on the performance of the monitored equipment according to the health value and the fault prediction time.

Preferably, the determining module includes:

calculating the health value as:

wherein: h_xIs the equipment health value;

m is the number of sections of the working condition life line;

and establishing a polynomial fitting regression formula based on the Mahalanobis distance of the monitored equipment and the time difference between the current monitoring time and the fault occurrence time:

wherein:ythe time difference between the monitored equipment at the monitoring moment and the fault occurrence time is obtained;

xthe mahalanobis distance of the monitored equipment at the monitoring moment is obtained;

a _n， a _n-1，…，a ₁，a ₀is a fitting coefficient;

Preferably, the edge side further comprises a filtering and correcting unit and a time-frequency processing unit,

the filtering and correcting unit is configured to, before caching: preprocessing the collected running information, wherein the preprocessing comprises filtering and correcting;

the time-frequency processing unit is configured to, after caching and before performing feature extraction on the operation information: and performing characteristic analysis on the operation information by adopting time domain analysis processing and frequency domain analysis processing.

As another aspect of the present invention, the present invention also provides a distributed storage medium, having stored therein a plurality of instructions,

set up in the edge side, is suitable for being loaded and carried out by the processor:

processing and caching the collected running information;

performing feature extraction on the cached operation information, wherein feature parameters obtained by the feature extraction comprise: at least one of amplitude, peak, time of action, frequency content, or the like;

set up in the high in the clouds, be applicable to by the processor loading and carry out:

and the terminal is provided with the following functions and is suitable for being loaded and executed by the processor:

The invention has the beneficial effects that:

according to the equipment state monitoring method based on edge cloud cooperation and the equipment performance monitoring system based on edge cloud cooperation, provided by the invention, the key information in the acquired information in the equipment monitoring requirement is not omitted, the application of too many cloud resources is not occupied, and the effective information density is higher, so that the abnormal positioning and checking can be quickly realized in massive data, the general health of the equipment can be outlined, the details can be further developed, the original field waveform of an abnormal field can be checked, the further analysis can be carried out based on the original field waveform, the current working condition of the monitored equipment can be comprehensively observed and researched, the future working condition can be predicted, and the health state and the working condition evaluation of the whole design can be realized; based on the framework, the edge cloud cooperation is realized, and comprehensive state monitoring is carried out; the edge side carries out new abnormal judgment and original data capture by adopting a judgment threshold value through cloud optimization, the accuracy of equipment state evaluation is improved, and equipment fault prediction is carried out;

the distributed storage medium provided by the invention can be distributed in a device state monitoring mode based on edge cloud cooperation at the edge side, the cloud side and the terminal, so that the device state monitoring can be comprehensively carried out on the basis of a set of hardware platform.

Drawings

Fig. 1 is a flowchart of an apparatus state monitoring method based on edge cloud coordination in embodiment 1 of the present invention;

fig. 2 is a data flow diagram in device status monitoring based on edge cloud coordination in embodiment 1 of the present invention;

fig. 3-1 is a system diagram of an edge side of the device state monitoring method based on edge cloud coordination in embodiment 1 of the present invention;

fig. 3-2 is a schematic flow chart illustrating generation of a data snapshot according to a device state in embodiment 1 of the present invention;

fig. 4-1 is a system diagram of a cloud of the device status monitoring method based on edge cloud coordination in embodiment 1 of the present invention;

FIG. 4-2 is a schematic diagram showing the structure of the event map of FIG. 4-1;

FIG. 5-1 is a diagram illustrating selection of a new decision threshold for a health value;

fig. 5-2 is another flowchart of the device status monitoring method based on edge cloud coordination in embodiment 1 of the present invention;

5-3 are flow diagrams of predictive reporting of performance of the monitored device of FIGS. 5-2;

FIGS. 5-4 are schematic diagrams of health value calculations;

fig. 6-1 to 6-5 are structural diagrams of a device status monitoring system based on edge cloud coordination in embodiment 2 of the present invention;

fig. 6-1 is an overall structure block diagram of an equipment state monitoring system based on edge cloud coordination;

FIG. 6-2 is a block diagram of the data processing unit of FIG. 6-1;

FIG. 6-3 is a block diagram of a data snapshot unit of FIG. 6-1;

FIG. 6-4 is a block diagram of the update decision threshold unit of FIG. 6-1;

FIG. 6-5 is a block diagram of the report unit of FIG. 6-1;

in the drawings, wherein: 1-edge side; 10-equipment; 11-a data acquisition unit; 11' -voltage sensors; 11 "-current sensor; 12-a data processing unit; 121-filtering and correcting module; 122-a time-frequency processing module; 123-a cache module; 13-a feature extraction unit; 14-a data snapshot unit; 141-a threshold acquisition module; 142-preset module; 143-a calculation module; 144-a judgment module; 145-a data snapshot generating unit; 15-an event reporting unit; 2-cloud end; 21-a data arrangement unit; 22-update decision threshold unit; 221-sample device matching module; 222-working condition lifeline matching module; 223-threshold setting module; 23-a reporting unit; 231-working condition lifeline fitting module; 232-device data extraction module; 233-point correspondence module; 234-a determination module; 235-a report generation module; 3-a terminal; 31-a mobile phone; 32-computer.

Detailed Description

In order to make those skilled in the art better understand the technical solution of the present invention, the method for monitoring the device status based on edge cloud coordination, the system for monitoring the device performance based on edge cloud coordination, and the distributed storage medium of the present invention are further described in detail with reference to the accompanying drawings and the detailed description. It should be understood that the detailed description and specific examples, while indicating the preferred embodiment of the invention, are intended for purposes of illustration and explanation only and are not intended to limit the scope of the invention.

It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention. The technical features mentioned in the different embodiments of the invention described below can be combined with each other as long as they do not conflict with each other.

The dilemma of the existing equipment monitoring technology lies in:

on one hand, to ensure the safety of equipment, firstly, a foreboding sign of abnormity is found, and the electrical problems of the motor, such as rotor broken bars, open phases, short circuits and the like, can be identified from the operation information of the equipment, such as the characteristics of current ripple; and identifying power quality problems of the power supply environment, such as potential problems of harmonic waves, three-phase imbalance, voltage sag and the like, and also identifying mechanical faults, such as misalignment, imbalance, bearing faults and the like. In order to establish a complete and comprehensive device operation pattern, the most ideal situation is to collect device operation information, such as current signals, without omission through high-precision and high-resolution signal capture, so that any instantaneous and weak abnormal change can be obtained without omission. On the other hand, even if cloud storage is adopted, resources are limited, and equipment operation information, such as current signals, is collected and stored without omission, so that a huge storage volume is formed inevitably, valuable information is submerged in the storage volume, and the fundamental problem of equipment operation health problem cannot be solved fundamentally.

The starting points of the invention are as follows: according to the rule of the collected equipment information data, under the condition that the density of valuable information of repeated information in the data is low, the collected data is subjected to feature extraction at the equipment end, whether the features are abnormal or not is judged, when the abnormality is judged, an abnormal data snapshot is stored, and the abnormal state of the operation of the equipment is captured and uploaded to the cloud, so that the valuable information is reserved, and the whole data volume is reduced; correspondingly, a data storage and display frame is arranged at the cloud end, so that the characteristics can be acquired from a database and organized in different levels, and massive but clear data support is provided for monitoring the equipment state; furthermore, abnormal positioning and checking can be rapidly realized, the health state and the working condition of the whole equipment can be evaluated, reference can be provided for the operation working condition of the similar equipment, and equipment fault prediction can be carried out.

Example 1:

in view of the above problems, the present embodiment provides a device state monitoring method based on edge cloud coordination, which performs device state monitoring in a coordination manner of edge side calculation, cloud storage, and autonomous learning by capturing operation information of a device including characteristic parameters such as current ripple.

As shown in fig. 1, and referring to the data flow diagram of the device status monitoring method based on edge cloud coordination shown in fig. 2, the device status monitoring method based on edge cloud coordination includes the steps of:

step S1): the method comprises the steps of collecting operation information of at least one piece of equipment, wherein the operation information comprises at least one of current, voltage, vibration, noise, temperature, strain and posture.

In the prior art, a PLC or an ammeter is adopted for collection, usually, a point is obtained through periodic collection, or an average value of a period of time is obtained, information is lost in the modes, instantaneous and weak changes cannot be obtained, and the changes are often precursors of abnormity. Industrial intelligence requires a more sensitive and comprehensive perception in the field, which means that the acquisition part requires a higher acquisition frequency, a higher resolution and a higher accuracy. The signal is acquired without distortion, so that any instantaneous and weak abnormal change can be obtained without omission. Wherein, at least one device is the same type of device or different type of device in the same working space, or the same type of device or different type of device in different working spaces. And a plurality of sensors can be comprehensively applied to one device, and an all-weather continuous acquisition mode is adopted.

Referring to an edge-side system block diagram of the device state monitoring method based on edge cloud coordination shown in fig. 3-1, in this step, state changes and abnormal information of the device 10 are collected and captured at high speed and without omission by sensors (for example, a voltage sensor 11 'and a current sensor 11' in fig. 2), so that comprehensive acquisition of operation information of the field device is realized, and the system also has a capability of capturing device abnormality. The data acquisition of the embodiment focuses on not missing information, and the data adopts a high-speed acquisition rate of not less than 2Msps, for example. It is easy to understand that, in the data acquisition process, in order to ensure the consistency and effectiveness of the data, the problems of channel synchronization, range configuration and the like are also considered, and the details are not described here.

Step S2): and processing and caching the collected running information.

In this step, a certain length of data segment is captured and buffered each time based on a processing mode of a time frame. The data segment should also contain descriptive contents capable of describing the device attributes and the data attributes thereof, and is used for describing the format of the acquired data, such as start time, sampling speed, value type, number of digits, value length, unit and the like.

Preferably, the processing the collected operation information includes:

before caching: the collected operation information is preprocessed, and the preprocessing includes, but is not limited to, filtering, correcting and the like. The collected information usually includes noise and other interferences, and needs to be filtered and other processing, and meanwhile, due to environmental changes and consistency problems of device hardware, the collected data needs to be corrected.

Step S3): and performing feature extraction on the cached operation information, wherein feature parameters obtained by the feature extraction comprise: at least one of amplitude, peak, action time, frequency component, or a combination thereof.

After caching and before extracting the characteristics of the running information, the method further comprises the following steps: and performing characteristic analysis on the operation information by adopting time domain analysis processing and frequency domain analysis processing. The feature analysis is the basis of subsequent feature extraction, and the time domain features such as period, average value, effective value, maximum/minimum value, rising time and the like can be extracted through time domain analysis; the distribution characteristics of different frequency components can be obtained through frequency domain analysis. The length of the data segment which is acquired and cached each time is determined by the length of the data required by the time domain processing or the frequency domain processing. The analysis objects include voltage, current, etc., and the analysis dimensions include amplitude, peak, action time, frequency components, etc.

In practical application, the data is cached after being subjected to filtering processing, correction processing, time domain processing or frequency domain processing, or is cached first and then subjected to filtering processing, correction processing, time domain processing or frequency domain processing, or is subjected to filtering processing, correction processing, time domain processing or frequency domain processing and caching processing alternately according to situations, and the sequence of the filtering processing, the correction processing, the time domain processing and the frequency domain processing is not limited.

In the step, the captured original data without omission is analyzed, various characteristic parameters are extracted, and the calculation method of the characteristic parameters can be updated through remote issuing, can also be judged by a fixed formula based on the relevant mechanism of the operation information, and can also be an inference model through autonomous learning. The feature extraction algorithm can be provided in the form of a micro service, so that version upgrading and replacement can be performed through the cloud. In order to run the cloud-trained model, deep learning model frameworks such as Caffe, PyTorch, TensorFlow, paddlepaddlel and the like can be simultaneously supported, which is not limited herein.

According to the characteristic parameters, whether the running state of the equipment is abnormal or not can be judged, and when the abnormal state is not judged, the characteristic parameter data can be stored and only stored; and when the abnormal data snapshot is judged to occur, capturing and uploading the abnormal state of the equipment operation to the cloud.

The characteristic parameters obtained or calculated after frequency domain processing and the related performance of the characteristic parameters and the abnormity are explained by combining the problems which usually occur in the operation process of the motor.

For example, the aforementioned rotor broken bar generally refers to a phenomenon that a copper bar of a rotor (the rotor comprises two end cast copper rings and a copper bar welded between the two cast copper rings) of a squirrel cage asynchronous motor is cracked or broken, and the detection of the rotor broken bar generally requires disassembling the motor and taking out the rotor by taking out a test resistor. The operation of the motor cannot be influenced in a short time when the rotor is broken, but the stress is uneven, so that the motor vibrates and generates heat, and the performance of the motor is influenced. In this embodiment, the current spectrum may be calculated by measuring the motor current, and the frequency of the current may be used as the characteristic parameter for the determination. When a rotor fault occurs, such as a crack or a broken bar, abnormal characteristic frequency components appear on the two side bands of the power frequency, and the frequency points appear at the electrode passing frequency f_PF。f_PFThe peak value is increased, which indicates that the deterioration condition of the rotor copper bar is increased. By mixing f_PFPeak value and power frequency peak value f_LThe ratio of (2) is considered as a combined characteristic parameter, the larger the ratio is, the more serious the abnormal condition is, when the ratio exceeds a set judgment threshold value, the fault is indicated to have occurred, and the judgment can also be carried out according to an empirical data range of the following table.

TABLE 1 rotor State evaluation Table

For the brushless motor, the problem of phase loss (that is, one phase of the three-phase motor cannot work) and the like often causes that the brushless motor shakes and cannot work, or the brushless motor cannot rotate and has high noise, and even is burnt out. Similarly, taking the measured motor current as an example, and taking the three-phase unbalance degree of the current amplitude as a characteristic parameter for inspection, the ideal situation should be 0, and the larger the characteristic parameter is, the larger the abnormal degree is, and the maximum value is reached when the phase is open.

Parameter formula (I)

Parameter formula (II)

Wherein:PRis the current imbalance;

I _avgis the average value of the current;

I _A、I _B、I _Ca, B, C three-phase current values, respectively.

The too big harmonic component of motor power frequency can lead to the useless power increase of motor to lead to the motor overheated, this embodiment surveys as characteristic parameter with the total harmonic distortion of electric current.

Formula of parameters (III)

Wherein:I _THDis harmonic current distortion;

I ₁is power frequency fundamental current;

I ₂、I ₃、I ₄…I _Ncurrent values of harmonic components of 2 to N orders respectively;

in a rotating electrical machine, due to manufacturing and assembling failure or a system is often in a high-speed and high-load state, a rotor imbalance or an unbalance or an intermediate fault is easily generated, and if the rotor imbalance or the unbalance or the intermediate fault is not diagnosed in time, secondary faults such as rubbing, foundation loosening and the like may be caused, and further, multiple faults are caused. For the prediction of rotor unbalance and misalignment, the present embodiment may still use the motor current as the collection object, the motor current eliminates the power frequency first, that is, performs Root Mean Square (RMS) demodulation, performs frequency analysis on the obtained data, when the misalignment or unbalance occurs, the position of the motor rotation speed will have a larger frequency component, examines the amplitude value at the frequency as a characteristic parameter, the normal condition should be 0, and the larger the characteristic parameter is, the larger the abnormal degree is.

Research shows that when the motor foundation is loosened, a characteristic frequency component appears at a half frequency of the motor rotating speed in a current frequency spectrum, so that an amplitude value at the frequency is considered as a characteristic parameter, the normal condition is 0, and the larger the characteristic parameter is, the larger the abnormal degree is.

For other abnormalities, the present embodiment does not explain the selection criteria of the characteristic parameters or the judgment of the characteristic parameters after time-frequency processing one by one, and can determine the characteristic parameters (at least one of amplitude, peak value, action time, frequency component or combined calculated amount) of the operation information (at least one of current, voltage, vibration, noise, temperature, strain, and attitude) according to the research direction of the application object and the test data, which is not limited herein nor described in detail herein.

Through the above processing, before the data snapshot is generated, in order to simplify the uploaded data and improve the valuable information density, the monitoring method of this embodiment further performs feature detection on the feature parameters, so as to determine whether the device operating state is abnormal. Based on feature detection, generating a data snapshot according to the equipment state by a multivariate regression analysis method based on Mahalanobis distance according to feature parameters, and the method comprises the following steps:

combining the characteristic parameters, the cache information of the collected frames corresponding to the characteristic parameters and the abnormal results to generate a data snapshot comprising a working condition deviation value, and stamping a time stamp on the data snapshot;

and generating a data snapshot for the characteristic parameters judged to be normal, and stamping a time stamp for the data snapshot.

As shown in fig. 3-2, the step of generating a data snapshot according to the device status by using a multivariate regression analysis method based on mahalanobis distance according to the characteristic parameters includes:

step S31): presetting a judgment threshold value, and setting a rated working condition of a certain characteristic parameter or calculating a mean value of the certain characteristic parameter as an expected value of Mahalanobis distance calculation.

In this step, the decision threshold may be set by the cloud, and the decision threshold is learned autonomously by the cloud 2. For example, after a large amount of abnormal data is collected, the abnormal data is obtained through a regression algorithm, such as a bayesian algorithm.

When the rated working condition of a certain characteristic parameter is set or the mean value of the certain characteristic parameter is calculated to be used as the expected value of the Mahalanobis distance calculation, under the condition that the values of all the characteristic parameters of the rated working condition are not known at the beginning of the test, a plurality of groups of data can be measured under the ideal working condition of the machine, and the mean value of all the characteristic parameters is calculated to be used as the mean value of the Mahalanobis distance calculation.

Step S32): and calculating a covariance matrix of the characteristic parameter according to the expected value, wherein at least one of covariance elements forming the covariance matrix is a vector of the characteristic parameter.

In this step, in the process of determining whether the device is abnormal, the system defined by the multi-feature parameters needs to perform abnormal determination on the multi-feature parameters respectively. In this embodiment, when the number of variables is more than two, the correlation between the variables can be measured by the covariance matrix. Let X be a column vector consisting of n random numbers (each of which is also a vector, here a row vector):

formula (1)

The ith, j term of the covariance matrix (the ith, j term is a covariance) is defined as follows:

formula (2)

Wherein the content of the first and second substances,μ _iis the expected value of the ith element, i.e.

。

I.e. the covariance matrix is:

formula (3)

The (i, j) th element in the covariance matrix isX _iAndX _jthe covariance of (a).

Step S33): and calculating to obtain the Mahalanobis distance as a working condition deviation value according to the covariance matrix.

In this step, the mahalanobis distance is calculated from the covariance matrix.

Mahalanobis distance represents the covariance distance of data, and is an effective method for calculating the similarity of two unknown sample sets. Unlike the euclidean distance, it takes into account the link between the various characteristics (e.g. height information brings weight information associated with it) and is scale independent, i.e. independent of the measurement scale. Therefore, in the embodiment, the characteristic parameters can be calculated by combining different characteristic parameters to obtain a plurality of reference dimensions of the same device, so that the multi-angle state evaluation device can be integrated conveniently.

For a mean value of

Multivariate with covariance matrix S

The mahalanobis distance is:

formula (4)

Step S34): judging the equipment state according to the working condition deviation and the judgment threshold, comprising:

if the working condition deviation is greater than the judgment threshold value, judging that the equipment is abnormal;

the data snapshot includes operating condition deviation values.

It should be understood that, the above-mentioned abnormality identification method in this embodiment is only an example, and according to the application field of the actual device, other abnormality identification algorithms may also be adopted to perform flexible and accurate abnormality capture, which is not limited herein.

Step S4): and generating a data snapshot at least comprising the characteristic parameters and the time stamp corresponding to the data information of the extracted characteristic parameters.

In the step, for the characteristic parameters judged to be abnormal, combining the characteristic parameters, the cache information of the collected frames corresponding to the characteristic parameters and the abnormal results to generate a data snapshot comprising a working condition deviation value, and stamping a time stamp on the data snapshot; and generating a data snapshot for the characteristic parameters judged to be normal, and stamping a time stamp for the data snapshot.

The data snapshot is a characteristic data frame structure with certain format and flexible mode. And if the abnormity of the characteristic parameters is captured by the abnormity judgment, merging the corresponding characteristic parameters and the abnormity description of the data in the data cache, merging to generate an abnormal data snapshot, and stamping a time stamp.

The format of the data snapshot includes: the device ID, the device type ID, the device original factory ID, the device user description, the exception record and the feature description, and the exception can further comprise a raw data curve. As shown in table 2:

table 2 data snapshot format

The data snapshot format adopts a json (javascript Object notification) format, which is a lightweight data exchange format, and the specific format is "start", "end". The terms/values are defined between "{", "}", the terms/values are defined between ": connecting, using between a plurality of" terms/values "," separating ", the specific format can be determined according to the requirement, and is not limited herein.

The device type ID in sequence number 2 in table 2 can be set according to table 3.

TABLE 3 device type ID

On the basis of table 3, the equipment can be further subdivided, for example, the "metal cutting machine" can be subdivided into a numerical control cutting machine, a lathe, a drilling machine, a milling machine, a boring machine, a grinding machine, a combined machine, a gear and thread processing machine, a cutting machine and the like; "forging equipment" may be subdivided into numerically controlled forging machines, hammers, casters, rolls, chillers, shears, truers, spring processors, and the like.

In the step, adding corresponding equipment description, description of a sensor for collecting data at the front end and data snapshot, and if the current data frame is found to be abnormal, packaging abnormal related data and field context data, namely the data snapshot; if normal event data, the data snapshot includes only the current feature parameter data, which may be one or more.

Step S5): and transmitting the index data set at least comprising the data snapshot and the device information and the event description corresponding to the data snapshot to a database of the cloud end 2 for storage.

The edge side 1 collects a large amount of high-quality industrial big data, and transmits the industrial big data to the database of the cloud side 2 through a 5G network, WIFI or other networks. In the step, only the data with high information density of the device event is sent to the cloud end 2, and a precious tag data set is continuously provided for the cloud end 2 to perform autonomous learning, so that the artificial intelligence model can be continuously evolved and optimized. By adopting the mode, the data on the edge side 1 can be compressed very efficiently, namely characteristic parameters are considered as objects, a large amount of data in repeated states in actual working conditions are removed, and the field data with state changes are sent back to the cloud end 2 only under the condition of state changes whether normal or abnormal, so that the transmission bandwidth and flow pressure of the state data from the edge side 1 to the cloud end 2 are greatly reduced, and the resource occupation of the cloud end 2 in storage, calculation and other aspects is reduced.

Step S1) -step S4) are processed at the edge side 1, and these processing functions may be integrated into an intelligent sensor (for example, an intelligent monitoring box for edge side equipment shown in fig. 3-1), and the intelligent sensor is disposed or fixed on the equipment during monitoring, so as to perform data acquisition, data processing and analysis. Step S5), the edge 1 and the cloud 2 are connected, and data is uploaded through the network (certainly, data exchange is both sides, and the cloud 2 also issues data to the edge 1, which will be described later).

Step S6): the cloud organizes a plurality of levels of data layout including devices, event maps, and signature waveforms according to the index data set.

Referring to a cloud system block diagram of the edge cloud coordination-based device status monitoring method shown in fig. 4-1, a hierarchy relationship of a working condition life line, an event map and a characteristic waveform is shown.

The working condition life line of the equipment is a curve of the comprehensive health index of the equipment changing along with time, and the current position of the equipment in the whole life cycle can be visually represented. Each operating condition lifeline represents a device, and switching different operating condition lifelines is equivalent to selecting different focal devices (i.e., the selected device).

As shown in fig. 4-2, the event map is a schematic diagram of an event map, where the event map is an abbreviated representation of the device status, the horizontal axis is time, and different areas indicate that the device is in different statuses, for example, horizontal stripes indicate that the current device is in a normal status, and diagonal stripes of other types indicate that the device is in different abnormal statuses, and the device may be classified according to the type of the abnormality or according to the severity level of the abnormality; the different areas can be distinguished by colors, patterns, characters and the like. Specifically, the event maps are state thumbnails of different parts or different dimensions of the currently focused equipment, and for one equipment, for example, there can be an external power supply event map for representing the state of an external power supply, an electrical event map for representing the internal electrical state of a motor, and a mechanical event map for representing the state of transmission and load. The event map can be used for rapidly positioning the time period for displaying the characteristic waveforms, if the interested state change is found in the event map in a certain time period, for example, a rotor broken bar event occurs in an electrical event, the region where the rotor broken bar event occurs in the event map is selected, and at the moment, the characteristic waveforms are also updated to the time period, so that the curve data of each characteristic waveform can be rapidly checked when the broken bar event occurs.

The characteristic waveform curve is a time-varying curve of a certain characteristic under a certain event map, for example, the vibration intensity characteristic of a bearing seat in a mechanical event map, and the curve of the intensity varying with time is drawn by a corresponding characteristic curve.

The cloud end 2 is substantially an intelligent monitoring package of cloud service equipment, wherein a database, namely a database maintained in the cloud service, stores index data sets from a plurality of intelligent edge sides 1 and different equipment, and is a data base for data organization and analysis of the cloud service. And selectively viewing the equipment information or further processing the data and indicating by a user, for example, the characteristic waveform curve corresponding to the selected equipment and the selected abnormity is the original waveform of the site where the abnormity occurs.

In addition, the preset decision threshold mentioned in step S31) may be completed in the cloud 2, and the updated decision threshold is substituted for the original decision threshold. The mode of obtaining the updated judgment threshold value by adopting the cloud 2 for autonomous learning is that the edge side 1 sends the data judged to be abnormal back to the cloud 2, and the cloud 2 can conduct autonomous learning training by means of the big data.

As shown in fig. 5-1, the specific process of the cloud 2 autonomous learning includes:

matching sample devices with the same type as the monitored devices, and acquiring working condition life lines of all finally failed sample devices, such as a curve (i) - (v) shown in fig. 5-1;

selecting a working condition life line with the maximum matching degree with the monitored equipment from the working condition life lines of the sample equipment, namely comparing the currently received characteristic parameters (curves) of the monitored equipment with sample equipment data (curves) with the same type in a database, and regarding one sample equipment with the highest similarity as the working condition life line with the maximum matching degree, wherein pattern recognition algorithms such as a Frechet algorithm, a Hausdorff algorithm and the like can be selected, and the matching curve of the monitored equipment (I) shown in figure 5-1 is a second;

taking the health value (the calculation of the health value will be described in detail in the subsequent health value determination method) at the position from the failure point of the selected working condition Life line to the Remaining service Life (RUL for short) point corresponding to the current latest acquired data as a new judgment threshold, wherein the Remaining service Life can be set according to the importance level of the equipment, and the more important equipment is, the longer the Remaining service Life is set.

Step S7): the data layout corresponding to the indexed data set is displayed at a terminal in communication with the database.

Through a 5G network, WIFI or other networks, the terminal accesses the cloud database, and therefore access, display and manual judgment of equipment information are achieved. On the basis of the device state monitoring method based on edge cloud cooperation, the device safety condition can be further predicted, and the method is used for evaluating the health condition of the device and predicting the failure occurrence time of the device.

That is, as shown in fig. 5-2, the method for monitoring the device status based on edge cloud coordination may further include step 8): and carrying out prediction reporting on the performance of the monitored equipment. As shown in fig. 5-3, the predicting and reporting the performance of the monitored device specifically includes:

step 81): fitting a working condition life line according to the working condition deviation value corresponding to the characteristic parameters of the same type of equipment to be monitored;

step 82): selecting monitored equipment in a database, and extracting an index data set of the monitored equipment;

step 83): performing M equal segmentation on the time axis of the working condition life line corresponding to the characteristic parameter from normal equipment to complete failure of the equipment, wherein the state index value of the normal point of the equipment corresponds to a Mahalanobis distance segment value obtained by bringing the expected value into the time axis, and the state index value of the complete failure point of the equipment corresponds to a Mahalanobis distance segment value obtained by bringing the characteristic parameter into the time axis when the equipment fails;

step 84): determining a health value and fault prediction time according to the position of the working condition deviation in the working condition life line;

step 85): and predicting and reporting the performance of the monitored equipment according to the health value and the fault prediction time.

As shown in fig. 5 to 4, in the step of determining the health value, since the mahalanobis distance does not necessarily have a linear relationship with the speed of equipment degradation, in order to avoid errors, the mahalanobis distance is used as a reference for health evaluation and evaluated in combination with a hierarchical evaluation method. Firstly, the normal working condition to the complete failure are divided into a plurality of stages, for example, M stages, the health value corresponding to the normal working condition is 100, the complete failure is defined as the health value is 0, and the width of the health interval in each stage is 100/M. The division points of each segment correspond to values D0, D1, D2, D3 … Dm-1, Dm. D0 is the value obtained by taking the mean value in the calculation of Mahalanobis distance, corresponding to 100 points, and M is the value between 1 and M. The larger M is, the finer the grading is, the finer the calculation result is, and the calculation result can be flexibly set according to requirements in practical application.

According to the position of the working condition deviation value falling into the time axis of the working condition life line, if the current Mahalanobis distance falls into the interval of [ Dm-1, Dm ], calculating the health value as follows:

formula (5)

Wherein H_xIs the equipment health value;

m is the number of sections of the working condition life line;

D_mis the deviation value of the end working condition of the falling section.

The failure occurrence time is obtained by calculating failure prediction time, and in the step of determining the failure prediction time, the method specifically comprises the following steps: and based on the Mahalanobis distance, predicting the fault occurrence time by adopting a regression algorithm. The time to failure prediction based on mahalanobis distance is obtained by a regression algorithm, for example, a polynomial regression algorithm.

Historical health reference line of the same type of equipment passing through the monitored equipment is composed of a plurality of points (x)₀,y₀),（x₁，y₁）,…(_xM-1And 0) fitting, wherein y is the time difference between the monitored equipment at the monitoring moment and the fault occurrence time, and x is the mahalanobis distance between the monitored equipment at the monitoring moment, so that the expression of the fault prediction time function is obtained as follows:

formula (6)

Wherein:a _n，a _n-1，…a ₁，a ₀are fitting coefficients.

And when the fault prediction time needs to be calculated, substituting the Mahalanobis distance of the current parameters calculated by the formula (4) into the formula (6), and obtaining the fault prediction time of the equipment. And predicting and reporting the performance of the monitored equipment according to the fault prediction time of the equipment, thereby realizing the performance or fault early warning of the equipment.

It should be understood here that the above-mentioned method is consistent in the processing manner mentioned in the different steps for the process of fitting the condition life line according to the condition deviation value corresponding to the characteristic parameter of the equipment. The working condition life line is a curve which is continuously updated and increased during the operation of the equipment (as long as the equipment is not completely failed), and the working condition life lines obtained by fitting different steps of the same equipment at the same time point are the same.

In the device state monitoring method based on edge cloud coordination in the embodiment, the time axis is used as a reference for expressing the event, the feature and the healthy device information, so that the comprehensive state of each time slice of the whole life cycle of the corresponding device can be stored and checked, and continuous use analysis and design optimization can be performed on the comprehensive data of a certain device in the future.

As can be seen, in the present embodiment, the edge side 1 has the following functions: high-speed acquisition, caching and preprocessing functions; the time domain analysis and spectrum analysis functions have a characteristic value extraction function, and further have the functions of anomaly detection and event reporting. On the basis of each function of the edge side 1, the function of the cloud 2 forms an equipment health management function: the full life cycle state database of the equipment can be constructed, working condition life line display and characteristic parameter change trend prediction are realized, and abnormal event management and health index indication are completed.

According to the embodiment, a side cloud cooperative architecture is constructed, so that the key information in the acquired information in the equipment monitoring requirement is not missed, and the application of too many resources of the cloud end 2 is not occupied. And at the edge side 1, performing feature extraction on the acquired data, and judging whether the feature parameters are abnormal or not. When an abnormality occurs, storing an abnormal data snapshot, capturing and uploading complete state data of the equipment abnormality only to the cloud 2, and recording characteristic parameters only by other non-abnormal data; at the cloud 2, data can be acquired from a database to check different devices or different feature data of the same device, effective information density is improved, and the data is organized in different levels of device-feature waveform-event map-comprehensive health, so that abnormal positioning and checking can be quickly realized in massive data, the general health of the device can be outlined, details can be deepened, an original field waveform of an abnormal field can be checked, further analysis can be carried out based on the original field waveform, the current working condition or the future working condition of the monitored device can be comprehensively known and researched, the health state and the working condition evaluation of the whole design can be realized, and the device fault prediction can be carried out.

Edge cloud cooperation is realized based on the framework: the edge side 1 only continuously sends the original data including the abnormity back to the cloud end 2, the cloud end 2 carries out continuous big data autonomous learning to obtain a judgment threshold value for judging the abnormity and sends the judgment threshold value to the edge side 1 at any time, so that the effective information density can be improved, the abnormity identification strategy can be corrected at any time, the abnormity of multiple states can be monitored for the same equipment, and comprehensive state monitoring can be carried out on the basis of a set of hardware platform; the edge side 1 performs new abnormal judgment and original data capture by optimizing a judgment threshold value through the cloud end 2, so that the judgment threshold value of the edge side 1 is more and more accurate finally, the accuracy of equipment state evaluation is higher and higher, and equipment fault prediction is performed.

Example 2:

the industrial artificial intelligence needs big data to carry out autonomous learning, however, if all the data are collected and sent to the cloud big data center through the industrial internet of things, the data volume is too large, and massive data are huge burdens on transmission, storage and calculation of the cloud. If the result after simple calculation is sent to the cloud, the data volume is reduced, but the original data of a specific event does not exist, so that sample data for autonomous learning cannot be formed.

The embodiment provides a monitoring system of an equipment state monitoring method based on edge cloud coordination, which corresponds to the embodiment.

As shown in fig. 6-1, an apparatus status monitoring system based on edge cloud coordination includes an edge side 1, a cloud side 2, and a terminal 3, where the terminal 3 may be, for example, a mobile phone 31 or a computer 32.

The edge side 1 includes a data acquisition unit 11, a data processing unit 12, a feature extraction unit 13, a data snapshot unit 14, and an event reporting unit 15, wherein:

the data acquisition unit 11 is configured to acquire operation information of at least one device, where the operation information includes at least one of current, voltage, vibration, noise, temperature, strain, and attitude. Wherein, at least one device is the same type of device or different type of device in the same working space, or the same type of device or different type of device in different working spaces. The data acquisition unit 11 acquires device status data, which often comes from sensors, such as a current sensor 11', a voltage sensor 11 ", a vibration sensor, and the like, and further performs signal conditioning, analog-to-digital conversion, and the like.

As shown in fig. 6-2, the data processing unit 12 of the edge side 1 stores the processed data, and includes a filtering module 121 and a time-frequency processing module 122 in addition to the buffer module 123.

And the data processing unit 12 is configured to process and cache the acquired running information. Wherein:

the filtering and correcting module 121, located before the buffering unit 123, is configured to, before performing buffering: preprocessing the collected operation information, wherein the preprocessing comprises filtering and correcting;

the time-frequency processing module 122, located behind the cache unit 123, is configured to, after caching and before performing feature extraction on the operation information: and performing characteristic analysis on the operation information by adopting time domain analysis processing and frequency domain analysis processing.

The buffer module 123 is configured to buffer data, where the data may be pre-processed data or post-processed data. Therefore, the connection relationship or the position relationship of the buffer module 123 with respect to the filtering module 121 and the time-frequency processing module 122 is not limited, and fig. 6-2 is only an example.

A feature extraction unit 13 configured to perform feature extraction on the cached operation information, where feature parameters obtained by the feature extraction include: at least one of amplitude, peak, action time, frequency component, or a combination thereof. The feature extraction unit 13 calculates feature parameters according to the raw data from the data acquisition unit 11 and the data processed by the data processing unit 12, and a group of data may be calculated to obtain a plurality of feature parameters.

And the data snapshot unit 14 is configured to generate a data snapshot at least including the characteristic parameters and the time stamp of the data information corresponding to the extracted characteristic parameters. The data snapshot unit 14 makes an exception decision on the characteristic parameters, if an exception condition is satisfied, the data segment is determined to be an abnormal data segment, the abnormal data segment is stored in the abnormal data snapshot, the data snapshot includes the original data, the characteristic parameters and the description of the exception, and a time stamp is printed on the abnormal data snapshot to indicate the occurrence time point of the abnormal data snapshot.

As shown in fig. 6-3, the data snapshot unit 14 includes a determining module 144, and the determining module 144 is configured to generate the data snapshot according to the device status by using a multiple regression analysis method based on mahalanobis distance according to the characteristic parameters, including:

merging the characteristic parameters, the cache information of the collected frames corresponding to the characteristic parameters and the abnormal results to generate data snapshots, and stamping time stamps on the data snapshots for the characteristic parameters judged to be abnormal;

The data snapshot unit 14 further includes a threshold value obtaining module 141, a preset module 142, and a calculating module 143, where:

a threshold obtaining module 141 configured to obtain a preset decision threshold;

the preset module 142 is configured to set a rated working condition of a certain characteristic parameter or calculate a mean value of the certain characteristic parameter as an expected value of mahalanobis distance calculation;

a calculation module 143 configured to calculate a covariance matrix of the feature parameter according to the expected value, wherein at least one of covariance elements forming the covariance matrix is a vector of the feature parameter; wherein the covariance matrix formula is:

wherein:μ _iis the expected value of the ith element;

the (i, j) th element in the matrix isX _iAndX _jthe covariance of (a).

And further, according to the covariance matrix, calculating to obtain the Mahalanobis distance as a working condition deviation value. The mahalanobis distance formula is:

wherein:

is an average value;

Sa covariance matrix that is multivariate.

Thus, the determining module 144 is configured to determine the device status according to the operating condition deviation and the decision threshold, including:

the data snapshot includes operating condition deviation values.

The event reporting unit 15 is configured to transmit the index data set at least including the data snapshot and the device information and the event description corresponding to the data snapshot to the database of the cloud 2 and store the index data set.

As shown in fig. 6-4, the cloud 2 includes a data layout unit 21, and the data layout unit 21 is configured to organize data layouts including multiple levels of devices, event maps, and feature waveforms according to the index data set. Preferably, the cloud 2 is further provided with an update decision threshold unit 22, and the update decision threshold unit 22 includes a sample device matching module 221, a working condition life line matching module 222, and a threshold setting module 223, where:

the sample equipment matching module 221 is used for matching sample equipment of which the type is the same as that of the monitored equipment and acquiring working condition life lines of all finally failed sample equipment;

the working condition life line matching module 222 is used for selecting a working condition life line with the maximum matching degree with the monitored equipment from the working condition life lines of the sample equipment;

and the threshold setting module 223 is configured to use the health value from the failure point of the selected working condition life line to the remaining service life point corresponding to the currently latest acquired data as a new decision threshold.

And the terminal 3 is configured to communicate with the database of the cloud 2 and display data arrangement corresponding to the index data set. Preferably, as shown in fig. 6-5, the cloud 2 further includes a reporting unit 23 configured to perform prediction reporting on the performance of the monitored device, where the reporting unit 23 includes an operating condition life line fitting module 231, a device data extracting module 232, a point correspondence module 233, a determining module 234, and a report generating module 235. In this way, the terminal 3 that establishes communication with the cloud 2 also has a report display function.

Wherein: and (3) according to the position of the time axis of the working condition life line of the working condition deviation value:

the health value was calculated as:

wherein: h_xIs the equipment health value;

m is the number of sections of the working condition life line;

a _n， a _n-1，…，a ₁，a ₀is a fitting coefficient;

A working condition lifeline fitting module 231 configured to fit a working condition lifeline according to a working condition deviation value corresponding to a characteristic parameter of the same type of equipment to be monitored;

a device data extraction module 232 configured to select a monitored device in the database and extract an index dataset of the monitored device;

the point correspondence module 233 is configured to perform M-grade segmentation on the time axis of the working condition life line corresponding to the characteristic parameter from normal equipment to complete equipment failure, where the state index value of the normal point of the equipment corresponds to a mahalanobis distance segment value obtained by bringing the state index value into an expected value, and the state index value of the complete equipment failure point corresponds to a mahalanobis distance segment value obtained by bringing the characteristic parameter into the failure of the equipment;

a determination module 234 configured to determine a health value and a fault prediction time based on a location of the condition deviation in the condition lifeline; wherein:

and the report generating module 235 is configured to perform prediction reporting on the performance of the monitored equipment according to the health value and the failure prediction time.

By applying the device state monitoring system based on the edge cloud cooperation of the embodiment, omission of any abnormal information of the device can be avoided, and too many resources of the cloud end 2 are not occupied. At the edge side 1, the collected data is subjected to feature extraction, whether feature parameters are abnormal or not is judged,

and when the abnormal data is judged to be abnormal, the abnormal data snapshot is stored, only the abnormal complete state data of the equipment is captured and uploaded to the cloud end 2, and only the characteristic parameters of other non-abnormal data are recorded. At the cloud 2, data can be acquired from a database to check different devices or different feature data of the same device, and the data is organized in different levels of device-feature waveform-event map-comprehensive health, so that the abnormal location and checking can be quickly realized in massive data, the general health of the device can be observed, details can be deepened, the original field waveform of an abnormal field can be checked, further analysis can be carried out based on the original field waveform, the current working condition of the monitored device can be comprehensively observed and researched, the health state and working condition evaluation of the whole design can be realized, and the device fault prediction can be carried out.

Edge cloud cooperation is realized based on the framework: the edge side 1 continuously sends the original data including the abnormity back to the cloud end 2, the cloud end 2 carries out continuous big data autonomous learning to obtain a judgment threshold value for judging the abnormity and sends the judgment threshold value to the edge side 1 at any time, so that the abnormity identification strategy can be corrected at any time, the abnormity of multiple states can be monitored for the same equipment, and comprehensive state monitoring is carried out on the basis of a set of hardware platform; the edge side 1 performs new abnormal judgment and original data capture by optimizing the judgment threshold value, so that the judgment threshold value of the edge side 1 is more and more accurate finally, and the accuracy of equipment state evaluation is more and more high.

Example 3:

the present embodiment provides a distributed storage medium, in which a plurality of instructions are stored, which can be set in different location spaces or areas according to different processing functions, including:

collecting operation information of at least one device, wherein the operation information comprises at least one of current, voltage, vibration, noise, temperature, strain and posture; wherein, at least one device is the same type of device or different type of device in the same working space, or the same type of device or different type of device in different working spaces;

processing and caching the collected running information;

and performing feature extraction on the cached operation information, wherein feature parameters obtained by the feature extraction comprise: at least one or a combination of amplitude, peak value, action time and frequency component;

generating a data snapshot at least comprising the characteristic parameters and the time stamps of the data information corresponding to the extracted characteristic parameters;

transmitting an index data set at least comprising a data snapshot, and device information and event description corresponding to the data snapshot to a cloud database and storing the index data set;

the cloud end organizes data arrangement of multiple levels including equipment, an event map and a characteristic waveform according to the index data set;

the data layout corresponding to the indexed data set is displayed at a terminal in communication with the database.

The storage medium shown in this embodiment may be a hard disk or a storage unit of a control system, and a computer program (i.e., a program product) is stored on the storage medium, and when the computer program is executed by a processor, according to different settings and locations, the steps described in the foregoing method embodiments are implemented, for example, the computer program is set in a cloud and is adapted to be loaded and executed by the processor: the cloud organizes a plurality of levels of data layout including devices, event maps, and signature waveforms according to the index data set. The specific implementation of each step is not repeated here.

It should be noted that examples of the storage medium may also include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory, or other optical and magnetic storage media, which are not described in detail herein.

The distributed storage medium provided in this embodiment stores the implementation program of the device state monitoring method based on edge cloud coordination provided in embodiment 1, and can implement the running and updating of control software logic in a device state monitoring manner based on edge cloud coordination, in a distribution manner of an edge side, a cloud side and a terminal, in cooperation with a sensor, a communication network and the like, thereby performing comprehensive device state monitoring on the basis of a set of hardware platform.

It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The specific embodiments are specific examples of implementing the technical solutions of the present invention. Also, the term "comprises/comprising" when used herein refers to the presence of a feature, integer or component, but does not preclude the presence or addition of one or more other features, integers or components.

It will be understood that the above embodiments are merely exemplary embodiments taken to illustrate the principles of the present invention, which is not limited thereto. It will be apparent to those skilled in the art that various modifications and improvements can be made without departing from the spirit and substance of the invention, and these modifications and improvements are also considered to be within the scope of the invention.

Claims

1. A device state monitoring method based on edge cloud cooperation is characterized by comprising the following steps:

processing and caching the collected running information;

2. The device status monitoring method based on edge cloud coordination according to claim 1, further comprising, before generating the data snapshot: generating the data snapshot according to the equipment state by a multivariate regression analysis method based on the Mahalanobis distance according to the characteristic parameters, wherein the method comprises the following steps:

wherein:

for the characteristic parameters judged to be abnormal, merging the characteristic parameters, the cache information of the collected frames corresponding to the characteristic parameters and the abnormal results to generate a data snapshot comprising the working condition deviation value, and stamping a time stamp on the data snapshot;

3. The device state monitoring method based on edge cloud coordination according to claim 2, wherein the preset decision threshold is:

4. The method for monitoring the state of the equipment based on the edge cloud coordination as claimed in claim 2, further comprising the step of predicting and reporting the performance of the monitored equipment:

5. The device status monitoring method based on edge cloud coordination according to claim 4, wherein in the step of determining the health value, the method comprises:

calculating the health value as:

wherein:

H_xis the equipment health value;

m is the number of sections of the working condition life line;

wherein:

ythe time difference between the monitored equipment and the fault occurrence time at the monitoring moment;

xmahalanobis distance of the monitored device at the monitoring time;

a _n， a _n-1，…，a ₁，a ₀is a fitting coefficient;

6. The utility model provides an equipment state monitoring system based on edge cloud is cooperative, which characterized in that, includes edge side, high in the clouds and terminal, the edge side includes data acquisition unit, data processing unit, feature extraction unit, data snapshot unit, incident report unit, wherein:

7. The edge cloud coordination-based equipment state monitoring system according to claim 6, wherein the data snapshot unit comprises a threshold acquisition module, a preset module, a calculation module, a judgment module, and a data snapshot generation module, wherein:

the data snapshot generating module is configured to combine the characteristic parameters, the cache information of the collected frames corresponding to the characteristic parameters and the abnormal results of the characteristic parameters which are judged to be abnormal to generate a data snapshot comprising the working condition deviation value, and stamp the data snapshot with a time stamp;

8. The device state monitoring system based on edge cloud coordination according to claim 7, wherein the cloud is further provided with an update decision threshold unit, and the update decision threshold unit includes a sample device matching module, a working condition life line matching module, and a threshold setting module, wherein:

9. The edge cloud coordination-based device state monitoring system according to claim 7, wherein the cloud further provides a reporting unit configured to perform prediction reporting on the performance of the monitored device, and the reporting unit includes a working condition life line fitting module, a device data extraction module, a point correspondence module, a determination module, and a report generation module, wherein:

10. A distributed storage medium having a plurality of instructions stored therein,

processing and caching the collected running information;