CN113433913B - System monitoring model generation and monitoring method, processor chip and industrial system - Google Patents
System monitoring model generation and monitoring method, processor chip and industrial system Download PDFInfo
- Publication number
- CN113433913B CN113433913B CN202110761540.1A CN202110761540A CN113433913B CN 113433913 B CN113433913 B CN 113433913B CN 202110761540 A CN202110761540 A CN 202110761540A CN 113433913 B CN113433913 B CN 113433913B
- Authority
- CN
- China
- Prior art keywords
- model
- monitored
- longicorn
- measurement
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 76
- 238000012544 monitoring process Methods 0.000 title claims abstract description 58
- 238000005259 measurement Methods 0.000 claims abstract description 115
- 241001481710 Cerambycidae Species 0.000 claims description 106
- 238000010845 search algorithm Methods 0.000 claims description 33
- 239000011159 matrix material Substances 0.000 claims description 31
- 230000002159 abnormal effect Effects 0.000 claims description 27
- 238000001514 detection method Methods 0.000 claims description 9
- 239000000126 substance Substances 0.000 claims description 6
- 238000012545 processing Methods 0.000 abstract description 8
- 238000004364 calculation method Methods 0.000 abstract description 7
- 230000006870 function Effects 0.000 description 13
- 238000010586 diagram Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 241000027431 Anoplophora Species 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 241001609692 Anoplophora chinensis Species 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
- G05B19/41875—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by quality surveillance of production
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32252—Scheduling production, machining, job shop
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Manufacturing & Machinery (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The embodiment of the invention provides a system monitoring model generation and monitoring method, a processor chip and an industrial system, and relates to the technical field of monitoring. The system monitoring model generation method comprises the following steps: constructing a measurement model and a discrimination model of the system to be monitored based on a data change model of the system to be monitored; obtaining the value of the measurement parameter in the measurement model according to the reference historical data of the system to be monitored; and obtaining the value of the discrimination parameter in the discrimination model based on the reference historical data and the value of the measurement parameter. In the invention, when the measurement model and the discrimination model of the system to be monitored are generated, the calculation force requirement is lower, and the method is suitable for being deployed in a processing chip with lower processing capacity, thereby reducing the cost and improving the deployment efficiency.
Description
Technical Field
The invention relates to the technical field of monitoring, in particular to a system monitoring model generation and monitoring method, a processor chip and an industrial system.
Background
In an industrial equipment system, input and output signals of industrial equipment of the system can be collected through a sensor generally. In order to monitor the running state of equipment in the industrial equipment system, abnormal data monitoring is carried out on output signals after the output signals in the industrial equipment system are obtained; the common data anomaly monitoring method comprises the following steps: the traditional methods such as a threshold value method and a box diagram also include machine learning algorithms such as clustering, isolated forest, random forest and neural network.
However, in the conventional data anomaly monitoring method, the current data read by the sensor is affected by the historical data of the system, so that the conventional data anomaly monitoring method is low in accuracy and high in misdiagnosis rate. Although the machine learning algorithm is high in precision and low in misdiagnosis rate, a large amount of early-stage data accumulation is needed, the operation computational power consumption is overlarge, and the problems of low deployment efficiency, high cost and the like exist in the actual scene application.
Disclosure of Invention
The invention aims to provide a system monitoring model generation and monitoring method, a processor chip and an industrial system, which can construct a measurement model and a discrimination model of a system to be monitored based on historical reference data of the system to be monitored, integrate the measurement and discrimination processes of the system to be monitored and achieve better detection precision; in addition, only a small amount of historical reference data of the system to be monitored is needed, the calculation force requirement is low, the method is suitable for being deployed in a processing chip with low processing capacity, the cost is reduced, and the deployment efficiency is improved.
In order to achieve the above object, the present invention provides a system monitoring model generation method, including: constructing a measurement model and a discrimination model of the system to be monitored based on a data change model of the system to be monitored; obtaining the value of the measurement parameter in the measurement model according to the reference historical data of the system to be monitored; and obtaining the value of the discrimination parameter in the discrimination model based on the reference historical data and the value of the measurement parameter.
The invention also provides a system monitoring method, which predicts the output predicted value of the system to be monitored based on the measurement model generated by the system monitoring model generation method and the output measured value measured by the sensor in the system to be monitored; determining whether the output predicted value is abnormal or not based on a discrimination model generated by the system monitoring model generation method; and if the output predicted value is determined to be abnormal, giving an alarm.
The invention also provides a processor chip, and the processor chip is used for the system monitoring model generation method and/or the system monitoring method.
The invention also provides an industrial system which comprises the processor chip, industrial equipment connected with the processor chip and a sensor, wherein the sensor is respectively connected with the processor chip and the industrial equipment.
In the embodiment of the invention, the measurement model and the discrimination model of the system to be monitored can be constructed based on the historical reference data of the system to be monitored, and the measurement and discrimination processes of the system to be monitored are integrated, so that better detection precision can be achieved; in addition, only a small amount of historical reference data of the system to be monitored is needed, the calculation force requirement is low, the method is suitable for being deployed in a processing chip with low processing capacity, the cost is reduced, and the deployment efficiency is improved.
In one embodiment, the constructing a measurement model and a discrimination model of the system to be monitored based on the data change model of the system to be monitored includes:
obtaining a regression model of a system to be monitored based on a data change model of the system to be monitored;
obtaining a state matrix updating equation of the system to be monitored according to the data change model and the regression model;
and updating an equation according to the state matrix, and constructing a measurement model and a discrimination model of the system to be monitored.
In one embodiment, the updating the equation according to the state matrix to construct the measurement model and the discrimination model of the system to be monitored includes:
based on the state matrix updating equation and the precision parameters of the sensors in the system to be monitored, constructing a measurement model of the system to be monitored;
and updating an equation and the constructed measurement model based on the state matrix to generate a discrimination model of the system to be monitored.
In one embodiment, the expression of the data change model is:
wherein, ω is i 、ω i ' and ω 0 Representing a measurement parameter in the data change model, x (n) representing an output measurement value measured by a sensor in the system to be monitored at n moments, r (n-i) representing an input value of the system to be monitored at n-i moments, and l representing a preset sliding window length;
the expression of the regression model is as follows:
wherein, ω is i 、ω i ' and ω 0 Representing the measured parameters in the regression model,representing the output predicted value of the regression model at the moment n, r (n-i) representing the input value of the system to be monitored at the moment n-i, x (n) representing the output measured value measured by a sensor in the system to be monitored at the moment n, and l representing the preset sliding window length;
the state matrix updating equation of the system to be monitored is as follows:
S(n+1)=(1-∈)*S(n)+∈*V 2 (n);
wherein S (n) represents a state matrix of the system to be monitored at n times, e represents an update coefficient of the state matrix, and V (n) represents a residual error between the data change model and the regression model at n times.
In one embodiment, the expression of the measurement model is:
y(n)=C*Z(n)+D*u(n)+V′(n);
wherein y (n) represents the real output value of the system to be monitored at the moment n, V′(n)~N(0,S′(n)),/>C=[1 0 ... 0],D=[0 0 ... 0],/> v '(n) represents a measurement error of a sensor in the system to be monitored, S' (n) represents an accuracy of the sensor in the system to be monitored, and S (n) represents a state matrix of the system to be monitored at n times.
In one embodiment, the expression of the discriminant model is:
wherein e is n Representing the real error of the system to be monitored at the moment n, k0 representing the discrimination parameter in the discrimination model,C=[1 0 ...]s' (n) denotes the accuracy of the sensor in the system to be monitored, based on>And S (n) represents a state matrix of the system to be monitored at the moment n.
In one embodiment, the obtaining the value of the measurement parameter in the measurement model according to the reference historical data of the system to be monitored includes:
and according to the reference historical data of the system to be monitored, recursively solving to obtain the value of the measurement parameter in the regression model as the value of the measurement parameter in the measurement model.
In one embodiment, the recursively solving the values of the measurement parameters in the regression model according to the reference historical data of the system to be monitored includes:
in each iteration process, acquiring an estimation error of the regression model and an overall system error based on historical reference data of current iteration time, and judging whether the iteration meets a preset iteration condition;
if the iteration meets the preset iteration condition, outputting the value of the current measurement parameter of the regression model;
and if the iteration does not meet the preset iteration condition, updating the iteration time and performing iteration.
In one embodiment, the obtaining the discriminant parameter in the discriminant model based on the reference historical data and the measured parameter includes:
and setting the discrimination parameters in the discrimination model by adopting a longicorn whisker search algorithm based on the reference historical data and the measurement parameters.
In one embodiment, the tuning discrimination parameters in the discrimination model by using a longicorn whisker search algorithm based on the reference historical data and the measurement parameters includes:
constructing the position of the longicorn by using the measurement parameters and the discrimination parameters in the discrimination model, setting a fitness function of the longicorn stigma search algorithm, and initializing the longicorn;
obtaining the fitness of the longicorn at the current position based on the historical reference data, the current position of the longicorn and the fitness function;
calculating the fitness of the left antenna and the right antenna of the longicorn, and updating the position of the longicorn;
updating the step length, the beard length and the iteration times of the longicorn, and judging whether preset iteration conditions are met;
if the preset iteration condition is met, terminating the longicorn stigma search algorithm, and outputting the value of each discrimination parameter based on the current longicorn position;
and if the preset iteration condition is not met, substituting the updated position of the longicorn into the fitness function, and repeating the longicorn stigma search algorithm.
In one embodiment, the fitness function of the longicorn whisker search algorithm is:
wherein the content of the first and second substances,representing the current position of the longicorn, R representing the recall ratio of the abnormal data points of the longicorn stigma search algorithm in the iteration, and N representing the false detection ratio of the abnormal data points of the longicorn stigma search algorithm in the iteration;
the position updating formula of the longicorn is as follows:
Drawings
Fig. 1 and fig. 2 are specific flowcharts of a system monitoring model generation method according to a first embodiment of the present invention;
FIG. 3 is a detailed flow chart of step 2013 of the system monitoring model generation method of FIG. 2;
FIG. 4 is a detailed flow chart of step 203 of the system monitoring model generation method of FIG. 2;
FIG. 5 is a schematic diagram of tuning a discriminative parameter in a discriminative model using a longicorn whisker search algorithm according to a first embodiment of the invention;
FIG. 6 is a schematic diagram of the discrimination of the first embodiment of the present invention using the Tianniu whisker search algorithm in an iterative process;
FIG. 7 is a detailed flow chart of a system monitoring method according to a second embodiment of the invention;
fig. 8 is a schematic diagram of an industrial system according to a fourth embodiment of the invention.
Detailed Description
The embodiments of the present invention will be described in detail below with reference to the accompanying drawings in order to more clearly understand the objects, features and advantages of the present invention. It should be understood that the embodiments shown in the drawings are not intended to limit the scope of the present invention, but are merely intended to illustrate the essential spirit of the technical solution of the present invention.
In the following description, for the purposes of illustrating various disclosed embodiments, certain specific details are set forth in order to provide a thorough understanding of the various disclosed embodiments. One skilled in the relevant art will recognize, however, that the embodiments may be practiced without one or more of the specific details. In other instances, well-known devices, structures and techniques associated with this application may not be shown or described in detail to avoid unnecessarily obscuring the description of the embodiments.
Throughout the specification and claims, the word "comprise" and variations thereof, such as "comprises" and "comprising," are to be understood as an open, inclusive meaning, i.e., as being interpreted to mean "including, but not limited to," unless the context requires otherwise.
Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
As used in this specification and the appended claims, the singular forms "a", "an", and "the" include plural referents unless the context clearly dictates otherwise. It should be noted that the term "or" is generally employed in its sense including "or/and" unless the context clearly dictates otherwise.
In the following description, for the purposes of clearly illustrating the structure and operation of the present invention, directional terms will be used, but terms such as "front", "rear", "left", "right", "outer", "inner", "outer", "inward", "upper", "lower", etc. should be construed as words of convenience and should not be construed as limiting terms.
The first embodiment of the invention relates to a system monitoring model generation method, which is applied to a processor chip, wherein the processor chip can execute the system monitoring model generation method in the embodiment to generate a measurement model and a discrimination model of a system to be monitored, the measurement model can be used for acquiring an output predicted value of the system to be monitored, and the discrimination model can be used for judging whether the output predicted value of the system to be monitored is abnormal or not, so that the real-time monitoring of the system to be monitored is realized; the system to be monitored can be an industrial system, such as a transformer system, a wind power system and the like.
Referring to fig. 1, the method for generating a system monitoring model of the present embodiment includes: step 101, constructing a measurement model and a discrimination model of a system to be monitored based on a data change model of the system to be monitored; 102, obtaining the value of a measurement parameter in a measurement model according to reference historical data of a system to be monitored; and 103, obtaining the value of the discrimination parameter in the discrimination model based on the reference historical data and the value of the measurement parameter.
By adopting the system monitoring model generation method in the embodiment, the measurement model and the discrimination model of the system to be monitored can be constructed based on the historical reference data of the system to be monitored, the measurement and discrimination processes of the system to be monitored are integrated, and better detection precision can be achieved; in addition, only a small amount of historical reference data of the system to be monitored is needed, the calculation force requirement is low, the method is suitable for being deployed in a processing chip with low processing capacity, the cost is reduced, and the deployment efficiency is improved.
The following detailed description, given with reference to specific mathematical models of the system to be monitored, is only an exemplary illustration and is not necessary to implement the present invention.
Fig. 2 shows a specific flow of the system monitoring model generation method according to the present embodiment.
in the substep 2011, a regression model of the system to be monitored is obtained based on the data change model of the system to be monitored.
The expression of the data change model is:
wherein, ω is i 、ω i ' and ω 0 The measurement parameters in the data change model are represented, x (n) represents output measurement values measured by a sensor in the system to be monitored at n moments, r (n-i) represents input values of the system to be monitored at n-i moments, and l represents a preset sliding window length. Wherein, ω is 0 For the bias of the data change model, taking the system to be monitored as a transformer system as an example, the input r (n) of the transformer system is the voltage input into the transformer system, and the output x (n) of the transformer system is the voltage output by the transformer system measured by the sensor in the transformer system.
Based on equation (1) of the above data change model, an expression of the regression model can be obtained:
wherein, ω is i 、ω i ' and ω 0 Representing the measured parameters in the regression model,representing the output predicted value of the regression model at the time n, r (n-i) representing the input value of the system to be monitored at the time n-i, x (n) representing the output measured value measured by a sensor in the system to be monitored at the time n, l representing the preset sliding window length, and omega i 、ω i ' and ω 0 Namely the parameters to be solved in the regression model.
And a substep 2012, obtaining a state matrix of the system to be monitored according to the data change model and the regression model.
Based on the above equation (1) of the data change model and equation (2) of the regression model, it is possible to obtain:
where V (n) represents the residual between the data change model and the regression model at time n.
Based on the above equations (2) and (3), it is possible to obtain:
based on the above equations (1), (2) and (3), the state matrix update equation of the system to be monitored can be obtained:
S(n+1)=(1-∈)*S(n)+∈*V 2 (n) formula (5)
Wherein S (n) represents a state matrix of the system to be monitored at n moments, and epsilon represents an updating coefficient of the state matrix.
From the above equation (5), V (N) follows a normal distribution with a mean value of 0 and a standard deviation of S (N), V (N) to N (0, S (N)).
Based on the above equation (4) and the matrix update equation (5), it can be obtained:
and a substep 2013 of constructing a measurement model and a discrimination model of the system to be monitored according to the state matrix.
Referring to fig. 3, sub-step 2013 includes the following sub-steps:
and a substep 20131, constructing a measurement model of the system to be monitored based on the state matrix updating equation and the precision parameters of the sensor in the system to be monitored.
Specifically, the relationship between the real output value of the system to be monitored and the measurement value of the sensor can be expressed as:
y (n) = x (n) + V' (n) formula (7)
Wherein y (n) represents the real output value of the system to be monitored at the time n, x (n) represents the measured output value measured by the sensor in the system to be monitored at the time n, and V' (n) represents the measurement error of the sensor.
The accuracy of the sensor in the system to be monitored is denoted by S ' (N), V ' (N) follows a normal distribution with a mean value of 0 and a standard deviation of S ' (N), V ' (N) to N (0, S ' (N)).
Based on the above equation (6), a measurement model of the system to be monitored can be obtained:
y (n) = C (n) + D (n) + u (n) + V' (n) formula (8)
Wherein y (n) represents the true output value of the system to be monitored at time n, V' (n) represents the measurement error of the sensor, C = [ 1.. 0], and D = [ 0.. 0].
And a substep 20132 of generating a discrimination model of the system to be monitored based on the state matrix updating equation and the constructed measurement model.
Specifically, the measured value of the sensor at the n moment is estimated according to the measured value of the sensor at the n-1 moment of the system to be monitoredThe specific expression is as follows:
covariance P of measured values of sensor at time n n|n-1 Can be expressed as:
estimating the real output value at the n moment according to the real output value at the n-1 moment of the system to be monitoredThe specific expression is as follows:
true error e of system to be monitored at moment n n Can be expressed as:
thus, a discriminant model of the system to be monitored can be obtained:
wherein k is 0 And the matrix updating coefficient in the matrix updating equation of the system to be monitored is also the discrimination parameter in the discrimination model.
Specifically, the measurement parameters included in the regression model are the same as those included in the measurement model, and are all ω i 、ω i ' and ω 0 Therefore, the values of the measurement parameters can be obtained by carrying out recursion solution on the regression model, in each iteration process, based on the historical reference data of the current iteration time, the estimation error and the system integral error of the regression model are obtained, whether the current iteration meets the preset iteration condition or not is judged, if the current iteration meets the preset iteration condition, the current value of the measurement parameters of the regression model is output, and the iteration is stopped; and if the iteration does not meet the preset iteration condition, updating the iteration time and continuing to perform the iteration. The specific iterative process is described in detail below.
The expression of the estimation error σ (t) of the regression model at time t is:
wherein x (t) represents the output measurement value measured by the sensor in the system to be monitored at the moment t,and the weight of the estimation error of the regression model at the t-1 moment in the overall system error of the regression model is represented, and l represents the preset sliding window length.
The expression of the overall system error delta (t) of the regression model at the time t is as follows:
δ (t) = g δ (t-1) + (1-g) × sign (σ (t)) formula (17)
Wherein g represents a moving average index (which is a preset constant), sign is used for determining whether the current data point is an abnormal point,τ represents a constant greater than 1.
The historical reference data of the system to be monitored comprises input and output data of the system to be monitored in a period of time, after an estimation error sigma (t) of the regression model at the time t is calculated by using the historical reference data of the system to be monitored, a system overall error delta (t) of the regression model at the time t can be calculated based on the estimation error sigma (t), and whether the iteration meets a preset iteration condition or not is judged, wherein the iteration condition is as follows: and the integral error of the current system of the regression model is smaller than a preset error threshold value, or the iteration times reach a preset maximum time threshold value.
When the current iteration is judged to meet the preset iteration condition, outputting the current test parameter omega in the regression model i 、ω i ' and ω 0 And stopping the iteration; and substituting the values of the three test parameters into the measurement model to obtain the measurement model of the system to be monitored.
When the iteration is judged not to meet the preset iteration condition, the iteration time is updated, the weight of the estimation error of the regression model in the system overall error of the regression model is updated based on the updated iteration time, and the weight of the estimation error of the regression model at the time t in the system overall error of the regression model is updatedThe expression of (a) is:
wherein the content of the first and second substances,represents the gain vector, <' > at time t of the regression model> Represents the inverse matrix of the regression model at time t, <' >> λ is a constant between 0 and 1.
Weight at the current time is updatedThen, based on the historical reference data and the updated weight->Recalculating the system integral error delta (t) of the regression model at the time t, judging whether the iteration meets the preset iteration condition or not, repeating the process until the iteration meets the preset iteration condition, and outputting the current test parameter omega (t) in the regression model i 、ω i ' and ω 0 So that the values of the three test parameters can be substituted into the measurement model to obtain the measurement model of the system to be monitored.
And step 203, setting the discrimination parameters in the discrimination model by adopting a longicorn stigma search algorithm based on the reference historical data and the measurement parameters.
Referring to fig. 4, step 203 includes the following sub-steps:
Specifically, based on the process of the expression (15) of the discrimination model constructed in step 201, the discrimination parameters to be set in the discrimination model include: e and k 0 Thus using the measurement parameter omega i 、ω i ' and ω 0 And the discrimination parameters in the discrimination model are belonged to and k 0 Constructing the position of a longicorn in a long horn search algorithm (BAS for short)
Fitness function (namely target function to be optimized) adopted in Tianniu whisker search algorithmComprises the following steps:
wherein, R represents the recall ratio of the abnormal data points of the longicorn stigma search algorithm in the iteration, and N represents the false detection ratio of the abnormal data points of the longicorn stigma search algorithm in the iteration. The historical reference data of the system to be monitored comprises input and output data of the system to be monitored in a period of time, wherein the input and output data comprise a plurality of abnormal data points and are marked, the recall ratio R represents the proportion of the number of the abnormal data points detected in one iteration process to the total number of the abnormal data points in the historical parameter data, and the false detection ratio N represents the proportion of the number of the abnormal data points false detected in one iteration process to the total number of the abnormal data points detected.
Initializing the longicorn, wherein the initialization process comprises the following steps: initialization of the position of the longicorn due to the measurement of the parameter omega i 、ω i ' and ω 0 The value of (c) has already been solved in step 202, and only e and k need to be solved when initializing the longicorn position 0 Carrying out initialization in a mode comprising: setting e and k 0 And then can be randomly generated from this rangeGenerate the values of e and k0, which in this embodiment are 0 The initial values of (a) are: e = rands (1), k 0 =rands(1)。
The initialization process further includes: and (4) initializing the iteration times, namely setting the iteration times m =1, and the length d and the step size step of the left and right tentacles of the longicorn.
And a substep 2032 of obtaining the fitness of the longicorn at the current position based on the historical reference data, the current position of the longicorn and the fitness function.
Specifically, the current position of the longicorn is determinedMiddle (omega) i ,ω i ′,ω 0 ,∈,k 0 ) The parameter values are respectively brought into a measurement model, a judgment model and a fitness function, then input signals at all times in historical reference data are substituted into the measurement model, the measurement model is used for predicting output predicted values of a system to be monitored at all times, the judgment model is used for carrying out abnormity judgment on the output predicted values at all times, judgment results of all output data values are recorded, statistics is carried out based on the judgment results of the data predicted values at all times and the abnormity marking conditions of data points at all times in the historical reference data, and then the current position of the longhorn is calculated>Is greater than or equal to>
Specifically, in the longicorn whisker search algorithm, the distance d0 between the left and right tentacles of the longicorn is fixed and can be usedRepresents the coordinates of the longicorn left tentacle, and is/are>Represents the coordinates of the right beard and>representing the barycentric coordinates of the longicorn (i.e. the current position of the longicorn), and generating a random vector dir = rands (n, 1) to represent the orientation of the head of the longicorn after the longicorn is flown next, i.e. the orientation of the vector of the right antenna pointing to the left antenna of the longicorn is arbitrary, where n is an integer greater than or equal to 1, and dir = rands (n, 1) is normalized to obtain:
dir=dir/norm(dir);
therefore, the temperature of the molten metal is controlled,can also be expressed in relation to the center of mass->The expression of (c):
then the current coordinates of the anoplophora sinica palpusIs brought into the fitness function->In the method, the fitness f of the longicorn left tentacle is calculated by using historical reference data left (ii) a In the same way, the current coordinate of the longicorn right tentacle is evaluated>Brought into fitness function>In the method, the fitness f of the anoplophora hamiltonii right palpus is calculated by using historical reference data right . The calculation method of the fitness of the left and right tentacles of the longicorn is substantially the same as that of the fitness of the longicorn, and is not repeated herein.
The fitness f of the longicorn left palpus left Fitness f of the anoplophora chinensis to the right tentacle right Comparing to obtain the next advancing direction of the longicorn, if f left <f right If the minimum value of the fitness of the longicorn is searched, the longicorn advances towards the left tentacle direction, and if f is left >f right In order to find the minimum value of the fitness of the longicorn, the longicorn advances towards the right tentacle direction, and the sign function sign can be represented as follows:
wherein when sign =1, the skynet of the next step advances towards the direction of the right tentacle; sign = -1, which means that the longicorn moves towards the left tentacle direction in the next step.
After the next advancing direction of the longicorn is determined, the position of the longicorn can be updated according to the current step length of the longicorn, and the position updating formula of the longicorn is
Wherein the content of the first and second substances,the position of the longicorn at the m iteration is shown, step is the current step size of the longicorn, and D is the direction of the longicorn to walk next.
D=sign*(dir+k 0 *rands(n,1))。
The updating formula of the step length of the longicorn is as follows:
step=step*eta;
wherein eta represents the step adjustment ratio of the longicorn, 0 < eta < 1, for example eta is 0.95.
The updated formula of the longicorn beard length P is as follows:
P m =r*P m -1;
wherein r represents the beard length attenuation coefficient of the longicorn, and m represents the iteration number.
The update mode of the number of iterations is m = m +1.
After updating the step length, the beard length and the iteration times of the longicorn, judging whether the iteration meets a preset iteration condition, wherein the iteration condition can be whether the fitness of the longicorn at the current position meets a preset fitness threshold or not, and/or whether the iteration times m of the longicorn reaches a preset iteration time threshold or not, when the iteration meets the iteration condition, terminating the longicorn beard search algorithm, and updating the position of the current longicornE and k in 0 The value of (a) is input into a discriminant model, which can be used to discriminate whether there is an anomaly in the output of the system to be monitored. And when the iteration does not meet the iteration condition, substituting the updated position of the longicorn into the fitness function, returning to the step 203, and repeating the process until the iteration meets the preset iteration condition.
Referring to fig. 5, it is a schematic diagram of setting the discriminating parameter in the discriminating model by using the longicorn whisker search algorithm, and as can be seen from fig. 5, the longicorn whisker search algorithm converges after 32 iterations, and the convergence speed is high.
Taking a system to be monitored as a transformer system as an example, the output voltage of the transformer is measured by a sensor, please refer to fig. 6, which is an abnormal data point marked in an iteration process by adopting a longicorn whisker search algorithm, a black dot is the marked abnormal data point, the detection rate is 100%, the false detection rate is 0%, and all the abnormal data points can be marked at the beginning of the abnormality.
In the embodiment, when a measurement model of the system to be monitored is constructed, the measurement parameters in the regression model are solved by using the linear recursive estimation with the sliding window, and the historical reference data and the sensor precision of the system to be monitored are combined, so that the generated measurement model is more consistent with the actual operation condition of the system to be monitored, therefore, when the measurement model is used for carrying out abnormal monitoring on the system to be monitored, a better and more accurate output predicted value can be obtained, the data prediction precision of the system to be monitored is improved, and the measurement model has better robustness. And the measurement parameters are solved by adopting recursive estimation, so that the computational power consumption is low, the speed is high, the method is suitable for being deployed in a processor chip with low computational power, and the cost is reduced.
In addition, the judgment model of the system to be monitored is constructed by adopting the longicorn stigma search algorithm in the embodiment, the number of judgment parameters to be set in the judgment model in the optimization process is small, the calculation force requirement is low, the convergence rate of the longicorn stigma search algorithm is high, the method is suitable for being deployed in a processor chip with low calculation force, the cost is reduced, and the precision is high; in addition, a long-time learning process is not needed, the requirement for early-stage data accumulation is low, and the deployment is easy.
A second embodiment of the present invention relates to a system monitoring method, which is applied to a processor chip, wherein a measurement model and a discrimination model generated based on the system monitoring model method in the first embodiment are deployed in the processor chip. Therefore, the processor chip can run the system monitoring method in this embodiment to monitor the system to be monitored, including: and predicting the output of the system to be monitored by using the measurement model, and monitoring the output predicted value of the measurement model by using the discrimination model.
The specific flow of the system monitoring method of the present embodiment is shown in fig. 7.
Specifically, the processor chip corrects the output measurement value by using the measurement model based on the output measurement value measured by a sensor in the system to be monitored to obtain the output prediction value of the system to be monitored, then performs abnormity monitoring on the output prediction value predicted by the measurement model by using the discrimination model, and sends out an alarm and marks an abnormal data point if data abnormity is judged to exist; the way of issuing the alarm may be: sound alarm, text alarm prompt and the like, so that the subsequent fault judgment of the working personnel is facilitated.
A third embodiment of the present invention relates to a processor chip for executing the system monitoring model generation method in the first embodiment and/or the system monitoring method in the second embodiment; the system to be monitored can be an industrial system, such as a transformer system, a wind power system and the like; the processor chip may be an MCU.
In this embodiment, a single processor chip may simultaneously generate a measurement model and a discrimination model of a system to be monitored, and perform anomaly monitoring on the system to be monitored by using the generated measurement model and discrimination model; or the two processor chips respectively execute the generation of the measurement model and the discrimination model of the system to be monitored and the abnormal monitoring of the system to be monitored by using the generated measurement model and the discrimination model.
Referring to fig. 8, a fourth embodiment of the present invention relates to an industrial system, including a processor chip 1, an industrial device 2 connected to the processor chip 1, and a sensor 3, where the sensor 3 is connected to the processor chip 1 and the industrial device 2 respectively; industrial equipment is for example transformer systems, wind power systems, etc.
The processor chip 1 has deployed therein a measurement model and a discrimination model generated based on the system monitoring model method in the first embodiment.
The processor chip 1 is connected with the sensor 3, the sensor 3 can send the measured output measurement value of the system to be monitored to the processor chip 1, the processor chip 1 can correct the output measurement value by using the measurement model to obtain the output prediction value of the system to be monitored, then the judgment model is used for carrying out abnormity monitoring on the output prediction value predicted by the measurement model, if the data abnormity is judged to exist, an alarm is given out, and an abnormal data point is marked; the way of issuing the alarm may be: sound alarm, text alarm prompt and the like, so that the subsequent fault judgment of the working personnel is facilitated.
While the preferred embodiments of the present invention have been described in detail above, it should be understood that aspects of the embodiments can be modified, if necessary, to employ aspects, features and concepts of the various patents, applications and publications to provide yet further embodiments.
These and other changes can be made to the embodiments in light of the above detailed description. In general, in the claims, the terms used should not be construed to be limited to the specific embodiments disclosed in the specification and the claims, but should be construed to include all possible embodiments along with the full scope of equivalents to which such claims are entitled.
Claims (13)
1. A system monitoring model generation method is characterized by comprising the following steps:
constructing a measurement model and a discrimination model of the system to be monitored based on a data change model of the system to be monitored;
obtaining the value of the measurement parameter in the measurement model according to the reference historical data of the system to be monitored;
obtaining a value of a discrimination parameter in the discrimination model based on the reference historical data and the value of the measurement parameter;
the method for constructing the measurement model and the discrimination model of the system to be monitored based on the data change model of the system to be monitored comprises the following steps:
obtaining a regression model of a system to be monitored based on a data change model of the system to be monitored;
obtaining a state matrix updating equation of the system to be monitored according to the data change model and the regression model;
and updating an equation according to the state matrix, and constructing a measurement model and a discrimination model of the system to be monitored.
2. The method according to claim 1, wherein the step of constructing a measurement model and a discrimination model of the system to be monitored according to the state matrix update equation comprises:
based on the state matrix updating equation and the precision parameters of the sensors in the system to be monitored, constructing a measurement model of the system to be monitored;
and updating an equation and the constructed measurement model based on the state matrix to generate a discrimination model of the system to be monitored.
3. The method of generating a system monitoring model according to claim 2, wherein the data change model is expressed as:
wherein, ω is i 、ω i ' and ω 0 Representing a measurement parameter in the data change model, x (n) representing an output measurement value measured by a sensor in the system to be monitored at n moments, r (n-i) representing an input value of the system to be monitored at n-i moments, and l representing a preset sliding window length;
the expression of the regression model is as follows:
wherein, ω is i 、ω i ' and ω 0 Representing the measured parameters in the regression model,expressing an output predicted value of the regression model at the moment n, r (n-i) expressing an input value of the system to be monitored at the moment n-i, x (n) expressing an output measured value measured by a sensor in the system to be monitored at the moment n, and l expressing a preset sliding window length;
the state matrix updating equation of the system to be monitored is as follows:
S(n+1)=(1-∈)*S(n)+∈*V 2 (n);
wherein S (n) represents a state matrix of the system to be monitored at n times, e represents an update coefficient of the state matrix, and V (n) represents a residual error between the data change model and the regression model at n times.
4. The system monitoring model generation method of claim 3, wherein the expression of the measurement model is:
y(n)=C*Z(n)+D*u(n)+V′(n);
wherein y (n) represents the real output value of the system to be monitored at the moment n, V′(n)~N(0,S′(n)),C=[1 0...0],D=[0 0...0], v '(n) represents a measurement error of a sensor in the system to be monitored, S' (n) represents an accuracy of the sensor in the system to be monitored, and S (n) represents a state matrix of the system to be monitored at time n.
5. The system monitoring model generation method according to claim 3, wherein the expression of the discriminant model is:
wherein e is n Representing the real error of the system to be monitored at the moment n, k0 representing the discrimination parameter in the discrimination model,C=[1 0...]s' (n) denotes the accuracy of the sensor in the system to be monitored,and S (n) represents a state matrix of the system to be monitored at the moment n.
6. The method according to claim 3, wherein obtaining the value of the measurement parameter in the measurement model according to the reference historical data of the system to be monitored comprises:
and according to the reference historical data of the system to be monitored, recursively solving to obtain the value of the measurement parameter in the regression model as the value of the measurement parameter in the measurement model.
7. The method according to claim 6, wherein the obtaining the value of the measurement parameter in the regression model by recursive solution according to the reference historical data of the system to be monitored comprises:
in each iteration process, acquiring an estimation error of the regression model and an overall system error based on historical reference data of current iteration time, and judging whether the iteration meets a preset iteration condition;
if the iteration meets the preset iteration condition, outputting the value of the current measurement parameter of the regression model;
and if the iteration does not meet the preset iteration condition, updating the iteration time and performing iteration.
8. The method according to any one of claims 1 to 7, wherein the obtaining a discriminant parameter in the discriminant model based on the reference historical data and the measurement parameter comprises:
and setting the discrimination parameters in the discrimination model by adopting a longicorn whisker search algorithm based on the reference historical data and the measurement parameters.
9. The method of claim 8, wherein the tuning the discriminant parameters in the discriminant model using a longitussimus whisker search algorithm based on the reference historical data and the measured parameters comprises:
constructing the position of the longicorn by using the measurement parameters and the discrimination parameters in the discrimination model, setting a fitness function of the longicorn stigma search algorithm, and initializing the longicorn;
obtaining the fitness of the longicorn at the current position based on the historical reference data, the current position of the longicorn and the fitness function;
calculating the fitness of the left antenna and the right antenna of the longicorn, and updating the position of the longicorn;
updating the step length, the beard length and the iteration times of the longicorn, and judging whether preset iteration conditions are met;
if the preset iteration condition is met, terminating the longicorn stigma search algorithm, and outputting the value of each discrimination parameter based on the current longicorn position;
and if the preset iteration condition is not met, substituting the updated position of the longicorn into the fitness function, and repeating the longicorn stigma search algorithm.
10. The method of generating a system monitoring model of claim 9, wherein the fitness function of the longicorn whisker search algorithm is:
wherein the content of the first and second substances,representing the current position of the longicorn, R representing the recall ratio of the abnormal data points of the longicorn stigma search algorithm in the iteration, and N representing the false detection ratio of the abnormal data points of the longicorn stigma search algorithm in the iteration;
the position updating formula of the longicorn is as follows:
11. A system monitoring method, comprising:
predicting an output predicted value of the system to be monitored based on the measurement model generated by the system monitoring model generation method according to any one of claims 1 to 10 and an output measured value measured by a sensor in the system to be monitored;
determining whether the output prediction value is abnormal based on a discriminant model generated by the system monitoring model generation method according to any one of claims 1 to 10;
and if the output predicted value is determined to be abnormal, giving an alarm.
12. A processor chip, wherein the processor chip is configured to perform the system monitoring model generating method of any one of claims 1 to 10, and/or the system monitoring method of claim 11.
13. An industrial system, comprising: the processor chip of claim 12, an industrial device coupled to the processor chip, and a sensor coupled to the processor chip and the industrial device, respectively.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110761540.1A CN113433913B (en) | 2021-07-06 | 2021-07-06 | System monitoring model generation and monitoring method, processor chip and industrial system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110761540.1A CN113433913B (en) | 2021-07-06 | 2021-07-06 | System monitoring model generation and monitoring method, processor chip and industrial system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113433913A CN113433913A (en) | 2021-09-24 |
CN113433913B true CN113433913B (en) | 2023-03-24 |
Family
ID=77759270
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110761540.1A Active CN113433913B (en) | 2021-07-06 | 2021-07-06 | System monitoring model generation and monitoring method, processor chip and industrial system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113433913B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113534679B (en) * | 2021-07-06 | 2023-04-28 | 上海新氦类脑智能科技有限公司 | System monitoring model generation method, processor chip and industrial system |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014155690A1 (en) * | 2013-03-29 | 2014-10-02 | 富士通株式会社 | Model updating method, device and program |
CN106019093A (en) * | 2016-05-15 | 2016-10-12 | 北华大学 | Online soft measurement method for three-phase arc furnace arc length |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101178235B1 (en) * | 2010-10-15 | 2012-08-30 | 한국전력공사 | Prediction and fault detection method and system for performance monitoring of plant instruments using principal component analysis, response surface method, Fuzzy Support Vector Regression and Generalized Likelihood Ratio Test |
US9483602B2 (en) * | 2010-10-27 | 2016-11-01 | Solido Design Automation Inc. | Method and system for identifying rare-event failure rates |
WO2015139709A1 (en) * | 2014-03-19 | 2015-09-24 | Northern Vo Aps | Method and system for monitoring a vehicle |
CN109902832B (en) * | 2018-11-28 | 2023-11-17 | 华为技术有限公司 | Training method of machine learning model, anomaly prediction method and related devices |
CN110880984B (en) * | 2019-10-18 | 2022-12-27 | 平安科技(深圳)有限公司 | Model-based flow anomaly monitoring method, device, equipment and storage medium |
CN111242272A (en) * | 2020-01-15 | 2020-06-05 | 东方红卫星移动通信有限公司 | Wireless sensor network anomaly detection method |
CN111444953B (en) * | 2020-03-24 | 2022-04-29 | 东南大学 | Sensor fault monitoring method based on improved particle swarm optimization algorithm |
-
2021
- 2021-07-06 CN CN202110761540.1A patent/CN113433913B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014155690A1 (en) * | 2013-03-29 | 2014-10-02 | 富士通株式会社 | Model updating method, device and program |
CN106019093A (en) * | 2016-05-15 | 2016-10-12 | 北华大学 | Online soft measurement method for three-phase arc furnace arc length |
Non-Patent Citations (2)
Title |
---|
Analysis and modeling of process monitoring in dynamic shop floor;Shuxia Li;<2008 IEEE International Conference on Service Operations and Logistics, and Informatics>;20081125;全文 * |
LSSVM动态软测量模型在磨煤机一次风量预测方面的应用;杨耀权等;《动力工程学报》;20160315(第03期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN113433913A (en) | 2021-09-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3929824A2 (en) | Robust multimodal sensor fusion for autonomous driving vehicles | |
CN110503071B (en) | Multi-target tracking method based on variational Bayesian label multi-Bernoulli superposition model | |
CN107292023B (en) | Bridge structure state diagnosis method based on narrow-area characteristics of damage index system | |
CN113534679B (en) | System monitoring model generation method, processor chip and industrial system | |
Li et al. | Data-driven bearing fault identification using improved hidden Markov model and self-organizing map | |
JP7154432B2 (en) | System and method for controlling a robot arm | |
JP6742554B1 (en) | Information processing apparatus and electronic apparatus including the same | |
CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
CN111949012B (en) | Intermittent process fault detection method based on double-weight multi-neighborhood preserving embedding algorithm | |
CN113433913B (en) | System monitoring model generation and monitoring method, processor chip and industrial system | |
CN112052802A (en) | Front vehicle behavior identification method based on machine vision | |
CN111352977A (en) | Time sequence data monitoring method based on self-attention bidirectional long-short term memory network | |
CN112115184A (en) | Time series data detection method and device, computer equipment and storage medium | |
CN111767183A (en) | Equipment abnormality detection method and device, electronic equipment and storage medium | |
Entezami et al. | Probabilistic damage localization by empirical data analysis and symmetric information measure | |
CN105894014B (en) | Abnormal behavior sequential detection method based on multi-factor inconsistency measurement | |
US20220004908A1 (en) | Information processing apparatus, information processing system, information processing method, and non-transitory computer readable medium storing program | |
Srivastava et al. | Detection and prognostics on low-dimensional systems | |
CN109509207B (en) | Method for seamless tracking of point target and extended target | |
CN116627116B (en) | Process industry fault positioning method and system and electronic equipment | |
CN113064489A (en) | Millimeter wave radar gesture recognition method based on L1-Norm | |
CN110675430B (en) | Unmanned aerial vehicle multi-target tracking method based on motion and appearance adaptation fusion | |
CN110779526B (en) | Path planning method, device and storage medium | |
US20230102866A1 (en) | Neural deep equilibrium solver | |
CN111610514B (en) | Inversion method and device for propagation characteristics of evaporation waveguide |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |