CN114416708A - Industrial multivariable alarm method and system based on search cone feasible working domain modeling - Google Patents

Industrial multivariable alarm method and system based on search cone feasible working domain modeling Download PDF

Info

Publication number
CN114416708A
CN114416708A CN202111509701.4A CN202111509701A CN114416708A CN 114416708 A CN114416708 A CN 114416708A CN 202111509701 A CN202111509701 A CN 202111509701A CN 114416708 A CN114416708 A CN 114416708A
Authority
CN
China
Prior art keywords
search
working domain
cone
feasible working
search cone
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.)
Pending
Application number
CN202111509701.4A
Other languages
Chinese (zh)
Inventor
王建东
王振
侯纪林
孙晓庆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong University of Science and Technology
Original Assignee
Shandong University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong University of Science and Technology filed Critical Shandong University of Science and Technology
Priority to CN202111509701.4A priority Critical patent/CN114416708A/en
Publication of CN114416708A publication Critical patent/CN114416708A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/211—Schema design and management
    • G06F16/212—Schema design and management with details for data modelling support
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/18—Status alarms

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Business, Economics & Management (AREA)
  • Emergency Management (AREA)
  • Quality & Reliability (AREA)
  • Artificial Intelligence (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention provides an industrial multivariable alarm method and system based on search cone feasible working domain modeling, wherein a correlation variable determined based on trend change is determined as a monitoring variable; preprocessing the data points of the acquired monitoring variables, searching whether a search cone can contain the preprocessed data points or not in a feasible working domain boundary model which is pre-constructed based on historical monitoring variable data, if so, determining the search cone to be a normal data point, otherwise, determining the search cone to be an abnormal data point; determining a normal data point corresponding to the abnormal data point; calculating a straight line corresponding to the dynamic alarm threshold, and calculating an intersection point of the straight line and the feasible working domain boundary model so as to determine the dynamic alarm threshold; and if the preprocessed data point is positioned outside the feasible working domain boundary model or a certain variable exceeds the operation range represented by the corresponding alarm threshold value, triggering an alarm. The invention can realize the design of multivariable alarm threshold values in modern industrial systems, has reasonable design and better universality.

Description

Industrial multivariable alarm method and system based on search cone feasible working domain modeling
Technical Field
The invention belongs to the technical field of industrial alarm, and particularly relates to an industrial multivariable alarm method and system based on search cone feasible working domain modeling.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
As the degree of automation of industrial processes increases, more and more variables are provided in plants with alarm systems. However, due to the unreasonable design of the alarm system, the problem of excessive interference alarm is caused, so that an operator cannot identify real alarm in time, and the timely treatment of abnormal working conditions is delayed. And because the variables in the system often have an association relationship, the design of the alarm system aiming at a single variable cannot take the association between the variables into consideration, so that the real abnormality of the associated variable cannot be detected to generate more interference alarms, which causes the abnormal deterioration to cause economic loss and casualties. Therefore, the design of the multivariable alarm system needs to be considered, and the problem of excessive interference alarm is solved.
According to the knowledge of the inventor, aiming at a multivariable alarm system, at present, in order to solve the problem of excessive disturbance alarm, methods such as a multivariate statistical thought, a transmission entropy, a Bayesian filter, alarm cluster analysis and the like are provided. However, in the actual industrial production process, in view of the movement of the material, the continuous conversion of energy and the like, the process variables involved are very many, and each monitored variable is closely related and has more or less some relation.
The existing multivariable alarm system has the following problems:
1) the alarm threshold value is not designed by fully utilizing the relation between monitoring variables, so that the alarm threshold value is unreasonable to design; 2) the design method of the multivariable alarm system based on the hyper-ellipsoid established feasible working domain can only be used for representing some process variables which completely accord with Gaussian distribution in any dimension space, and has weak universality; 3) the design method of the multivariable alarm system based on the convex hull model for establishing the feasible working domain has complex calculation process, the calculation resource consumption is overlarge during the solution optimization calculation, the under-fitting condition is easy to occur, the higher the dimension is, the more serious the condition is, the even the non-solution condition can occur, and the practicability is lacked.
Disclosure of Invention
The invention provides an industrial multivariable alarm method and system based on search cone feasible working domain modeling, aiming at solving the problems.
It should be noted that the industrial multivariable alarm method/system provided by the invention can be applied to equipment related to industrial production, such as a three-container water tank, a transformer and the like, can also be applied to large-scale systems related to industrial production, such as a coal mill powder making system and the like, and can also be applied to detection and alarm of industrial production environment, such as fire detection, gas leakage detection and the like; the alarm system or the design method used by the application object or the application scene should fall within the protection scope of the present invention if the technical scheme provided by the present invention is directly applied or simply changed.
According to some embodiments, the invention adopts the following technical scheme:
an industrial multivariable alarm method based on search cone feasible working domain modeling comprises the following steps:
determining an associated variable determined based on the trend change as a monitoring variable;
preprocessing the data points of the acquired monitoring variables, searching whether a search cone can contain the preprocessed data points or not in a feasible working domain boundary model which is pre-constructed based on historical monitoring variable data, if so, determining the search cone to be a normal data point, otherwise, determining the search cone to be an abnormal data point;
determining a normal data point corresponding to the abnormal data point;
calculating a straight line corresponding to the dynamic alarm threshold according to the normal data points, and calculating an intersection point of the straight line and the feasible working domain boundary model so as to determine the dynamic alarm threshold;
and if the preprocessed data point is positioned outside the feasible working domain boundary model or a certain variable exceeds the operation range represented by the corresponding alarm threshold value, triggering an alarm.
As an alternative embodiment, the specific process of determining the associated variable determined based on the trend change as the monitoring variable includes: and performing piecewise linear representation on the sample data by using a bottom-up piecewise linear representation method to obtain the trend change relation of each variable, and further determining the monitoring variable according to the trend change relation and the actual relation among the variables.
As an alternative embodiment, the specific process of constructing a feasible working domain boundary model in advance based on historical monitoring variable data includes:
preprocessing a monitoring variable normal data point in historical monitoring data, regarding a feasible working domain formed by monitoring variables as being formed by a plurality of search cones, and determining the side edge length of each search cone based on the preprocessed data;
setting the optimal step angle of the search cone, respectively selecting the value of each angular coordinate of the search cone, obtaining the vertex coordinate of the infinitesimal according to the arrangement and combination mode of all the angular coordinates, and forming a vertex set together with the origin;
and determining a search cone expression according to the vertex set, and obtaining a feasible working domain boundary model in a traversal search mode.
As a further limitation, the specific process of performing the pretreatment includes: and expressing the historical normal data points of the monitoring variables by using vectors, and standardizing the data of the monitoring variables by adopting a Z-score standardization method.
By way of further limitation, the search cones are all represented by convex hulls in a high-dimensional space, and the vertex set of the convex hulls is the origin and the vertices of the hypercube at the bottom of the search cones.
As a further limitation, the specific process of setting the step angle of the optimal search cone includes:
defining the vertex angle of two adjacent edges of the search cone as the step angle alpha of the search cone, and determining the value of alpha when the minimum value is taken by the following objective function F (alpha):
F(α)＝(1-η)σcεr
wherein eta represents a fitness index of the feasible working domain boundary model; sigmacA composite standard deviation representing an alarm threshold for adjacent monitored data points; epsilonrRepresenting the radius difference index of adjacent search cones.
As a further limitation, the specific process of respectively selecting the values of the angular coordinates of each of the search cones and obtaining the vertex coordinates of the infinitesimal elements according to the arrangement and combination mode of all the angular coordinates includes: respectively making each angular coordinate equal to 0 and stepping angle alpha, searching angular coordinate [ phi ] of each edge of cone in n-dimensional space1,φ2,…,φn-1]In total 2n-1One of them is corresponding to 2 of search cone bottom surface infinitesimaln-1And obtaining the coordinate of the vertex of the infinitesimal according to the arrangement and combination mode of all the angular coordinates.
As a further limitation, the specific process of determining the search cone expression according to the vertex set includes: data points on all hyperplanes of the convex hull are found through a fast convex hull algorithm, an expression is used for representing the ith hyperplane which is enclosed into the convex hull, n data points on the ith hyperplane are indexed one by one and substituted into the expression, the normal vector and the hyperplane offset are obtained through solving, all hyperplanes enclosed into the convex hull are calculated one by one, the complete expression of the hyperplane is obtained, the search cone is represented, and the search cone expression is obtained.
As an alternative embodiment, the specific process of calculating a straight line corresponding to the dynamic alarm threshold, calculating an intersection point of the straight line and the feasible working domain boundary model, and determining the dynamic alarm threshold includes: according to a linear expression corresponding to a certain monitoring variable dynamic alarm threshold, the linear expression and a feasible working domain boundary model are combined to obtain the intersection point of the linear expression and the feasible working domain boundary model, and a set Z is formedjAccording to the set ZjThe maximum value and the minimum value determine the upper limit and the lower limit of the dynamic alarm threshold value.
An industrial multivariable alarm system based on search cone feasible working domain modeling, comprising:
a monitoring variable determination module configured to determine an associated variable determined based on the trend change as a monitoring variable;
the data point judgment module is configured to preprocess the data points of the acquired monitoring variables, search whether a search cone can contain the preprocessed data points in a feasible working domain boundary model which is pre-constructed based on historical monitoring variable data, if so, the search cone is a normal data point, and otherwise, the search cone is an abnormal data point;
the dynamic alarm threshold value determining module is configured to determine a normal data point corresponding to the abnormal data point, calculate a straight line corresponding to the dynamic alarm threshold value according to the normal data point, calculate an intersection point of the straight line and the feasible working domain boundary model, and further determine the dynamic alarm threshold value;
and the alarm module is configured to trigger an alarm if the preprocessed data point is positioned outside the feasible working domain boundary model or a certain variable exceeds an operation range represented by the corresponding alarm threshold value, otherwise, the alarm is not triggered.
Compared with the prior art, the invention has the beneficial effects that:
according to the method, the abnormal data points are calculated by judging whether the abnormal data points are in the search cone model or not, so that the calculation process is simplified, and the practicability of the model is improved; the method solves the problems that the abnormal data points exist in the alarm method based on the convex hull model, the calculation resource consumption is overlarge during the solution optimization calculation, the situation that the dimension is higher is more serious, even the situation that no solution exists possibly occurs, the practical application is not facilitated, and the under-fitting condition exists.
The method adopts a mode of preprocessing the data to screen the data, fully utilizes the direct relation of each variable data, only needs the historical normal data point of the process variable to form the closed region of the convex hull model in the high-dimensional space, overcomes the defects that the process variable is limited in the boundary model of the feasible work domain of the hyper-ellipsoid to meet the Gaussian distribution and the historical normal data point of the process variable is limited in the convex hull model to form the convex closed region in the high-dimensional space, and increases the universality.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a flow diagram of a multivariate alarm method based on search cone feasible working domain modeling in accordance with at least one embodiment of the present disclosure;
FIG. 2 is a diagram illustrating various indexes and objective functions of a boundary model of a feasible working domain of a three-tank system according to the first embodiment;
FIG. 3 is a feasible working domain boundary model of the three-tank water tank and its fitting situation according to the first embodiment;
FIG. 4 is a view showing the operation status of the monitoring variable data points of the three-tank water tank and the dynamic alarm lines thereof according to the first embodiment;
FIG. 5 is a graph of the process variable trend of the portion of the coal pulverizer of the second embodiment;
FIG. 6 is a graph of the variation trend of another part of the process variables of the coal pulverizer of the second embodiment;
FIG. 7 is the raw normal data of the monitored variables of the coal pulverizer of the second embodiment;
FIG. 8 is a diagram illustrating various indexes and objective functions of a boundary model of a feasible working area of a coal pulverizer according to a second embodiment;
FIG. 9 is a graph of a fit of a feasible working domain boundary model based on a search cone to historical normal data points of a coal pulverizer according to a second embodiment;
FIG. 10 is a graph of monitored operational values of process variables and their dynamic alarm thresholds for the second embodiment.
The specific implementation mode is as follows:
the invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
As shown in FIG. 1, the invention provides a multivariate alarm method or alarm system design method based on search cone feasible working domain modeling, comprising the following steps:
step 1: and preprocessing the basic monitoring variables, and determining the associated variables determined based on the trend change as the monitoring variables of the multivariable alarm system. And performing piecewise linear representation on the sample data by using a bottom-up Piecewise Linear Representation (PLR) method to obtain the trend change relation of each variable, and further obtaining a basic monitoring variable with monitoring value according to the trend change relation and the actual mechanism relation among the variables.
Step 2: and establishing a feasible working domain boundary model based on the search cone model.
Obtaining all associated process variables that should be included in a feasible working domain by the method described in step 1
Figure BDA0003404765200000081
Wherein
Figure BDA0003404765200000082
Representing the nth related variable, and representing all historical normal data points contained in the normal working state in the production process by using an equation (1):
Figure BDA0003404765200000083
in the formula (1), the reaction mixture is,
Figure BDA0003404765200000084
represents all historical normal data points; t represents the sampling time of all historical normal data points, and in some embodiments, the value can be 1 s; n represents the number of all historical normal data points;
Figure BDA0003404765200000085
representing sampled data of a multivariate process at a time. Then, the specific steps of establishing a feasible working domain boundary model based on the search cone model according to the preprocessed data obtained in the previous step are as follows:
step 2.1: the raw data of the relevant variables were normalized using the Z-score normalization methodPreprocessing normalizes the raw data to normalized data with a mean of 0 and a variance of 1. For the original data set
Figure BDA0003404765200000086
The corresponding normalized data set should be:
X(t)＝[x1(t),x2(t),…,xn(t)] (2)
in the formula (2), xjThe formula for calculation of (t) is:
Figure BDA0003404765200000087
in the formula (3), the reaction mixture is,
Figure BDA0003404765200000088
which represents the mean value of the original data,
Figure BDA0003404765200000089
denotes the standard deviation of the original data, and n denotes the nth original data.
Step 2.2: all normal data points x (t) in the n-dimensional space may be wrapped by a plurality of n-dimensional pyramids with side edges of length r, the n-dimensional pyramids (with equal side edges) are search cones (with isosceles triangle waist length r in the two-dimensional space and regular quadrangular pyramid side edge length r in the three-dimensional space), each search cone may be represented by a convex hull in the high-dimensional space, and the vertex set of the convex hull is composed of the origin O and the vertex of the hypercube at the bottom of the search cone. The hypercube at the bottom of the cone searched by the concept of infinitesimal can be expressed as Δ s ═ Δ1Δ2…Δn-1In total, 2n-1A vertex and 2n-1And (6) strip edges. Wherein ΔjRepresenting the edges of the hypercube at the bottom of the search cone in a cartesian coordinate system. The feasible working domain is composed of an infinite number of tiny search cones, and the side edge length r can be obtained by adopting the formula (4) and the formula (5) based on the standardized data X (t) falling in a certain search cone. For convenience of calculation, the rounding-up value of the data point farthest from the origin O in x (t) may be taken as r.
Figure BDA0003404765200000091
In the formula (4), X (t) represents a normalized normal data point,
Figure BDA0003404765200000092
expressing to round up the operand, returning a minimum integer greater than the operand, | · | expressing to take the modulus value to the vector expressed by a certain process variable, the calculation formula for an n-dimensional vector is:
Figure BDA0003404765200000093
step 2.3: defining the vertex angle of two adjacent edges of the search cone as the step angle alpha of the search cone, and respectively making each angular coordinate phijJ-1, 2, …, n-1 equals 0 and α. The value of alpha is represented by formula (6)
The medium objective function F (α) is determined when it takes the minimum value.
F(α)＝(1-η)σcεr (6)
Eta in the formula (6) represents the fitness index of the feasible working domain boundary model, and the calculation formula is shown in the formula (7); sigmacThe comprehensive standard deviation of the alarm threshold values of the adjacent monitoring data points is represented, and the calculation formula is shown as a formula (13); epsilonrThe radius difference index of adjacent search cones is expressed, and the calculation formula is shown in formula (17).
Figure BDA0003404765200000094
The fitness index eta in the formula (7) is used for measuring the fitting condition of a feasible working domain boundary model to an original feasible working domain, for a determined feasible working domain consisting of historical normal data points, the model fitness index is defined as the proportion of data points X (t) filling the feasible working domain boundary model, and for a multivariable X, each process variable X of the multivariable XjValue range of
Figure BDA0003404765200000101
The operation domain in the dimension can be represented, and the calculation expression is as follows:
Figure BDA0003404765200000102
in formula (8), min (x)j) And max (x)j) Respectively representing the minimum and maximum values of all historical normal data for the process variable. Then each process variable x is calculatedjAfter the range of values for j ═ 1,2, …, n, the entire feasible working domain boundary model can be divided into a myriad of equal meshes (denoted as supervolumes in high dimensional space), each of which can be denoted as:
Figure BDA0003404765200000103
in the formula (9), n represents the number of process variables and also represents the dimension of the super-volume grid; k is a radical ofjDenotes the k-th dimension in the j-th dimensionjA grid whose value range can be calculated by:
Figure BDA0003404765200000104
in the formula (10), cj(kj) The minimum value interval is expressed as the minimum value interval of the jth process variable, which needs to be given by the prior knowledge of field operators or reasonably valued by historical data, and depends on the actual requirements of the operators. The center coordinates of each super-body grid can be obtained by the grid expression, namely:
C(k1,k2,…,kn)＝[c1(k1),c2(k2),…,cn(kn)] (11)
in the formula (11), cj(kj) Representing the coordinates of the hyper-volume grid in the j dimension, the expression is
Figure BDA0003404765200000105
If the center coordinates of a certain hyper-body grid are located in the feasible working domain boundary model, namely the center coordinates satisfy a certain equation (25) in the feasible working domain boundary model, the hyper-body grid is defined as an internal grid. With kjVarying in each process variable dimension, the number n of all internal meshes can be derivedi. For an internal supersome mesh, if at least one historical normal data point falls within the supersome mesh (k'1,k′2,…,k′n) In the formula:
Figure BDA0003404765200000114
then, the current hyper-volume grid is defined as the active count grid, taking all n's asiAll the internal grids are checked and calculated, and finally n is obtainedcA valid count grid.
σc＝σh+σl (13)
σ in formula (13)hHigh alarm threshold integrated standard deviation index, σ, representing adjacent monitored data pointslThe low alarm threshold composite standard deviation index, which represents the adjacent monitored data points, can be calculated by equations (14) and (15), respectively:
Figure BDA0003404765200000111
Figure BDA0003404765200000112
n in the formulae (14) and (15)iNumber of data points representing boundary model of working domain where test is feasible, fi,h,fi,lThe amplitude fluctuation conditions of the adjacent data points of the ith test data point x relative to the high alarm threshold and the low alarm threshold of the reference data point are respectively represented by the following formula (16):
Figure BDA0003404765200000113
h in formula (16)x,lxHigh and low alarm thresholds, h, representing test data points x, respectivelyi(t),li(t) represents the high alarm threshold and the low alarm threshold, respectively, for the ith adjacent data point of the test data point x.
Figure BDA0003404765200000121
In the formula (17), NrRepresenting the number of search cones, r, used to build a feasible working domain boundary modeliIndicating the side edge length of the ith search cone.
Step 2.4: each search cone can be represented by a convex hull in a high-dimensional space, and the vertex set of the convex hull is composed of the origin O and the vertex of the hypercube at the bottom of the search cone. Searching angular coordinate [ phi ] of each edge of cone in n-dimensional space1,φ2,…,φn-1]In total 2n-1One of the different permutation and combination modes corresponds to the search of the cone bottom surface infinitesimal delta s-delta1Δ2…Δ n-12 of (2)n-1And (4) obtaining the vertex coordinates of the infinitesimal points by a formula (18) one by one according to the arrangement and combination modes of all the angular coordinates, and forming a vertex set V of the n-dimensional space search cone together with the origin O.
Figure BDA0003404765200000122
Step 2.5: according to the vertex set V in the step 2.4, finding data points on all hyperplanes of the convex hull through a fast convex hull algorithm, and enclosing the ith hyperplane p of the convex hull in a high-dimensional space(i)Is expressed as:
Figure BDA0003404765200000123
in the formula (19), Xp(t) represents a data point on the hyperplane, i.e., the point satisfies the above formula on the face. So as to enclose the ith hyperplane p of the convex hull(i)Will pass through the data points on n planes, respectively denoted as X (I)i,1),X(Ii,2),…,X(Ii,n). And the data points are calculated by a fast convex hull algorithm from a search cone vertex set V. And (3) indexing n data points on the ith hyperplane one by one and substituting the formula (19) to obtain a formula (20):
a(i)XT(Ii,j)-b(i)＝0,j＝1,2,…,n (20)
a in formula (20)(i)Is the hyperplane p(i)The unit normal vector of (2). Then the vertical type (21) and the formula (22) are connected to obtain a normal vector
Figure BDA0003404765200000131
Offset from hyperplane b(i)。
Figure BDA0003404765200000132
Figure BDA0003404765200000133
Then, let i be 1,2, …, m, and solve one by one to obtain all hyperplanes surrounding the convex hull, so as to obtain the complete expression (23) of the hyperplane, i.e. the search cone represented by the expression (23).
AV′-B≤0 (23)
V' in equation (23) represents a result obtained by searching the cone vertex set V through a fast convex hull (Quick-hull) algorithm, which is located on each hyperplane of the convex hull. In addition, a and B represent:
Figure BDA0003404765200000134
for search cones in a component high-dimensional space represented by a convex hull m in formula (24)The number of hyperplanes (which degenerate to a straight line in two-dimensional space),
Figure BDA0003404765200000135
k ∈ {1,2, …, m } denotes the unit normal vector of the kth hyperplane in the convex hull, b(k)And k e {1,2, …, m } represents the offset distance of the kth hyperplane in the convex hull to the origin O.
Step 2.6: searching a data point X positioned in the current search cone by the formula (23) according to all historical normal data pointssAnd X issThe data point farthest from the origin O is defined as the boundary data point Xs,bThen the part of the feasible working domain that belongs to the current search cone can be represented by equation (25), and an expression that can describe the current part of the feasible working domain is temporarily stored in the set M.
Figure BDA0003404765200000136
Step 2.7: the n-1 th angular coordinate phin-1Step a. When two adjacent angular coordinates [ phi ]1,φ2,…,φn-1]Is smaller than its upper limit value pi (or 180 deg.) and an angular coordinate phi1,φ2,…,φn-2]When the deflection angle is smaller than the upper limit value of 2 pi (or 360 degrees), jumping to the step 2.4 to continue calculating; otherwise let phi j0 and alpha, respectively, and the offset angle phij-1Is stepped by a and at phi1And ending the traversal process when the constraint is not satisfied.
Step 2.8: and obtaining a feasible working domain boundary model M according to the calculation sequence.
And step 3: calculating a dynamic alarm threshold and determining an alarm state.
Step 3.1: for a new monitoring data point
Figure BDA0003404765200000141
The raw monitoring data points are compared according to equation (3)
Figure BDA0003404765200000142
Pretreatment ofIs a normalized data point X (t)0)。
Step 3.2: judging the normalized data point X (t)0) Whether it is inside the feasible working domain boundary model: searching for a search cone meeting the condition in the feasible working domain boundary model according to the formula (23), if a certain search cone exists, the data point X (t) can be searched0) If the data point is included, the data point is a normal data point, and the next step 3.3 is carried out to calculate the dynamic alarm threshold value; the data point X (t) can be divided in case no search cone exists0) If it is included, the data point is an abnormal data point, and step 3.6 is performed to calculate the dynamic alarm threshold.
Step 3.3: calculating x from equations (26) and (28)j(t0) Straight line Y corresponding to dynamic alarm thresholdjThen the straight line Y isjThe intersection point of the feasible working domain boundary model and the feasible working domain boundary model can be obtained by combining and temporarily storing the intersection point in the set ZjIn (1).
Figure BDA0003404765200000143
In the formula (26), p1,p2Respectively representing a straight line YjThe two points above are shown in equation (27).
Figure BDA0003404765200000144
C′DT+E＝0 (28)
In the formula (28), D represents a straight line YjAny of the data points on the data stream above,
Figure BDA0003404765200000151
represents the unit normal vector of the ith hyperplane in the hyperplane cluster H, e(i)And (3) representing the offset distance from the ith hyperplane in the hyperplane cluster H to the origin O, as shown in formula (29).
Figure BDA0003404765200000152
Step 3.4: the current process variable x can be obtained by the formula (30)j(t0) Normalized dynamic alarm threshold hj(t0) And lj(t0) Then, the dynamic alarm threshold of the original data can be obtained from the formula (31)
Figure BDA0003404765200000153
And
Figure BDA0003404765200000154
Figure BDA0003404765200000155
Figure BDA0003404765200000156
step 3.5: let j equal 1,2, …, n, respectively, and then solve the data point X (t) in steps 3.3 and 3.40) All process variables xjThe dynamic alarm threshold of (2).
Step 3.6: if the data point X (t)0) For an abnormal data point, the data point X (t) is calculated by equations (26), (28) and (32)0) Straight line Y with origin OXAnd in combination with formula (33) to obtain X (t)0) Corresponding normal data point XMAnd then step 3.3 is carried out, and calculation is carried out according to the dynamic alarm line calculation mode of the normal number data points.
Figure BDA0003404765200000157
Y in the formula (32)j,dRepresents a straight line YjC' is the unit normal vector representation of the hyperplane cluster H and the straight line YjUnit direction vector Y ofj,dPerpendicular n-1 non-identical vectors.
Figure BDA0003404765200000161
Step 3.7: finishing X (t)0) Dynamic alarm thresholds for all process variables.
Step 3.8: obtaining all process variables
Figure BDA0003404765200000162
After the dynamic alarm threshold value, judging an alarm variable XaWhether or not to issue an alarm signal, i.e. when
Figure BDA0003404765200000163
Within the feasible working domain boundary model, or each process variable is within the operating range represented by its corresponding alarm threshold, then XaThe value is '0', and no alarm is triggered; when in use
Figure BDA0003404765200000164
Outside the feasible working domain boundary model, or if some process variable exceeds the operation range represented by the corresponding alarm threshold value, XaAnd (5) taking the value as '1', and triggering an alarm. As shown in equation (34):
Figure BDA0003404765200000165
in order to make the implementation of the method of the present invention more obvious to those skilled in the art, the following description is given by way of specific embodiments. It should be noted, however, that the scope of the present invention is not limited to the following examples.
Example one
The implementation of the method of the present invention using a three-tank water tank as an example is as follows. Firstly, the proposed method is applied to establish a feasible working domain boundary model, wherein the size of the search cone step angle alpha can be determined according to the three indexes eta, sigma shown in FIG. 2c,εrAnd the target function F (alpha) is obtained through optimization, as can be seen from (d) in FIG. 2, when the step angle of the search cone is 1.4 degrees, the target function obtains the minimum value, and at the moment, the other three indexes eta, sigma are obtainedc,εrIs gotValues of 0.996, 0.0286 and 6.6753, respectively, are shown in FIG. 3 for the calculation of a feasible working domain boundary model.
Calculating a dynamic alarm threshold of a test data point according to the established feasible working domain boundary model, wherein in (a) in fig. 4, a monitoring variable is designed to firstly run in a normal state, namely inside the feasible working domain, then gradually run to the outside of the feasible working domain, and finally return to the inside of the feasible working domain through proper adjustment, and in (b) and (c) in fig. 4, a three-container water tank monitoring variable h is respectively used as a three-container water tank monitoring variable h1,h2The dynamic alarm threshold of (2). The liquid level monitoring variable h can be seen from the figure2The system passes through the upper limit of the dynamic alarm threshold value in 1371s and enters a non-feasible working domain, but the liquid level monitoring variable h is changed at the moment1Still within the dynamic alarm threshold range, it is clear that this multivariable system should trigger an alarm, but only monitor the variable h1Or a static alarm line is adopted to judge whether alarm occurs, so that false alarm and false alarm are easy to generate.
Example two
Taking data of a No. three unit A mill of the Huanengzhou power plant in 2018 years as an example, according to the analysis of industrial operation data of a coal mill, the relationship among the current of a discharging motor, the current of the grinding motor, the negative pressure of an inlet of a mill exhauster, the negative pressure of an inlet of the mill, the differential pressure of an inlet and an outlet of the mill, the coal feeding rate, the opening of a hot air valve of the mill, the opening of a circulating air valve of the mill, the outlet temperature of the mill and the inlet temperature of the mill exhauster is shown in the graph of FIG. 5 and FIG. 6 (each segment of data has two hours to 7200 data points).
The first step is as follows: preprocessing each variable in the graph, performing PLR segmentation on the 10 sample data by adopting a bottom-up piecewise linear representation method, and selecting five associated variables of coal feeding rate, coal mill inlet and outlet differential pressure, coal mill outlet temperature, motor discharge current and mill motor current to establish a feasible working domain boundary model of the coal mill pulverizing system according to a final judgment result of a trend change relation and an actual mechanism relation among the variables.
The second step is that: firstly, historical normal data of each associated variable is obtained, and motor current x is discharged1Current x of motor for grinding2Coal mill outlet temperature x3Inlet and outlet differential pressure of coal millx4Four monitoring variables and a certain coal feed rate x5After unreasonable data parameters deviating from the coal feeding rate set value are processed and removed from the (mean value 29.13t/h) variable, historical normal data points for establishing a feasible working domain boundary model can be obtained, and the result is shown in FIG. 7. The raw data was then normalized to normalized data with a mean of 0 and a variance of 1 using the Z-score normalization method.
And (5) calculating the hypersphere radius r capable of wrapping all normal data points by adopting the formula (4) and the formula (5). The step angle of the search cone is set to α, and then each index is calculated and the calculation result of the objective function F (α) in equation (6) is shown in fig. 8. Therefore, the deflection angle (step angle) of the optimal search cone is obtained to be 6.8 degrees, the model fitness index is 0.9092, the comprehensive standard deviation index value is 2.9637, and the radius difference index value is 41.2553. Then respectively making each angular coordinate phijAnd j is 1,2, …, n-1 is equal to 0 and alpha, the arrangement and combination modes of all the angular coordinates are gradually integrated into an expression (18) to obtain the vertex coordinates of the infinitesimal, and the vertex coordinates and the origin O form a vertex set V of the n-dimensional space search cone. The search cone expression is obtained by solving the data in the vertex set V into the equations (20), (21) and (22). Then by setting the (n-1) th angular coordinate phin-1And obtaining a feasible working domain boundary model M in a traversing search mode of the stepping alpha. The feasible working domain boundary model obtained finally is shown in fig. 9.
The third step: first, the original monitoring data point is obtained according to the formula (3)
Figure BDA0003404765200000181
Preprocessing as a normalized data point X (t)0). And (4) dividing the data into normal data points and abnormal data points according to the formula (35). Solving the abnormal data points in the joint type (26), (28), (32) and (33) to obtain the normal data point X corresponding to the abnormal data pointM. Then x is calculated according to equations (26) and (28)j(t0) Straight line Y corresponding to dynamic alarm thresholdjThen the straight line Y isjThe intersection point of the two is obtained by simultaneous solution with a feasible working domain boundary model, and the dynamic alarm threshold of the original data is obtained by substituting equations (30) and (31)
Figure BDA0003404765200000182
And
Figure BDA0003404765200000183
repeating the above steps to obtain data points X (t) by respectively changing j to 1,2, …, n0) All process variables xjThe dynamic alarm threshold of (2). The resulting dynamic alarm threshold is shown in fig. 10.
In fig. 10, the middle curve of each variable is the standardized monitoring operation value of the process variable, the upper and lower curves respectively correspond to the high and low alarm thresholds, and the thickened curve on the right side is in an abnormal working state. It can thus be seen that the monitoring data points first run within the feasible operating range and then start to gradually exceed their normal operating range at time t 4226s, i.e. the alarm system should issue an alarm signal at time t 4226 s.
The invention also provides the following product examples:
an industrial multivariable alarm system based on search cone feasible working domain modeling, comprising:
a monitoring variable determination module configured to determine an associated variable determined based on the trend change as a monitoring variable;
the data point judgment module is configured to preprocess the data points of the acquired monitoring variables, search whether a search cone can contain the preprocessed data points in a feasible working domain boundary model which is pre-constructed based on historical monitoring variable data, if so, the search cone is a normal data point, and otherwise, the search cone is an abnormal data point;
the dynamic alarm threshold value determining module is configured to determine a normal data point corresponding to the abnormal data point, calculate a straight line corresponding to the dynamic alarm threshold value according to the normal data point, calculate an intersection point of the straight line and the feasible working domain boundary model, and further determine the dynamic alarm threshold value;
and the alarm module is configured to trigger an alarm if the preprocessed data point is positioned outside the feasible working domain boundary model or a certain variable exceeds an operation range represented by the corresponding alarm threshold value, otherwise, the alarm is not triggered.
In the embodiment, the abnormal data points are calculated by judging whether the abnormal data points are in the search cone model, so that the calculation process is simplified, and the practicability of the model is improved.
Meanwhile, the data is screened by preprocessing the data, the direct relation of variable data is fully utilized, only the historical normal data points of the process variables are required to form the closed region of the convex hull model in the high-dimensional space, the defects that the process variables are limited in the boundary model of the feasible super-ellipsoid working domain to meet Gaussian distribution and the historical normal data points of the process variables are limited in the convex hull model to form the convex closed region in the high-dimensional space are overcome, and the universality is improved.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.

Claims (10)

1. An industrial multivariable alarm method based on search cone feasible working domain modeling is characterized in that: the method comprises the following steps:
determining an associated variable determined based on the trend change as a monitoring variable;
preprocessing the data points of the acquired monitoring variables, searching whether a search cone can contain the preprocessed data points or not in a feasible working domain boundary model which is pre-constructed based on historical monitoring variable data, if so, determining the search cone to be a normal data point, otherwise, determining the search cone to be an abnormal data point;
determining a normal data point corresponding to the abnormal data point;
calculating a straight line corresponding to the dynamic alarm threshold according to the normal data points, and calculating an intersection point of the straight line and the feasible working domain boundary model so as to determine the dynamic alarm threshold;
and if the preprocessed data point is positioned outside the feasible working domain boundary model or a certain variable exceeds the operation range represented by the corresponding alarm threshold value, triggering an alarm.
2. The industrial multivariable alarm method based on the search cone feasible working domain modeling as claimed in claim 1, wherein: the specific process of determining the associated variable determined based on the trend change as the monitoring variable includes: and performing piecewise linear representation on the sample data by using a bottom-up piecewise linear representation method to obtain the trend change relation of each variable, and further determining the monitoring variable according to the trend change relation and the actual relation among the variables.
3. The industrial multivariable alarm method based on the search cone feasible working domain modeling as claimed in claim 1, wherein: the specific process of pre-constructing the feasible working domain boundary model based on the historical monitoring variable data comprises the following steps:
preprocessing a monitoring variable normal data point in historical monitoring data, regarding a feasible working domain formed by monitoring variables as being formed by a plurality of search cones, and determining the side edge length of each search cone based on the preprocessed data;
setting the optimal step angle of the search cone, respectively selecting the value of each angular coordinate of the search cone, obtaining the vertex coordinate of the infinitesimal according to the arrangement and combination mode of all the angular coordinates, and forming a vertex set together with the origin;
and determining a search cone expression according to the vertex set, and obtaining a feasible working domain boundary model in a traversal search mode.
4. The industrial multivariable alarm method based on the search cone feasible working domain modeling as claimed in claim 3, characterized in that: the specific process of pretreatment comprises: and expressing the historical normal data points of the monitoring variables by using vectors, and standardizing the data of the monitoring variables by adopting a Z-score standardization method.
5. The industrial multivariable alarm method based on the search cone feasible working domain modeling as claimed in claim 3, characterized in that: the search cones are all represented by convex hulls in a high-dimensional space, and the vertex set of the convex hulls is composed of an origin and the vertices of a hypercube at the bottom of the search cones.
6. The industrial multivariable alarm method based on the search cone feasible working domain modeling as claimed in claim 3, characterized in that: the specific process of setting the step angle of the optimal search cone comprises the following steps:
defining the vertex angle of two adjacent edges of the search cone as the step angle alpha of the search cone, and determining the value of alpha when the minimum value is taken by the following objective function F (alpha):
F(α)＝(1-η)σcεr
wherein eta represents a fitness index of the feasible working domain boundary model; sigmacA composite standard deviation representing an alarm threshold for adjacent monitored data points; epsilonrRepresenting the radius difference index of adjacent search cones.
7. An industrial multivariable alarm method based on search cone feasible working domain modeling as claimed in claim 3 or 6, characterized in that: the specific process of respectively selecting the values of the angular coordinates of each pyramid of the search cone and obtaining the infinitesimal vertex coordinates according to the arrangement and combination mode of all the angular coordinates comprises the following steps: respectively making each angular coordinate equal to 0 and stepping angle alpha, searching angular coordinate [ phi ] of each edge of cone in n-dimensional space1,φ2,…,φn-1]In total 2n-1One of them is corresponding to 2 of search cone bottom surface infinitesimaln-1And obtaining the coordinate of the vertex of the infinitesimal according to the arrangement and combination mode of all the angular coordinates.
8. The industrial multivariable alarm method based on the search cone feasible working domain modeling as claimed in claim 3, characterized in that: according to the vertex set, the specific process for determining the search cone expression comprises the following steps: data points on all hyperplanes of the convex hull are found through a fast convex hull algorithm, an expression is used for representing the ith hyperplane which is enclosed into the convex hull, n data points on the ith hyperplane are indexed one by one and substituted into the expression, the normal vector and the hyperplane offset are obtained through solving, all hyperplanes enclosed into the convex hull are calculated one by one, the complete expression of the hyperplane is obtained, the search cone is represented, and the search cone expression is obtained.
9. The industrial multivariable alarm method based on the search cone feasible working domain modeling as claimed in claim 1, wherein: calculating a straight line corresponding to the dynamic alarm threshold, calculating an intersection point of the straight line and the feasible working domain boundary model, and further determining the dynamic alarm threshold comprises the following specific processes: according to a linear expression corresponding to a certain monitoring variable dynamic alarm threshold, the linear expression and a feasible working domain boundary model are combined to obtain the intersection point of the linear expression and the feasible working domain boundary model, and a set Z is formedjAccording to the set ZjThe maximum value and the minimum value determine the upper limit and the lower limit of the dynamic alarm threshold value.
10. An industrial multivariable alarm system based on search cone feasible working domain modeling is characterized in that: the method comprises the following steps:
a monitoring variable determination module configured to determine an associated variable determined based on the trend change as a monitoring variable;
the data point judgment module is configured to preprocess the data points of the acquired monitoring variables, search whether a search cone can contain the preprocessed data points in a feasible working domain boundary model which is pre-constructed based on historical monitoring variable data, if so, the search cone is a normal data point, and otherwise, the search cone is an abnormal data point;
the dynamic alarm threshold value determining module is configured to determine a normal data point corresponding to the abnormal data point, calculate a straight line corresponding to the dynamic alarm threshold value according to the normal data point, calculate an intersection point of the straight line and the feasible working domain boundary model, and further determine the dynamic alarm threshold value;
and the alarm module is configured to trigger an alarm if the preprocessed data point is positioned outside the feasible working domain boundary model or a certain variable exceeds an operation range represented by the corresponding alarm threshold value, otherwise, the alarm is not triggered.
CN202111509701.4A 2021-12-10 2021-12-10 Industrial multivariable alarm method and system based on search cone feasible working domain modeling Pending CN114416708A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111509701.4A CN114416708A (en) 2021-12-10 2021-12-10 Industrial multivariable alarm method and system based on search cone feasible working domain modeling

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111509701.4A CN114416708A (en) 2021-12-10 2021-12-10 Industrial multivariable alarm method and system based on search cone feasible working domain modeling

Publications (1)

Publication Number Publication Date
CN114416708A true CN114416708A (en) 2022-04-29

Family

ID=81266347

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111509701.4A Pending CN114416708A (en) 2021-12-10 2021-12-10 Industrial multivariable alarm method and system based on search cone feasible working domain modeling

Country Status (1)

Country Link
CN (1) CN114416708A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115905974A (en) * 2022-11-04 2023-04-04 北京科技大学 Method for detecting abnormal furnace condition of blast furnace
US20240005771A1 (en) * 2023-08-10 2024-01-04 Chengdu Qinchuan Iot Technology Co., Ltd. Methods and systems for gas leakage safety warning based on internet of things (iot) of smart gas

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115905974A (en) * 2022-11-04 2023-04-04 北京科技大学 Method for detecting abnormal furnace condition of blast furnace
CN115905974B (en) * 2022-11-04 2024-02-27 北京科技大学 Method for detecting abnormal furnace condition of blast furnace
US20240005771A1 (en) * 2023-08-10 2024-01-04 Chengdu Qinchuan Iot Technology Co., Ltd. Methods and systems for gas leakage safety warning based on internet of things (iot) of smart gas

Similar Documents

Publication Publication Date Title
CN114416708A (en) Industrial multivariable alarm method and system based on search cone feasible working domain modeling
CN105631596B (en) Equipment fault diagnosis method based on multi-dimensional piecewise fitting
Dou et al. An unsupervised online monitoring method for tool wear using a sparse auto-encoder
CN111648992B (en) Gas turbine compressor fault identification early warning method
CN114813105A (en) Gear box fault early warning method and system based on working condition similarity evaluation
CN110009126B (en) Online alarm analysis method based on fusion of PLS model and PCA contribution degree
Strączkiewicz et al. Supervised and unsupervised learning process in damage classification of rolling element bearings
CN114273977A (en) MES-based cutter wear detection method and system
Vishnu et al. Recurrent neural networks for online remaining useful life estimation in ion mill etching system
Guleria et al. Classification of surface roughness during turning of forged EN8 steel using vibration signal processing and support vector machine
CN114326486A (en) Process monitoring method based on probability slow feature analysis and elastic weight consolidation
CN108596364B (en) Dynamic early warning method for major hazard source in chemical industry park
Jiang et al. A SVDD and K‐Means Based Early Warning Method for Dual‐Rotor Equipment under Time‐Varying Operating Conditions
Ding et al. Deep Forest‐Based Fault Diagnosis Method for Chemical Process
Kühnert Data-driven methods for fault localization in process technology
CN114997309A (en) Water feed pump fault early warning method and device
Kong et al. Dynamic early warning method for major hazard installation systems in chemical industrial park
Lu et al. Visual monitoring of industrial operation states based on kernel fisher vector and self-organizing map networks
Wang et al. SPA-based modified local reachability density ratio wSVDD for nonlinear multimode process monitoring
Amiri et al. Economic-statistical design of adaptive X-bar control chart: a Taguchi loss function approach
Xu et al. An efficient method for online identification of steady state for multivariate system
CN110543908B (en) Control chart pattern recognition method based on dynamic observation window
Pei et al. Fault detection in chemical processes using discriminant analysis and control chart
Zhang et al. Equipment health assessment and fault-early warning algorithm based on improved SVDD
Castillo et al. A hybrid fuzzy‐fractal approach for time series analysis and plant monitoring

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