CN112749370A - Fault tracking and positioning method and system based on Internet of things - Google Patents

Fault tracking and positioning method and system based on Internet of things Download PDF

Info

Publication number
CN112749370A
CN112749370A CN202110364634.5A CN202110364634A CN112749370A CN 112749370 A CN112749370 A CN 112749370A CN 202110364634 A CN202110364634 A CN 202110364634A CN 112749370 A CN112749370 A CN 112749370A
Authority
CN
China
Prior art keywords
fault
variable
tracking
algorithm
random forest
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.)
Granted
Application number
CN202110364634.5A
Other languages
Chinese (zh)
Other versions
CN112749370B (en
Inventor
高美婵
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.)
Guangdong International Science And Technology Co ltd
Original Assignee
Guangdong International Science And Technology Co ltd
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 Guangdong International Science And Technology Co ltd filed Critical Guangdong International Science And Technology Co ltd
Priority to CN202110364634.5A priority Critical patent/CN112749370B/en
Publication of CN112749370A publication Critical patent/CN112749370A/en
Application granted granted Critical
Publication of CN112749370B publication Critical patent/CN112749370B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
    • G16Y40/00—IoT characterised by the purpose of the information processing
    • G16Y40/10—Detection; Monitoring
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
    • G16Y40/00—IoT characterised by the purpose of the information processing
    • G16Y40/20—Analytics; Diagnosis
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
    • G16Y40/00—IoT characterised by the purpose of the information processing
    • G16Y40/60—Positioning; Navigation

Abstract

The invention provides a fault tracking and positioning method based on the Internet of things, which comprises the following steps: constructing a multi-dynamic sensor network; coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data; establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable; carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section; constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section; the variable importance measurement of the process variable is obtained through a random forest classification regression algorithm, and the fault variable is determined according to the variable importance measurement to carry out fault positioning.

Description

Fault tracking and positioning method and system based on Internet of things
Technical Field
The invention relates to the field of analysis of the Internet of things, in particular to a fault tracking and positioning method and system based on the Internet of things.
Background
The internet of things is a concept that has been gradually developed in recent years, and it connects people and things in the real world through various sensing devices and networks. Due to the ubiquitous nature of equipment and the target sensing capability of the equipment, the internet of things can be applied to aerospace, intelligent buildings, medical health, production and manufacturing, military, environment monitoring, entertainment multimedia, transportation and the like. The intelligent traffic system based on the internet of things technology can avoid traffic jam and utilize a road network to the maximum extent. One basis of the above various applications is to realize target positioning and tracking based on the internet of things, for example, in the medical health field, the geographical position of a special patient needs to be positioned; in the monitoring field, the action track of personnel entering a monitoring area needs to be tracked; in the military field, various military forces of enemies, such as tanks, infantries and the like, are positioned and tracked.
However, in the present stage, based on radar and a sensor network, people propose a plurality of effective target positioning and tracking algorithms, but the situation requires that a target is constantly monitored by sensing equipment, and the positioning and tracking accuracy is greatly reduced under the condition that the sensing equipment is sparse or damaged.
Disclosure of Invention
The invention mainly aims to overcome the defects in the prior art, provides a fault tracking and positioning method and system based on the Internet of things, constructs a multi-dynamic sensor network, enlarges the sensing and observation space, provides a coupled anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm, and realizes comprehensive fault search and tracking; and the method provided by the invention has higher tracking and positioning precision and has obvious superiority by combining the weight clustering analysis model and the random forest classification regression algorithm to accurately position the fault.
The invention adopts the following technical scheme:
a fault tracking and positioning method based on the Internet of things comprises the following steps:
constructing a multi-dynamic sensor network;
coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data;
establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable;
carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section;
constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section;
and obtaining variable importance measurement of the process variable through a random forest classification regression algorithm, and determining a fault variable according to the variable importance measurement to perform fault positioning.
Specifically, the constructing of the multi-dynamic sensor network specifically includes:
the observation space of a single sensor is
Figure 571041DEST_PATH_IMAGE002
；
Figure 34996DEST_PATH_IMAGE004
Single sensor acquisition
Figure 932545DEST_PATH_IMAGE006
The number of the measured values is,
Figure 521789DEST_PATH_IMAGE007
are integers, each is
Figure 911313DEST_PATH_IMAGE009
；
The observation space of the multi-dynamic sensor network constructed by S sensors is
Figure 40943DEST_PATH_IMAGE011
：
Figure 954452DEST_PATH_IMAGE013
Wherein
Figure 714598DEST_PATH_IMAGE015
Representing mutual exclusion of each observation space, wherein S is an integer;
Figure 450472DEST_PATH_IMAGE017
respectively representing the detection spaces of the sensors;
the total measurement set was:
Figure 993580DEST_PATH_IMAGE019
。
specifically, the coupling anti-bee-crowding distributed collaborative search algorithm and the multi-bernoulli filtering algorithm realize fault tracking and acquire real-time sensor data, and specifically include:
in the coupled anti-bee-crowding distributed collaborative search algorithm and the multi-Bernoulli filtering algorithm, the construction of the sensor nodes in the multi-dynamic sensor network comprises two modes of searching and tracking:
when the sensor network estimates a new target, the sensor node executing the search mode sets the type of the sensor node as an estimated position, and enters a real-time tracking mode;
when the sensor network does not detect a new target, the search mode is switched to, the type of the sensor node is set as the origin, the position is updated according to the rules of the coupled anti-bee-crowding distributed collaborative search algorithm and the multi-Bernoulli filtering algorithm, and an unknown area is continuously searched to detect the target which possibly appears in the area.
Specifically, a weight cluster analysis model is established for the sensor data of the historical normal samples, and the control limit of the controlled variable is calculated, wherein the controlled variable specifically comprises:
SPE control quantity:
Figure 537825DEST_PATH_IMAGE021
Figure 669205DEST_PATH_IMAGE023
control amount:
Figure 361217DEST_PATH_IMAGE025
Figure 973595DEST_PATH_IMAGE027
control amount:
Figure 372347DEST_PATH_IMAGE029
wherein the content of the first and second substances,
Figure 411978DEST_PATH_IMAGE031
in order to be a sample of the sample,
Figure 526040DEST_PATH_IMAGE033
is the average of the samples and is,
Figure 942109DEST_PATH_IMAGE035
is the projection matrix of the residual space, P is the principal component vector,
Figure 523263DEST_PATH_IMAGE037
is a diagonal matrix formed by the principal elements,
Figure 671479DEST_PATH_IMAGE039
is the control limit of the SPE control quantity,
Figure 741679DEST_PATH_IMAGE041
is composed of
Figure 227018DEST_PATH_IMAGE023
The control limit of (a) is set,
Figure 193837DEST_PATH_IMAGE043
to disrupt the variability of normal process dependence,
Figure 44112DEST_PATH_IMAGE045
Is a measure of the magnitude of linear correlation between variables.
Specifically, a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section is constructed, and the random forest classification regression algorithm specifically comprises the following steps:
the random forest classification regression algorithm comprises at least two classification decision trees;
resampling a training sample set and a corresponding external data set by using bootstrap, wherein the external data set is used for carrying out model test and variable importance measurement;
generating a classification decision tree for each group of training samples in the training sample set by using a node random splitting technology;
and the generated classification decision trees form a random forest classification regression algorithm model, and when a sample to be detected is predicted, the mean value of the predicted values of the classification decision trees is used as a final prediction result.
Specifically, variable importance measurement of a process variable is obtained through a random forest classification regression algorithm, a fault variable is determined according to the variable importance measurement for fault positioning, and the variable importance measurement calculation specifically comprises the following steps:
to the established random forest classification regression algorithm model
Figure 932434DEST_PATH_IMAGE047
H is the parameter in the model, b is the number of decision trees, the external data set data is brought into the corresponding decision tree for prediction to obtain the mean square error
Figure 952955DEST_PATH_IMAGE049
The mean square error is defined as:
Figure 508701DEST_PATH_IMAGE051
in the formula (I), the compound is shown in the specification,
Figure 529878DEST_PATH_IMAGE053
and
Figure 436654DEST_PATH_IMAGE055
i =1,2.. n, n being the number of samples;
centralizing variables in the extrinsic data
Figure 263796DEST_PATH_IMAGE057
Randomly replacing, and then introducing into corresponding decision tree for prediction to obtain the replaced mean square error
Figure 546485DEST_PATH_IMAGE059
Wherein
Figure 128776DEST_PATH_IMAGE007
Is a variable number;
computing decision tree variables
Figure 100002_DEST_PATH_IMAGE061
Difference of mean square error before and after random permutation VI:
Figure DEST_PATH_IMAGE063
variables of
Figure 70318DEST_PATH_IMAGE064
The variable importance measure VI':
Figure 100002_DEST_PATH_IMAGE066
。
specifically, variable importance measurement of a process variable is obtained through a random forest classification regression algorithm, and a fault variable is determined according to the variable importance measurement to perform fault location, wherein the method specifically comprises the following steps:
acquiring a variable importance measurement coefficient value of the process variable through a random forest classification regression algorithm;
and determining the variable with the maximum variable importance measurement coefficient value as a fault variable to realize fault positioning.
In another aspect, an embodiment of the present invention provides a fault tracking and positioning system based on the internet of things, including:
a sensor network construction unit: constructing a multi-dynamic sensor network;
a fault tracking unit: coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data;
a weight cluster analysis model establishing unit: establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable;
screening fault units: carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section;
a classification algorithm construction unit: constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section;
a fault location unit: and obtaining variable importance measurement of the process variable through a random forest classification regression algorithm, and determining a fault variable according to the variable importance measurement to perform fault positioning.
Specifically, the sensor network constructing unit constructs a multi-dynamic sensor network, specifically: the observation space of a single sensor is
Figure 104746DEST_PATH_IMAGE002
；
Figure 103926DEST_PATH_IMAGE004
Single sensor acquisition
Figure 529222DEST_PATH_IMAGE006
The number of the measured values is,
Figure 285957DEST_PATH_IMAGE007
are integers, each is
Figure 451971DEST_PATH_IMAGE009
；
Observation of multiple dynamic sensor networks constructed with S sensorsSpace is
Figure 243341DEST_PATH_IMAGE011
：
Figure 980484DEST_PATH_IMAGE067
Wherein
Figure 21252DEST_PATH_IMAGE015
Representing mutual exclusion of each observation space, wherein S is an integer;
Figure 279974DEST_PATH_IMAGE017
respectively representing the detection spaces of the sensors;
the total measurement set was:
Figure 457009DEST_PATH_IMAGE068
。
specifically, the fault positioning unit acquires variable importance measurement of a process variable through a random forest classification regression algorithm, and determines a fault variable according to the variable importance measurement to perform fault positioning; the variable importance measure calculation specifically includes:
to the established random forest classification regression algorithm model
Figure 552004DEST_PATH_IMAGE047
H is the parameter in the model, b is the number of decision trees, the external data set data is brought into the corresponding decision tree for prediction to obtain the mean square error
Figure 17752DEST_PATH_IMAGE049
The mean square error is defined as:
Figure 856395DEST_PATH_IMAGE069
in the formula (I), the compound is shown in the specification,
Figure 353847DEST_PATH_IMAGE053
and
Figure 619744DEST_PATH_IMAGE055
i =1,2.. n, n being the number of samples;
centralizing variables in the extrinsic data
Figure 900683DEST_PATH_IMAGE057
Randomly replacing, and then introducing into corresponding decision tree for prediction to obtain the replaced mean square error
Figure 418384DEST_PATH_IMAGE070
Wherein
Figure 101169DEST_PATH_IMAGE007
Is a variable number;
computing decision tree variables
Figure 803546DEST_PATH_IMAGE061
Difference of mean square error before and after random permutation VI:
Figure 100002_DEST_PATH_IMAGE071
variables of
Figure 178639DEST_PATH_IMAGE064
Of the variable importance measure
Figure 100002_DEST_PATH_IMAGE073
：
Figure 296767DEST_PATH_IMAGE074
。
As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following advantages:
1) the invention provides a fault tracking and positioning method based on the Internet of things, which specifically comprises the steps of constructing a multi-dynamic sensor network; coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data; establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable; carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section; constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section; obtaining variable importance measurement of the process variable through a random forest classification regression algorithm, and determining a fault variable according to the variable importance measurement to perform fault positioning; the method constructs a multi-dynamic sensor network, enlarges the sensing and observation space, provides a coupled anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm, and realizes the comprehensive fault search and tracking; and combining a weight clustering analysis model and a random forest classification regression algorithm to accurately position faults.
2) According to the invention, a multi-dynamic sensor network is introduced, and when a target is searched in a given interested area, a plurality of dynamic sensor nodes can be optimally controlled to realize the maximization of the coverage area through the multi-dynamic sensor network, so that a basis is provided for subsequent accurate tracking and positioning.
3) The invention provides a coupled anti-bee-crowding distributed collaborative searching algorithm and a multi-Bernoulli filtering algorithm, even if the scale of a multi-dynamic sensor network is increased, the coverage rate of a target can still be stable, and the comprehensive and effective searching and tracking of multiple targets in multiple areas can be realized.
4) The invention provides a method for combining a weight clustering analysis model and a random forest classification regression algorithm, wherein a causal relationship coefficient of a process variable to a control variable is obtained by using variable importance measurement of the random forest classification regression algorithm, and a variable with the maximum value is identified as a fault variable; parameters do not need to be optimized and selected, a good model can be established for a small sample set, and the positioning effect has obvious superiority.
Drawings
Fig. 1 is a flowchart of a fault tracking and positioning method based on the internet of things according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of tracking simulation implemented by coupling an anti-bee-crowding distributed collaborative search algorithm with a multi-Bernoulli filtering algorithm;
fig. 3 is a comparative example of fault location implemented by the method of the present invention and a conventional method according to an embodiment of the present invention, where (a) is a schematic diagram of a result of fault location implemented by the method of the present invention, and (b) is a schematic diagram of a result of fault location implemented by the conventional method;
fig. 4 is a structural diagram of a fault tracking and positioning system based on the internet of things according to an embodiment of the present invention.
The invention is described in further detail below with reference to the figures and specific examples.
Detailed Description
The invention provides a fault tracking and positioning method and system based on the Internet of things, a multi-dynamic sensor network is constructed, sensing and observation spaces are enlarged, a coupled anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm are provided, and comprehensive fault searching and tracking are realized; and the method provided by the invention has higher tracking and positioning precision and has obvious superiority by combining the weight clustering analysis model and the random forest classification regression algorithm to accurately position the fault.
As shown in fig. 1, an embodiment of the present invention provides a flow chart of a fault tracking and locating method based on the internet of things, including the following steps:
s101: constructing a multi-dynamic sensor network;
specifically, the constructing of the multi-dynamic sensor network specifically includes:
the observation space of a single sensor is
Figure 771742DEST_PATH_IMAGE002
；
Figure 582703DEST_PATH_IMAGE004
Single sensor acquisition
Figure 772989DEST_PATH_IMAGE006
The number of the measured values is,
Figure 694808DEST_PATH_IMAGE007
are integers, each is
Figure 961973DEST_PATH_IMAGE009
；
The observation space of the multi-dynamic sensor network constructed by S sensors is
Figure 147098DEST_PATH_IMAGE011
：
Figure DEST_PATH_IMAGE075
Wherein
Figure 700045DEST_PATH_IMAGE015
Representing mutual exclusion of each observation space, wherein S is an integer;
Figure 487873DEST_PATH_IMAGE076
respectively representing the detection spaces of the sensors;
the total measurement set was:
Figure 671861DEST_PATH_IMAGE068
。
s102: coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data;
specifically, the coupling anti-bee-crowding distributed collaborative search algorithm and the multi-bernoulli filtering algorithm realize fault tracking and acquire real-time sensor data, and specifically include:
in the coupled anti-bee-crowding distributed collaborative search algorithm and the multi-Bernoulli filtering algorithm, the construction of the sensor nodes in the multi-dynamic sensor network comprises two modes of searching and tracking:
when the sensor network estimates a new target, the sensor node executing the search mode sets the type of the sensor node as an estimated position, and enters a real-time tracking mode;
when the sensor network does not detect a new target, the search mode is switched to, the type of the sensor node is set as the origin, the position is updated according to the rules of the coupled anti-bee-crowding distributed collaborative search algorithm and the multi-Bernoulli filtering algorithm, and an unknown area is continuously searched to detect the target which possibly appears in the area.
As shown in fig. 2, a schematic diagram for implementing tracking simulation by coupling the anti-bee-congestion distributed collaborative search algorithm and the multi-bernoulli filtering algorithm, wherein a large circle represents a field range of the node, and "+" represents an object to be searched; the octagon represents a mobile sensor node.
S103: establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable;
specifically, a weight cluster analysis model is established for the sensor data of the historical normal samples, and the control limit of the controlled variable is calculated, wherein the controlled variable specifically comprises:
SPE control quantity:
Figure 621362DEST_PATH_IMAGE077
Figure 255081DEST_PATH_IMAGE023
control amount:
Figure DEST_PATH_IMAGE078
Figure 659649DEST_PATH_IMAGE027
control amount:
Figure 353935DEST_PATH_IMAGE079
wherein the content of the first and second substances,
Figure 677601DEST_PATH_IMAGE031
in order to be a sample of the sample,
Figure 64195DEST_PATH_IMAGE033
is the average of the samples and is,
Figure 131508DEST_PATH_IMAGE035
is the projection matrix of the residual space, P is the principal component vector,
Figure 680301DEST_PATH_IMAGE037
is a diagonal matrix formed by the principal elements,
Figure 174868DEST_PATH_IMAGE039
is the control limit of the SPE control quantity,
Figure 51688DEST_PATH_IMAGE041
is composed of
Figure 922692DEST_PATH_IMAGE023
The control limit of (a) is set,
Figure 325991DEST_PATH_IMAGE043
to disrupt the variability of normal process dependence,
Figure 660633DEST_PATH_IMAGE045
Is a measure of the magnitude of linear correlation between variables.
S104: carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section;
s105: constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section;
specifically, a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section is constructed, and the random forest classification regression algorithm specifically comprises the following steps:
the random forest classification regression algorithm comprises at least two classification decision trees;
resampling a training sample set and a corresponding external data set by using bootstrap, wherein the external data set is used for carrying out model test and variable importance measurement;
generating a classification decision tree for each group of training samples in the training sample set by using a node random splitting technology;
and the generated classification decision trees form a random forest classification regression algorithm model, and when a sample to be detected is predicted, the mean value of the predicted values of the classification decision trees is used as a final prediction result.
S106: and obtaining variable importance measurement of the process variable through a random forest classification regression algorithm, and determining a fault variable according to the variable importance measurement to perform fault positioning.
Specifically, variable importance measurement of a process variable is obtained through a random forest classification regression algorithm, a fault variable is determined according to the variable importance measurement for fault positioning, and the variable importance measurement calculation specifically comprises the following steps:
to the established random forest classification regression algorithm model
Figure 883804DEST_PATH_IMAGE047
H is the parameter in the model, b is the number of decision trees, the external data set data is taken into the corresponding decision tree for prediction to obtain the mean square error
Figure 558499DEST_PATH_IMAGE049
The mean square error is defined as:
Figure 222830DEST_PATH_IMAGE080
in the formula (I), the compound is shown in the specification,
Figure 689494DEST_PATH_IMAGE081
and
Figure 806485DEST_PATH_IMAGE055
i =1,2.. n, n being the number of samples;
centralizing variables in the extrinsic data
Figure 753713DEST_PATH_IMAGE057
Randomly replacing, and then introducing into corresponding decision tree for prediction to obtain the replaced mean square error
Figure DEST_PATH_IMAGE082
Wherein
Figure 82669DEST_PATH_IMAGE007
Is a variable number;
computing decision tree variables
Figure 824360DEST_PATH_IMAGE061
Difference of mean square error before and after random permutation VI:
Figure DEST_PATH_IMAGE083
variables of
Figure 428648DEST_PATH_IMAGE064
The variable importance measure VI':
Figure DEST_PATH_IMAGE084
。
specifically, variable importance measurement of a process variable is obtained through a random forest classification regression algorithm, and a fault variable is determined according to the variable importance measurement to perform fault location, wherein the method specifically comprises the following steps:
acquiring a variable importance measurement coefficient value of the process variable through a random forest classification regression algorithm;
and determining the variable with the maximum variable importance measurement coefficient value as a fault variable to realize fault positioning.
Fig. 3 is a schematic diagram of a result of fault location implemented by the method of the present invention and a schematic diagram of a result of fault location implemented by the conventional method, where fig. (a) is a schematic diagram of a result of fault location implemented by the method of the present invention, and fig. (b) is a schematic diagram of a result of fault location implemented by the conventional method, and it can be seen from the diagrams that, although both methods implement identification of a fault variable 9, the effect of locating the variable 9 by the method of the present invention is more obvious and more prominent.
As shown in fig. 4, another aspect of the embodiments of the present invention provides a fault tracking and locating system based on the internet of things, including:
the sensor network construction unit 401: constructing a multi-dynamic sensor network;
specifically, the sensor network constructing unit 401 constructs a multi-dynamic sensor network, specifically: the observation space of a single sensor is
Figure 989686DEST_PATH_IMAGE002
；
Figure 690926DEST_PATH_IMAGE004
Single sensor acquisition
Figure 806781DEST_PATH_IMAGE006
The number of the measured values is,
Figure 491840DEST_PATH_IMAGE007
are integers, each is
Figure 46449DEST_PATH_IMAGE009
；
The observation space of the multi-dynamic sensor network constructed by S sensors is
Figure 336616DEST_PATH_IMAGE011
：
Figure 886022DEST_PATH_IMAGE013
Wherein
Figure 996060DEST_PATH_IMAGE015
Representing mutual exclusion of each observation space, wherein S is an integer;
Figure 416677DEST_PATH_IMAGE017
respectively representing the detection spaces of the sensors;
the total measurement set was:
Figure 295772DEST_PATH_IMAGE019
。
the fault tracking unit 402: coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data;
weight cluster analysis model creation unit 403: establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable;
screening failure unit 404: carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section;
classification algorithm construction unit 405: constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section;
the fault locating unit 406: and obtaining variable importance measurement of the process variable through a random forest classification regression algorithm, and determining a fault variable according to the variable importance measurement to perform fault positioning.
Specifically, the fault location unit 406 obtains a variable importance measure of the process variable through a random forest classification regression algorithm, and determines a fault variable according to the variable importance measure to perform fault location; the variable importance measure calculation specifically includes:
to the established random forest classification regression algorithm model
Figure DEST_PATH_IMAGE085
H is the parameter in the model, b is the number of decision trees, the external data set data is brought into the corresponding decision tree for prediction to obtain the mean square error
Figure 691112DEST_PATH_IMAGE049
The mean square error is defined as:
Figure DEST_PATH_IMAGE086
in the formula (I), the compound is shown in the specification,
Figure 567408DEST_PATH_IMAGE081
and
Figure 994978DEST_PATH_IMAGE055
i =1,2.. n, n being the number of samples;
centralizing variables in the extrinsic data
Figure 603945DEST_PATH_IMAGE057
Randomly replacing, and then introducing into corresponding decision tree for prediction to obtain the replaced mean square error
Figure DEST_PATH_IMAGE087
Wherein
Figure DEST_PATH_IMAGE088
Is a variable number;
computing decision tree variables
Figure 511465DEST_PATH_IMAGE061
Difference of mean square error before and after random permutation VI:
Figure 392833DEST_PATH_IMAGE083
variables of
Figure 762110DEST_PATH_IMAGE064
The variable importance measure VI':
Figure 412534DEST_PATH_IMAGE084
。
the invention provides a fault tracking and positioning method based on the Internet of things, which specifically comprises the steps of constructing a multi-dynamic sensor network; coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data; establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable; carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section; constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section; obtaining variable importance measurement of the process variable through a random forest classification regression algorithm, and determining a fault variable according to the variable importance measurement to perform fault positioning; the method constructs a multi-dynamic sensor network, enlarges the sensing and observation space, provides a coupled anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm, and realizes the comprehensive fault search and tracking; and combining a weight clustering analysis model and a random forest classification regression algorithm to accurately position faults.
According to the invention, a multi-dynamic sensor network is introduced, and when a target is searched in a given interested area, a plurality of dynamic sensor nodes can be optimally controlled to realize the maximization of the coverage area through the multi-dynamic sensor network, so that a basis is provided for subsequent accurate tracking and positioning.
The invention provides a coupled anti-bee-crowding distributed collaborative searching algorithm and a multi-Bernoulli filtering algorithm, even if the scale of a multi-dynamic sensor network is increased, the coverage rate of a target can still be stable, and the comprehensive and effective searching and tracking of multiple targets in multiple areas can be realized.
The invention provides a method for combining a weight clustering analysis model and a random forest classification regression algorithm, wherein a causal relationship coefficient of a process variable to a control variable is obtained by using variable importance measurement of the random forest classification regression algorithm, and a variable with the maximum value is identified as a fault variable; parameters do not need to be optimized and selected, a good model can be established for a small sample set, and the positioning effect has obvious superiority.
The above description is only an embodiment of the present invention, but the design concept of the present invention is not limited thereto, and any insubstantial modifications made by using the design concept should fall within the scope of infringing the present invention.

Claims (10)

1. A fault tracking and positioning method based on the Internet of things is characterized by comprising the following steps:
constructing a multi-dynamic sensor network;
coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data;
establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable;
carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section;
constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section;
and obtaining variable importance measurement of the process variable through a random forest classification regression algorithm, and determining a fault variable according to the variable importance measurement to perform fault positioning.
2. The method for fault tracking and locating based on the internet of things as claimed in claim 1, wherein the multi-dynamic sensor network is constructed, in particular
The observation space of a single sensor is
Figure 54491DEST_PATH_IMAGE002
；
Figure 257940DEST_PATH_IMAGE004
Single sensor acquisition
Figure 113507DEST_PATH_IMAGE006
The number of the measured values is,
Figure 353995DEST_PATH_IMAGE007
are integers, each is
Figure 898109DEST_PATH_IMAGE009
；
The observation space of the multi-dynamic sensor network constructed by S sensors is
Figure 734609DEST_PATH_IMAGE011
：
Figure 290355DEST_PATH_IMAGE013
Wherein
Figure 560800DEST_PATH_IMAGE015
Representing mutual exclusion of each observation space, wherein S is an integer;
Figure 973237DEST_PATH_IMAGE017
respectively representing the detection spaces of the sensors;
the total measurement set was:
Figure 597116DEST_PATH_IMAGE019
。
3. the method for tracking and locating the fault based on the internet of things as claimed in claim 1, wherein the coupling anti-bee-crowding distributed collaborative search algorithm and the multi-bernoulli filtering algorithm are used for realizing fault tracking and acquiring real-time sensor data, and specifically comprises:
in the coupled anti-bee-crowding distributed collaborative search algorithm and the multi-Bernoulli filtering algorithm, the construction of the sensor nodes in the multi-dynamic sensor network comprises two modes of searching and tracking:
when the sensor network estimates a new target, the sensor node executing the search mode sets the type of the sensor node as an estimated position, and enters a real-time tracking mode;
when the sensor network does not detect a new target, the search mode is switched to, the type of the sensor node is set as the origin, the position is updated according to the rules of the coupled anti-bee-crowding distributed collaborative search algorithm and the multi-Bernoulli filtering algorithm, and an unknown area is continuously searched to detect the target which possibly appears in the area.
4. The method for tracking and positioning the fault based on the internet of things as claimed in claim 1, wherein a weight cluster analysis model is established for sensor data of historical normal samples, and a control limit of a controlled quantity is calculated, wherein the controlled quantity specifically comprises:
SPE control quantity:
Figure 256637DEST_PATH_IMAGE021
Figure 573348DEST_PATH_IMAGE023
control amount:
Figure 327940DEST_PATH_IMAGE025
Figure 4778DEST_PATH_IMAGE027
control amount:
Figure 3958DEST_PATH_IMAGE029
wherein the content of the first and second substances,
Figure 504953DEST_PATH_IMAGE031
in order to be a sample of the sample,
Figure 245376DEST_PATH_IMAGE033
is the average of the samples and is,
Figure 476637DEST_PATH_IMAGE035
is the projection matrix of the residual space, P is the principal component vector,
Figure 612215DEST_PATH_IMAGE037
is a diagonal matrix formed by the principal elements,
Figure 5150DEST_PATH_IMAGE039
is the control limit of the SPE control quantity,
Figure 232869DEST_PATH_IMAGE041
is composed of
Figure 15624DEST_PATH_IMAGE023
A control limit of (d);
Figure 910767DEST_PATH_IMAGE043
to disrupt the variability of normal process dependence,
Figure 474604DEST_PATH_IMAGE045
Is a measure of the magnitude of linear correlation between variables.
5. The method for tracking and positioning the fault based on the internet of things as claimed in claim 1, wherein a random forest classification regression algorithm based on the process variable and the control quantity of the fault data section is constructed, and the random forest classification regression algorithm specifically comprises the following steps:
the random forest classification regression algorithm comprises at least two classification decision trees;
resampling a training sample set and a corresponding external data set by using bootstrap, wherein the external data set is used for carrying out model test and variable importance measurement;
generating a classification decision tree for each group of training samples in the training sample set by using a node random splitting technology;
and the generated classification decision trees form a random forest classification regression algorithm model, and when a sample to be detected is predicted, the mean value of the predicted values of the classification decision trees is used as a final prediction result.
6. The method for tracking and positioning the fault based on the internet of things as claimed in claim 1, wherein a variable importance measure of a process variable is obtained through a random forest classification regression algorithm, and the fault variable is determined according to the variable importance measure for fault positioning, wherein the variable importance measure specifically comprises the following steps:
to the established random forest classification regression algorithm model
Figure 815718DEST_PATH_IMAGE047
H is the parameter in the model, b is the number of decision trees, the external data set data is taken into the corresponding decision tree for prediction to obtain the mean square error
Figure 388781DEST_PATH_IMAGE049
The mean square error is defined as:
Figure 607273DEST_PATH_IMAGE051
in the formula (I), the compound is shown in the specification,
Figure 355393DEST_PATH_IMAGE053
and
Figure 433070DEST_PATH_IMAGE055
i =1,2.. n, n being the number of samples;
centralizing variables in the extrinsic data
Figure 465617DEST_PATH_IMAGE057
Randomly replacing, and then introducing into corresponding decision tree for prediction to obtain the replaced mean square error
Figure DEST_PATH_IMAGE059
Wherein
Figure 23769DEST_PATH_IMAGE007
Is a variable number;
computing decision tree variables
Figure DEST_PATH_IMAGE061
Difference of mean square error before and after random permutation VI:
Figure 818156DEST_PATH_IMAGE063
variables of
Figure DEST_PATH_IMAGE064
Of the variable importance measure
Figure DEST_PATH_IMAGE066
：
Figure DEST_PATH_IMAGE068
。
7. The Internet of things-based fault tracking and positioning method according to claim 1, characterized in that variable importance measurement of process variables is obtained through a random forest classification regression algorithm, fault variables are determined according to the variable importance measurement for fault positioning, and specifically:
acquiring a variable importance measurement coefficient value of the process variable through a random forest classification regression algorithm;
and determining the variable with the maximum variable importance measurement coefficient value as a fault variable to realize fault positioning.
8. A fault tracking and positioning system based on the Internet of things is characterized by comprising:
a sensor network construction unit: constructing a multi-dynamic sensor network;
a fault tracking unit: coupling an anti-bee-crowding distributed collaborative search algorithm and a multi-Bernoulli filtering algorithm to realize fault tracking and acquire real-time sensor data;
a weight cluster analysis model establishing unit: establishing a weight clustering analysis model for the sensor data of the historical normal sample, and calculating to obtain a control limit of the controlled variable;
screening fault units: carrying out fault detection on the acquired real-time sensor data according to the established weight clustering analysis model and the control quantity, screening out a fault data section and obtaining a control quantity corresponding to the fault data section;
a classification algorithm construction unit: constructing a random forest classification regression algorithm based on the process variables and the control quantity of the fault data section;
a fault location unit: and obtaining variable importance measurement of the process variable through a random forest classification regression algorithm, and determining a fault variable according to the variable importance measurement to perform fault positioning.
9. The system according to claim 8, wherein the sensor network constructing unit constructs a multi-dynamic sensor network, specifically:
the observation space of a single sensor is
Figure 117550DEST_PATH_IMAGE002
；
Figure 655586DEST_PATH_IMAGE004
Single sensor acquisition
Figure 989615DEST_PATH_IMAGE006
The number of the measured values is,
Figure 581002DEST_PATH_IMAGE007
are integers, each is
Figure 524950DEST_PATH_IMAGE009
；
The observation space of the multi-dynamic sensor network constructed by S sensors is
Figure 977928DEST_PATH_IMAGE011
：
Figure 291097DEST_PATH_IMAGE069
Wherein
Figure 551921DEST_PATH_IMAGE015
Representing mutual exclusion of each observation space, wherein S is an integer;
Figure DEST_PATH_IMAGE070
respectively representing the detection spaces of the sensors;
the total measurement set was:
Figure DEST_PATH_IMAGE071
。
10. the Internet of things-based fault tracking and positioning system according to claim 8, wherein the fault positioning unit obtains variable importance measurement of process variables through a random forest classification regression algorithm, and determines fault variables according to the variable importance measurement to perform fault positioning; the variable importance measure calculation specifically includes:
to the established random forest classification regression algorithm model
Figure DEST_PATH_IMAGE072
H is the parameter in the model, b is the number of decision trees, the external data set data is taken into the corresponding decision tree for prediction to obtain the mean square error
Figure 576640DEST_PATH_IMAGE049
The mean square error is defined as:
Figure DEST_PATH_IMAGE073
in the formula (I), the compound is shown in the specification,
Figure 174587DEST_PATH_IMAGE053
and
Figure 217629DEST_PATH_IMAGE055
i =1,2.. n, n being the number of samples;
centralizing variables in the extrinsic data
Figure 291765DEST_PATH_IMAGE057
Randomly replacing, and then introducing into corresponding decision tree for prediction to obtain the replaced mean square error
Figure 803780DEST_PATH_IMAGE059
Wherein
Figure 864140DEST_PATH_IMAGE007
Is a variable number;
computing decision tree variables
Figure 886322DEST_PATH_IMAGE061
Difference of mean square error before and after random permutation VI:
Figure 512386DEST_PATH_IMAGE063
variables of
Figure 760965DEST_PATH_IMAGE064
Of the variable importance measure
Figure 608704DEST_PATH_IMAGE066
：
Figure DEST_PATH_IMAGE074
。
CN202110364634.5A 2021-04-06 2021-04-06 Fault tracking and positioning method and system based on Internet of things Active CN112749370B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110364634.5A CN112749370B (en) 2021-04-06 2021-04-06 Fault tracking and positioning method and system based on Internet of things

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110364634.5A CN112749370B (en) 2021-04-06 2021-04-06 Fault tracking and positioning method and system based on Internet of things

Publications (2)

Publication Number Publication Date
CN112749370A true CN112749370A (en) 2021-05-04
CN112749370B CN112749370B (en) 2021-07-02

Family

ID=75651714

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110364634.5A Active CN112749370B (en) 2021-04-06 2021-04-06 Fault tracking and positioning method and system based on Internet of things

Country Status (1)

Country Link
CN (1) CN112749370B (en)

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1719803A (en) * 2005-07-28 2006-01-11 武汉大学 Correcting method and application of multi-size router for extensible large-scale sensor network
CN103336813A (en) * 2013-06-27 2013-10-02 南京邮电大学 Data integrated management scheme for Internet of Things based on middleware framework
US20140032175A1 (en) * 2012-05-18 2014-01-30 International Business Machines Corporation Quality of information assessment in dynamic sensor networks
US20150317284A1 (en) * 2014-04-30 2015-11-05 International Business Machines Corporation Sensor output change detection
CN106321318A (en) * 2016-08-18 2017-01-11 河南职业技术学院 Starting control system and control method of automobile engine
CN108616367A (en) * 2016-12-12 2018-10-02 华为技术有限公司 Fault Locating Method and the network equipment
CN110018670A (en) * 2019-03-28 2019-07-16 浙江大学 A kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules
US10558207B1 (en) * 2019-04-11 2020-02-11 Sas Institute Inc. Event monitoring system
CN111413587A (en) * 2020-03-30 2020-07-14 山东理工大学 Method and system for determining installation position of power distribution network fault monitoring device
CN111639304A (en) * 2020-06-02 2020-09-08 江南大学 CSTR fault positioning method based on Xgboost regression model

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1719803A (en) * 2005-07-28 2006-01-11 武汉大学 Correcting method and application of multi-size router for extensible large-scale sensor network
US20140032175A1 (en) * 2012-05-18 2014-01-30 International Business Machines Corporation Quality of information assessment in dynamic sensor networks
CN103336813A (en) * 2013-06-27 2013-10-02 南京邮电大学 Data integrated management scheme for Internet of Things based on middleware framework
US20150317284A1 (en) * 2014-04-30 2015-11-05 International Business Machines Corporation Sensor output change detection
CN106321318A (en) * 2016-08-18 2017-01-11 河南职业技术学院 Starting control system and control method of automobile engine
CN108616367A (en) * 2016-12-12 2018-10-02 华为技术有限公司 Fault Locating Method and the network equipment
CN110018670A (en) * 2019-03-28 2019-07-16 浙江大学 A kind of industrial process unusual service condition prediction technique excavated based on dynamic association rules
US10558207B1 (en) * 2019-04-11 2020-02-11 Sas Institute Inc. Event monitoring system
CN111413587A (en) * 2020-03-30 2020-07-14 山东理工大学 Method and system for determining installation position of power distribution network fault monitoring device
CN111639304A (en) * 2020-06-02 2020-09-08 江南大学 CSTR fault positioning method based on Xgboost regression model

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
张颖 等: "传感器网络同步态的节点故障诊断算法", 《重庆大学学报》 *
欧阳成 等: "改进的多贝努利滤波检测前跟踪算法", 《系统工程与电子技术》 *

Also Published As

Publication number Publication date
CN112749370B (en) 2021-07-02

Similar Documents

Publication Publication Date Title
CN116186634B (en) Intelligent management system for construction data of building engineering
CN108777873A (en) The wireless sensor network abnormal deviation data examination method of forest is isolated based on weighted blend
CN110087207B (en) Method for reconstructing missing data of wireless sensor network
CN107707417B (en) Wireless sensor network abnormal node detection and positioning method based on subgraph processing
CN109348416B (en) Fingerprint indoor positioning method based on binary k-means
CN115269357A (en) Micro-service abnormity detection method based on call chain
CN108257365B (en) Industrial alarm design method based on global uncertainty evidence dynamic fusion
KR20140068650A (en) Method for detecting overlapping communities in a network
CN105512011B (en) A kind of electronics testability modeling appraisal procedure
Dianati Tilaki et al. Rangelands production modeling using an artificial neural network (ANN) and geographic information system (GIS) in Baladeh rangelands, North Iran
CN115342814B (en) Unmanned ship positioning method based on multi-sensor data fusion
CN110930541B (en) Method for analyzing working condition state of agricultural machine by using GPS information
CN117251798A (en) Meteorological equipment anomaly detection method based on two-layer progressive process
CN112749370B (en) Fault tracking and positioning method and system based on Internet of things
CN104021199A (en) Function module detecting method based on node domination capacity similarity
CN112333147B (en) Nuclear power plant DCS platform network operation situation sensing method and system
CN110751201B (en) SAR equipment task failure cause reasoning method based on textural feature transformation
CN112949735A (en) Liquid hazardous chemical substance volatile concentration abnormity discovery method based on outlier data mining
CN110110771B (en) Saline soil salinity estimation method based on surface image
CN110266527B (en) Sensor node fault classification alarm method and device based on spatial correlation
CN116597294A (en) SLAM map topology evaluation method and device, electronic equipment and storage medium
CN111479235A (en) Formalization method for detecting wireless sensor network node
CN110717507B (en) Soil moisture content sensor optimization layout method based on APDJ algorithm
CN112613371A (en) Hyperspectral image road extraction method based on dense connection convolution neural network
CN112036403A (en) Intelligent detection method for missing of bolt pin of power transmission tower based on attention mechanism

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