CN115048959B - Method for diagnosing faults of gun anti-squat device based on RMSD-DS - Google Patents
Method for diagnosing faults of gun anti-squat device based on RMSD-DS Download PDFInfo
- Publication number
- CN115048959B CN115048959B CN202210649020.6A CN202210649020A CN115048959B CN 115048959 B CN115048959 B CN 115048959B CN 202210649020 A CN202210649020 A CN 202210649020A CN 115048959 B CN115048959 B CN 115048959B
- Authority
- CN
- China
- Prior art keywords
- evidence
- fault
- rmsd
- squat
- reliability
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 239000008186 active pharmaceutical agent Substances 0.000 title claims abstract description 32
- 238000003745 diagnosis Methods 0.000 claims abstract description 77
- 230000004927 fusion Effects 0.000 claims abstract description 32
- 239000011159 matrix material Substances 0.000 claims description 31
- 238000005299 abrasion Methods 0.000 claims description 20
- 238000006073 displacement reaction Methods 0.000 claims description 20
- 230000010354 integration Effects 0.000 claims description 18
- 238000005315 distribution function Methods 0.000 claims description 10
- 150000001875 compounds Chemical class 0.000 claims description 4
- 230000007547 defect Effects 0.000 claims description 2
- 238000009472 formulation Methods 0.000 claims description 2
- 239000000203 mixture Substances 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 208000037408 Device failure Diseases 0.000 claims 3
- OVSKIKFHRZPJSS-UHFFFAOYSA-N 2,4-D Chemical compound OC(=O)COC1=CC=C(Cl)C=C1Cl OVSKIKFHRZPJSS-UHFFFAOYSA-N 0.000 claims 2
- 208000032368 Device malfunction Diseases 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 4
- 238000012935 Averaging Methods 0.000 abstract 1
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000007500 overflow downdraw method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010304 firing Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Landscapes
- Locating Faults (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a fault diagnosis method for an artillery anti-squat device based on RMSD-DS, and belongs to the field of artillery fault diagnosis. The method comprises the steps of firstly determining a typical fault mode and fault characteristic signals of an artillery anti-squat device, and acquiring probability distribution of corresponding evidence of each fault characteristic signal through a Gaussian model; then quantitatively describing the importance degree of each evidence in the fusion decision Guo Zong by constructing RMSD similarity coefficients and solving the reliability of each evidence, and accordingly, distributing weights to the evidence so as to eliminate the conflict influence between information; and finally, solving the integrated evidence after weighted averaging, and carrying out self fusion on the integrated evidence by applying a DS fusion rule to obtain a final fusion result, thereby realizing fault diagnosis of the gun anti-squat device. The method for solving the basic probability distribution value of the evidence corresponding to each fault characteristic signal is simple and has excellent fault diagnosis effect. The invention can improve the fault diagnosis efficiency and the fault diagnosis precision of the cannon anti-squat device when the conflict information is fused.
Description
Technical Field
The invention relates to a fault diagnosis method for an artillery anti-squat device based on RMSD-DS, and belongs to the technical field of artillery fault diagnosis.
Background
The anti-recoil device of the gun is used as a key component of the gun and plays roles of dissipating and storing recoil energy and resetting the gun body when the gun shoots. In the battlefield, the complexity and uncertainty of the battlefield environment cause the faults of the cannon anti-squat device to frequently occur, and if the faults of the cannon anti-squat device are not diagnosed and solved in time, the firing efficiency and the firing precision of the cannon can be seriously affected, even precious fighters are mushed, and serious battlefield accidents are caused. Therefore, the method has important practical significance for researching the fault diagnosis method of the gun anti-squat device.
The cannon anti-squat device often carries out fault diagnosis work based on multi-source information fusion based on four signal sources of maximum squat displacement, maximum squat speed, maximum recoil speed and recoil speed. Due to the complexity of the battlefield environment, the data sensor of the cannon anti-squat device is easily damaged or the data acquisition process is disturbed, so that the information output by the sensors are mutually conflicted. At this time, if the traditional multi-source information fusion method is applied, similar to the neural network method, the conditions of low diagnosis efficiency and low diagnosis precision are easy to appear when the fault diagnosis mode of the gun anti-squat device is identified, and the fault diagnosis requirement of the gun anti-squat device cannot be met.
Disclosure of Invention
In order to solve the problems of low efficiency, low precision and the like of the conventional gun recoil device fault diagnosis method when the conflict information is fused, the main purpose of the invention is to provide a gun recoil device fault diagnosis method based on RMSD-DS, which determines a typical fault mode and a fault characteristic signal of the gun recoil device; acquiring a fault characteristic signal corresponding to the cannon anti-squat device in a typical fault mode; the method comprises the steps of obtaining a fault characteristic signal corresponding to a typical fault mode, and classifying the fault characteristic signal into fault training sample data and fault sample data to be detected; solving the average value and standard deviation of training samples belonging to different fault modes of the gun anti-squat device on different fault characteristic signals, and then constructing Gaussian models of the training samples belonging to different fault modes on different fault signals; according to the constructed gun anti-squat device fault mode Gaussian model, basic probability distribution of evidence corresponding to a to-be-detected sample fault characteristic signal of the gun anti-squat device is solved; according to the basic probability distribution of evidence corresponding to the fault characteristic signals of the sample to be detected of the cannon recoil device obtained by solving, under the fault mode frame of the cannon recoil device, defining and solving conflict factors among all the evidences, and constructing a conflict factor matrix according to all the conflict factors solved; defining and solving root mean square deviation (Root Mean Square Deviation, RMSD) distances among all evidences under the fault mode framework of the gun anti-squat device; taking the geometric mean of the conflict factors and the normalized RMSD distance as the value of the RMSD conflict coefficient to construct the RMSD conflict coefficient; according to the constructed RMSD conflict coefficient, solving and constructing an RMSD similarity coefficient; defining the reliability of each evidence as the sum of RMSD similarity coefficients of the evidence and other evidence, and determining the reliability of each evidence according to the definition; analyzing the reliability of each evidence, defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, and solving the weight of each evidence; weight distribution is carried out on all evidences according to the reliability of each evidence, the conflict between information is reduced, and then the integrated evidence is obtained after weighted average; the integrated evidence is subjected to self-fusion by using a Dempster-Shafer (DS) evidence theory method under the gun fault mode framework, so that the occurrence probability of a fault mode of a corresponding gun anti-squat device is obtained; traversing the occurrence probability of the failure modes of all the gun anti-squat devices, and determining the failure mode corresponding to the maximum basic probability distribution value as the failure mode of final diagnosis, namely realizing high-precision and efficient diagnosis of the failure of the gun anti-squat devices based on RMSD-DS.
The aim of the invention is achieved by the following technical scheme.
The invention discloses a fault diagnosis method for an artillery anti-squat device based on RMSD-DS, which comprises the following steps:
Step one: and determining a typical fault mode and a fault characteristic signal of the gun anti-squat device.
Step two: and acquiring fault characteristic signals corresponding to the cannon anti-squat device in a typical fault mode.
Step three: analyzing the fault characteristic signals which are obtained in the second step and correspond to the fault characteristic signals in the typical fault mode, and classifying the fault characteristic signals into fault training sample data and fault sample data to be detected; solving the average value and standard deviation of training samples belonging to different fault modes of the gun anti-squat device on different fault characteristic signals, and then constructing Gaussian models of the training samples belonging to different fault modes on different fault signals.
Step four: and (3) solving the basic probability distribution of the evidence corresponding to the fault characteristic signal of the sample to be detected of the cannon anti-squat device according to the Gaussian model of the fault mode of the cannon anti-squat device constructed in the step (III).
Step five: according to the basic probability distribution of evidence corresponding to the fault characteristic signals of the sample to be detected of the gun recoil device obtained by solving in the step four, under the fault mode frame of the gun recoil device, defining and solving conflict factors among all the evidence, and constructing a conflict factor matrix according to all the solved conflict factors; defining and solving the RMSD distance between each evidence under the fault mode framework of the gun anti-squat device; taking the geometric mean of the conflict factors and the normalized RMSD distance as the value of the RMSD conflict coefficient to construct the RMSD conflict coefficient; and according to the constructed RMSD conflict coefficient, solving and constructing the RMSD similarity coefficient, and defining the reliability of the following step six by the constructed RMSD similarity coefficient.
Step six: defining the reliability of each evidence as the sum of RMSD similarity coefficients of the evidence and other evidence, and determining the reliability of each evidence according to the definition; analyzing the reliability of each evidence, defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, and solving the weight of each evidence; and (3) carrying out weight distribution on all evidences according to the reliability of each evidence, reducing the conflict between information, and obtaining integrated evidence after weighted average, so that the integration of the following step seven is facilitated, and the fault diagnosis accuracy is improved.
Step seven: in the gun fault mode framework, the integration evidence obtained in the step six is subjected to self fusion by using a Dempster-Shafer (DS) evidence theory method, and the occurrence probability of the corresponding gun anti-squat device fault mode is obtained; traversing the occurrence probability of the failure modes of all the gun anti-squat devices, and determining the failure mode corresponding to the maximum basic probability distribution value as the failure mode of final diagnosis, namely realizing high-precision and efficient diagnosis of the failure of the gun anti-squat devices based on RMSD-DS.
The method also comprises the step eight of: substituting the data of the sample to be detected of the faults of the gun recoil device determined in the step three into the fault mode Gaussian model constructed in the step four, and solving the basic probability distribution of the evidence corresponding to each fault characteristic signal; and D, re-distributing weights to each evidence by using the reliability determined in the step six, reducing the influence caused by conflict information, and improving the fault diagnosis performance of the anti-squat device. The improvement of the diagnosis performance of the anti-recoil device of the gun comprises the improvement of the diagnosis efficiency and the diagnosis precision of the anti-recoil device.
Step one: and determining a typical fault mode and a fault characteristic signal of the gun anti-squat device.
Determining three typical failure modes of the gun anti-squat device, namely, a check ring abrasion X, a compound feed machine air leakage Y and a brake rod piston abrasion Z, wherein the failure mode frame of the gun anti-squat device is expressed as Θ= { X, Y, Z }; the fault characteristic signals for determining the anti-squat device of the gun are respectively as follows: maximum squat displacement Xmax, maximum squat speed Vmax, maximum recoil speed Umax, and recoil to position speed unend.
Step two: and acquiring fault characteristic signals corresponding to the cannon anti-squat device in a typical fault mode.
When the cannon works, four fault characteristic signals of the anti-squat device under each fault mode are collected by a sensor arranged on the anti-squat device, the obtained data are represented by F i, wherein F=X, Y and Z respectively represent three fault modes of abrasion of a check ring, air leakage of a re-entry machine and abrasion of a brake rod piston; i=1, 2,3,4, representing four fault signature signals of maximum squat displacement Xmax, maximum squat speed Vmax, maximum return speed Umax and return to position speed Uend, respectively; the set of sample data collected is represented as (F 1,F2,F3,F4),F1 represents the maximum squat displacement Xmax signal data corresponding to failure mode F, F 2 represents the maximum squat speed Vmax signal data corresponding to failure mode F, F 3 represents the maximum recoil speed Umax signal data corresponding to failure mode F, and F 4 represents the recoil to bit speed Uend signal data corresponding to failure mode F.
Step three: analyzing the fault characteristic signals which are obtained in the second step and correspond to the fault characteristic signals in the typical fault mode, and classifying the fault characteristic signals into fault training sample data and fault sample data to be detected; solving the average value and standard deviation of training samples belonging to different fault modes of the gun anti-squat device on different fault characteristic signals, and then constructing Gaussian models of the training samples belonging to different fault modes on different fault signals.
Step 3.1: and (3) classifying the fault characteristic signals obtained in the step two and corresponding to the typical fault mode into fault training sample data and fault sample data to be detected.
Based on the fault characteristic signals corresponding to the typical fault modes obtained in the second step, sample data with preset proportion are selected from four fault characteristic signal data corresponding to each fault mode respectively to serve as fault training samples, and the remaining sample data serve as fault samples to be detected.
Step 3.2: and (3) solving the average value and standard deviation of the training samples belonging to different fault modes of the gun anti-squat device on different fault characteristic signals for the fault training samples selected in the step (3.1).
For the selected fault training samples, solving an average value mu (F i) and a standard deviation sigma (F i) of the training samples belonging to different fault modes on different fault characteristic signals, wherein a solving formula of the average value mu (F i) is shown as a formula (1), and a solving formula of the standard deviation sigma (F i) is shown as a formula (2):
In the formulas (1) and (2), f=x, Y, Z, representing three failure modes; i=1, 2,3,4, representing four fault signature signals; j=1, 2, …, N, representing the data sequence;
step 3.3: and (3) constructing Gaussian models of training samples belonging to different fault modes on different fault signals according to the average value and the standard deviation obtained by solving in the step (3.2).
Constructing a Gaussian model (membership function) of training samples belonging to different fault modes on different fault signals according to the average mu (F i) and the standard deviation sigma (F i) obtained in the step3.2, wherein the Gaussian model (membership function) is shown in a formula (3):
In formula (3), f=x, Y, Z, represents three failure modes; i=1, 2,3,4, representing four fault signature signals.
When the fault characteristic signal is the maximum squat displacement (Xmax), the gaussian model on the fault mode, throttle ring wear (X), the re-feeder blow-by (Y) and the brake lever piston wear (Z) is:
When the fault characteristic signal is the maximum squat speed (Vmax), the fault mode is a gaussian model on the throttle ring wear (X), the air leakage (Y) of the rewinder and the brake lever piston wear (Z) is: and/>
When the fault characteristic signal is the maximum re-advancing speed (Umax), the fault mode is a gaussian model on the throttle ring wear (X), the re-advancing machine air leakage (Y) and the backing rod piston wear (Z) and is: and/>
When the fault characteristic signal is the return-to-position speed (usend), the fault mode is a Gaussian model on the throttle ring abrasion (X), the air leakage (Y) of the return-to-machine and the abrasion (Z) of the braking rod piston, and the Gaussian model is as follows: and/>
And (4) to (15) are Gaussian models of training samples of different fault modes on different fault signals.
Step four: according to the gun anti-squat device fault mode Gaussian model constructed in the third step, the basic probability distribution of evidence corresponding to the sample fault characteristic signals to be detected of the gun anti-squat device is solved, and the fault diagnosis efficiency is improved on the basis of guaranteeing the fault diagnosis precision of the gun anti-squat device in the subsequent step.
Step 4.1: and (3) solving the ordinate of the intersection point of the sample to be detected and the Gaussian models of different fault modes under each fault signal according to the Gaussian model of the gun anti-squat device constructed in the step (3.3).
For a set of samples to be inspected for which the failure mode is unknown, the corresponding data may be represented as (F 1,F2,F3,F4), where f= X, Y, Z; the subscript 1 represents the maximum squat displacement (Xmax), 2 represents the maximum squat speed (Vmax), 3 represents the maximum return speed (Umax), and 4 represents the return to bit speed (Uend).
When the fault characteristic signal is the maximum squat displacement (Xmax), the ordinate solving formula of the intersection point of the sample to be detected and the Gaussian model of different fault modes is as follows:
when the fault characteristic signal is the maximum squat speed (Vmax), the ordinate of the intersection point of the sample to be detected and the Gaussian model of different fault modes is as follows:
When the fault characteristic signal is the maximum complex advance speed (Umax), the ordinate of the intersection point of the sample to be detected and the Gaussian model of different fault modes is as follows:
When the fault characteristic signal is the return-to-bit speed (usend), the ordinate of the intersection point of the sample to be detected and the Gaussian model of different fault modes is as follows:
Step 4.2: and representing the basic probability distribution of evidence corresponding to the four fault characteristic signals.
Each sample to be detected contains four fault characteristic signals (Xmax, vmax, umax and Uend), each fault characteristic signal corresponds to a group of evidences, and then the basic probability distribution function of the evidences corresponding to each fault characteristic signal (Xmax, vmax, umax and Uend) can be expressed as m i (i=1, 2,3, 4); in the first step, the failure modes of the gun recoil device include three types of X, Y and Z, and the basic probability distribution of a group of evidences comprises m i(X)、mi (Y) and m i (Z), wherein m i (X) represents the basic probability distribution of the sample to be detected belonging to the failure mode X under the evidence m i, m i (Y) represents the basic probability distribution of the sample to be detected belonging to the failure mode Y under the evidence m i, and m i (Z) represents the basic probability distribution of the sample to be detected belonging to the failure mode Z under the evidence m i.
Step 4.3: and 4.1, acquiring the ordinate of the intersection point of the to-be-detected sample and the Gaussian model of different fault modes under each fault signal, and respectively solving the basic probability distribution of the evidence corresponding to each fault characteristic signal of the to-be-detected sample.
The basic probability distribution function solution formula of the corresponding evidence m 1 when the fault signature is the maximum squat displacement (Xmax) is:
the basic probability distribution function of the corresponding evidence m 2 for the maximum squat speed (Vmax) of the fault signature is: m 2(X),m2(Y),m2 (Z);
The basic probability distribution function of the corresponding evidence m 3 when the fault signature is the maximum complex advance speed (Umax) is: m 3(X),m3(Y),m3 (Z);
The basic probability distribution function of the corresponding evidence m 4 when the fault signature is a return to bit velocity (usend) is: m 4(X),m4(Y),m4 (Z).
Step five: according to the basic probability distribution of evidence corresponding to the fault characteristic signals of the sample to be detected of the gun recoil device obtained by solving in the step four, under the fault mode frame of the gun recoil device, defining and solving conflict factors among all the evidence, and constructing a conflict factor matrix according to all the solved conflict factors; defining and solving the RMSD distance between each evidence under the fault mode framework of the gun anti-squat device; taking the geometric mean of the conflict factors and the normalized RMSD distance as the value of the RMSD conflict coefficient to construct the RMSD conflict coefficient; and according to the constructed RMSD conflict coefficient, solving and constructing the RMSD similarity coefficient, and defining the reliability of the following step six by the constructed RMSD similarity coefficient.
Step 5.1: under the fault mode framework of the cannon anti-squat device, conflict factors among all evidences are defined and solved, and a conflict factor matrix is constructed according to all the solved conflict factors. The conflict factor between the evidence is the conflict factor between every two groups of evidence.
To facilitate the formulation of the solution to the conflict factor, under the cannon failure mode framework Θ= { X, Y, Z }, m 1 and m 2 are defined as two groups of evidence, the corresponding failure modes are denoted as F ' (F ' =x, Y, Z) and F "(F ' =x, Y, Z), respectively, and the conflict factors of evidence m 1 and m 2 are shown as formula (40):
The collision factor matrix between evidence is obtained according to the formula (40):
However, the conflict factor has a defect, and according to the formula (40), the conflict factor obtained by solving when the two evidences are identical is not 0, so that the conflict factor needs to be corrected by introducing the RMSD distance between the evidences in the subsequent step 5.2, and the conflict factor, that is, the RMSD conflict factor, is constructed in the subsequent step 5.3.
Step 5.2: and defining and solving the RMSD distance between each evidence under the fault mode framework of the gun anti-squat device, constructing an RMSD distance matrix according to all the solved RMSD distances, and solving the normalized RMSD distance matrix.
Defining m 1 and m 2 as two sets of evidence, the corresponding failure modes are denoted as F '(F' =x, Y, Z) and F "(F" =x, Y, Z), respectively, then the root mean square offset (Root Mean Square Deviation, RMSD) distance between evidence m 1 and m 2 is defined as:
Solving according to a formula (42) to obtain an RMSD distance matrix between evidences, wherein the RMSD distance matrix is shown as a formula (43):
Then find the maximum RMSD max =max { RMSD }, in the RMSD distance matrix, then normalize the RMSD distance matrix, i.e., divide each element of the RMSD distance matrix by RMSD max, to obtain the RMSD distance matrix after normalization, as shown in equation (44):
step 5.3: defining the geometric mean value of the conflict factor in the step 5.1 and the normalized RMSD distance in the step 5.2 as the value of the RMSD conflict coefficient, and constructing the RMSD conflict coefficient.
Taking the geometric mean of the conflict factor obtained in the step 5.1 and the normalized RMSD distance obtained in the step 5.2 as the value of the RMSD conflict coefficient, the RMSD conflict coefficients of the definition evidences m 1 and m 2 are expressed as:
In the formula (45), K (m 1,m2) represents a conflict factor of evidence m 1 and m 2, and RMSD (m 1,m2) represents a normalized RMSD distance of evidence m 1 and m 2;
the RMSD collision coefficient matrix is obtained according to equation (45):
Step 5.4: the RMSD similarity coefficients are solved and constructed based on the RMSD collision coefficients described in step 5.3.
The RMSD collision coefficient constructed in step 5.3 represents the collision degree between evidences, the value range is [0,1], and subtracting the RMSD collision coefficient from 1 can represent the similarity degree between evidences, and the RMSD similarity coefficient defining evidences m 1 and m 2 is expressed as:
SimRMSD(m1,m2)＝1-ConRMSD(m1,m2) (47)
The RMSD similarity coefficient matrix may be obtained according to equation (47) as:
Step six: defining the reliability of each evidence as the sum of RMSD similarity coefficients of the evidence and other evidence, and determining the reliability of each evidence according to the definition; analyzing the reliability of each evidence, defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, and solving the weight of each evidence; and (3) carrying out weight distribution on all evidences according to the reliability of each evidence, reducing the conflict between information, and obtaining integrated evidence after weighted average, so that the integration of the following step seven is facilitated, and the fault diagnosis accuracy is improved.
Step 6.1: defining the reliability of each evidence as the sum of the RMSD similarity coefficients of the evidence and other evidence, and determining the reliability of each evidence according to the definition.
Defining the reliability of each evidence as the sum of the RMSD similarity coefficients of that evidence and other evidence, the reliability of evidence m 1 is formulated as:
Rel(m1)＝SimRMSD(m1,m2)+SimRMSD(m1,m3)+SimRMSD(m1,m4) (49)
The reliability of evidence m 2 is formulated as:
Rel(m2)＝SimRMSD(m2,m1)+SimRMSD(m2,m3)+SimRMSD(m2,m4) (50)
the reliability of evidence m 3 is formulated as:
Rel(m3)＝SimRMSD(m3,m1)+SimRMSD(m3,m2)+SimRMSD(m3,m4) (51)
The reliability of evidence m 4 is formulated as:
Rel(m4)＝SimRMSD(m4,m1)+SimRMSD(m4,m2)+SimRMSD(m4,m3) (52)
Wherein, the reliability of the evidence represents the supporting degree of other evidence on the evidence;
the greater the reliability of the evidence, the higher the importance of the evidence in the fusion decision process, and the weight distributed in the subsequent step 6.2 is preferably greater;
The smaller the reliability of the evidence, the lower the importance of the evidence in the fusion decision process, and the smaller the weight assigned in the subsequent step 6.2.
Step 6.2: analyzing the reliability of each evidence determined in the step 6.1, defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, and solving the weight of each evidence.
Defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, the weight of evidence m 1 is expressed as:
The reliability of evidence m 2 is formulated as:
the reliability of evidence m 3 is formulated as:
The reliability of evidence m 4 is formulated as:
step 6.3: and (3) carrying out weight distribution on all evidences according to the reliability of each evidence in the step (6.2), reducing the conflict between information, and obtaining integrated evidence after weighted average, so that the integration of the subsequent step (seven) is facilitated, and the fault diagnosis accuracy is improved.
According to the weight distributed in the step 6.2, the integration evidence is obtained after weighted average, and is expressed as follows:
further, a basic probability distribution is obtained when the failure modes are X, Y and Z under the integrated evidence, and is expressed as:
Step seven: in the gun fault mode framework, the integration evidence obtained in the step six is subjected to self fusion by using a Dempster-Shafer (DS) evidence theory method, and the occurrence probability of the corresponding gun anti-squat device fault mode is obtained; traversing the occurrence probability of the failure modes of all the gun anti-squat devices, and determining the failure mode corresponding to the maximum basic probability distribution value as the failure mode of final diagnosis, namely realizing high-precision and efficient diagnosis of the failure of the gun anti-squat devices based on RMSD-DS.
Step 7.1: defining m 1 and m 2 as two sets of evidence, the corresponding failure modes are denoted F '(F' =x, Y, Z) and F "(F" =x, Y, Z), respectively, giving DS fusion rules for evidence m 1 and m 2.
For convenience in giving DS fusion rules, under the cannon failure mode framework Θ= { X, Y, Z } m 1 and m 2 are two groups of evidence, the corresponding failure modes are denoted as F '(F' =x, Y, Z) and F "(F" =x, Y, Z), respectively, and DS fusion rules of evidence m 1 and m 2 are shown as formula (61):
Step 7.2: and 3, according to the DS fusion rule given in the step 7.1, fusing the integration evidence for 3 times, and fusing the integration evidence by utilizing the DS fusion rule to obtain the occurrence probability of the fault mode of the corresponding gun anti-squat device.
Step 7.3: and 7.2, traversing to obtain occurrence probabilities of failure modes of all gun anti-squat devices, and determining the failure mode corresponding to the maximum basic probability distribution value as the failure mode of final diagnosis, namely realizing high-precision and efficient diagnosis of gun anti-squat device failures based on the RMSD-DS method.
The beneficial effects are that:
1. Compared with the traditional fusion method such as a neural network, the method for diagnosing the faults of the gun anti-squat device based on the RMSD-DS, disclosed by the invention, has the advantages that under the framework of the fault mode of the gun anti-squat device, the importance degree of each evidence is quantitatively described by constructing the RMSD similarity coefficient and solving the reliability of each evidence, weight distribution is carried out on all the evidence according to the reliability of each evidence, the influence caused by conflict information is reduced, the integrated evidence is obtained after weighted average, the DS method is utilized for integrating the evidence, and the occurrence probability of the fault mode of the corresponding gun anti-squat device is obtained; and traversing the occurrence probability of the fault modes of all the cannon anti-squat devices, determining the fault mode corresponding to the maximum basic probability distribution value as the final diagnosis fault mode, and improving the fault diagnosis precision of the cannon anti-squat devices.
2. According to the method for diagnosing the faults of the cannon squat device based on the RMSD-DS, disclosed by the invention, the Gaussian model of the fault mode of the cannon squat device is established by analyzing the fault history data of the cannon squat device, and the basic probability distribution of the evidence corresponding to the fault characteristic signals of the cannon squat device is solved by the Gaussian model of the fault mode of the cannon squat device, so that the fault diagnosis efficiency of the cannon squat device is improved.
3. The method for diagnosing the faults of the gun anti-squat device based on the RMSD-DS, disclosed by the invention, realizes high-precision and efficient diagnosis of the faults of the gun anti-squat device based on the RMSD-DS on the basis of realizing the beneficial effects 1 and 2.
Drawings
FIG. 1 is a flow chart of a fault diagnosis method of an artillery anti-squat device based on RMSD-DS of the invention.
FIG. 2 is a Gaussian model of various modes for different fault signature signals, wherein: fig. 2 (a) shows gaussian models of various fault modes when the fault characteristic signal is Xmax, fig. 2 (b) shows gaussian models of various fault modes when the fault characteristic signal is Vmax, fig. 2 (c) shows gaussian models of various fault modes when the fault characteristic signal is Umax, and fig. 2 (d) shows gaussian models of various fault modes when the fault characteristic signal is Uend.
FIG. 3 is a flow chart for solving for RMSD similarity coefficients between two pieces of evidence.
Detailed Description
For a better illustration of the objects and advantages of the present invention, the summary of the invention is further illustrated in the following figures and examples, which are provided by comparing the conventional neural network fusion method with the DS evidence theory method.
As shown in fig. 1; the fault diagnosis method for the gun anti-squat device based on the RMSD-DS, disclosed by the embodiment, comprises the following specific implementation steps:
Step one: and determining a typical fault mode and a fault characteristic signal of the gun anti-squat device.
Determining three typical failure modes of the gun anti-squat device, namely, a check ring abrasion X, a compound feed machine air leakage Y and a brake rod piston abrasion Z, wherein the failure mode frame of the gun anti-squat device is expressed as Θ= { X, Y, Z }; the fault characteristic signals for determining the anti-squat device of the gun are respectively as follows: maximum squat displacement Xmax, maximum squat speed Vmax, maximum recoil speed Umax, and recoil to position speed unend.
Step two: and acquiring fault characteristic signals corresponding to the cannon anti-squat device in a typical fault mode.
When the cannon works, four fault characteristic signals of the anti-squat device under each fault mode are collected by a sensor arranged on the anti-squat device, the obtained data are represented by F i, wherein F=X, Y and Z respectively represent three fault modes of abrasion of a check ring, air leakage of a re-entry machine and abrasion of a brake rod piston; i=1, 2,3,4, representing four fault signature signals of maximum squat displacement Xmax, maximum squat speed Vmax, maximum return speed Umax and return to position speed Uend, respectively; the set of sample data collected is denoted (F 1,F2,F3,F4), i.e., F 1 represents the maximum squat displacement Xmax signal data corresponding to failure mode F, F 2 represents the maximum squat speed Vmax signal data corresponding to failure mode F, F 3 represents the maximum return speed Umax signal data corresponding to failure mode F, and F 4 represents the return-to-bit speed Uend signal data corresponding to failure mode F.
In each failure mode, 100 sets of failure data were acquired, totaling 300 sets of data.
Step three: analyzing the fault characteristic signals which are obtained in the second step and correspond to the fault characteristic signals in the typical fault mode, and classifying the fault characteristic signals into fault training sample data and fault sample data to be detected; solving the average value and standard deviation of training samples belonging to different fault modes of the gun anti-squat device on different fault characteristic signals, and then constructing Gaussian models of the training samples belonging to different fault modes on different fault signals.
Step 3.1: and (3) classifying the fault characteristic signals obtained in the step two and corresponding to the typical fault mode into fault training sample data and fault sample data to be detected.
Based on the fault characteristic signals corresponding to the typical fault modes obtained in the second step, 80% of sample data are selected from four fault characteristic signal data corresponding to each fault mode to serve as fault training samples, and the remaining 20% of sample data serve as fault samples to be detected.
In order to verify the effectiveness of the method, the sensor of the gun anti-squat device is simulated to be damaged by exchanging the maximum squat displacement data corresponding to the fault mode Y and the fault mode Z; the purpose of this process is to allow for conflicts between the information output by the sensors.
Step 3.2: and (3) solving the average value and standard deviation of the training samples belonging to different fault modes of the gun anti-squat device on different fault characteristic signals for the fault training samples selected in the step (3.1).
The results obtained by the solution are shown in table 1.
Table 1 means and standard deviations of training samples (different failure modes)
Step 3.3: and (3) constructing Gaussian models of training samples belonging to different fault modes on different fault signals according to the average value and the standard deviation obtained by solving in the step (3.2).
When the fault characteristic signal is the maximum squat displacement (Xmax), the gaussian model on the fault mode, throttle ring wear (X), the re-feeder blow-by (Y) and the brake lever piston wear (Z) is:
/>
When the fault characteristic signal is the maximum squat speed (Vmax), the fault mode is a gaussian model on the throttle ring wear (X), the air leakage (Y) of the rewinder and the brake lever piston wear (Z) is: and/>
When the fault characteristic signal is the maximum re-advancing speed (Umax), the fault mode is a gaussian model on the throttle ring wear (X), the re-advancing machine air leakage (Y) and the backing rod piston wear (Z) and is: and/>
When the fault characteristic signal is the return-to-position speed (usend), the fault mode is a Gaussian model on the throttle ring abrasion (X), the air leakage (Y) of the return-to-machine and the abrasion (Z) of the braking rod piston, and the Gaussian model is as follows: and/>
After solving the gaussian model, the gaussian model of various fault modes is drawn under each fault characteristic signal, as shown in fig. 2.
Step four: and (3) solving the basic probability distribution of the evidence corresponding to the fault characteristic signal of the sample to be detected of the cannon anti-squat device according to the Gaussian model of the fault mode of the cannon anti-squat device constructed in the step (III).
In each failure mode, the samples to be detected have 20 groups, and only the basic probability distribution of the corresponding evidence of each failure characteristic signal of one group of sample data to be detected is given from each failure mode due to text space limitation, as shown in tables 2-4.
TABLE 2 basic probability assignment for evidence corresponding to each failure characteristic signal for an actual failure of X
For the group of samples to be detected, as can be seen from table 2, when the actual fault is X, the evidence m 1、m2 and m 3 support the fault mode X to occur, the evidence m 4 supports the fault mode Y to occur, and a conflict exists between the evidences.
TABLE 3 assignment of base probability of evidence for each failure characteristic signal when the actual failure is Y
For the group of samples to be detected, as can be seen from table 3, when the actual fault is Y, the evidence m 2、m3 and m 4 support the fault mode Y to occur, the evidence m 1 supports the fault mode Z to occur, and a conflict exists between the evidences.
TABLE 4 basic probability assignment for evidence corresponding to each failure characteristic signal for Z actual failure
For the set of samples to be tested, as can be seen from table 4, when the actual fault is Z, evidence m 1 supports fault mode X occurrence, evidence m 2、m3 and m 4 support fault mode Z occurrence, and a conflict situation exists between the evidences.
Step five: according to the basic probability distribution of evidence corresponding to the fault characteristic signals of the sample to be detected of the gun recoil device obtained by solving in the step four, under the fault mode frame of the gun recoil device, defining and solving conflict factors among all the evidence, and constructing a conflict factor matrix according to all the solved conflict factors; defining and solving the RMSD distance between each evidence under the fault mode framework of the gun anti-squat device; taking the geometric mean of the conflict factors and the normalized RMSD distance as the value of the RMSD conflict coefficient to construct the RMSD conflict coefficient; and according to the constructed RMSD conflict coefficient, solving and constructing the RMSD similarity coefficient, and defining the reliability of the following step six by the constructed RMSD similarity coefficient.
A flow chart for solving for RMSD similarity coefficients between two evidences is shown in fig. 3.
Based on the data given in the step four, solving the obtained RMSD similarity coefficient to be,
When the actual fault is X, the RMSD similarity coefficient matrix among evidences of a group of sample data to be detected is as follows:
when the actual fault is Y, the RMSD similarity coefficient matrix among evidences of one group of sample data to be detected is as follows:
When the actual fault is Z, the RMSD similarity coefficient matrix among evidences of a group of sample data to be detected is as follows:
Step six: defining the reliability of each evidence as the sum of RMSD similarity coefficients of the evidence and other evidence, and determining the reliability of each evidence according to the definition; analyzing the reliability of each evidence, defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, and solving the weight of each evidence; and (3) carrying out weight distribution on all evidences according to the reliability of each evidence, reducing the conflict between information, and obtaining integrated evidence after weighted average, so that the integration of the following step seven is facilitated, and the fault diagnosis accuracy is improved.
When the actual fault is X, the integration evidence of one group of sample data to be detected is as follows:
when the actual fault is Y, the integration evidence of one group of sample data to be detected is as follows:
when the actual fault is Z, the integration evidence of one group of sample data to be detected is as follows:
Step seven: DS fusion of the integrated evidence and output of the diagnosis result.
And according to the DS fusion rule, fusing the integration evidence for 3 times to obtain a final fusion result.
When the actual fault is X, the final fusion result of one group of sample data to be detected is as follows:
m(X)＝0.9999，m(Y)＝3.44e-05，m(Z)＝4.83e-05；
The final diagnostic mode is: x is a group; the diagnosis result is correct.
When the actual fault is Y, the final fusion result of one group of sample data to be detected is as follows:
m(X)＝0.0081，m(Y)＝0.9886，m(Z)＝0.0033；
the final diagnostic mode is: y; the diagnosis result is correct.
When the actual fault is Z, the final fusion result of one group of sample data to be detected is as follows:
m(X)＝9.16e-04，m(Y)＝1.27e-05，m(Z)＝0.9991；
the final diagnostic mode is: z; the diagnosis result is correct.
According to the same method, the diagnosis results of all the samples to be detected are obtained through solving, as shown in table 5.
Table 5 diagnostic results of all samples to be tested (method of this patent)
The results in Table 5 show that the method provided by the patent has 100% of fault diagnosis accuracy for fault modes X and Y, 95% of fault diagnosis accuracy for fault mode Z, and 98.3% of total fault diagnosis accuracy, and the method provided by the patent has outstanding diagnosis effect and excellent diagnosis accuracy.
Step seven: in the gun fault mode framework, the integration evidence obtained in the step six is subjected to self fusion by using a Dempster-Shafer (DS) evidence theory method, and the occurrence probability of the corresponding gun anti-squat device fault mode is obtained; traversing the occurrence probability of the failure modes of all the gun anti-squat devices, and determining the failure mode corresponding to the maximum basic probability distribution value as the failure mode of final diagnosis, namely realizing high-precision and efficient diagnosis of the failure of the gun anti-squat devices based on RMSD-DS.
To further highlight the diagnostic effect of the method of this patent, diagnostic results obtained by applying DS fusion rules and BP neural network methods are presented as shown in tables 6-7.
TABLE 6 diagnostic results (DS) of all samples to be examined
The results in table 6 show that the accuracy of the DS method for fault diagnosis of fault mode X reaches 100%, 1%, 95% and 68.3% of the total error diagnosis is far lower than that of the method proposed in this patent.
TABLE 7 diagnostic results of all samples to be examined (BP neural network)
The results in table 7 show that the accuracy of the BP neural network method for fault diagnosis of the fault mode X reaches 70%, the accuracy of the fault diagnosis of the fault mode Y reaches 100%, the accuracy of the fault diagnosis of the fault mode Y reaches 75%, and the total accuracy of the fault diagnosis is 81.67%, which is far lower than that of the method proposed by the present patent.
As can be seen from a combination of Table 6 and Table 7, the present example has excellent diagnostic effects and higher diagnostic efficiency and accuracy.
Step eight: substituting the data of the sample to be detected of the faults of the gun recoil device determined in the step three into the fault mode Gaussian model constructed in the step four, and solving the basic probability distribution of the evidence corresponding to each fault characteristic signal; and D, re-distributing weights to each evidence by using the reliability determined in the step six, reducing the influence caused by conflict information, and improving the fault diagnosis performance of the anti-squat device. The improvement of the diagnosis performance of the anti-recoil device of the gun comprises the improvement of the diagnosis efficiency and the diagnosis precision of the anti-recoil device.
While the foregoing disclosure has been presented in a specific form for purposes of illustration, description and description, it should be understood that the invention is not limited to the particular embodiments disclosed, but is intended to cover modifications within the spirit and scope of the invention.
Claims (6)
1. The fault diagnosis method of the gun anti-squat device based on the RMSD-DS is characterized by comprising the following steps of:
step one: determining a typical fault mode and a fault characteristic signal of the gun anti-squat device;
Step two: acquiring a fault characteristic signal corresponding to the cannon anti-squat device in a typical fault mode;
Step three: analyzing the fault characteristic signals corresponding to the typical fault modes obtained in the second step, and classifying the fault characteristic signal data into fault training sample data and fault sample data to be detected; solving the average value and standard deviation of training samples belonging to different fault modes of the gun anti-squat device on different fault characteristic signals, and then constructing Gaussian models of the training samples belonging to different fault modes on different fault signals;
The third step comprises the following steps:
Step 3.1: classifying the fault characteristic signals obtained in the second step and corresponding to the typical fault mode into fault training sample data and fault sample data to be detected;
based on the fault characteristic signals corresponding to the typical fault modes obtained in the second step, sample data with preset proportion are selected from four fault characteristic signal data corresponding to each fault mode to serve as fault training samples, and the rest sample data serve as fault samples to be detected;
step 3.2: solving the average value and standard deviation of training samples belonging to different fault modes of the gun anti-squat device on different fault characteristic signals for the fault training samples selected in the step 3.1;
For the selected fault training samples, solving an average value mu (F i) and a standard deviation sigma (F i) of the training samples belonging to different fault modes on different fault characteristic signals, wherein a solving formula of the average value mu (F i) is shown in a formula 1, and a solving formula of the standard deviation sigma (F i) is shown in a formula 2:
In the formulas (1) and (2), f=x, Y, Z, representing three failure modes; i=1, 2,3,4, representing four fault signature signals; j=1, 2, …, N, representing the data sequence;
Step 3.3: constructing Gaussian models of training samples belonging to different fault modes on different fault signals according to the average value and the standard deviation obtained by solving in the step 3.2;
According to the mean value mu (F i) and the standard deviation sigma (F i) obtained in the step 3.2, constructing a Gaussian model of training samples belonging to different fault modes on different fault signals, wherein the Gaussian model is shown in a formula 3:
in formula 3, f=x, Y, Z, representing three failure modes; i=1, 2,3,4, representing four fault signature signals; x represents the data of the sample and, Representing the probability of the sample data being x;
When the fault characteristic signal is the maximum squat displacement Xmax, the fault mode is the gaussian model on the throttle ring wear X, the compound machine blow-by Y and the brake lever piston wear Z:
When the fault characteristic signal is the maximum squat speed Vmax, the gaussian model on the fault mode, which is the throttle ring wear X, the air leakage Y of the rewinder and the piston wear Z of the backing rod, is: and/>
When the fault characteristic signal is the maximum return speed Umax, the fault mode is a Gaussian model on the throttle ring abrasion X, the return machine leakage Y and the brake rod piston abrasion Z, and the Gaussian model is as follows: and/>
When the fault characteristic signal is the return-to-position speed Uend, the fault mode is a Gaussian model on the throttle ring abrasion X, the air leakage Y of the return-to-machine and the abrasion Z of the braking rod piston, and the Gaussian model is as follows: and/>
Formulas 4 to 15 are gaussian models of training samples of different fault modes on different fault signals;
Step four: according to the gun recoil device fault mode Gaussian model constructed in the step three, solving the basic probability distribution of evidence corresponding to the gun recoil device to-be-detected sample fault characteristic signal;
The fourth step comprises the following steps:
Step 4.1: solving the ordinate of the intersection point of the sample to be detected and the Gaussian models with different fault modes under each fault signal according to the Gaussian model with the fault mode of the gun anti-squat device constructed in the step 3.3;
For a set of samples to be inspected for which the failure mode is unknown, the corresponding data may be represented as { F 1,F2,F3,F4 }, where f= X, Y, Z; 1 in the subscript represents the maximum squat displacement Xmax,2 represents the maximum squat speed Vmax,3 represents the maximum return speed Umax, and 4 represents the return to bit speed Uend;
when the fault characteristic signal is the maximum squat displacement Xmax, the longitudinal coordinate solving formula of the intersection point of the sample to be detected and the Gaussian model of different fault modes is as follows:
when the fault characteristic signal is the maximum squat speed Vmax, the ordinate of the intersection point of the sample to be detected and the Gaussian model of different fault modes is as follows:
when the fault characteristic signal is the maximum complex advance speed Umax, the ordinate of the intersection point of the sample to be detected and the Gaussian model of different fault modes is as follows:
When the fault characteristic signal is the return-to-bit speed Uend, the ordinate of the intersection point of the sample to be detected and the Gaussian model of different fault modes is as follows:
step 4.2: representing the basic probability distribution of evidence corresponding to the four fault characteristic signals;
Each sample to be detected contains four fault characteristic signals, each fault characteristic signal corresponds to a group of evidence, and the basic probability distribution function of the evidence corresponding to each fault characteristic signal can be expressed as m i (i=1, 2,3, 4); in the first step, the failure modes of the gun recoil device include three types of X, Y and Z, and the basic probability distribution of a group of evidences comprises m i(X)、mi (Y) and m i (Z), wherein m i (X) represents the basic probability distribution of the sample to be detected belonging to the failure mode X under the evidence m i, m i (Y) represents the basic probability distribution of the sample to be detected belonging to the failure mode Y under the evidence m i, and m i (Z) represents the basic probability distribution of the sample to be detected belonging to the failure mode Z under the evidence m i;
step 4.3: according to the step 4.1, acquiring the longitudinal coordinates of the intersection points of the sample to be detected and the Gaussian models of different fault modes under each fault signal, and respectively solving the basic probability distribution of the evidence corresponding to each fault characteristic signal of the sample to be detected;
The basic probability distribution function solution formula of the corresponding evidence m 1 when the fault characteristic signal is the maximum squat displacement Xmax is:
the basic probability distribution function of the corresponding evidence m2 when the fault signature is the maximum squat speed Vmax is: m 2(X),m2(Y),m2 (Z);
The basic probability distribution function of the corresponding evidence m 3 when the fault characteristic signal is the maximum complex advance speed Umax is as follows: m 3(X),m3(Y),m3 (Z);
the basic probability distribution function of the corresponding evidence m 4 when the fault characteristic signal is the complex advance to the bit speed Uend is as follows: m 4(X),m4(Y),m4 (Z);
step five: according to the basic probability distribution of evidence corresponding to the fault characteristic signals of the sample to be detected of the gun recoil device obtained by solving in the step four, under the fault mode frame of the gun recoil device, defining and solving conflict factors among all the evidence, and constructing a conflict factor matrix according to all the solved conflict factors; defining and solving Root Mean Square Deviation (RMSD) distances among all evidences under the fault mode framework of the gun anti-squat device; taking the geometric mean of the conflict factors and the normalized RMSD distance as the value of the RMSD conflict coefficient to construct the RMSD conflict coefficient; according to the constructed RMSD conflict coefficient, solving and constructing an RMSD similarity coefficient, and defining the reliability of the following step six conveniently through the constructed RMSD similarity coefficient;
the fifth step comprises the following steps:
Step 5.1: under the fault mode framework of the gun anti-squat device, defining and solving conflict factors among all evidences, and constructing a conflict factor matrix according to all the solved conflict factors; the conflict factors among the evidences are the conflict factors among every two groups of evidences;
to facilitate the formulation of the solution to the conflict factor, under the cannon failure mode framework Θ= { X, Y, Z } m 1 and m 2 are defined as two groups of evidence, the corresponding failure modes are denoted as F '{ F' =x, Y, Z } and F "{ F" =x, Y, Z } respectively, and the conflict factors of evidence m 1 and m 2 are shown as formula 40:
The collision factor matrix between evidence is obtained according to equation 40:
However, the conflict factor has defects, and according to the formula 40, the conflict factor obtained by solving when two evidences are the same is not 0, so that the conflict factor needs to be corrected by introducing the RMSD distance between the evidences in the subsequent step 5.2, and the conflict factor, namely the RMSD conflict factor, is constructed in the subsequent step 5.3;
Step 5.2: defining and solving the RMSD distance between each evidence under the fault mode framework of the gun anti-squat device, constructing an RMSD distance matrix according to all the solved RMSD distances, and solving the normalized RMSD distance matrix;
Defining m 1 and m 2 as two sets of evidence, the corresponding failure modes are denoted as F '{ F' =x, Y, Z } and F "{ F" =x, Y, Z }, respectively, then defining the RMSD distance between evidence m 1 and m 2 as:
solving for the RMSD distance matrix between evidence according to equation 42, as shown in equation 43:
Then find the maximum RMSD max =max { RMSD }, in the RMSD distance matrix, then normalize the RMSD distance matrix, i.e., divide each element of the RMSD distance matrix by RMSD max, to obtain the RMSD distance matrix after normalization, as shown in equation 44:
step 5.3: defining a geometric mean value of the conflict factor in the step 5.1 and the normalized RMSD distance in the step 5.2 as a value of an RMSD conflict coefficient, and constructing the RMSD conflict coefficient;
Taking the geometric mean of the conflict factor obtained in the step 5.1 and the normalized RMSD distance obtained in the step 5.2 as the value of the RMSD conflict coefficient, the RMSD conflict coefficients of the definition evidences m 1 and m 2 are expressed as:
In formula 45, K (m 1,m2) represents the collision factor of evidence m 1 and m 2, Representative evidence m 1 and m 2 normalized RMSD distances;
The RMSD collision coefficient matrix is obtained according to equation 45:
step 5.4: solving and constructing RMSD similarity coefficients based on the RMSD collision coefficients in the step 5.3;
The RMSD collision coefficient constructed in step 5.3 represents the collision degree between evidences, the value range is [0,1], and subtracting the RMSD collision coefficient from 1 can represent the similarity degree between evidences, and the RMSD similarity coefficient defining evidences m 1 and m 2 is expressed as:
The RMSD similarity coefficient matrix may be obtained according to equation 47 as:
Step six: defining the reliability of each evidence as the sum of RMSD similarity coefficients of the evidence and other evidence, and determining the reliability of each evidence according to the definition; analyzing the reliability of each evidence, defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, and solving the weight of each evidence; weight distribution is carried out on all evidences according to the reliability of each evidence, so that the conflict between information is reduced, then integrated evidence is obtained after weighted average, fusion in the subsequent step seven is facilitated, and the fault diagnosis accuracy is improved;
Step seven: in the gun fault mode framework, the DS evidence theory method is utilized to carry out self fusion on the integration evidence obtained in the step six, and the occurrence probability of the corresponding gun anti-squat device fault mode is obtained; traversing the occurrence probability of the failure modes of all the gun anti-squat devices, and determining the failure mode corresponding to the maximum basic probability distribution value as the failure mode of final diagnosis, namely realizing high-precision and efficient diagnosis of the failure of the gun anti-squat devices based on RMSD-DS.
2. The RMSD-DS based gun recoil device failure diagnosis method of claim 1, further comprising step eight: substituting the data of the sample to be detected of the faults of the gun recoil device determined in the step three into the fault mode Gaussian model constructed in the step four, and solving the basic probability distribution of the evidence corresponding to each fault characteristic signal; the reliability determined in the step six is used for reassigning weights to each evidence, so that the influence caused by conflict information is reduced, and the fault diagnosis performance of the anti-squat device is improved; the improvement of the diagnosis performance of the anti-squat device comprises the improvement of the diagnosis efficiency and the diagnosis precision of the anti-squat device.
3. The method for diagnosing faults of an artillery anti-squat device based on RMSD-DS as claimed in claim 1 or 2, wherein the implementation method of the first step is as follows:
Determining three typical failure modes of the gun anti-squat device, namely, a check ring abrasion X, a compound feed machine air leakage Y and a brake rod piston abrasion Z, wherein the failure mode frame of the gun anti-squat device is expressed as Θ= { X, Y, Z }; the fault characteristic signals for determining the anti-squat device of the gun are respectively as follows: maximum squat displacement Xmax, maximum squat speed Vmax, maximum recoil speed Umax, and recoil to position speed unend.
4. The method for diagnosing faults of an artillery anti-squat device based on RMSD-DS as claimed in claim 3, wherein the implementation method of the second step is as follows:
When the cannon works, four fault characteristic signals of the anti-squat device under each fault mode are collected by a sensor arranged on the anti-squat device, the obtained data are represented by F i, wherein F=X, Y and Z respectively represent three fault modes of abrasion of a check ring, air leakage of a re-entry machine and abrasion of a brake rod piston; i=1, 2,3,4, representing four fault signature signals of maximum squat displacement Xmax, maximum squat speed Vmax, maximum return speed Umax and return to position speed Uend, respectively; the acquired set of sample data is represented as { F 1,F2,F3,F4},F1 representing maximum squat displacement Xmax signal data corresponding to failure mode F, F 2 representing maximum squat speed Vmax signal data corresponding to failure mode F, F 3 representing maximum recoil speed Umax signal data corresponding to failure mode F, and F 4 representing recoil to bit speed Uend signal data corresponding to failure mode F.
5. The RMSD-DS based gun recoil device malfunction diagnosis method as defined in claim 4, wherein the sixth step includes the steps of:
Step 6.1: defining the reliability of each evidence as the sum of RMSD similarity coefficients of the evidence and other evidence, and determining the reliability of each evidence according to the definition;
Defining the reliability of each evidence as the sum of the RMSD similarity coefficients of that evidence and other evidence, the reliability of evidence m 1 is formulated as:
The reliability of evidence m 2 is formulated as:
the reliability of evidence m 3 is formulated as:
The reliability of evidence m 4 is formulated as:
Wherein, the reliability of the evidence represents the supporting degree of other evidence on the evidence;
the greater the reliability of the evidence, the higher the importance of the evidence in the fusion decision process, and the weight distributed in the subsequent step 6.2 is preferably greater;
The smaller the reliability of the evidence is, the lower the importance of the evidence in the fusion decision process is, and the weight distributed in the subsequent step 6.2 is preferably small;
Step 6.2: analyzing the reliability of each evidence determined in the step 6.1, defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, and solving the weight of each evidence;
defining the weight of each evidence as the ratio of the reliability of the evidence to the sum of the reliability of all the evidence, the weight of evidence m 1 is expressed as:
The reliability of evidence m 2 is formulated as:
the reliability of evidence m 3 is formulated as:
The reliability of evidence m 4 is formulated as:
Step 6.3: weight distribution is carried out on all evidences according to the reliability of each evidence in the step 6.2, the conflict between information is reduced, then integrated evidence is obtained after weighted average, fusion of the subsequent step seven is facilitated, and the fault diagnosis accuracy is improved;
according to the weight distributed in the step 6.2, the integration evidence is obtained after weighted average, and is expressed as follows:
the basic probability distribution for the failure modes X, Y and Z under the integrated evidence is obtained and expressed as:
6. the RMSD-DS based gun recoil device failure diagnosis method of claim 5, wherein the seventh step includes the steps of:
Step 7.1: defining m 1 and m 2 as two groups of evidence, and respectively representing corresponding fault modes as F '{ F' =x, Y, Z } and F "{ F" =x, Y, Z }, and giving DS fusion rules of evidence m 1 and m 2;
For convenience in giving DS fusion rules, under the cannon fault mode framework Θ= { X, Y, Z }, m 1 and m 2 are two groups of evidence, and the corresponding fault modes are respectively denoted as F '{ F' =x, Y, Z } and F "{ F" =x, Y, Z }, and then DS fusion rules of evidence m 1 and m 2 are shown in formula 61:
Step 7.2: according to the DS fusion rule given in the step 7.1, the integration evidence is fused for 3 times, and the DS fusion rule is utilized to fuse the integration evidence, so that the occurrence probability of the fault mode of the corresponding gun anti-squat device is obtained;
Step 7.3: and 7.2, traversing to obtain occurrence probabilities of failure modes of all gun anti-squat devices, and determining the failure mode corresponding to the maximum basic probability distribution value as the failure mode of final diagnosis, namely realizing high-precision and efficient diagnosis of gun anti-squat device failures based on the RMSD-DS method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210649020.6A CN115048959B (en) | 2022-06-09 | 2022-06-09 | Method for diagnosing faults of gun anti-squat device based on RMSD-DS |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210649020.6A CN115048959B (en) | 2022-06-09 | 2022-06-09 | Method for diagnosing faults of gun anti-squat device based on RMSD-DS |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115048959A CN115048959A (en) | 2022-09-13 |
| CN115048959B true CN115048959B (en) | 2024-06-21 |
Family
ID=83162215
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210649020.6A Active CN115048959B (en) | 2022-06-09 | 2022-06-09 | Method for diagnosing faults of gun anti-squat device based on RMSD-DS |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115048959B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108520266A (en) * | 2018-03-01 | 2018-09-11 | 西北工业大学 | A kind of Time Domain Fusion method for diagnosing faults based on DS evidence theories |
| CN109540520A (en) * | 2018-11-29 | 2019-03-29 | 中国船舶重工集团海装风电股份有限公司 | A kind of rolling bearing fault fusion diagnosis method based on improvement D-S evidence theory |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110297141A (en) * | 2019-07-01 | 2019-10-01 | 武汉大学 | Fault Locating Method and system based on multilayer assessment models |
| CN113063314B (en) * | 2021-03-23 | 2022-03-22 | 哈尔滨工程大学 | Fault diagnosis method for gun launching system based on SVM (support vector machine) and GA-SVM (genetic algorithm-support vector machine) |
| CN114444585A (en) * | 2022-01-13 | 2022-05-06 | 北京理工大学 | Multi-source information fusion method for conflict evidence |
-
2022
- 2022-06-09 CN CN202210649020.6A patent/CN115048959B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108520266A (en) * | 2018-03-01 | 2018-09-11 | 西北工业大学 | A kind of Time Domain Fusion method for diagnosing faults based on DS evidence theories |
| CN109540520A (en) * | 2018-11-29 | 2019-03-29 | 中国船舶重工集团海装风电股份有限公司 | A kind of rolling bearing fault fusion diagnosis method based on improvement D-S evidence theory |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115048959A (en) | 2022-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111259927B (en) | Rocket engine fault diagnosis method based on neural network and evidence theory | |
| CN108346144B (en) | Automatic bridge crack monitoring and identifying method based on computer vision | |
| CN112308381B (en) | Equipment contribution data analysis method, system, storage medium and computer device | |
| CN110895878B (en) | Traffic state virtual detector generation method based on GE-GAN | |
| CN109063242B (en) | Guidance tool error identification method based on particle swarm optimization | |
| CN111950627A (en) | Multi-source information fusion method and application thereof | |
| CN111222095B (en) | Rough difference judging method, device and system in dam deformation monitoring | |
| CN113065223B (en) | Multi-level probability correction method for digital twin model of tower mast cluster | |
| CN110633790A (en) | Method and system for measuring residual oil quantity of airplane oil tank based on convolutional neural network | |
| CN112860952A (en) | Method and system for positioning and repairing parts of inspection robot | |
| CN110874509A (en) | Multidimensional information fusion state evaluation method for high-end numerical control equipment | |
| CN115048959B (en) | Method for diagnosing faults of gun anti-squat device based on RMSD-DS | |
| CN115546223A (en) | Method and system for detecting loss of fastening bolt of equipment under train | |
| CN109635389A (en) | A kind of electric steering engine stiffness test data processing method | |
| CN116412995A (en) | Method for determining wind tunnel strain balance measurement uncertainty among partitions | |
| CN112131752B (en) | Super-collapse pollution rate tolerance estimation algorithm based on quasi-calibration | |
| CN111753465B (en) | Method for diagnosing leakage fault in buffering oil cylinder of anti-collision buffering hydraulic system of ship lift | |
| CN117745096A (en) | Evaluation method and system for control screen cloth measurement scheme | |
| CN114993604B (en) | Wind tunnel balance static calibration and measurement method based on deep learning | |
| CN116385292A (en) | IMU-assisted LiDAR dynamic point cloud eliminating method | |
| CN112906746B (en) | Multi-source track fusion evaluation method based on structural equation model | |
| CN113624192B (en) | Large diesel engine hole coaxiality measuring method based on multiple optimization | |
| CN114722695A (en) | FADS resolving system and method based on dimensionless input and output neural network | |
| CN111738288B (en) | Bullet warhead fast matching method based on single-point laser detection | |
| CN114065379A (en) | Model reliability evaluation method based on small sample data |
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 |