CN111597655A - Product health judgment method based on fault occurrence probability - Google Patents

Product health judgment method based on fault occurrence probability Download PDF

Info

Publication number
CN111597655A
CN111597655A CN202010423177.8A CN202010423177A CN111597655A CN 111597655 A CN111597655 A CN 111597655A CN 202010423177 A CN202010423177 A CN 202010423177A CN 111597655 A CN111597655 A CN 111597655A
Authority
CN
China
Prior art keywords
product
probability
data
failure
distribution density
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
CN202010423177.8A
Other languages
Chinese (zh)
Other versions
CN111597655B (en
Inventor
秦鹏举
贾东明
鞠宏艳
张翔宇
阎涛
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.)
Xian Aerospace Propulsion Institute
Original Assignee
Xian Aerospace Propulsion Institute
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 Xian Aerospace Propulsion Institute filed Critical Xian Aerospace Propulsion Institute
Priority to CN202010423177.8A priority Critical patent/CN111597655B/en
Publication of CN111597655A publication Critical patent/CN111597655A/en
Application granted granted Critical
Publication of CN111597655B publication Critical patent/CN111597655B/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
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • 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
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04—Ageing analysis or optimisation against ageing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/12—Timing analysis or timing optimisation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Abstract

The invention provides a product health judgment method based on fault occurrence probability, which can be used for judging product health by converting data into a probability distribution density function and then carrying out statistic limit distribution analysis when no big data exists. The time sequence data obtained in the test is converted into the probability distribution density functions, then a statistic is constructed, the limit distribution of the statistic is determined, and whether the two probability distribution density functions are from the same parent is judged according to the limit distribution at a certain significance level, so that the technical problem in the field of health monitoring and detection can be solved by means of research results from the field of statistics, fault probability calculation is completed without depending on big data, a large amount of test expenditure and test time are saved, the same effect can be realized by only needing two tests (one normal product and one fault product), and the economic benefit is very obvious.

Description

Product health judgment method based on fault occurrence probability
Technical Field
The invention belongs to the technical field of time sequence data processing, and particularly relates to a product health judgment method based on fault occurrence probability.
Background
The solid rocket engine has the characteristics of simple structure, reliable work, simple maintenance, convenient use and the like, and is widely applied to the military field. However, the engine has the particularity of 'long-term storage, disposable use and test-free storage', so that the engine has important significance in health monitoring.
According to the research results at home and abroad, the ultrasonic guided wave is one of the stress waves, and the guided wave detection not only has a series of advantages of a wave information detection method, but also has the characteristics of long detection distance, high detection efficiency, high speed and the like. The composite material is not influenced by low-frequency vibration of the surrounding environment, is suitable for a plurality of types, and can generate stronger response to common defects such as debonding, delamination, cracks and the like in a multilayer bonding structure (such as a solid rocket engine: comprising a metal/composite material shell, a heat insulation layer, a grain structure and the like).
When the product health of the solid engine is broken through the ultrasonic guided waves, the change data (namely time series data) of the obtained signals along with time needs to be processed, and then the fault occurrence rate of the product is obtained. However, the existing time series data processing is mainly to convert the time series data into a frequency domain for feature analysis through methods such as fourier analysis and wavelet analysis, and is only suitable for time series data with relatively long response time, and the response time of the time series data obtained by an ultrasonic guided wave monitoring mode is very short (1ms), so that the response of the natural frequency of a product cannot be activated, and effective frequency domain information cannot be obtained, and thus, the product health judgment cannot be effectively carried out.
In addition, in the research process, the difficulty of judging whether a product is healthy or not is found to be the number of tests, if the number of tests is enough, the distribution of parameters under a normal condition can be judged by a big data method, and then whether a certain product has a fault or not is judged according to the significance level.
Disclosure of Invention
In view of the above, the present invention provides a method for judging product health based on failure occurrence probability, which can convert data into a probability distribution density function and then perform statistic limit distribution analysis for judging product health when there is no big data.
In order to achieve the above object, the product health judgment method based on the fault occurrence probability of the invention is as follows:
measuring a normal product and a fault product at the same measuring point by using the same testing method to obtain two groups of time sequence data; acquiring distribution density function f of two groups of time sequence data1And f2；
Calculating the distance d (f) between two distribution density functions1,f2) The probability P (d (f)) of occurrence of a failure at the measurement point is obtained1,f2) N), where n is an equivalent degree of freedom;
according to P (d (f)1,f2),n)＝P0Solving for the equivalent degree of freedom n at this time, where P0Is a known probability of failure occurrence;
solving d according to the required tolerance fault occurrence probability P and according to P (d, n) ═ P, wherein d is a distance critical value when the fault occurrence probability is P;
obtaining time sequence data of a product to be tested and a distribution density function f thereof at the same measuring point by the same testing method3D (f) is calculated1,f3)；
When d (f)1,f3)>d, the probability that the product is in failure is greater than the normal probability of the product; when d (f)1,f3)<d, the probability that the product is normal is greater than the product failure probability; d (f)1,f3) When d, the probability of failure and normality is half each.
Wherein the product is a solid rocket engine.
Wherein a distribution density function f of two sets of time series data is obtained1And f2The specific mode is as follows:
firstly, carrying out null shift correction processing on the two groups of time sequence data;
calculating the envelope curves of the two groups of data after the zero drift correction processing, and then carrying out normalization processing on the envelope curves of the two groups of data to obtain a distribution density function f1And f2。
Wherein the content of the first and second substances,
Figure BDA0002497711600000031
when is, P (d (f)1,f2),n))＝χ2(d(f1,f2) N); where τ represents time, and T represents the termination time of the health monitoring time-series data.
Wherein the product health monitoring time series data is obtained by ultrasonic guided wave monitoring.
Has the advantages that:
finding a qualified product to perform a health monitoring test to obtain time sequence data under a normal condition, then finding a non-qualified product to perform the health monitoring test to obtain the time sequence data under a fault condition, wherein the test quantity is too small to perform statistical analysis, the invention converts the time sequence data obtained in the test into probability distribution density functions, then constructs a statistic to determine the limit distribution of the statistic, and then judges whether the two probability distribution density functions are from the same parent body according to the limit distribution at a certain significance level, so that the technical problem in the health monitoring test field can be solved by means of research results from the statistical field, the fault probability calculation can be completed without depending on large data, a large amount of test expenditure and test time are saved, and the fault criterion of the test data is completed in the traditional method, at least more than 20 effective sample amounts are needed, and the method can achieve the same effect only by two tests (one normal product and one fault product), and has obvious economic benefit.
The method is not only suitable for time series data obtained by ultrasonic guided wave monitoring, but also suitable for processing test data obtained by various short-time vibration tests and short-time strain tests on the surface of a product shell or a component, and can be applied to occasions of judging whether the product shell or the component is abnormal or not according to the time series data obtained by the various short-time test tests on the surface of the product shell or the component.
Drawings
FIG. 1 is a graph showing the results of an engine charge test according to the present invention;
FIG. 2 is a graph showing the results of the engine case test of the present invention;
FIG. 3 is the results of the present invention after engine charge zero drift correction;
FIG. 4 shows the results of the null shift correction of the engine of the present invention;
FIG. 5 shows the results of the zero drift correction of the engine charge and its envelope;
FIG. 6 shows the results of the null shift correction of the engine case and its envelope curve according to the present invention;
FIG. 7 is a graph of normalized engine charge results according to the present invention;
FIG. 8 shows the normalized results of the engine case of the present invention.
FIG. 9 is a flow chart of obtaining a distance threshold according to the present invention.
Detailed Description
The invention is described in detail below by way of example with reference to the accompanying drawings.
The invention uses the technology of the limit distribution of the statistics for judging health, and grafts the research results in two originally unrelated fields into a new method for solving the technical problem that whether the product is healthy or not is judged by a probability method without big data.
The embodiment provides a test to normal state and fault state product through methods such as supersound guided wave, obtains two sets of time sequence data, handles two sets of time sequence data and obtains the distance between two sets of time sequence data, and then obtains the critical value of distance, through the contrast with this critical value, can calculate the probability that the product that awaits measuring breaks down.
The product health judgment method based on the fault occurrence probability in the embodiment specifically comprises the following steps:
step 1, obtaining two groups of time sequence data under normal state and fault state products;
the normal state product and the fault state product refer to two products with the same technical state, and can also refer to the same product in different time period states.
In the embodiment, under a normal state product, an ultrasonic guided wave excitation signal with a specific frequency and a specific waveform is generated at a signal excitation position through a waveform generator, and the signal is received at a signal receiving position.
In a fault state product, the excitation signal is generated at the same frequency and waveform at the same signal excitation as in a normal state product, and the signal is received at the same signal reception.
And 2, performing null shift correction processing on the two groups of data.
And 3, calculating the envelope curves of the two groups of data after the null shift correction.
Step 4, normalization processing is carried out on the envelope curves of the two groups of data, and after normalization, the data can be regarded as a distribution density function f1And f2；
Step 5, calculating a formula according to the distance
Figure BDA0002497711600000041
Calculating the distance between two distribution density functions according toThe probability theory research result shows that the limit distribution has the same limit distribution as the log likelihood ratio function and is x2(d(f1,f2) N), i.e. the probability of failure at the measuring point is χ2(d(f1,f2) N), where n is the equivalent degree of freedom.
Step 6, according to chi2(d(f1,f2),n)＝P0Solving for the equivalent degree of freedom n at this time, where P0The probability of failure under the condition of known failure is that the theoretical value is 1, but 1 cannot be calculated, so that data close to 1 needs to be found, and P can be recommended to a certain appointed value0Is 0.99.
Step 7, according to the required tolerance fault occurrence probability P and the X2And (d, n) ═ P, and d is obtained by solving, wherein d is the distance critical value when the fault occurrence probability is P. In general, P is 0.5. The flow chart for obtaining the distance threshold is shown in fig. 9.
Step 8, obtaining time sequence data of the product to be detected at the same measuring point, and obtaining a distribution density function f of the time sequence data of the product to be detected in the same way as the step 1-43D (f) is calculated1,f3)；
When d (f)1,f3)>d, the probability that the product is in failure is greater than the normal probability of the product; when d (f)1,f3)<d, the probability that the product is normal is greater than the product failure probability; d (f)1,f3) When d, it means that the probability of failure and the probability of normality are each half, and the failure should be handled generally.
Wherein f is1And f2The normalized functions are respectively normal products and fault products, tau represents time, and T represents the termination time of the health monitoring time sequence data.
In this embodiment, health monitoring is performed on a product by an ultrasonic guided wave method, and time series data is obtained.
The concrete case is as follows: the data processing method is described by taking test data in two extreme states (loaded and empty (i.e. not loaded)) of a solid rocket engine as an example. The solid rocket engine under two states is respectively tested by the same method and the same measuring point, wherein the excitation signal is a sine signal of 50kHz, and the sampling time is as follows: [ -100.5 μ s,899.4 μ s ], for 1ms, and the test results are shown in FIGS. 1 and 2.
The data processing process is as follows:
1) zero drift correction processing is carried out on data
For experimental data, among others, order
Figure BDA0002497711600000051
Then
y1(t)＝y(t)-y0(2)
Thus, the results after zero drift correction of the engine charge and the empty case are shown in FIGS. 3 and 4.
2) Calculating envelope curve of each group of data by Hilbert method
For data y1(t) performing a Hibert transform:
Figure BDA0002497711600000052
having an envelope of
y2(t)＝|y1(t)+iH[y1(t)]| (4)
The envelope of the engine charge and the empty case is shown in figures 5 and 6.
4) Normalizing each group of data
Figure BDA0002497711600000061
After normalization, the data can be viewed as a distribution density function.
The results after normalization of the engine charge and the empty case are shown in fig. 7 and 8.
5) And calculating the distance between the two distribution density functions, defining the distance between the two distribution density functions, if the distance value is small, proving that the two curves are relatively close, and considering that the two curves have no difference or have small difference, and if the distance value is large, proving that the two curves have large difference.
The distance is defined as follows:
Figure BDA0002497711600000062
wherein f is1And f2Normalized functions for the charged and empty case engines, respectively.
d(f1,f2) The same limit distribution as the log-likelihood function is χ2Distribution, the process of which is demonstrated as follows: handle f1,f2Considering a distribution density function from a certain distribution family (Θ), there is an n-dimensional parameter θ1,θ2∈ Θ, such that
f1(t)＝f(t；θ1)；f2(t)＝f(t；θ2)
Theta is measured1Consider a true value, let θ2Regarded as theta1The property of the estimated maximum likelihood value accords with various properties of the estimated maximum likelihood, and the asymptotic distribution of the known log likelihood ratio function is chi2Distributed with n degrees of freedom, and should have
Figure BDA0002497711600000063
Wherein the degree of freedom n is an unknown parameter.
The calculated distance between the shell and the charge engine was 0.915, noting that χ2Since (0.915,1) is 0.661 max, the degree of freedom n is 1.
For a given probability of failure P, the threshold d satisfies χ2D can be calculated from (d,1) ═ P. For P-0.5, d-0.455.
6) Arranging measuring points on the outer surface of the engine to be measured, and then performing a test to obtain a response curve under the specified excitation frequency;
7): preprocessing the data according to 1) to 4), obtaining f x (t) curves, and calculating to obtain
Figure BDA0002497711600000071
Then d (f) at the measurement point1,f*)>The probability of failure at 0.455 is greater than normal, when d (f) is at the measurement point1,f*)<The probability of normal at 0.455 is greater than the probability of failure.
In summary, the above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (5)

1. A product health judging method based on fault occurrence probability is characterized in that two groups of time sequence data are obtained by measuring a normal product and a fault product at the same measuring point by the same testing method; acquiring distribution density function f of two groups of time sequence data1And f2；
Calculating the distance d (f) between two distribution density functions1,f2) The probability P (d (f)) of occurrence of a failure at the measurement point is obtained1,f2) N), where n is an equivalent degree of freedom;
according to P (d (f)1,f2),n)＝P0Solving for the equivalent degree of freedom n at this time, where P0Is a known probability of failure occurrence;
solving d according to the required tolerance fault occurrence probability P and according to P (d, n) ═ P, wherein d is a distance critical value when the fault occurrence probability is P;
obtaining time sequence data of a product to be tested and a distribution density function f thereof at the same measuring point by the same testing method3D (f) is calculated1,f3)；
When d (f)1,f3)>d, the probability that the product is in failure is greater than the normal probability of the product; when d (f)1,f3)<d, the probability that the product is normal is greater than the product failure probability; d (f)1,f3) When d, the probability of failure and normality is half each.
2. The method of claim 1, wherein the product is a solid rocket engine.
3. The method of claim 1, wherein the distribution density function f of two sets of time series data is obtained1And f2The specific mode is as follows:
firstly, carrying out null shift correction processing on the two groups of time sequence data;
calculating the envelope curves of the two groups of data after the zero drift correction processing, and then carrying out normalization processing on the envelope curves of the two groups of data to obtain a distribution density function f1And f2。
4. The method for product health judgment based on failure occurrence probability according to claim 1,
Figure FDA0002497711590000011
when is, P (d (f)1,f2),n))＝χ2(d(f1,f2) N); where τ represents time, and T represents the termination time of the health monitoring time-series data.
5. The method for product health assessment based on probability of failure according to claim 1, wherein the product health monitoring time series data is obtained by ultrasound guided wave monitoring.
CN202010423177.8A 2020-05-19 2020-05-19 Product health judging method based on fault occurrence probability Active CN111597655B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010423177.8A CN111597655B (en) 2020-05-19 2020-05-19 Product health judging method based on fault occurrence probability

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010423177.8A CN111597655B (en) 2020-05-19 2020-05-19 Product health judging method based on fault occurrence probability

Publications (2)

Publication Number Publication Date
CN111597655A true CN111597655A (en) 2020-08-28
CN111597655B CN111597655B (en) 2023-04-21

Family

ID=72182585

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010423177.8A Active CN111597655B (en) 2020-05-19 2020-05-19 Product health judging method based on fault occurrence probability

Country Status (1)

Country Link
CN (1) CN111597655B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070220298A1 (en) * 2006-03-20 2007-09-20 Gross Kenny C Method and apparatus for providing fault-tolerance in parallel-processing systems
CN102411106A (en) * 2011-11-18 2012-04-11 广东电网公司广州供电局 Fault monitoring method and device for power transformer
CN105894027A (en) * 2016-03-31 2016-08-24 华北电力科学研究院有限责任公司 Principal element degree of association sensor fault detection method and apparatus based on density clustering
CN107132310A (en) * 2017-03-28 2017-09-05 浙江大学 Transformer equipment health status method of discrimination based on gauss hybrid models

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070220298A1 (en) * 2006-03-20 2007-09-20 Gross Kenny C Method and apparatus for providing fault-tolerance in parallel-processing systems
CN102411106A (en) * 2011-11-18 2012-04-11 广东电网公司广州供电局 Fault monitoring method and device for power transformer
CN105894027A (en) * 2016-03-31 2016-08-24 华北电力科学研究院有限责任公司 Principal element degree of association sensor fault detection method and apparatus based on density clustering
CN107132310A (en) * 2017-03-28 2017-09-05 浙江大学 Transformer equipment health status method of discrimination based on gauss hybrid models

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
王俨剀等: "航空发动机健康等级综合评价方法", 《航空动力学报》 *
辛业春等: "基于多变量多尺度云样本熵和模糊支持向量机的开关柜故障分类", 《电网技术》 *

Also Published As

Publication number Publication date
CN111597655B (en) 2023-04-21

Similar Documents

Publication Publication Date Title
Ramasso et al. Unsupervised consensus clustering of acoustic emission time-series for robust damage sequence estimation in composites
Kim et al. Estimation of porosity content of composite materials by applying discrete wavelet transform to ultrasonic backscattered signal
CN114235241B (en) Nonlinear ultrasonic stress detection method and device based on Moire circle stress principle and composite material plane stress detection method
CN114061805A (en) Air coupling Lamb wave nonlinear ultrasonic stress detection method, system and device for fiber reinforced composite material
CN109596710A (en) The device and method of sleeve grouting defect ultrasound detection based on wavelet-packet energy
CN114330435A (en) Composite material defect detection method based on dynamic weight wavelet coefficient deep residual network
Moctezuma et al. Performance analysis of acoustic emission hit detection methods using time features
Wang et al. Acoustic emission characteristics of used 70 MPa type IV hydrogen storage tanks during hydrostatic burst tests
Li et al. State estimation of a lithium-ion battery based on multi-feature indicators of ultrasonic guided waves
Song et al. An improved structural health monitoring method utilizing sparse representation for acoustic emission signals in rails
Lei et al. Ultrasonic pig for submarine oil pipeline corrosion inspection
CN117214294A (en) Ultrasonic guided wave detection device for pipeline damage and three-dimensional reconstruction method for damage signal of ultrasonic guided wave detection device
Wang Wavelet Transform Based Feature Extraction for Ultrasonic Flaw Signal Classification.
Bai et al. Crack location in beams by data fusion of fractal dimension features of laser-measured operating deflection shapes
CN111597655A (en) Product health judgment method based on fault occurrence probability
Sodagar et al. An alternative approach for measuring the scattered acoustic pressure field of immersed single and multiple cylinders
Ding et al. Spacecraft leakage detection using acoustic emissions based on empirical mode decomposition and support vector machine
Wang et al. The optimal wavelet threshold de-nosing method for acoustic emission signals during the medium strain rate damage process of concrete
CN113640394B (en) Fatigue crack identification method and system
CN113884567B (en) Rail weld damage detection method and device based on ultrasonic Lamb wave
Al-Jumaili Damage assessment in complex structures using acoustic emission
Yumurtaci et al. Determining damages in ceramic plates by using discrete wavelet packet transform and support vector machine
CN111122704A (en) Method for quantitatively predicting damage area based on impact of higher harmonic on composite material
Rajendra et al. Physics-based classification of acoustic emission waveforms
CN111506973A (en) Product health judgment method based on product health monitoring time series 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