CN116522089A - Method and system for predicting early fatigue crack and residual life of metal structure - Google Patents

Method and system for predicting early fatigue crack and residual life of metal structure Download PDF

Info

Publication number
CN116522089A
CN116522089A CN202310483173.2A CN202310483173A CN116522089A CN 116522089 A CN116522089 A CN 116522089A CN 202310483173 A CN202310483173 A CN 202310483173A CN 116522089 A CN116522089 A CN 116522089A
Authority
CN
China
Prior art keywords
fatigue
sample
fatigue crack
test piece
damage
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310483173.2A
Other languages
Chinese (zh)
Inventor
吕珊珊
姜明顺
魏钧涛
张雷
张法业
隋青美
贾磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong University
Original Assignee
Shandong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shandong University filed Critical Shandong University
Priority to CN202310483173.2A priority Critical patent/CN116522089A/en
Publication of CN116522089A publication Critical patent/CN116522089A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4472—Mathematical theories or simulation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4409—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison
    • G01N29/4418—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison with a model, e.g. best-fit, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0475—Generative networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Signal Processing (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Pathology (AREA)
  • Immunology (AREA)
  • Biochemistry (AREA)
  • Analytical Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Algebra (AREA)
  • Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)

Abstract

The invention provides a method and a system for predicting early fatigue cracks and residual life of a metal structure, wherein the method comprises the following steps: based on the multidimensional nonlinear ultrasonic response characteristics, establishing a fatigue crack damage actual measurement sample of a typical thickness test piece; generating a fatigue damage virtual sample with the distribution characteristic height consistent with that of the actual measurement sample, and expanding the data of the actual measurement sample; calibrating fatigue parameters of the virtual samples, and constructing a fatigue crack damage sample library by combining the fatigue crack damage actual measurement samples; based on the fatigue crack damage sample library, an intelligent prediction model is established and used for quantitatively evaluating the fatigue crack extension length and the residual service life of the test piece.

Description

Method and system for predicting early fatigue crack and residual life of metal structure
Technical Field
The invention belongs to the technical field of structural health monitoring, and particularly relates to a method and a system for predicting early fatigue cracks and residual life of a metal structure.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The metal structure has excellent strength and workability, and is widely applied to the fields of aerospace, rail transit, nuclear power wind power and the like. However, under the influence of severe service environment and complex alternating load, fatigue microcracks are easily generated at stress concentration positions, and the microcracks are continuously expanded to form macrocracks, so that fatigue failure is finally caused. It is researched that the early fatigue property degradation of the material accounts for 80% -90% of the service life of the structure, but once macroscopic damage is formed, the macroscopic damage can rapidly expand and lead to sudden fracture failure of the structure, and even serious safety accidents are caused. As can be seen, microcracking has become a significant potential hazard affecting the proper operation of equipment structures.
The nonlinear ultrasonic guided wave technology can go deep to a microscopic level, evaluate the performance degradation of the material based on nonlinear interaction between waves and damage, essentially reflect the influence of micro defects on the nonlinearity of the material, is very sensitive to the change of an early microstructure (micron level) of the material, has the advantages of no damage, high detection efficiency and the like, and is an effective means for realizing quantification of early fatigue cracks of the structure and prediction of the residual life.
The existing nonlinear ultrasonic guided wave fatigue crack prediction technology mainly comprises a model-based prediction method and a data-driven prediction method. The model prediction method predicts the current running condition of the structure based on physical failure theory by deeply analyzing dynamics, material characteristics and the like of the structure, and generally needs to comprehensively consider physical, chemical, pneumatic-thermal and other processes experienced in the service process of the structure to establish a complex failure mechanism mathematical model, but the complexity of modeling and analysis limits the application and popularization of the method. In contrast, the data driving method does not depend on structural knowledge, and can be used for predicting structural fatigue crack damage by establishing a sample database and extracting features corresponding to structural states in sample monitoring data, so that the method is a research hotspot in the current structural health monitoring field.
Although the data-driven prediction method has higher application flexibility, the method requires the input of massive nonlinear response information, however, a single test piece in a real fatigue experiment can only obtain one piece of complete nonlinear response data, the experiment time consumption is obviously prolonged along with the increase of the thickness of the test piece, and therefore, a large amount of structural nonlinear response data is difficult to obtain through the experiment. In addition, the difference of data between the training sample and the actually measured sample is also an important factor affecting the prediction accuracy, but the actual sample acquired through experiments is difficult to cover the thickness range of the actual used workpiece.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a method for predicting early fatigue cracks and residual life of a metal structure, which can improve the accuracy and reliability of predicting early fatigue crack damage of the structure.
To achieve the above object, one or more embodiments of the present invention provide the following technical solutions:
in a first aspect, a method for predicting early fatigue cracks and residual life of a metal structure is disclosed, comprising:
based on the multidimensional nonlinear ultrasonic response characteristics, establishing a fatigue crack damage actual measurement sample of a typical thickness test piece;
generating a fatigue damage virtual sample with the distribution characteristic height consistent with that of the actual measurement sample, and expanding the data of the actual measurement sample;
calibrating fatigue parameters of the virtual samples, and constructing a fatigue crack damage sample library by combining the fatigue crack damage actual measurement samples;
based on the fatigue crack damage sample library, an intelligent prediction model is established and used for quantitatively evaluating the fatigue crack extension length and the residual service life of the test piece.
As a further technical scheme, based on multidimensional nonlinear ultrasonic response characteristics, a fatigue crack damage actual measurement sample of a typical thickness test piece is established, and the specific steps are as follows:
in the breaking process of a test piece, performing fast Fourier transform on the collected ultrasonic guided wave signal, and extracting signal amplitude values under fundamental frequency, double frequency, triple frequency and/or difference frequency of guided wave excitation frequency;
calculating typical multidimensional nonlinear response characteristics of the fatigue crack based on the signal amplitude of the extracted guided wave excitation frequency;
and constructing a structural fatigue crack damage actual measurement sample based on the typical multidimensional nonlinear response characteristics and the fatigue crack length and the residual life of the test piece.
As a further technical scheme, before collecting the ultrasonic guided wave signal of the test piece, the method further comprises:
fixing the test piece on a fatigue testing machine, and setting fatigue loading parameters;
and (3) exciting and collecting ultrasonic guided wave signals once every set period, and recording the detected fatigue crack extension length until the test piece breaks.
As a further technical scheme, the fatigue damage virtual sample with the distribution characteristics highly consistent with those of the actually measured sample is generated, specifically:
taking a course curve of nonlinear response characteristics of all test pieces in the fatigue damage actual measurement sample along with the fatigue loading period as real data;
taking random noise as input, utilizing a generator to obtain virtual nonlinear response characteristics similar to actual measurement sample distribution, and recording the virtual nonlinear response characteristics as virtual generated sample characteristics;
judging whether the current sample characteristic is false data made by a generator or not by utilizing a discriminator to discriminate the input actually measured sample characteristic and the virtually generated sample characteristic;
the generator and the discriminator are mutually opposed, and through continuous iteration and updating, the final generator generates a virtual generated sample with the height consistent with the distribution of the measured sample.
As a further technical scheme, the virtual sample is subjected to fatigue parameter calibration, and a fatigue crack damage sample library is constructed by combining the fatigue crack damage actual measurement sample, specifically:
taking a multidimensional nonlinear response characteristic curve of an actual measurement sample as an MLSTM network input, taking a 'test piece thickness, crack length and residual life' corresponding to nonlinear characteristics as a model output, and establishing a 'nonlinear response characteristic-fatigue parameter' mapping model by using the MLSTM network;
inputting the nonlinear response characteristic curve in the generated sample into the MLSTM model to obtain fatigue parameters corresponding to the nonlinear characteristics of the generated sample, wherein the method comprises the following steps: test piece thickness, crack length, fatigue life;
and constructing a fatigue crack damage sample library based on the actually measured sample and the generated sample, wherein the thickness and nonlinear response characteristics of the test piece are sample library data characteristics, and the fatigue crack extension length and the residual life of the test piece are sample library data labels.
As a further technical scheme, based on the fatigue crack damage sample library, the intelligent prediction model is established, and specifically comprises the following steps:
and taking the fatigue crack damage sample library as a training sample, taking the data characteristics in the fatigue crack damage sample library as input, taking the data labels as output, and establishing a mapping model between the 'test piece thickness and nonlinear characteristics' and the 'crack extension length and the test piece residual life' by utilizing a BPNN network.
As a further technical scheme, the data characteristics in the fatigue crack damage sample library comprise test piece thickness and nonlinear response characteristics;
the data tag includes the fatigue crack growth length and the remaining life of the test piece.
As a further technical solution, the method further comprises a prediction step: when the structural fatigue damage state is predicted, the thickness of a test piece to be detected and typical nonlinear response characteristics under the current state calculated according to the collected ultrasonic guided waves are input into a trained BPNN prediction model, and the extension length and the residual service life of the fatigue crack under the current state of the structure are obtained.
In a second aspect, a system for predicting early fatigue cracking and residual life of a metal structure is disclosed, comprising:
a fatigue crack damage actual measurement sample setup module configured to: based on the multidimensional nonlinear ultrasonic response characteristics, establishing a fatigue crack damage actual measurement sample of a typical thickness test piece;
a damage measured sample data expansion module configured to: generating a fatigue damage virtual sample with the distribution characteristic height consistent with that of the actual measurement sample, and expanding the data of the actual measurement sample;
a fatigue crack damage sample library building module configured to: calibrating fatigue parameters of the virtual samples, and constructing a fatigue crack damage sample library by combining the fatigue crack damage actual measurement samples;
an intelligent predictive model creation module configured to: based on the fatigue crack damage sample library, an intelligent prediction model is established and used for quantitatively evaluating the fatigue crack extension length and the residual service life of the test piece.
The one or more of the above technical solutions have the following beneficial effects:
the technical scheme of the invention is a data driving prediction method based on an artificial intelligent network, and can improve the accuracy and reliability of early fatigue crack damage prediction of the structure.
According to the technical scheme, the GAN and the MLSTM network are utilized to rapidly obtain any number of virtual generation samples with the characteristic distribution height consistent with that of the actual measurement fatigue damage samples, so that a rich fatigue crack damage sample library is established, and the requirement of a data driving prediction method on mass data driving is met; by utilizing parameters such as multidimensional nonlinear response characteristics, structure thickness, fatigue crack extension length, residual service life and the like under different fatigue loading periods, the fatigue damage prediction model is established based on the BPNN network, the dependence on the input characteristics on time sequence continuity can be reduced, and the real-time prediction precision and efficiency of the fatigue crack damage state can be improved.
Additional aspects of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a schematic illustration of a fatigue crack detection system for a metal structure in accordance with an embodiment of the present invention;
FIG. 2 is a diagram illustrating the quantitative evaluation of fatigue crack growth length and the prediction of residual life of a structure according to an embodiment of the present invention.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention.
Embodiments of the invention and features of the embodiments may be combined with each other without conflict.
The invention provides a general idea:
the method utilizes an artificial intelligent network to establish a data-driven structure early fatigue crack growth length quantitative evaluation and residual service life prediction model. Firstly, a fatigue loading experiment is carried out to obtain multidimensional nonlinear response characteristics such as higher harmonic wave, frequency mixing and the like of a small amount of typical thickness test pieces, and a fatigue damage actual measurement sample is constructed by combining crack extension length and residual life change process. Then, based on a generated countermeasure network (Generative adversarial network, GAN) and a multi-layer long and short memory (Multiple long short-term memory, MLSTM) network, a generated sample which is highly consistent with the characteristic distribution of the actually measured sample is obtained, so that fatigue damage data expansion containing multidimensional nonlinear response characteristics is realized, and a fatigue damage sample database is established. Finally, a fatigue crack growth length quantitative evaluation and residual service life prediction model based on the thickness of the test piece and the multidimensional nonlinear response characteristic is established by combining a counter-propagating neural network (Back propagation neural network, BPNN) and a fatigue damage sample. In practical application, the fatigue crack extension length and the residual service life of the structure in the current state can be obtained by only inputting the thickness of the test piece to be tested and the damage characteristics obtained by measurement into the BPNN prediction model.
Example 1
The embodiment discloses a metal structure fatigue crack detection system, referring to fig. 1, mainly comprising an aluminum alloy test piece, a piezoelectric sensor (Piezoelectric Transducer, PZT), a universal fatigue testing machine, a digital microscope, a nonlinear ultrasonic guided wave detection instrument, an upper computer and the like:
(1) Selecting a 6061-T6 aluminum alloy test piece with typical thickness (2 mm, 4mm and 8 mm) as a research object, and manually prefabricating an initial crack with the length of 5mm at the middle position of the test piece before the fatigue test starts;
(2) Three PZT sensors are attached to each test piece surface, two of which are used as high frequency (frequency w a ) Low frequency (frequency is w b ) The exciter is used for modulating the guided wave in a sine way, and one exciter is used as a guided wave receiver;
(3) The universal fatigue testing machine is responsible for applying cyclic load to the aluminum alloy test piece;
(4) The digital microscope is used for observing the fatigue crack extension length in real time;
(5) The nonlinear ultrasonic guided wave detection instrument is connected with the PZT sensor and is responsible for transmitting guided wave signals;
(6) The upper computer is responsible for generating excitation signals, displaying and storing received signals and analyzing subsequent processing.
Example two
Referring to fig. 2, based on the above metal structure fatigue crack detection system, the embodiment discloses a method for predicting early fatigue crack and residual life of a metal structure, which includes:
step one: based on multidimensional nonlinear ultrasonic response characteristics, establishing a fatigue crack damage actual measurement sample of a typical thickness aluminum alloy test piece:
(1) Fixing a test piece on a universal fatigue testing machine, and setting fatigue loading parameters such as a fatigue loading period (Hz), a waveform, a stress ratio, a maximum stress and the like;
(2) Every 2×10 4 Performing excitation and collection of ultrasonic guided wave signals for one time, and recording the fatigue crack extension length detected by a digital microscope until a test piece breaks;
(3) Performing a fast fourier transform (Fast Fourier transform, FFT) on the acquired ultrasonic guided wave signal and extracting the fundamental frequency (w a And w is equal to b ) Frequency doubling (2 w) a And 2w b ) Frequency tripled (3 w) a And 3w b ) Sum and difference frequency (w) a+b And w is equal to a-b ) The following signal amplitudes are respectively recorded as:and (3) with
Wherein the sum frequency is the sum of two fundamental frequencies, i.e. w a+b The method comprises the steps of carrying out a first treatment on the surface of the The difference frequency is the difference between the two fundamental frequencies, i.e. w a-b 。
(4) Calculating typical multidimensional nonlinear response characteristics of fatigue cracks: and->
Numerous studies have shown that these six non-linear parameters change significantly as the crack length changes, so the six non-linear characteristics described above are used to characterize the crack length. The advantages are that: the nonlinear response parameter is strongly related to the crack propagation length, so that the model built by the nonlinear response parameter has high prediction accuracy.
(5) Multidimensional nonlinear ultrasonic response characteristics calculated under different stress ratios based on typical thickness test pieceAnd measuring parameters such as the obtained fatigue crack extension length, the residual life of the test piece, the typical thickness test piece and the like to construct a structural fatigue crack damage actual measurement sample.
Wherein, the residual life of the test piece is the fatigue loading period which can be born by the test piece-the current loading period; the fatigue loading period bearable by the test piece is the total loading period from the moment of breaking the test piece, and the test piece can be determined when the test piece breaks.
The data in the sample mainly comprises two major parts, wherein one part is a damage characteristic and consists of the thickness of a test piece and nonlinear ultrasonic response characteristics (comprising six parameters); part is a damaged tag, consisting of crack length and remaining life. The damage characteristic is input of the model, and the damage label is output of the model. There is a one-to-one relationship between input and output. Each row in the sample table has a thickness, a characteristic, a length, and a lifetime, and different rows represent different stages.
Step two: by utilizing strong characteristic learning and generating capability of the GAN network, the fatigue damage virtual sample which is highly consistent with the measured damage sample multidimensional nonlinear ultrasonic response characteristic is generated by continuously resisting and iterating through the generator and the identifier, and the damage sample data expansion is realized:
(1) Nonlinear response characteristics of all test pieces in fatigue damage actual measurement sample And->Load with fatigueThe periodic variation course curve is used as real data, namely the actual measurement sample in fig. 2, for training a GAN discriminator, training an MLSTM model and training a BPNN model;
(2) Using random noise as input generator, using GAN network generator to obtain virtual nonlinear response characteristic similar to actual measurement sample distributionAnd->Recording as virtual generated sample characteristics;
(3) Judging nonlinear response characteristics and generated sample characteristics of all test pieces in an input actual measurement sample by utilizing a discriminator of the GAN network, and judging whether the current sample characteristics are false data made by a generator or not;
(4) The generator and the discriminator are mutually opposed, through continuous iteration and updating, the generator of the final GAN network can generate virtual generated sample characteristics which are highly consistent with the distribution of the actually measured sample characteristics, and each sample data comprisesAnd->And typical nonlinear response history curves.
Step three: based on the high-precision recognition capability of the MLSTM network on time series data characteristics, fatigue parameter calibration is carried out on virtual generated sample characteristics, and then an abundant fatigue crack damage sample library is constructed by combining the fatigue crack damage actual measurement samples:
(1) Multidimensional nonlinear response characteristics of measured samplesAnd->Curve asThe method comprises the steps of inputting an MLSTM network, outputting a model of 'test piece thickness, crack length and residual life' corresponding to nonlinear characteristics, and establishing a 'nonlinear response characteristic-fatigue parameter' mapping model by using the MLSTM network;
the mapping model has a fixed structure, parameters are only required to be initially set, nonlinear corresponding characteristics in an actual measurement sample are given to the input end of the model, and the output end is the thickness of a test piece, the crack length and the residual life, wherein the crack length and the residual life can be collectively called as fatigue parameters, various parameters of the model can be obtained, and the parameters and the structure of the model correspond to one model.
(2) Nonlinear response characteristics in the sample will be generatedAnd (3) withInputting the curve into an MLSTM model, namely a mapping model, and obtaining fatigue parameters such as test piece thickness, crack length, fatigue life and the like corresponding to nonlinear characteristics of a generated sample;
in the fatigue loading process of the test piece, each moment has a group of nonlinear corresponding characteristics (comprising six values), then in the whole loading process, N groups of nonlinear response characteristics (comprising N multiplied by 6 values) exist, N values of the same parameter are converted into a curve according to time history, and six curves, namely nonlinear response characteristic curves, can be obtained by the six parameters.
(3) Constructing a fatigue crack damage sample library based on the measured sample and the generated sample, wherein the thickness of the test piece and the nonlinear response characteristicsAnd->And the fatigue crack extension length and the residual life of the test piece are sample library data labels.
The thickness and nonlinear response characteristics of the test piece are input values of the BPNN model in fig. 2, the fatigue crack extension length and the residual life of the test piece are output values of the BPNN model, and finally the function of the whole patent is to predict the fatigue crack extension length and the residual life of the test piece based on two known parameters of the thickness and nonlinear response characteristics of the test piece.
Step four: utilizing excellent self-learning and generalization capability of a BPNN network, establishing an intelligent prediction model based on quantitative assessment of the fatigue crack growth length and residual service life of the test piece thickness and multidimensional nonlinear ultrasonic response characteristics:
(1) The fatigue crack damage sample library is used as a training sample, and the data characteristics (test piece thickness, nonlinear response characteristicsAnd->) The method comprises the steps of (1) taking a data tag (fatigue crack extension length and test piece residual life) as output, and establishing a mapping model between a 'test piece thickness and nonlinear characteristics' and a 'crack extension length and test piece residual life' by utilizing a BPNN network;
the mapping model is composed of an input layer, an hidden layer and an output layer. The input layer comprises seven neurons, and the seven neurons correspond to the thickness of the test piece and six nonlinear response characteristics respectively. The hidden layer contains 100 neurons. The output layer contains 2 neurons, which correspond to crack propagation length and test piece residual life, respectively. Model loss function e=quadratic root number (|crack prediction length and crack actual length| 2 Predictive value of residual life +| -actual value of residual life | 2 ). The model training goal is to get the loss function E to meet<Model parameters of 0.001 condition, connection weight and threshold between input layer and hidden layer, connection weight and threshold between hidden layer and output layer.
(2) When the structural fatigue damage state is predicted, the thickness of a test piece to be detected and typical nonlinear response characteristics under the current state calculated according to the collected ultrasonic guided waves are input into a trained BPNN prediction model, namely a mapping model, so that the extension length and the residual service life of the fatigue crack under the current state of the structure can be obtained.
Aiming at the problems of quantitative evaluation of fatigue crack propagation length and prediction of residual life, the technical scheme of the invention provides an ultrasonic detection method based on multidimensional nonlinear characteristics, and the reliability and accuracy of crack damage detection are improved by fusing higher harmonic wave and mixed wave response of guided wave signals;
according to the technical scheme, by utilizing the strong characteristics and learning capability of the GAN network, any number of virtual generated samples with the characteristic distribution height consistent with that of the actual measurement samples of the fatigue damage of the test piece with typical thickness can be obtained quickly at low cost, and the expansion of the fatigue crack damage sample data under different states is realized;
according to the technical scheme, the strong interpretation capability of the MLSTM network on time sequence characteristics is utilized, a nonlinear characteristic-fatigue parameter mapping model is established by combining the fatigue damage actual measurement sample, the fatigue parameters corresponding to the nonlinear response characteristic curve in the generated sample can be calibrated with high precision, and then a rich fatigue crack damage sample library is established by combining the actual measurement sample characteristics and the fatigue parameters;
according to the technical scheme, abundant fatigue crack damage sample data are utilized, a crack length and life prediction model is established based on the BPNN network, dependence of a prediction result on a characteristic time sequence can be reduced, and crack extension length assessment and residual life prediction can be realized only according to nonlinear damage characteristics of a test piece with a specific thickness at the current moment.
Example two
It is an object of the present embodiment to provide a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, which processor implements the steps of the above method when executing the program.
Example III
An object of the present embodiment is to provide a computer-readable storage medium.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the above method.
Example IV
It is an object of the present embodiment to provide a metal structure early fatigue crack and residual life prediction system, comprising:
a fatigue crack damage actual measurement sample setup module configured to: based on the multidimensional nonlinear ultrasonic response characteristics, establishing a fatigue crack damage actual measurement sample of a typical thickness test piece;
a damage measured sample data expansion module configured to: generating a fatigue damage virtual sample with the distribution characteristic height consistent with that of the actual measurement sample, and expanding the data of the actual measurement sample;
a fatigue crack damage sample library building module configured to: calibrating fatigue parameters of the virtual samples, and constructing a fatigue crack damage sample library by combining the fatigue crack damage actual measurement samples;
an intelligent predictive model creation module configured to: based on the fatigue crack damage sample library, an intelligent prediction model is established and used for quantitatively evaluating the fatigue crack extension length and the residual service life of the test piece.
The steps involved in the devices of the second, third and fourth embodiments correspond to those of the first embodiment of the method, and the detailed description of the embodiments can be found in the related description section of the first embodiment. The term "computer-readable storage medium" should be taken to include a single medium or multiple media including one or more sets of instructions; it should also be understood to include any medium capable of storing, encoding or carrying a set of instructions for execution by a processor and that cause the processor to perform any one of the methods of the present invention.
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented by general-purpose computer means, alternatively they may be implemented by program code executable by computing means, whereby they may be stored in storage means for execution by computing means, or they may be made into individual integrated circuit modules separately, or a plurality of modules or steps in them may be made into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.

Claims (10)

1. The method for predicting the early fatigue crack and the residual life of the metal structure is characterized by comprising the following steps:
based on the multidimensional nonlinear ultrasonic response characteristics, establishing a fatigue crack damage actual measurement sample of a typical thickness test piece;
generating a fatigue damage virtual sample with the distribution characteristic height consistent with that of the actual measurement sample, and expanding the data of the actual measurement sample;
calibrating fatigue parameters of the virtual samples, and constructing a fatigue crack damage sample library by combining the fatigue crack damage actual measurement samples;
based on the fatigue crack damage sample library, an intelligent prediction model is established and used for quantitatively evaluating the fatigue crack extension length and the residual service life of the test piece.
2. The method for predicting early fatigue crack and residual life of a metal structure as claimed in claim 1, wherein the method for establishing the actual measurement sample of fatigue crack damage of a test piece with typical thickness based on multidimensional nonlinear ultrasonic response characteristics comprises the following specific steps:
in the breaking process of a test piece, performing fast Fourier transform on the collected ultrasonic guided wave signal, and extracting signal amplitude values under fundamental frequency, double frequency, triple frequency and/or difference frequency of guided wave excitation frequency;
calculating typical multidimensional nonlinear response characteristics of the fatigue crack based on the signal amplitude of the extracted guided wave excitation frequency;
and constructing a structural fatigue crack damage actual measurement sample based on the typical multidimensional nonlinear response characteristics and the fatigue crack length and the residual life of the test piece.
3. The method for predicting early fatigue crack and residual life of a metal structure of claim 1, further comprising, prior to collecting ultrasonic guided wave signals of the test piece:
fixing the test piece on a fatigue testing machine, and setting fatigue loading parameters;
and (3) exciting and collecting ultrasonic guided wave signals once every set period, and recording the detected fatigue crack extension length until the test piece breaks.
4. The method for predicting early fatigue crack and residual life of a metal structure as set forth in claim 1, wherein the generating of the virtual sample of fatigue damage highly consistent with the distribution characteristics of the measured sample is specifically:
taking a course curve of nonlinear response characteristics of all test pieces in the fatigue damage actual measurement sample along with the fatigue loading period as real data;
taking random noise as input, utilizing a generator to obtain virtual nonlinear response characteristics similar to actual measurement sample distribution, and recording the virtual nonlinear response characteristics as virtual generated sample characteristics;
judging whether the current sample characteristic is false data made by a generator or not by utilizing a discriminator to discriminate the input actually measured sample characteristic and the virtually generated sample characteristic;
the generator and the discriminator are mutually opposed, and through continuous iteration and updating, the final generator generates a virtual generated sample with the height consistent with the distribution of the measured sample.
5. The method for predicting early fatigue cracks and residual life of a metal structure according to claim 1, wherein the virtual sample is subjected to fatigue parameter calibration, and a fatigue crack damage sample library is constructed by combining a fatigue crack damage actual measurement sample, specifically:
taking a multidimensional nonlinear response characteristic curve of an actual measurement sample as an MLSTM network input, taking a 'test piece thickness, crack length and residual life' corresponding to nonlinear characteristics as a model output, and establishing a 'nonlinear response characteristic-fatigue parameter' mapping model by using the MLSTM network;
inputting the nonlinear response characteristic curve in the generated sample into the MLSTM model to obtain fatigue parameters corresponding to the nonlinear characteristics of the generated sample, wherein the method comprises the following steps: test piece thickness, crack length, fatigue life;
and constructing a fatigue crack damage sample library based on the actually measured sample and the generated sample, wherein the thickness and nonlinear response characteristics of the test piece are sample library data characteristics, and the fatigue crack extension length and the residual life of the test piece are sample library data labels.
6. The method for predicting early fatigue cracks and residual life of a metal structure according to claim 1, wherein the method for establishing an intelligent prediction model based on a fatigue crack damage sample library comprises the following steps:
and taking the fatigue crack damage sample library as a training sample, taking the data characteristics in the fatigue crack damage sample library as input, taking the data labels as output, and establishing a mapping model between the 'test piece thickness and nonlinear characteristics' and the 'crack extension length and the test piece residual life' by utilizing a BPNN network.
7. The method for predicting early fatigue crack and life remaining in a metal structure as in claim 1, further comprising the step of predicting: when the structural fatigue damage state is predicted, the thickness of a test piece to be detected and typical nonlinear response characteristics under the current state calculated according to the collected ultrasonic guided waves are input into a trained BPNN prediction model, and the extension length and the residual service life of the fatigue crack under the current state of the structure are obtained.
8. A metal structure early fatigue crack and residual life prediction system, comprising:
a fatigue crack damage actual measurement sample setup module configured to: based on the multidimensional nonlinear ultrasonic response characteristics, establishing a fatigue crack damage actual measurement sample of a typical thickness test piece;
a damage measured sample data expansion module configured to: generating a fatigue damage virtual sample with the distribution characteristic height consistent with that of the actual measurement sample, and expanding the data of the actual measurement sample;
a fatigue crack damage sample library building module configured to: calibrating fatigue parameters of the virtual samples, and constructing a fatigue crack damage sample library by combining the fatigue crack damage actual measurement samples;
an intelligent predictive model creation module configured to: based on the fatigue crack damage sample library, an intelligent prediction model is established and used for quantitatively evaluating the fatigue crack extension length and the residual service life of the test piece.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method of any of the preceding claims 1-7 when the program is executed by the processor.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, performs the steps of the method of any of the preceding claims 1-7.
CN202310483173.2A 2023-04-24 2023-04-24 Method and system for predicting early fatigue crack and residual life of metal structure Pending CN116522089A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310483173.2A CN116522089A (en) 2023-04-24 2023-04-24 Method and system for predicting early fatigue crack and residual life of metal structure

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310483173.2A CN116522089A (en) 2023-04-24 2023-04-24 Method and system for predicting early fatigue crack and residual life of metal structure

Publications (1)

Publication Number Publication Date
CN116522089A true CN116522089A (en) 2023-08-01

Family

ID=87395527

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310483173.2A Pending CN116522089A (en) 2023-04-24 2023-04-24 Method and system for predicting early fatigue crack and residual life of metal structure

Country Status (1)

Country Link
CN (1) CN116522089A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117761160A (en) * 2023-12-22 2024-03-26 北京航力安太科技有限责任公司 Nondestructive testing system based on ultrasonic guided waves

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117761160A (en) * 2023-12-22 2024-03-26 北京航力安太科技有限责任公司 Nondestructive testing system based on ultrasonic guided waves

Similar Documents

Publication Publication Date Title
Karve et al. Digital twin approach for damage-tolerant mission planning under uncertainty
Mousavi et al. Structural damage detection method based on the complete ensemble empirical mode decomposition with adaptive noise: A model steel truss bridge case study
Yanez-Borjas et al. Statistical time features for global corrosion assessment in a truss bridge from vibration signals
CN101706355A (en) NExT/ARMA-based structural response analysis method
KR102241879B1 (en) Duct Void Detection System and Learning Apparatus of Duct Void Detection Model Using Artificial Intelligence Learning Model
CN105067703A (en) Plate-like metal member/material remaining life non-destructive detection and evaluation method
CN110990978B (en) Bolt state monitoring method and device
CN116522089A (en) Method and system for predicting early fatigue crack and residual life of metal structure
CN114357594A (en) Bridge abnormity monitoring method, system, equipment and storage medium based on SCA-GRU
Paulraj et al. Structural steel plate damage detection using non destructive testing, frame energy based statistical features and artificial neural networks
CN109409271A (en) Testing of Ferromagnetic Material Hardness prediction algorithm based on BP neural network innovatory algorithm
Vazirizade et al. Online nonlinear structural damage detection using Hilbert Huang transform and artificial neural networks
Li et al. Magnetic barkhausen noise technique for early-stage fatigue prediction in martensitic stainless-steel samples
Tefera et al. Challenges in applying vibration-based damage detection to highway bridge structures
Pham-Bao et al. A novel approach to investigate the mechanical properties of the material for bridge health monitoring using convolutional neural network
Fu et al. Automatic bolt tightness detection using acoustic emission and deep learning
Dackermann Vibration-based damage identification methods for civil engineering structures using artificial neural networks
Kanwar et al. Monitoring of RCC structures affected by earthquakes
Han et al. Multiscale convolutional generative adversarial network for anchorage grout defect detection
Liu et al. Combined two-level guided wave structural health monitoring strategy using multifeature integration and machine learning: application to early-age grouted sleeves
Hakim et al. A sensitivity analysis on the damage identification capability of artificial neural networks
Sofge Structural health monitoring using neural network based vibrational system identification
Vazirizade et al. Structural damage detection using ensemble empirical mode decomposition, hilbert transform and artificial neural networks
Liao et al. Long Short-Term Memory Neural Networks for Predicting Dynamic Response of Structures of High Complexities
Cianetti et al. How to experimentally monitor the fatigue behaviour of vibrating mechanical systems

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