SiC-cmc material fatigue failure early warning system based on staf and r-ct combination

Abstract

The application discloses a SiC-CMC material fatigue failure early warning system based on STAF and R-CT combination, a STAF model carries out time domain signal difference reference combined with multistage filtering processing on received digital acoustic emission signals, and outputs effective damage information under a damage state; in the R-CT model, information is first processed by using R-type factor analysis method, the relationship between data quantity and factors in acoustic emission characteristic parameters is obtained, and thus comprehensive factor scores are obtained; based on and combined with dovetail mutation theory, characteristic information of transition points of damage states in the SiC-CMC material fatigue failure process is obtained. The application adopts time domain reference-multistage filtering technology, removes noise signals by comparing signal parameter differences of different time domain intervals, and further carries out multistage filtering on waveforms to remove noise signals in the waveforms. The comprehensive factor scores are obtained by carrying out factor analysis on ringing counts, energy and absolute energy of damage signals to carry out dovetail mutation calculation, and thus the life prediction and early warning of SiC-CMC structural fatigue damage are realized.

Description

Technical Field

[0001] This invention relates to fatigue failure analysis and early warning processing of continuous silicon carbide fiber reinforced silicon carbide matrix composites (SiC-CMC), and more specifically, to a fatigue failure early warning system designed for SiC-CMC materials that combines the STAF model and the R-CT model.

[0002] The STAF model refers to a filtering method that uses time-domain signal difference reference to filter the acquired digital acoustic emission signals of SiC-CMC materials.

[0003] The R-CT model refers to a method that first uses R-type factor analysis to obtain a comprehensive factor score, and then uses the swallowtail catastrophe theory to obtain the transition point of the damage state of SiC-CMC materials. Background Technology

[0004] Continuous silicon carbide fiber-reinforced silicon carbide matrix composites (SiC-CMC), as a ceramic material with excellent properties such as high strength, high hardness, and high temperature resistance, maintain good structural and performance stability under extreme temperatures and harsh environments. This makes them ideal materials for hot-end components of aerospace engines and turbine gas turbines, including tail nozzles, combustion chambers, turbine outer rings, and guide vanes. The application of ultra-high temperature composite materials in the aerospace field is referenced in *Science & Technology in China*, Vol. 8, 2004, pp. 48-49, by Qiu Haipeng, Beijing Aeronautical Manufacturing Engineering Research Institute. During service, SiC-CMC materials are often subjected to cyclic loading, leading to continuous internal damage accumulation and ultimately catastrophic fracture. Therefore, how to better identify fatigue failure and predict the lifespan of SiC-CMC materials is a pressing technical problem that needs to be solved.

[0005] Acoustic emission (AE) technology, as a non-destructive testing technique, offers advantages such as real-time online detection, full-process visualization, and integrated damage monitoring, and is widely used in damage detection of structural components in engineering fields. Experimental studies show that when materials are subjected to cyclic loading, deformation damage occurs, and the damaged area acts as an acoustic emission source, generating acoustic emission signals. Furthermore, the waveform and characteristic parameters of these acoustic emission signals change continuously with the damage state. Acoustic emission characteristic parameters include amplitude, duration, rise time, count, energy, and absolute energy, among others.

[0006] An early warning mechanism refers to a system that issues warnings in advance. This system, comprised of institutions, regulations, networks, and measures, provides timely alerts, enabling proactive information feedback and laying the foundation for timely deployment and risk prevention. Currently, high-temperature structural components of aero-engines are generally processed using SiC-CMC materials. Identifying the damage state or damage mechanism of SiC-CMC materials is beneficial for testing the performance of high-temperature structural components in aero-engines. Summary of the Invention

[0007] In this invention, SiC-CMC materials are monitored using an acoustic emission instrument during service. The sampled information is converted into digital acoustic emission signals after A / D conversion. Since the changes in acoustic emission characteristic parameters (amplitude, duration, rise time, count, energy, absolute energy, etc.) contained in the digital acoustic emission signals are related to fatigue damage in the time domain, the application of acoustic emission technology provides source data for damage assessment, safety reliability, and life prediction of SiC-CMC materials used in high-temperature structural components of aero-engines during service, enabling safe service and failure early warning.

[0008] To address the service safety issues of SiC-CMC materials used in high-temperature structural components of aero-engines under cyclic loading during service, this invention proposes an early warning system for fatigue failure of SiC-CMC materials in service, based on a combination of the STAF model and the R-CT model. This fatigue failure early warning system monitors and provides early warning of the fatigue failure process of SiC-CMC materials during service. Specifically: Firstly, the STAF model performs time-domain signal difference comparison combined with multi-level filtering on the received digital acoustic emission signal to output effective damage information under the damaged state. Secondly, in the R-CT model, firstly... The information was processed using R-factor analysis to obtain the relationship between the amount of data and factors in the acoustic emission characteristic parameters, thereby obtaining the comprehensive factor score. Thirdly, based on Furthermore, by combining the swallowtail catastrophe theory, the characteristics of the damage state transition point during the fatigue failure process of SiC-CMC materials were obtained, namely, the damage-transition point information. Fourthly, the early warning information visualization unit is used to demonstrate the fatigue failure state of SiC-CMC materials in service in real time. The monitoring results output by the fatigue failure early warning system of this invention can enable fatigue life prediction and failure early warning for high-temperature structural components of aero-engines.

[0009] The advantages of using the fatigue failure early warning system of this invention for fatigue life prediction and early warning of SiC-CMC materials used in high-temperature structural components of aero engines are as follows:

[0010] (1) The present invention uses the STAF model to extract the damage signal by time domain and multi-level filtering. The obtained fatigue acoustic emission characteristic parameter information can reflect the actual characterization of fatigue damage of SiC-CMC material in service, so as to comprehensively and accurately assess the life of SiC-CMC material and provide early warning reminders for failure.

[0011] (2) The R-CT model is used in this invention to quickly identify the damage transition time point of SiC-CMC material and to accurately evaluate the fatigue life of high-temperature structural components of aero-engines under load.

[0012] (3) In the R-CT model designed in this invention, fatigue damage early warning is carried out simultaneously through R-type factor analysis, swallowtail mutation and acoustic emission signal change law, providing two early warning signals for the damage study of SiC-CMC materials. Attached Figure Description

[0013] Figure 1 This is a structural block diagram of the SiC-CMC material fatigue failure early warning system based on the combination of STAF and R-CT according to the present invention.

[0014] Figure 2 This is a diagram showing the change of acoustic emission energy over time, as well as the division of time regions and levels according to the present invention.

[0015] Figure 3 This is a schematic diagram illustrating the calculation of the fit degree in this invention.

[0016] Figure 4 This is a flowchart of the STAT model processing of the present invention.

[0017] Figure 5 This is a flowchart of the R-CT model processing of the present invention.

[0018] Figure 6 This is a graph showing the cumulative comprehensive factor score and mutation amount of this invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0020] See Figure 1 As shown, the present invention discloses a fatigue failure early warning system for SiC-CMC materials based on the combination of STAF and R-CT, comprising a STAF model and an R-CT model. The STAF model and the R-CT model are developed using Matlab language (version 2014 b) and embedded in the memory of the acoustic emission instrument.

[0021] Acoustic emission sensors are used to collect acoustic emission signals from high-temperature structural components of aero engines and then output them to the STAF model.

[0022] The early warning information visualization unit is used to demonstrate the fatigue failure state of SiC-CMC materials in service in real time. The early warning information visualization unit can be a computer system pre-installed with operating software.

[0023] SiC-CMC materials, as ideal materials for high-temperature structural components of aero-engines, are subject to thermo-mechanical field coupling during service, leading to continuous internal damage accumulation and even catastrophic accidents. Therefore, to monitor and provide early warning analysis of the safe service process of high-temperature structural components of aero-engines, this invention utilizes acoustic emission technology to study the life prediction and failure warning of high-temperature structural components in fatigue environments. Due to its high sensitivity, acoustic emission technology is susceptible to noise interference during detection. To address this, a time-domain reference-multi-stage filtering technique (STAF model) is employed. The STAF model eliminates noise signals by comparing the differences in signal parameters across different time domain intervals, and further performs multi-stage filtering on the waveform to identify noise signals. In the R-CT model, factor analysis is performed on the ring count, energy, and absolute energy of the damage signal to obtain a comprehensive factor score for calculating the swallowtail mutation, thereby achieving life prediction and early warning of fatigue damage in high-temperature structural components of aero-engines.

[0024] In this invention, the acoustic emission instrument uses a SAMOS-type detection system from Physical Acoustics, Inc., with a PCI-8 acoustic emission function card. The acoustic emission parameters are as follows: filter threshold: 35dB; sampling rate: 2MSPS; PDT: 300μs, HDT: 600μs, HLT: 1000μs; acoustic emission probe: Nano30 piezoelectric sensor, resonant frequency 140kHz, bandpass filter 150-400kHz. The test sample is a high-temperature structural component of an aero-engine, and the material system of the high-temperature structural component is SiC-CMC. More specifically, the test sample is a tail nozzle of a certain type of aero-engine, and the composite material system of the tail nozzle contains silicon carbide fiber-toughened silicon carbide (SiC). f / SiC).

[0025] Digital acoustic transmission information

[0026] In this invention, each acoustic emission sensor collects sampling information during the sampling time, which is then converted into digital acoustic emission information via A / D conversion. Generally, the digital acoustic emission information includes amplitude... (Unit: dB), Duration (Unit: μs), rise time (Unit: μs), Peak frequency (Unit: kHz), Energy (unit: Absolute energy (Unit: aJ) and ring count (Unit: dimensionless constant).

[0027] The digital acoustic emission information collected by the first acoustic emission sensor during the sampling time is denoted as . ,and .

[0028] The digital acoustic emission information collected by the second acoustic emission sensor during the sampling time is denoted as... ,and .

[0029] Among them, the The digital acoustic emission information collected by each acoustic emission sensor during the sampling time is denoted as . ,and subscript This indicates the identification number of the acoustic emission sensor.

[0030] For the first Road digital sound transmission information The amplitude.

[0031] For the first Road digital sound transmission information The duration.

[0032] For the first Road digital sound transmission information The rise time.

[0033] For the first Road sound transmission information The peak frequency.

[0034] For the first Road digital sound transmission information Energy.

[0035] For the first Road digital sound transmission information Absolute energy.

[0036] For the first Road digital sound transmission information The ringing count.

[0037] The digital acoustic emission information collected by the last acoustic emission sensor during the sampling time is denoted as... ,and subscript This represents the total number of acoustic emission sensors distributed on the test sample.

[0038] In this invention, the acoustic emission information collected by the multi-channel acoustic emission sensors distributed on the test sample during the sampling time forms a digital acoustic emission information set, denoted as... ,and .

[0039] For ease of explanation, the first Road digital sound transmission information Also known as arbitrary digital acoustic transmission information To simplify the above The same steps are used to process the 7 elements in the data. Instead, there is acoustic emission information. .

[0040] A hierarchical partitioning method for digital acoustic emission information (i.e., the CJJ method).

[0041] During the information acquisition process, the acoustic emission sensor contains noise due to environmental factors. To eliminate the interference error in the acoustic emission waveform signal caused by noise, this invention employs a hierarchical partitioning method to perform different levels of noise reduction processing on each element of the digital acoustic emission information.

[0042] In this invention, the hierarchy identifier is denoted as... The number of levels is denoted as and the smallest The value is assigned to 1.

[0043] In this invention, any channel of digital acoustic transmission information If we label the elements in the hierarchy, then we have The hierarchical range value is , The hierarchical range value is , The hierarchical range value is , The hierarchical range value is , The hierarchical range value is , The hierarchical range value is , The hierarchical range value is .

[0044] For example, the initial set hierarchical interval values ​​are as follows: =70、 =9999 =2113、 =350, =65535、 =1608000000 (GAICHENG 10 9 )aJ、 =23534

[0045] In this invention, the initial hierarchical interval value Afterwards, The reduction is called the reduced hierarchical interval value, denoted as . Repeat execution The reduction, until arrive The interval value reduction ends when the minimum value of any element in the interval is reached.

[0046] In this invention, the hierarchical digital acoustic emission information is denoted as... .

[0047] A method for dividing digital acoustic emission information into time periods (i.e., the STT method).

[0048] See Figure 2 As shown, in one sampling period of this invention, the acoustic emission information of the test sample collected by the acoustic emission instrument sensor is recorded at the sampling start time point as... The sampling end time is denoted as . .from to Divided into four time periods, each designated as the first time period. The second time period The third time period The fourth time period .

[0049] In the process of predicting and warning of fatigue damage in test samples using acoustic emission technology, based on a sampling period... The energy accumulation is used to determine the end point of each time period. Start to accumulated energy ( The point in time when the trend becomes flat is denoted as . ;from Start to accumulated energy ( The point in time when the number of cases increases rapidly is denoted as . ;from Start to accumulated energy ( The point in time when the trend becomes flat is denoted as . .

[0050] In this invention, from to The four time periods are as follows:

[0051] First time period It refers to from arrive For a period of time.

[0052] Second time period It refers to from arrive For a period of time.

[0053] The third time period It refers to from arrive For a period of time.

[0054] The fourth time period It refers to from arrive For a period of time.

[0055] Goodness of fit calculation

[0056] In this invention, in order to determine whether the noise reduction threshold is appropriate, the appropriate threshold is selected. Filtering is performed, and the energy in the filtered noise signal is accumulated to obtain the noise-accumulated energy. Then, a linear... The noise-accumulated energy is fitted, and the linear fit value of the digital acoustic emission information representing the energy is obtained, denoted as R. See [reference needed]. Figure 3 As shown.

[0057] During the acquisition of fatigue damage signals from in-service SiC-CMC samples using acoustic emission instruments, the equipment is affected by environmental factors such as ambient noise, aircraft engine operating noise, and electromagnetic noise. To obtain more accurate digital acoustic emission information to reflect SiC-CMC sample damage and reduce data processing time, this invention requires... Therefore, a highly efficient damage data extraction method based on time-domain interval signal difference reference was first designed to perform filtering processing. Furthermore, a multi-stage filtering method, including wavelet threshold denoising, modal empirical decomposition filtering, and other techniques, was combined to obtain acoustic emission data reflecting the damage state of SiC-CMC samples during service, i.e., effective damage information. The It can effectively predict and analyze the fatigue damage life of SiC-CMC samples during service.

[0058] In this invention, In The damage is related to crack initiation and propagation, and is also directly proportional to the degree of damage to the test sample. To better determine the degree of damage to the test sample during fatigue damage, R-type factor analysis is used to obtain the comprehensive factor score of the test sample. This invention combines R-type factor analysis with the swallowtail catastrophe theory model to identify the transition point of the damage state during the fatigue failure process of the test sample, providing early warning of failure and ensuring its safety and reliability during service.

[0059] In this invention, a distributed approach is first used to process the acoustic emission waveform signal of one channel, and then a centralized approach is used to evaluate the damage status.

[0060] STAF model

[0061] In this invention, the STAF model is... The digital acoustic emission information is processed in a step-by-step manner, processing each channel individually. However, for... All elements in the model employ the same multi-stage filtering technique, and the STAF model processes digital acoustic emission information. .

[0062] See Figure 4 As shown, the STAF model of the present invention includes the following processing steps:

[0063] Step A: Acquisition of time-domain acoustic emission information at different time periods;

[0064] Step A1: Dividing acoustic emission information into four time periods;

[0065] In this invention, the STT method is used to analyze acoustic emission information. The time period is divided into four time periods.

[0066] For example, the sampling period is 500 seconds.

[0067] Step A2, time domain division;

[0068] See Figure 2 As shown, the present invention covers four time periods ( , , , In the time domain, 10 seconds are selected as the time for extracting acoustic emission information.

[0069] First time domain Time period From the start time of the test The first 10 seconds of counting.

[0070] Second time domain Time period Any 10-second interval.

[0071] Third time domain Time period Any 10-second interval.

[0072] 4th time domain Time period Any 10-second interval.

[0073] In this invention, acoustic emission information in the time domain (10 seconds) is used as the filtering information source, which fully considers the application of acoustic emission instruments on SiC-CMC materials.

[0074] Step A3: Acquisition of acoustic emission information in the time domain;

[0075] Acoustic emission information The information in each of the four time periods is represented as follows:

[0076] The acoustic emission information for the first time period is denoted as ; in the Extracting the time domain The acoustic emission information below is denoted as time-domain acoustic emission information. ;

[0077] The acoustic emission information for the second time period is denoted as ; in the Extracting the time domain The acoustic emission information below is denoted as time-domain acoustic emission information. ;

[0078] The acoustic emission information for the third time period is denoted as ; in the Extracting the time domain The acoustic emission information below is denoted as time-domain acoustic emission information. ;

[0079] The acoustic emission information for the fourth time period is denoted as ; in the Extracting the time domain The acoustic emission information below is denoted as time-domain acoustic emission information. .

[0080] Step B, hierarchical division of acoustic emission information;

[0081] To eliminate interference errors in the acoustic emission waveform signal caused by noise, this invention employs hierarchical division of acoustic emission information. Different levels of noise reduction are performed.

[0082] Step B1: Set the initial level range values;

[0083] Initially, the hierarchical interval values ​​are denoted as .

[0084] Step B2: Obtain acoustic emission information within the hierarchical interval;

[0085] In this invention, the CJJ method is used from The acoustic emission information under the hierarchical intervals is extracted and denoted as the hierarchical noise reduction information set in the first time period. .

[0086] In this invention, the CJJ method is used from The acoustic emission information under the hierarchical intervals is extracted and denoted as the hierarchical noise reduction information set in the second time period. .

[0087] In this invention, the CJJ method is used from The acoustic emission information under the hierarchical intervals is extracted and denoted as the hierarchical noise reduction information set in the third time period. .

[0088] In this invention, the CJJ method is used from The acoustic emission information under the hierarchical intervals is extracted and denoted as the hierarchical noise reduction information set in the fourth time period. .

[0089] For example, time-domain energy information Energy - Noise Reduction Threshold The distribution is as follows Figure 2 As shown, due to the discontinuity of acoustic emission waveform information during monitoring, this invention employs different level intervals ( ) to improve early warning accuracy. , , , , , , By extracting acoustic emission waveform information at different time periods, various noise reduction thresholds for acoustic emission characteristic parameters are obtained.

[0090] In this invention, the hierarchical noise reduction of the amplitude after steps A to B are respectively , , and .

[0091] In this invention, the hierarchical noise reduction for the duration after steps A to B are respectively , , and .

[0092] In this invention, the hierarchical noise reduction of the rise time after steps A to B is respectively... , , and .

[0093] In this invention, the hierarchical noise reduction of the peak frequency after steps A to B are respectively , , and .

[0094] In this invention, the hierarchical noise reduction of energy after steps A to B are respectively , , and .

[0095] In this invention, the hierarchical noise reduction of absolute energy after steps A to B are respectively , , and .

[0096] In this invention, the hierarchical noise reduction of ring count after steps A to B are respectively , , and .

[0097] Step C: Calculate the fitting degree of the hierarchical acoustic emission information;

[0098] See Figure 3 As shown, application Linear relationship for accumulation Filtered energy The fitting process is performed to determine the hierarchical interval division values ​​output from step B.

[0099] Step C1: Merge first, then submit;

[0100] First Perform union processing to obtain hierarchical acoustic emission information – union information. Then Intersection processing is performed to obtain hierarchical acoustic emission information – intersection information. .

[0101] In this invention, after processing in step C1, the amplitude-level acoustic emission information-intersection information is... Duration – Hierarchical acoustic emission information – Intersection information Rise time – hierarchical acoustic emission information – intersection information are Peak frequency – hierarchical acoustic emission information – intersection information are Energy-level acoustic emission information-intersection information is Absolute energy – hierarchical acoustic emission information – intersection information is Ringing count – hierarchical acoustic emission information – intersection information .

[0102] For example, , , and Empty , and Not empty.

[0103] For example, There are 8 pieces of information in it.

[0104] For example, There are 118 pieces of information in it.

[0105] For example, There are 558 pieces of information in it.

[0106] Step C2, determining the intersection information;

[0107] like If empty, then according to right Filter each element in the list and remove elements less than or equal to 10. Acoustic emission information, retaining more than The acoustic emission information is used as the filtered acoustic emission information – hierarchical information set. Proceed to step C3;

[0108] For example, due to If empty, then the filtered amplitude - acoustic emission information - hierarchical information set .

[0109] For example, due to If empty, then the filtered energy-acoustic emission information-hierarchical information set is... .

[0110] For example, due to If empty, then the filtered absolute energy – acoustic emission information – hierarchical information set .

[0111] For example, due to If empty, then the filtered ring count – acoustic emission information – hierarchy information set .

[0112] like If not empty, return to step B1 and decrease the level interval value. The adjusted hierarchical interval value is denoted as... .

[0113] In this invention, the increment / decrease rate of the hierarchical interval values ​​is 5% to 15%.

[0114] In this invention, after the initial hierarchical interval value, the hierarchical interval value is repeatedly reduced. until arrive The interval value reduction ends when the minimum value of any element in the interval is reached.

[0115] For example, Non-empty indicates that the duration is within Filtering performed at the hierarchical interval values ​​will filter out some fatigue damage information.

[0116] For example, Non-empty indicates that the rise time is within Filtering performed at the hierarchical interval values ​​will filter out some fatigue damage information.

[0117] For example, Non-empty indicates that at the peak frequency Filtering performed at the hierarchical interval values ​​will filter out some fatigue damage information.

[0118] Step C3: Calculate the fitting degree of the acoustic emission information;

[0119] Accumulation Filtered energy , obtain filtered energy accumulation Then on Perform linear fitting to obtain the cumulative energy linearity, denoted as . .

[0120] like Then select As the acoustic emission information – effective threshold, then proceed to step D;

[0121] like If so, return to step B.

[0122] In this invention, the calculation of the goodness of fit is used to measure the hierarchical interval values. Is the setting optimal?

[0123] Step D: Wavelet denoising process for acoustic emission information – effective threshold;

[0124] In this invention, wavelet denoising coefficients are used. Filtered acoustic emission information – hierarchical information set Wavelet decomposition is performed on the acoustic emission signal to obtain detail components (high-frequency part) and approximate components (low-frequency components); wavelet denoising is only performed on all waveforms with energy greater than 2000.

[0125] Then, an inverse wavelet transform is performed on the approximate components to obtain the acoustic emission information after wavelet denoising. .

[0126] This represents the wavelet denoising coefficient.

[0127] This represents the acoustic emission information after wavelet denoising.

[0128] This represents the wavelet scaling variable.

[0129] This represents the wavelet displacement variable.

[0130] This represents the discrete wavelet basis function.

[0131] Represents the differential symbol.

[0132] This represents the wavelet denoising coefficient.

[0133] Step E: Empirical Modal Decomposition (EMD) noise reduction;

[0134] In this invention, the modal empirical decomposition is based on pages 59-62 of "Research on Acoustic Emission Signal Processing Algorithms" published by Chemical Industry Press in August 2017; author: Yu Jintao.

[0135] In this invention, an EMD filter pair is selected. EMD decomposition was performed to extract multiple IMF components, and the low-frequency noise components extracted from the EMD decomposition were removed. Then, the remaining IMF components are reconstructed to obtain the noise-removed acoustic emission information. .

[0136] These are the intrinsic mode functions.

[0137] denoted as the order of the IMF component.

[0138] The order of the IMF component of the damage signal after EMD decomposition.

[0139] The index is the order number.

[0140] It represents the residual.

[0141] Step F, average frequency noise reduction;

[0142] In this invention, waveform peak frequency filtering affects the waveform after EMD decomposition and noise reduction. Identify the waveform signal and the average frequency. Analysis was performed, and noise waveforms with an average frequency below 35kHz were filtered out to obtain a set of fatigue damage information, denoted as . .

[0143] In this invention, steps A to F are used to... The processing yields fatigue damage information for seven elements, denoted as... .

[0144] This indicates fatigue damage amplitude information.

[0145] This indicates fatigue damage duration information.

[0146] This indicates fatigue damage rise time information.

[0147] This indicates fatigue damage – peak frequency information.

[0148] This indicates fatigue damage-energy information.

[0149] This indicates fatigue damage – absolute energy information.

[0150] This indicates fatigue damage - ring count information.

[0151] R-CT model

[0152] See Figure 5 As shown, the R-CT model of the present invention includes the following processing steps:

[0153] Step 1: Analysis based on the R-type factor;

[0154] In this invention, the R-type factor analysis method is based on pages 173-176 of "Multivariate Statistical Analysis" published by China Statistics Press in August 1999; authors: Yu Xiulin and Ren Xuesong.

[0155] Step 101: Extract acoustic emission information characterizing the degree of damage;

[0156] Acoustic emission signals of fatigue damage obtained The nth channel Energy that reflects the degree of material damage Absolute energy And ring count Factor analysis was performed on the three parameters of each acoustic emission signal.

[0157] Step 102: Confirm whether the damage parameters are suitable for factor analysis;

[0158] Normalize the above three variables to obtain the energy. Absolute energy And ring count The normalized data are respectively , , The KMO test method was used to... , , Validation was performed to obtain the correlation between the three variables, and dimensionality enhancement was performed to obtain high-dimensional variables. , , .

[0159] Step 103, Solving for the factor loading matrix;

[0160] R-factor analysis was used to analyze high-dimensional variables. , , The solution is performed to transform the high-dimensional variables into linear functions of the common factors, and the loading matrix of the common factors is calculated.

[0161] In this invention, high-dimensional variables are expressed as linear functions of common factors, denoted as: ;

[0162] in, Represented as The degree of damage represented by each acoustic emission signal, q is Total number of acoustic emission signal events, factor load matrix , The elements in the middle are the calculated factor loading matrix. These are represented as common factors.

[0163] Step 104, Calculate common factor scores;

[0164] Common factors can reflect the correlation between initial variables. Using common factors to represent initial variables is more conducive to describing the characteristics of the research subjects. Therefore, it is necessary to reverse this process and express common factors as linear combinations of initial variables. This function can be used to obtain... The energy of each acoustic emission signal Absolute energy And ring count The factor scores are energy-factor scores. Absolute Energy - Factor Score Ring count - factor score .

[0165] In this invention, the common factor is represented by the linear combination formula of the initial variables. .

[0166] As a factor In scalar The score.

[0167] Composite factor score Accumulate the damage to obtain the cumulative damage parameters. Perform mutation theory calculations and lifetime prediction analysis.

[0168] Step 2, calculations based on swallowtail mutations;

[0169] In this invention, the swallowtail mutation theory is referenced from pages 34-36 of "Control and Application Based on Mutation Theory" published by Harbin Institute of Technology Press in April 2013; authors: Zhao Xinhua and Cao Wei.

[0170] In this invention, the swallowtail mutation theory method is used to calculate the common factor score in step one. By performing potential function analysis and surface configuration analysis, the critical points on the surface plot can be obtained, such as... Figure 6 As shown.

[0171] Furthermore, the function corresponding to the swallowtail mutation function involved is set as follows: ;in, , , , , , These are the coefficients in the multiple regression analysis.

[0172] If the equilibrium surface equation and singularity set of the swallowtail catastrophe are the first and second derivatives of the swallowtail catastrophe function, respectively, then the potential function of the swallowtail catastrophe is: ;in, , , These are the system's control parameters.

[0173] Furthermore, in order to obtain the singularity set of the swallowtail catastrophe model and thus pick out the catastrophe points, the potential function is further calculated: .

[0174] In this invention, the nth channel is obtained by calculating the cumulative comprehensive factor score based on the swallowtail mutation theory. The abrupt change point reflecting material damage is denoted as the damage-transition point information. The abrupt change time representing the near-failure of SiC-CMC materials is denoted as […]. It can provide early warning for the safe application of materials.

[0175]

[0176] Step 3, mutational fusion of multiple channels;

[0177] In this invention, the R-CT model is used for N channels. Swallowtail mutation calculations were performed on all data, and the critical point information nearing failure was obtained on the surface plots calculated from N channels by data analysis. It is used for failure early warning judgment.

[0178] Example 1

[0179] The tail nozzle of a certain type of aero-engine: total experimental length 50mm, working section length 30mm, working section width 4mm, and sample thickness 2mm.

[0180] The composition of the SiC-CMC material used in the structural components is shown in Table 1:

[0181] Table 1 Properties of silicon carbide fibers

[0182]

[0183] The testing equipment includes:

[0184] (A) Two Nano-30 acoustic emission sensors with a resonant frequency of 140kHz and a peak frequency of 351kHz.

[0185] (B) The acoustic emission instrument is a fully digital 16-channel DiSP acoustic emission system from PAC Corporation, USA. The threshold value for acoustic emission detection is 35dB, the peak emission time (PDT) is 30μs, the acoustic emission impact time (HDT) is 60μs, and the acoustic emission impact lockout time (HLT) is 200μs.

[0186] Factor analysis was performed on the acoustic emission information of fatigue damage parameters of the tail nozzle using three parameters: count, energy, and absolute energy, to obtain the contribution rate of each parameter. Then, the comprehensive factor score was calculated. The copper drum swallowtail catastrophe theory was used to calculate the cumulative comprehensive factor score, thereby identifying the relationship between the catastrophe time and fracture time of the SiC-CMC structural component.

[0187] This invention establishes a method for predicting the fatigue life of SiC-CMC materials based on acoustic emission technology. It filters the obtained acoustic emission waveform information of fatigue damage through time-domain reference-multi-stage filtering to obtain information reflecting the fatigue damage of structural components, and then performs catastrophe theory calculations to predict fatigue life. Applying this invention enables filtering of acoustic emission noise signals obtained during fatigue testing, and provides an intuitive, quantitative, and real-time assessment of the fatigue life status of SiC-CMC, thereby issuing early warnings and reducing losses such as equipment damage and personnel casualties.

Claims

1. A fatigue failure early warning system for SiC-CMC materials based on a combination of STAF and R-CT, characterized in that: This includes the STAF model and the R-CT model; The STAF model is... The digital acoustic transmission information is processed on a per-channel basis, i.e., a step-by-step information processing method; however, for... All elements in the model employ the same multi-stage filtering method, and the STAF model processes digital acoustic emission information. ; The digital acoustic emission information collected by the first acoustic emission sensor during the sampling time is denoted as... The digital acoustic emission information collected by the second acoustic emission sensor during the sampling time is denoted as... ;No. The digital acoustic emission information collected by each acoustic emission sensor during the sampling time is denoted as . The digital acoustic emission information collected by the last acoustic emission sensor during the sampling time is denoted as... ; For the first Road digital sound transmission information The amplitude; For the first Road digital sound transmission information The duration; For the first Road digital sound transmission information The rise time; For the first Road sound transmission information Peak frequency; For the first Road digital sound transmission information Energy; For the first Road digital sound transmission information Absolute energy; For the first Road digital sound transmission information The ring count; The STAF model includes the following processing steps: Step A: Acquisition of time-domain acoustic emission information at different time periods; Step A1: Dividing acoustic emission information into four time periods; Using the STT method to analyze acoustic emission information The process involves dividing the data into four time periods. The STT method, which stands for Digital Sound Transmission Information Time Period Division Method, comprises the following steps: The method for dividing digital acoustic emission information into time periods refers to the process of using the acoustic emission sensor to collect acoustic emission information of the test sample within a sampling period, and recording the sampling start time as _____. The time point at which sampling ends is denoted as... ;from to Divided into four time periods, each designated as the first time period. The second time period The third time period The fourth time period ; In the process of predicting and warning of fatigue damage in test samples using acoustic emission technology, based on a sampling period... The energy accumulation is used to determine the end point of each time period; from Start to accumulated energy The point at which the price level begins to flatten out is denoted as . ;from Start to accumulated energy The point in time when the increase is rapid is denoted as ;from Start to accumulated energy The point at which the price level begins to flatten out is denoted as . ; from to The four time periods are as follows: First time period It refers to from arrive A period of time; Second time period It refers to from arrive A period of time; The third time period It refers to from arrive A period of time; The fourth time period It refers to from arrive A period of time; Step A2, time domain division; from , , , Ten seconds were selected from each of the four time periods as the time for extracting acoustic emission information in the time domain. First time domain Time period From the start time of the test The first 10 seconds of counting; Second time domain Time period Any 10-second interval; Third time domain Time period Any 10-second interval; 4th time domain Time period Any 10-second interval; Step A3: Acquisition of acoustic emission information in the time domain; Acoustic emission information The information in each of the four time periods is represented as follows: The acoustic emission information for the first time period is denoted as ; in the Extracting the time domain The acoustic emission information below is denoted as time-domain acoustic emission information. ; The acoustic emission information for the second time period is denoted as ; in the Extracting the time domain The acoustic emission information below is denoted as time-domain acoustic emission information. ; The acoustic emission information for the third time period is denoted as ; in the Extracting the time domain The acoustic emission information below is denoted as time-domain acoustic emission information. ; The acoustic emission information for the fourth time period is denoted as ; in the Extracting the time domain The acoustic emission information below is denoted as time-domain acoustic emission information. ; Step B, hierarchical division of acoustic emission information; To eliminate interference errors in the acoustic emission waveform signal caused by noise, this invention employs hierarchical division of acoustic emission information. Perform noise reduction at different levels; Step B1: Set the initial level range values; Initially, the hierarchical interval values ​​are denoted as ; Step B2: Obtain acoustic emission information within the hierarchical interval; The CJJ method, or Digital Acoustic Emissions Method, involves the following steps: The hierarchical partitioning method for digital acoustic emission information refers to using a hierarchical partitioning method to perform noise reduction processing on each element of digital acoustic emission information at different levels; Hierarchical identifier is denoted as The number of levels is denoted as and the smallest The value is assigned to 1; For any channel of digital acoustic transmission information If we label the elements in the hierarchy, then we have The hierarchical range value is , The hierarchical range value is , The hierarchical range value is , The hierarchical range value is , The hierarchical range value is , The hierarchical range value is , The hierarchical range value is ; The attached layer of digital acoustic emission information is denoted as ; Using the CJJ method from The acoustic emission information under the hierarchical intervals is extracted and denoted as the hierarchical noise reduction information set in the first time period. ; Using the CJJ method from The acoustic emission information under the hierarchical intervals is extracted and denoted as the hierarchical noise reduction information set in the second time period. ; Using the CJJ method from The acoustic emission information under the hierarchical intervals is extracted and denoted as the hierarchical noise reduction information set in the third time period. ; Using the CJJ method from The acoustic emission information under the hierarchical intervals is extracted and denoted as the hierarchical noise reduction information set in the fourth time period. ; After steps A to B, the amplitude is denoised in stages. , , and ; The hierarchical noise reduction for the duration after steps A to B are respectively , , and ; The hierarchical noise reduction of rise time after steps A to B are respectively , , and ; The hierarchical noise reduction of the peak frequency after steps A to B are respectively , , and ; The hierarchical noise reduction of energy after steps A to B are respectively , , and ; The hierarchical noise reduction of absolute energy after steps A to B are respectively , , and ; After steps A to B, the ring count is reduced in stages. , , and ; Step C: Calculate the fitting degree of the hierarchical acoustic emission information; application Linear relationship for accumulation Filtered energy The fitting process is performed to determine the hierarchical interval division values ​​output from step B. Step C1: Merge first, then submit; First Perform union processing to obtain hierarchical acoustic emission information – union information. Then Intersection processing is performed to obtain hierarchical acoustic emission information – intersection information. ; After processing in step C1, the amplitude-level acoustic emission information-intersection information is: Duration – Hierarchical acoustic emission information – Intersection information Rise time – hierarchical acoustic emission information – intersection information are Peak frequency – hierarchical acoustic emission information – intersection information are Energy-level acoustic emission information-intersection information is Absolute energy – hierarchical acoustic emission information – intersection information is Ringing count – hierarchical acoustic emission information – intersection information ; Step C2, determining the intersection information; like If empty, then according to right Filter each element in the list and remove elements less than or equal to 10. Acoustic emission information, retaining more than The acoustic emission information is used as the filtered acoustic emission information – hierarchical information set. ;Execute step C3; like If not empty, return to step B1 and decrease the level interval value. The adjusted hierarchical interval value is denoted as... ; After the initial hierarchical interval value, the hierarchical interval value is repeatedly reduced. until arrive The minimum value of any element in the interval is reached, and the interval value reduction ends. Step C3: Calculate the fitting degree of the acoustic emission information; Accumulation Filtered energy , obtain filtered energy accumulation Then on Perform linear fitting to obtain the cumulative energy linearity, denoted as . ; like Then select As the acoustic emission information – effective threshold, then proceed to step D; like Then return to step B; Step D: Wavelet denoising process for acoustic emission information – effective threshold; Using wavelet noise reduction coefficients Filtered acoustic emission information – hierarchical information set Wavelet decomposition is performed on the acoustic emission signal to obtain detail components and approximate components; wavelet denoising is only performed on all waveforms with energy greater than 2000. Then, an inverse wavelet transform is performed on the approximate components to obtain the wavelet-denoised acoustic emission information. ; Indicates the wavelet denoising coefficients; This represents the acoustic emission information after wavelet denoising. Represents wavelet scaling variables; Represents wavelet displacement variables; Represents the discrete wavelet basis functions; Represents the differential symbol; Indicates the wavelet denoising coefficients; Step E: Empirical Modal Decomposition (EMD) noise reduction; Select EMD filter pair EMD decomposition was performed to extract multiple IMF components, and the low-frequency noise components extracted from the EMD decomposition were removed. Then, the remaining IMF components are reconstructed to obtain the noise-removed acoustic emission information. ; These are the intrinsic mode functions; The order of the IMF component; The order of the IMF component of the damage signal after EMD decomposition; The ordinal number is the order of the number; For residuals; Step F, average frequency noise reduction; Waveform peak frequency filtering after EMD decomposition and noise reduction Identify the waveform signal and the average frequency. Analysis was performed, and noise waveforms with an average frequency below 35kHz were filtered out to obtain a set of fatigue damage information, denoted as . ; Use steps A through F to perform The processing yields fatigue damage information for seven elements, denoted as... ; Indicates fatigue damage amplitude information; Indicates fatigue damage duration information; This indicates fatigue injury rise time information; Indicates fatigue damage - peak frequency information; Indicates fatigue damage - energy information; Indicates fatigue damage – absolute energy information; This indicates fatigue damage – ring count information; The R-CT model includes the following processing steps: Step 1: Analysis based on R-type factors; Step 101: Extract acoustic emission information characterizing the degree of damage; Acoustic emission signals of fatigue damage obtained The nth channel Energy that reflects the degree of material damage Absolute energy And ring count Factor analysis was performed on the three parameters of each acoustic emission signal; Step 102: Confirm whether the damage parameters are suitable for factor analysis; Normalize the above three variables to obtain the energy. Absolute energy And ring count The normalized data are respectively , , The KMO test method was used to... , , Validation was performed to obtain the correlation between the three variables, and dimensionality enhancement was performed to obtain high-dimensional variables. , , ; Step 103, Solving for the factor loading matrix; R-factor analysis was used to analyze high-dimensional variables. , , The solution is performed to transform the high-dimensional variables into linear functions of the common factors, and the loading matrix of the common factors is calculated. Step 104, Calculate common factor scores; This function can be used to obtain The energy of each acoustic emission signal Absolute energy And ring count The factor scores are energy-factor scores. Absolute Energy - Factor Score Ring count - factor score ; The common factors are expressed as a linear combination of the initial variables. ; As a factor In scalar The score on; Composite factor score Accumulate the damage to obtain the cumulative damage parameters. Perform catastrophe theory calculations and lifetime prediction analyses; Step 2, calculations based on swallowtail mutations; Using the swallowtail mutation theory method to Potential function analysis and surface configuration analysis are performed to obtain the critical points on the surface plot; Furthermore, the function corresponding to the swallowtail mutation function involved is set as follows: ;in, , , , , , These are the coefficients for multiple regression analysis; The equilibrium surface equation and singularity set of the swallowtail catastrophe are the first and second derivatives of the swallowtail catastrophe function, respectively. Therefore, the potential function of the swallowtail catastrophe is: ;in, , , These are the system's control parameters; To obtain the set of singularities in the swallowtail catastrophe model and thus pick out the catastrophe points, the potential function is further calculated: ; Step 3, mutational fusion of multiple channels; R-CT model for N channels Swallowtail mutation calculations were performed on all data, and the critical point information nearing failure was obtained on the surface plots calculated from N channels by data analysis. It is used for failure early warning judgment.

2. The fatigue failure early warning system for SiC-CMC materials based on the combination of STAF and R-CT according to claim 1, characterized in that: Initial hierarchical interval values Afterwards, The reduction is called the reduced hierarchical interval value, denoted as . Repeat execution The reduction, until arrive The interval value reduction ends when the minimum value of any element in the interval is reached.

3. The fatigue failure early warning system for SiC-CMC materials based on the combination of STAF and R-CT according to claim 1, characterized in that: The STAF model and R-CT model are embedded in the acoustic emission instrument's memory.

Images