A torsion shaft health degree evaluation method based on optical fiber ring acoustic emission

Abstract

The application discloses a torsion shaft health degree evaluation method based on optical fiber ring acoustic emission, which comprises the following steps: collecting energy data of acoustic emission signals of a torsion shaft in a fatigue failure process by using an acoustic emission sensor; realizing performance degradation analysis of the torsion shaft based on singular spectrum analysis; and realizing the judgment of the health degree of the torsion shaft in the whole life cycle based on the distance-based health degree function on the basis of the performance degradation analysis. The application measures the acoustic emission signals of the torsion shaft by using the optical fiber ring acoustic emission sensor, and extracts the fault factors by using the method combining data and knowledge; the health degree function is defined according to the fault factors, the real-time health degree of the torsion shaft is obtained, the health degree monitoring of the torsion shaft in the whole life cycle is realized, and thus the early warning can be given before the fault occurs, and unnecessary loss is avoided.

Description

Technical Field

[0001] This invention relates to the field of acoustic emission nondestructive testing technology, and in particular to a method for assessing the health of a torsion shaft based on optical fiber loop acoustic emission. Background Technology

[0002] Regarding the assessment of the health of torsion shafts, existing technologies generally use ultrasonic testing to detect residual stress in the torsion shaft and then roughly judge the health status of the torsion shaft based on the waveform changes of the tested signal during the torsion shaft fracture. However, most of these methods only describe the waveform changes and lack quantitative analysis of failure factors, and cannot provide a judgment on the health of the torsion shaft.

[0003] Therefore, how to provide a method for assessing the health of torsion shafts based on fiber optic ring acoustic emission has become a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] In view of the above background technology, this invention provides a method for assessing the health of a torsion shaft based on fiber optic loop acoustic emission. Addressing the problem of real-time health monitoring of torsion shafts, a torsional fatigue test method for torsion shafts is developed, utilizing a fiber optic loop acoustic emission sensor to collect acoustic emission wave signals generated by crack propagation on the surface and inside the torsion shaft. Singular spectrum decomposition is used to decompose the acoustic emission signals into components, and fault factors are injected based on the characteristics of different components. Based on these fault factors, a torsion shaft health function is designed to accurately describe the changes in the degree of damage to the torsion shaft.

[0005] The present invention solves the technical problem by adopting the following technical solution:

[0006] A method for assessing the health of a torsion shaft based on optical fiber loop acoustic emission includes using an acoustic emission sensor to collect energy data of acoustic emission signals during the fatigue failure process of the torsion shaft, performing performance degradation analysis of the torsion shaft based on singular spectrum analysis, and determining the health status of the torsion shaft throughout its entire life cycle based on a distance-based health function.

[0007] Furthermore, the performance degradation analysis of the torsion shaft based on singular spectrum analysis includes: preprocessing energy data to determine the characteristic values ​​corresponding to different components, decomposing singular spectrum components to extract fault characteristics, and determining the fatigue failure period.

[0008] Furthermore, the method for preprocessing energy data is as follows:

[0009] The energy data is preprocessed to remove redundant data and fill in missing data. The rule for filling in missing data is to set a fixed time interval of 10... -ns, round the time term in the data to n decimal places; then fill in the missing data between adjacent data points, using the energy value from the previous time step; if the time difference between two adjacent data points is less than 10 -n If the time values ​​are the same after rounding, then the two time values ​​are combined, and the energy value is taken as the larger of the two values.

[0010] Furthermore, the method for decomposing singular spectral components includes the following steps:

[0011] (1) Embedding:

[0012] The analysis object of singular spectral components is a finite-length one-dimensional time series [x1, x2, ..., x...]. N ], N is the sequence length; first, a suitable window length L needs to be selected to lag the original time series to obtain the trajectory matrix:

[0013]

[0014] Pick Let K = N - L + 1, then the trajectory matrix X is an L × K matrix:

[0015]

[0016] (2) Decomposition:

[0017] Perform singular value decomposition on the trajectory matrix, that is, decompose X into the following form:

[0018] X=UΣV T (3)

[0019] Where U is the left matrix; Σ is the diagonal matrix, and the values ​​on the main diagonal are the singular values; V is the right matrix; U and V are both identity orthogonal matrices, satisfying UU T =I,VV T =I;

[0020] First, calculate the covariance matrix of the trajectory matrix:

[0021] S = XX T (4)

[0022] Eigenvalue decomposition of S yields eigenvalues ​​λ1>λ2>...>λ L ≥0 and the corresponding eigenvectors U1, U2, ..., U L , at this time U=[U1,U2,...,U L ], The singular spectrum of the original sequence; and we have:

[0023]

[0024] Where, λ i The corresponding eigenvector U i It reflects the evolution of time series and is called the time empirical orthogonal function;

[0025] (3) Grouping:

[0026] Calculate the hysteresis sequence X i in U m Projection on:

[0027]

[0028] X i This represents the i-th column of the trajectory matrix X. It is X i The time evolution pattern reflected in the original sequence x i+1 ,x i+2 ,...,x i+L The weights of different time periods are called temporal principal components; The resulting matrix is ​​an unnormalized right matrix, i.e.

[0029] (4) Reconstruction:

[0030] Reconstruction is performed using time-based empirical orthogonal functions and time-based principal components. The specific reconstruction process is as follows:

[0031]

[0032] in, It is x i-j The time evolution pattern reflected in the original sequence x i+1 ,x i+2 ,...,x i+L Time period weight, U k,j It is λ i Corresponding eigenvector U i The value of the j-th row;

[0033] Thus, the sum of all reconstructed sequences should equal the original sequence, that is:

[0034]

[0035] Furthermore, the method for determining the fatigue failure period is as follows:

[0036] In the four stages of crack nucleation, microcrack propagation, macrocrack propagation, and instantaneous fracture, the torsion shaft exhibits a relatively rapid crack initiation rate and a high frequency of high-energy acoustic emission signals. Conversely, between two adjacent fatigue failure stages, crack propagation is relatively slow, and the energy value of acoustic emission is relatively low. Based on this characteristic, the following rules are used for judgment:

[0037] Entering the fatigue failure stage: Previously not in the fatigue failure stage, current energy trend value E t The energy threshold for acoustic emission is higher than E. th And in the future t last Within a time period, E t Greater than E th Time exceeds t th1 ;

[0038] End of fatigue failure phase: Previously in the fatigue failure phase, and the current energy trend value E t Below the high energy threshold value E for acoustic emission th The duration is greater than t th2 ;

[0039] This allows us to determine the time intervals for the four fatigue failure stages.

[0040] Furthermore, based on the distance-based health function, the method for determining the health level of the torque shaft throughout its entire life cycle is as follows: determine the health baseline value, determine the slope coefficient of the health function, construct the health function, obtain the health function value at each time point, and determine the health level of the torque shaft throughout its entire life cycle through the health function value at each time point.

[0041] Furthermore, the method for determining health benchmark values ​​is as follows:

[0042] Based on the determination of the time interval where the fatigue failure stage occurs, all acoustic emission energy trend values ​​E are... t The set is divided into two sets: the High set and the Low set; the High set corresponds to the E value of the fatigue failure stage being determined. t The Low set corresponds to E outside the High set. t Because E in the Low set t The value is low, indicating a stationary sequence with a slow crack propagation rate. Therefore, a baseline value 'a' is derived using the Low set, assuming that when E... t When E is greater than a, the health of the torque shaft fluctuates downwards. t When the value is less than 'a', the health of the torsion shaft fluctuates upwards.

[0043] To highlight E t The impact of deviation from the reference value 'a' on the health of the torque shaft, let (E) t -a) 2 sgn(aE t The slope of the change in health is proportional to the energy level. Since the low-energy trend segment corresponds to a stationary health level sequence, the baseline value 'a' satisfies:

[0044] ∑(E t (t)-a) 2sgn(aE t (t))=0,t∈Low (9)

[0045] This yields the baseline value a.

[0046] Furthermore, the method for determining the slope coefficient of the health function is as follows:

[0047] After obtaining the baseline value a, due to (E t -a) 2 sgn(aE t The slope of the health function is proportional to the slope of the change in health, therefore the slope k of the health function is:

[0048] k = b(E) t (t)-a) 2 sgn(aE t (t)), t∈All (10).

[0049] Furthermore, the method for constructing the health function is as follows:

[0050] Let the health of the torque shaft be F. Since the torque shaft eventually breaks, its health becomes 0. Therefore:

[0051] F(0)=∑b(E t (t)-a) 2 sgn(aE t (t)) (11)

[0052] Since F(0) is known, b can be solved, so the health function F(t) has the following relationship:

[0053] F(t)=F(t-Δt)+b(E t (t)-a) 2 sgn(aE t (t)), t∈All (12)

[0054] Where Δt is the sampling interval after data preprocessing.

[0055] Beneficial effects:

[0056] This invention addresses the problem of torque shaft performance degradation, specifically the difficulty in identifying universal performance degradation patterns, predicting torque shaft fracture, and assessing the overall health of the torque shaft throughout its lifespan using existing technologies. This invention utilizes a fiber optic loop acoustic emission sensor to measure the acoustic emission signals of the torque shaft and extracts fault factors through a combination of data and knowledge analysis. A health function is defined based on these fault factors to derive the real-time health of the torque shaft. This enables full lifespan health monitoring of the torque shaft, ensuring early warning before failures occur and preventing unnecessary losses. Attached Figure Description

[0057] Figure 1 This is a flowchart of the method of the present invention.

[0058] Figure 2 This is a schematic diagram of the layout of the fiber optic ring acoustic emission sensor for the present invention, which addresses the inherent defects.

[0059] Figure 3 This is a flowchart of the torsion shaft performance degradation analysis based on singular spectrum analysis according to the present invention.

[0060] Figure 4 This is a flowchart of the distance-based health function solution of the present invention.

[0061] Figure 5 This is a photograph of the fracture of the torsion shaft at a point defect, as an experimental example of the present invention.

[0062] Figure 6 This is a time series of acoustic emission energy from a torsion shaft experiment, an experimental example of this invention.

[0063] Figure 7 The data curves for the experimental examples of this invention are completed at time intervals of 0.1s.

[0064] Figure 8 The data curves for the experimental examples of this invention are completed at time intervals of 0.01s.

[0065] Figure 9 The experimental figures of this invention are filled with data curves at time intervals of 0.001s.

[0066] Figure 10 This is a time series of the energy trend term in an experimental example of the present invention.

[0067] Figure 11 This is the health curve of the torsion shaft in the experimental example of the present invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] The torsion shaft health evaluation method proposed in this invention collects data through a real-time torsion shaft monitoring system. This system collects and monitors the acoustic emission signals generated by the development of cracks in the torsion shaft through an optical fiber ring acoustic emission sensor.

[0070] When the microstructure of a material is insufficient to withstand forces from all directions, deformation and fracture occur. At the moment of deformation and fracture, the material rapidly releases energy and propagates it as transient elastic waves—this is the phenomenon of acoustic emission. A fiber optic loop acoustic emission sensor is a sensor composed of a loop of single-mode fiber wound around itself, used to detect acoustic emission wave signals. With the continuous advancement of modern information technology and new sensing technologies, optical fibers can serve not only as information transmission units but also as excellent information sensing units. The fiber optic loop, acting as a sensor, senses the stress wave of the acoustic emission signal. Under the action of the stress wave, it undergoes stretching or compression, causing a change in the optical path of the light transmitted within it. This, in turn, modulates the phase difference between the sensing arm and the reference arm using the acoustic emission signal. The fiber optic loop acoustic emission sensor utilizes this acoustic emission signal acting on the phase change of the light wave propagating within the sensing fiber optic loop, and then detects the acoustic emission signal by observing the change in the phase interference of the light waves.

[0071] The energy of the acoustic emission signal of the torsion shaft during fatigue failure is collected by an acoustic emission sensor, and the health evaluation of the torsion shaft is achieved by analyzing the collected data.

[0072] This invention patent is divided into two parts: performance degradation analysis and health evaluation of the torsion shaft. Energy data during the fatigue failure process of the torsion shaft is collected using an acoustic emission sensor. Data preprocessing is performed to remove redundant data and fill in missing data. Then, singular spectrum analysis is used to extract the energy data change trend. Combining fatigue failure theory with the analysis of the characteristics of each component, corresponding failure characteristic factors are extracted, and the fatigue failure stage corresponding to different moments is determined, thus achieving performance degradation analysis of the torsion shaft. Based on the performance degradation analysis, a performance degradation benchmark value is derived according to the determination of the fatigue failure stage. A health function of the torsion shaft is constructed based on the deviation between the energy trend value and the benchmark value, realizing the determination of the health level of the torsion shaft throughout its entire life cycle, achieving the goal of using monitoring data for torsion shaft health management.

[0073] refer to Figure 1 This invention discloses a method for assessing the health of a torsion shaft based on optical fiber loop acoustic emission. The method includes using an acoustic emission sensor to collect energy data of acoustic emission signals during the fatigue failure process of the torsion shaft, performing performance degradation analysis of the torsion shaft based on singular spectrum analysis, and determining the health status of the torsion shaft throughout its entire life cycle based on a distance-based health function.

[0074] The present invention is applied to torsion shafts, which can generally be used in the suspension system of armored vehicles, turbine disks of aircraft engines, and control sticks of aircraft to maintain system stability and provide torque.

[0075] This invention designs a fatigue failure test for a torsion shaft, using a torsion shaft with product model 5008SY and product number 8S26. During the experiment, the torsion shaft undergoes periodic torsional motion, with a maximum torsion angle of 90° and a load application cycle of 45 cycles / min. The experiment ends when the torsion shaft breaks.

[0076] To accelerate the fatigue failure process of the torsion shaft, three artificial defects were created on it. To facilitate the location of the fault area, sensors were deployed at both ends of the torsion shaft. The defects and sensor layout of the torsion shaft are as follows. Figure 2 As shown.

[0077] After setting up the defect and acoustic emission sensors, to prevent metal material from splashing when the torsion shaft breaks and to ensure the safety of the experimental operators, the metal outer cover should be kept tightly closed throughout the experiment.

[0078] refer to Figure 2 The performance degradation analysis of the torsion shaft based on singular spectrum analysis includes: preprocessing energy data to determine the characteristic values ​​corresponding to different components, decomposing singular spectrum components to extract fault characteristics, and determining the fatigue failure period.

[0079] The method for preprocessing energy data is as follows:

[0080] The acoustic emission sensor used has a very high sampling frequency of 1MHz. In order to ensure efficiency, the difference between the signal value and the previous time step is recorded only when it reaches a threshold value. Therefore, the actual collected data is not evenly spaced and has too many significant digits (as shown in Table 1), which is not conducive to the analysis of time series. This problem needs to be solved by interpolating the data.

[0081]

[0082] Table 1

[0083] The energy data is preprocessed to remove redundant data and fill in missing data. The rule for filling in missing data is to set a fixed time interval of 10... -n s, round the time term in the data to n decimal places; then fill in the missing data between adjacent data points, using the energy value from the previous time step; if the time difference between two adjacent data points is less than 10 -n If the time values ​​are the same after rounding, then these two values ​​are combined, and the energy value is taken as the larger of the two. Since the fatigue failure test of the torsion shaft has a long cycle, the value of n should not be too large in order to take into account the computing power of the computer.

[0084] Singular Spectrum Analysis (SSA) is a method for processing nonlinear time series data. It extracts different component sequences (long-term trend, seasonal trend, noise, etc.) from the trajectory matrix of the time series under study through decomposition and reconstruction, thereby enabling analysis or denoising of the time series and its application in other tasks. SSA mainly includes four steps: embedding, decomposition, grouping, and reconstruction.

[0085] The method for decomposing singular spectral components includes the following steps:

[0086] (1) Embedding:

[0087] The analysis object of singular spectral components is a finite-length one-dimensional time series [x1, x2, ..., x...]. N ], N is the sequence length; first, a suitable window length L needs to be selected to lag the original time series to obtain the trajectory matrix:

[0088]

[0089] Pick Let K = N - L + 1, then the trajectory matrix X is an L × K matrix:

[0090]

[0091] (2) Decomposition:

[0092] Perform singular value decomposition on the trajectory matrix, that is, decompose X into the following form:

[0093] X=UΣV T (3)

[0094] Where U is the left matrix; Σ is the diagonal matrix, and the values ​​on the main diagonal are the singular values; V is the right matrix; U and V are both identity orthogonal matrices, satisfying UU T =I,VV T =I;

[0095] Since direct decomposition of the trajectory matrix is ​​difficult, we first calculate the covariance matrix of the trajectory matrix:

[0096] S = XX T (4)

[0097] Eigenvalue decomposition of S yields eigenvalues ​​λ1>λ2>...>λ L ≥0 and the corresponding eigenvectors U1, U2, ..., U L , at this time U=[U1,U2,...,U L ], The singular spectrum of the original sequence; and we have:

[0098]

[0099] Where, λ i The corresponding eigenvector U i It reflects the evolution of time series and is called the time empirical orthogonal function (T-EOF);

[0100] (3) Grouping:

[0101] Calculate the hysteresis sequence X i in U m Projection on:

[0102]

[0103] X i This represents the i-th column of the trajectory matrix X. It is X i The time evolution pattern reflected in the original sequence x i+1 ,x i+2 ,...,x i+L The weights of time periods are called time principal components (TPCs). The resulting matrix is ​​an unnormalized right matrix, i.e.

[0104] (4) Reconstruction:

[0105] Reconstruction is performed using time-based empirical orthogonal functions and time-based principal components. The specific reconstruction process is as follows:

[0106]

[0107] in, It is x i-j The time evolution pattern reflected in the original sequence x i+1 ,x i+2 ,...,x i+L Time period weight, U k,j It is λ i Corresponding eigenvector U i The value of the j-th row;

[0108] Thus, the sum of all reconstructed sequences should equal the original sequence, that is:

[0109]

[0110] Based on the contribution of singular values, the sequences corresponding to the singular values ​​are summed to obtain separate sequences for the trend term, periodic term, and noise term. This patent extracts the energy trend term for performance degradation analysis.

[0111] According to fatigue failure theory, the fatigue failure process of a torsion shaft can be roughly divided into four stages: crack nucleation stage, microcrack propagation stage, macrocrack propagation stage, and instantaneous fracture stage.

[0112] In the first stage (crack nucleation stage), the torsion shaft is subjected to alternating stress, which generates slip lines inside, and the metal material is squeezed in and extruded, forming the nucleus of microcracks.

[0113] The second stage (microcrack propagation stage) is when cracks nucleate and propagate along the slip surface, forming a large number of small, non-single microcracks with a depth of less than 0.005 mm within the torsion shaft. This stage produces a very large number of cracks.

[0114] The third stage (macro-crack propagation stage) is in which the crack propagation direction is basically perpendicular to the principal stress, and the crack size is greater than 0.01 mm.

[0115] In the fourth stage (instantaneous fracture stage), the macroscopic crack on the torsion shaft expands beyond the critical size, causing unstable propagation and rapid fracture, with very high instantaneous energy of acoustic emission.

[0116] The method for determining the fatigue failure period is as follows:

[0117] In the four stages of crack nucleation, microcrack propagation, macrocrack propagation, and instantaneous fracture, the torsion shaft exhibits a relatively rapid crack initiation rate and a high frequency of high-energy acoustic emission signals. Conversely, between two adjacent fatigue failure stages, crack propagation is relatively slow, and the energy value of acoustic emission is relatively low. Based on this characteristic, the following rules are used for judgment:

[0118] Entering the fatigue failure stage: Previously not in the fatigue failure stage, current energy trend value E t The energy threshold for acoustic emission is higher than E. th And in the future t last Within a time period, E t Greater than E th Time exceeds t th1 ;

[0119] End of fatigue failure phase: Previously in the fatigue failure phase, and the current energy trend value E t Below the high energy threshold value E for acoustic emission th The duration is greater than t th2 ;

[0120] This allows us to determine the time intervals for the four fatigue failure stages.

[0121] refer to Figure 4The method for determining the health status of a torque shaft throughout its entire life cycle based on a distance-based health function is as follows: determine the health baseline value, determine the slope coefficient of the health function, construct the health function, obtain the health function value at each time point, and determine the health status of the torque shaft throughout its entire life cycle through the health function value at each time point.

[0122] The method for determining health baseline values ​​is as follows:

[0123] Based on the determination of the time interval where the fatigue failure stage occurs, all acoustic emission energy trend values ​​E are... t The set is divided into two sets: the High set and the Low set; the High set corresponds to the E value of the fatigue failure stage being determined. t The Low set corresponds to E outside the High set. t Because E in the Low set t The value is low, indicating a stationary sequence with a slow crack propagation rate. Therefore, a baseline value 'a' is derived using the Low set, assuming that when E... t When E is greater than a, the health of the torque shaft fluctuates downwards. t When the value is less than 'a', the health of the torsion shaft fluctuates upwards.

[0124] To highlight E t The impact of deviation from the reference value 'a' on the health of the torque shaft, let (E) t -a) 2 sgn(aE t The slope of the change in health is proportional to the energy level. Since the low-energy trend segment corresponds to a stationary health level sequence, the baseline value 'a' satisfies:

[0125] ∑(E t (t)-a) 2 sgn(aE t (t))=0,t∈Low (9)

[0126] This yields the baseline value a.

[0127] The method for determining the slope coefficient of the health function is as follows:

[0128] After obtaining the baseline value a, due to (E t -a) 2 sgn(aE t The slope of the health function is proportional to the slope of the change in health, therefore the slope k of the health function is:

[0129] k = b(E) t (t)-a) 2 sgn(aE t (t)), t∈All (10).

[0130] The method for constructing the health function is as follows:

[0131] Let the health of the torque shaft be F. Since the torque shaft eventually breaks, its health becomes 0. Therefore:

[0132] F(0)=∑b(E t (t)-a) 2 sgn(aE t (t)) (11)

[0133] Since F(0) is known, b can be solved, so the health function F(t) has the following relationship:

[0134] F(t)=F(t-Δt)+b(E t (t)-a) 2 sgn(aE t (t)), t∈All (12)

[0135] Where Δt is the sampling interval after data preprocessing.

[0136] This invention addresses the problem of torque shaft performance degradation, specifically the difficulty in identifying universal performance degradation patterns, predicting torque shaft fracture, and assessing the overall health of the torque shaft throughout its lifespan using existing technologies. This invention utilizes a fiber optic loop acoustic emission sensor to measure the acoustic emission signals of the torque shaft and extracts fault factors through a combination of data and knowledge analysis. A health function is defined based on these fault factors to derive the real-time health of the torque shaft. This enables full lifespan health monitoring of the torque shaft, ensuring early warning before failures occur and preventing unnecessary losses.

[0137] Since torsion shaft fractures often occur without warning, the performance degradation patterns of torsion shafts are unclear, making it impossible to provide early warnings before fracture. This invention proposes a method for studying the health status of torsion shafts based on fiber optic loop acoustic emission detection technology. The method utilizes a fiber optic loop acoustic emission sensor to collect acoustic emission wave signals generated by crack propagation on the surface and inside the torsion shaft. Singular spectral decomposition is used to decompose the acoustic emission signals into components, and fault factors are injected based on the characteristics of different components. A torsion shaft health function is designed to accurately describe the changes in the degree of damage to the torsion shaft.

[0138] Experimental Example

[0139] Fatigue tests were conducted on the torsion shaft according to the technical plan, and energy data of acoustic emission signals emitted during the entire life cycle of the torsion shaft were obtained.

[0140] Because the fatigue test of the torsion shaft was lengthy, it was conducted in multiple stages, divided into 11 phases over a total of 7 hours and 20 minutes. The torsion shaft fractured at the exact midpoint of the defect, with a clean fracture surface. Figure 5 As shown. Observe the condition of the torsion shaft near the fracture point. Due to the poor heat dissipation conditions in the middle of the torsion shaft, the temperature accumulated too high during the experiment, and the local color changed significantly. The fracture of the torsion shaft at this point was greatly affected by temperature.

[0141] The experimental data for the torsion shaft are as follows: Figure 6 (a)- Figure 6 (k) Energy-time is shown.

[0142] The energy range is [0, 7 × 10⁻⁶]. 4 The unit is ms·mv. From Figure 5 The relationship between performance degradation and energy cannot be directly observed in the data, so further processing and analysis of the data are required.

[0143] 1. Performance degradation analysis of torsion shafts based on singular spectrum analysis and fatigue failure theory

[0144] 1) Data preprocessing

[0145] Depend on Figure 4 It can be seen that only a very small number of adjacent data have a time interval of less than 0.001s, therefore a fixed time interval of 10 is set. -3 s, 10 -2 s and 10 -1 Taking energy data as an example, the obtained local preprocessing results (0-50s) are as follows: Figure 7 , Figure 8 and Figure 9 As shown:

[0146] It can be seen that the preprocessing result obtained with a time interval of 0.01s is very similar to that obtained with 0.001s. However, 0.1s loses some data features compared to 0.001s and cannot reflect the changes in energy in a timely manner. Therefore, considering the computational load and whether to retain the original data attributes, 0.01s is selected as the preprocessing time interval.

[0147] 2) Singular spectrum analysis and determination of fatigue failure period

[0148] Singular spectrum analysis was performed on the eleven stages of the torsion shaft experiment data. λ was selected corresponding to the trend term, and the energy trend terms from the first to the eleventh stage all correspond to λ1.

[0149] Thus, the time series plot of the acoustic emission signal energy trend term is obtained as follows: Figure 10 (a)- Figure 10 (k):

[0150] Setting E th =5000msmv is the high-energy trend threshold value. If it was not in the fatigue failure stage before, the current energy trend value E t Higher than Eth And in the future t last =Within 1000s, E t Greater than E th Time exceeds t th1 =10s, then it is considered to have entered the fatigue failure stage. If it was previously in the fatigue failure stage, and the current energy trend value E t Below E th The duration is greater than t th2 =1330s, then the fatigue failure stage is considered to have ended. Therefore, four regions can be defined ( Figure 10 The colors (sky blue, deep blue, red, and black) correspond to the crack nucleation stage, microcrack propagation stage, macrocrack propagation stage, and instantaneous fracture stage in fatigue failure.

[0151] 2. Distance-based torsion shaft health function

[0152] The health function F(t) is set as a function of time t. Since there are defects in the settings of both experiments, the range of health variation is set to (0, 0.95), F(0) = 0.95. The smaller the health value, the higher the degree of performance degradation of the torsion shaft.

[0153] Since fatigue failure has been divided into four stages, the energy trend values ​​of the four stages are used to form a High set, and the remaining energy trend values ​​are used to form a Low set. A baseline value 'a' is found when E... t >a, health level fluctuates downwards, when E t <a, health level fluctuates upwards. According to formulas (9) and (11), we can obtain...

[0154] a = 1158.58, b = 3.68 × 10 -14 .

[0155] Therefore, the formula for the health function of the torsion shaft is:

[0156] F(t)=F(t-0.01)+3.68×10 -14 (E t (t)-1158.58) 2 sgn(1158.58-E t (t)), t∈All

[0157] The health status of the torque shaft can be calculated using the above formula, and the result is as follows: Figure 11 As shown.

[0158] Figure 11 The red dots in the graph represent the end times of the crack nucleation stage, the microcrack propagation stage, the macrocrack propagation stage, and the instantaneous fracture stage. The coordinates represent the time of the end time and the corresponding health level. Therefore, we can conclude that:

[0159] Under the fatigue failure test conditions of the torsion shaft, the health threshold at the end of the crack nucleation period is 0.60, the health threshold at the end of the microcrack propagation period is 0.57, and the health threshold at the end of the macrocrack propagation period is 0.093. When the health value is lower than 0.093, it is considered that the instantaneous fracture period is about to begin.

[0160] Therefore, it can be concluded that the performance degradation analysis method based on singular spectrum analysis and fatigue failure theory proposed in this patent can effectively extract the fault characteristics of the torsion shaft and determine the start time of each fatigue failure stage. The distance-based health function constructed based on the determination of fatigue failure stages can derive the health status of the torsion shaft at various moments. The trend of health status change conforms to fatigue failure theory, verifying the effectiveness of the method.

[0161] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the health of a torsion shaft based on fiber optic loop acoustic emission, characterized in that, This includes using acoustic emission sensors to collect energy data of acoustic emission signals during the fatigue failure process of a torsion shaft, performing performance degradation analysis of the torsion shaft based on singular spectrum analysis, and determining the health status of the torsion shaft throughout its entire life cycle based on a distance-based health function, based on the performance degradation analysis. The method for determining the health status of a torque shaft throughout its entire life cycle based on a distance-based health function is as follows: determine the health baseline value, determine the slope coefficient of the health function, construct the health function, obtain the health function value at each time point, and determine the health status of the torque shaft throughout its entire life cycle through the health function value at each time point. The method for determining health baseline values ​​is as follows: Based on the determination of the time interval where the fatigue failure stage occurs, all acoustic emission energy trend values ​​E are... t The set is divided into two sets: the High set and the Low set; the High set corresponds to the E value of the fatigue failure stage being determined. t The Low set corresponds to E outside the High set. t Because E in the Low set t The value is low, indicating a stationary sequence with a slow crack propagation rate. Therefore, a baseline value 'a' is derived using the Low set, assuming that when E... t When E is greater than a, the health of the torque shaft fluctuates downwards. t When the value is less than 'a', the health of the torsion shaft fluctuates upwards. To highlight E t The impact of deviation from the reference value 'a' on the health of the torque shaft, let (E) t -a) 2 sgn(aE t The slope of the change in health is proportional to the energy level. Since the low-energy trend segment corresponds to a stationary health level sequence, the baseline value 'a' satisfies: ∑ ( E t (t)-a) 2 sgn(a-E t (t))=0,t∈Low (9) This yields the baseline value a; The method for determining the slope coefficient of the health function is as follows: After obtaining the baseline value a, due to (E t -a) 2 sgn(aE t The slope of the health function is proportional to the slope of the change in health, therefore the slope k of the health function is: k=b(E t (t)-a) 2 sgn(a-E t (t)),t∈All (10) The method for constructing the health function is as follows: Let the health of the torque shaft be F. Since the torque shaft eventually breaks, its health becomes 0. Therefore: F(0)=∑b(E t (t)-a) 2 sgn(a-E t (t)) (11) Since F(0) is known, b can be solved, so the health function F(t) has the following relationship: F(t)=F(t-Δt)+b(E t (t)-a) 2 sgn(a-E t (t)),t∈All (12) Where Δt is the sampling interval after data preprocessing.

2. The method for assessing the health of a torsion shaft based on fiber optic loop acoustic emission according to claim 1, characterized in that, The performance degradation analysis of torsion shafts based on singular spectrum analysis includes: preprocessing energy data to determine the characteristic values ​​corresponding to different components, decomposing singular spectrum components to extract fault characteristics, and determining the fatigue failure period.

3. The method for assessing the health of a torsion shaft based on fiber optic loop acoustic emission according to claim 2, characterized in that, The method for preprocessing energy data is as follows: The energy data is preprocessed to remove redundant data and fill in missing data. The rule for filling in missing data is to set a fixed time interval of 10... -n s, round the time term in the data to n decimal places; then fill in the missing data between adjacent data points, using the energy value from the previous time step; if the time difference between two adjacent data points is less than 10 - n If the time values ​​are the same after rounding, then the two time values ​​are combined, and the energy value is taken as the larger of the two values.

4. The method for assessing the health of a torsion shaft based on fiber optic loop acoustic emission according to claim 3, characterized in that, The method for decomposing singular spectral components includes the following steps: (1) Embedding: The analysis object of singular spectral components is a finite-length one-dimensional time series [x1, x2, ..., x...]. N ], N is the sequence length; first, a suitable window length L needs to be selected to lag the original time series to obtain the trajectory matrix: Pick Let K = N - L + 1, then the trajectory matrix X is an L × K matrix: (2) Decomposition: Perform singular value decomposition on the trajectory matrix, that is, decompose X into the following form: X=UΣV T (3) Where U is the left matrix; Σ is the diagonal matrix, and the values ​​on the main diagonal are the singular values; V is the right matrix; U and V are both identity orthogonal matrices, satisfying UU T =I,VV T =I; First, calculate the covariance matrix of the trajectory matrix: S=XX T (4) Eigenvalue decomposition of S yields eigenvalues ​​λ1>λ2>...>λ L ≥0 and the corresponding eigenvectors U1, U2, ..., U L , at this time U=[U1,U2,...,U L ], The singular spectrum of the original sequence; and we have: Where, λ i The corresponding eigenvector U i It reflects the evolution of time series and is called the time empirical orthogonal function; (3) Grouping: Calculate the hysteresis sequence X i in U m Projection on: X i This represents the i-th column of the trajectory matrix X. It is X i The time evolution pattern reflected in the original sequence x i+1 ,x i+2 ,...,x i+L The weights of different time periods are called temporal principal components; The resulting matrix is ​​an unnormalized right matrix, i.e. (4) Reconstruction: Reconstruction is performed using time-based empirical orthogonal functions and time-based principal components. The specific reconstruction process is as follows: in, It is x i-j The time evolution pattern reflected in the original sequence x i+1 ,x i+2 ,...,x i+L Time period weight, U k,j It is λ i Corresponding eigenvector U i The value of the j-th row; Thus, the sum of all reconstructed sequences should equal the original sequence, that is:

5. The method for assessing the health of a torsion shaft based on fiber optic loop acoustic emission according to claim 4, characterized in that, The method for determining the fatigue failure period is as follows: In the four stages of crack nucleation, microcrack propagation, macrocrack propagation, and instantaneous fracture, the torsion shaft exhibits a relatively rapid crack initiation rate and a high frequency of high-energy acoustic emission signals. Conversely, between two adjacent fatigue failure stages, crack propagation is relatively slow, and the energy value of acoustic emission is relatively low. Based on this characteristic, the following rules are used for judgment: Entering the fatigue failure stage: Previously not in the fatigue failure stage, current energy trend value E t The energy threshold for acoustic emission is higher than E. th And in the future t last Within a time period, E t Greater than E th Time exceeds t th1 ; End of fatigue failure phase: Previously in the fatigue failure phase, and the current energy trend value E t Below the high energy threshold value E for acoustic emission th The duration is greater than t th2 ; This allows us to determine the time intervals for the four fatigue failure stages.

Images