A data acquisition and analysis method and system for rudder stock mechanism pull-off test

Abstract

The application discloses a kind of rudder stock mechanism pull-off test data acquisition and analysis method and system, it is related to data acquisition analysis technical field, including the load of acquisition pull-off test whole process, displacement and strain data, form multi-source synchronous data stream;Real-time monitoring is judged based on load drop slope and fracture event occurs and records fracture time;Intercept fracture window data section, extract strain transient sequence by low-pass filtering and wavelet decomposition;Identify strain rate rapid increase point, strain rebound starting point and strain zero point, divide crack initiation stage, rapid expansion and penetration stage and post-fracture residual stage, extract dynamic characteristics and construct fracture dynamics characteristic vector;Estimate crack propagation speed by combining load displacement curve and high-speed video frame sequence, generate dynamic mechanics parameter set;Mark to corresponding video frame and output visual analysis report.The application has the effect of realizing the accurate time synchronization of mechanics waveform data and high-speed camera image.

Description

Technical Field

[0001] This application relates to the field of data acquisition and analysis technology, and in particular to a data acquisition and analysis method and system for a rudder lever mechanism breakage test. Background Technology

[0002] As a key load-bearing component of a ship's steering mechanism, the rudder stock's ultimate load-bearing capacity and fracture behavior directly determine the ship's maneuverability in extreme sea conditions, especially with the increasing size of ships and the frequency of extreme sea states. Tensile testing, which applies a monotonically increasing axial tensile force to the rudder stock until fracture, obtains mechanical data such as load-displacement and stress-strain, serving as a fundamental method for evaluating material properties and verifying design safety margins. In rudder stock testing for large ships, in addition to obtaining the ultimate load and total elongation, it is also necessary to capture the high-frequency dynamic response in the final moments before fracture to assess strain rate sensitivity, fracture toughness, and rapid crack propagation behavior. Current industry standard procedures typically employ load, displacement, and strain gauge sensors with sampling frequencies set between 1 kHz and 10 kHz, combined with high-speed video recording of the fracture process.

[0003] Note that the time from crack initiation to through-fracture in a rudder stock typically lasts from tens to hundreds of milliseconds. In ductile materials, significant necking and localized strain concentration occur before fracture, with the strain rate increasing exponentially during this stage, accompanied by millisecond-level stress fluctuations and strain rebound. These high-frequency transient signals contain crucial information for assessing dynamic fracture toughness, such as peak strain rate, strain rebound amplitude, and critical indicators of elastic-to-plastic instability. However, practical testing is limited by storage bandwidth and sampling depth, and traditional methods often treat high-frequency fluctuations at the moment of fracture as noise, filtering or discarding them. Even with complete recording, there is a lack of algorithms to automatically identify and extract millisecond-level strain fluctuation patterns. Furthermore, the clocks of high-speed cameras and data acquisition systems are independent, resulting in a misalignment of several milliseconds to tens of milliseconds between the temporal location of the fracture event on the load curve and the image, making it difficult to accurately correlate strain data with fracture morphology.

[0004] Current data processing solutions for rudder stock fracture tests generally focus on quasi-static stress-strain curves and fracture morphology, lacking a systematic design for the acquisition, synchronization, and feature mining of millisecond-level transient data before fracture. This technological gap has resulted in the long-term absence of key data dimensions regarding the dynamic response of materials at the moment of fracture. Testers are unable to answer the core question of what exactly happened to the material in the last few milliseconds before fracture, thus hindering the refined evaluation of the ultimate performance of rudder stock materials. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides a data acquisition and analysis method and system for rudder lever mechanism tensile testing.

[0006] Firstly, this application provides a data acquisition and analysis method for a rudder lever mechanism tensile test, comprising the following steps: The load, displacement, and strain data of the rudder lever mechanism were collected throughout the tensile test, and a multi-source synchronous data stream was formed based on the load, displacement, and strain data. The load data descent slope is determined based on the load data in the multi-source synchronous data stream. The load data descent slope is compared with a preset trigger threshold. When the load data descent slope exceeds the preset trigger threshold, a fracture event is determined to have occurred and is recorded as the fracture time. Based on the fracture moment, the fracture window data segment is identified from the multi-source synchronous data stream, and fracture window data is generated based on the fracture window data segment; the fracture window data is processed by low-pass filtering and wavelet decomposition to generate the strain transient sequence. Stage boundary point identification is performed based on the strain transient sequence to obtain a boundary point set, which includes the strain rate surge point, strain rebound initiation point, and strain zeroing point. Based on the boundary point set, the crack initiation stage, rapid propagation and penetration stage, and post-fracture residual stage are divided, and dynamic feature parameters are extracted from the crack initiation stage, rapid propagation and penetration stage, and post-fracture residual stage. Based on the extraction results, a fracture dynamic feature vector is constructed. The load and displacement data in the multi-source synchronous data stream are analyzed together to obtain the load-displacement curve. The fracture dynamics feature vector and the load-displacement curve are analyzed together, and the crack propagation rate is estimated by combining the high-speed camera video frame sequence. A dynamic mechanical parameter set is generated based on the crack propagation rate. The dynamic mechanical parameter set is labeled to the corresponding high-speed camera video frame sequence, and a full-process fracture visualization analysis report is output.

[0007] Secondly, this application provides a data acquisition and analysis system for a rudder lever mechanism tensile test, including: The data acquisition module is used to collect load data, displacement data and strain data throughout the entire tensile test of the rudder mechanism, and to form a multi-source synchronous data stream based on the load data, displacement data and strain data; The fracture event triggering module is used to identify the load data descent slope from the load data in the multi-source synchronous data stream, compare the load data descent slope with a preset trigger threshold, and determine that a fracture event has occurred when the load data descent slope exceeds the preset trigger threshold, and record it as the fracture time. The data processing and transformation module is used to identify the fracture window data segment from the multi-source synchronous data stream, generate fracture window data based on the fracture window data segment, and perform low-pass filtering and wavelet decomposition on the fracture window data to generate strain transient sequence. The feature extraction module is used to identify stage boundary points in the strain transient sequence to obtain a set of boundary points, which includes the strain rate surge point, the strain rebound initiation point, and the strain zero point. Based on the set of boundary points, the module divides the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage, and extracts dynamic feature parameters for the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage. Based on the extraction results, a fracture dynamic feature vector is constructed. The joint analysis module is used to analyze the joint relationship between load data and displacement data in multi-source synchronous data streams to obtain load-displacement curves; it performs joint analysis on fracture dynamics feature vectors and load-displacement curves, and estimates crack propagation velocity by combining high-speed camera video frame sequences, and generates a dynamic mechanical parameter set based on crack propagation velocity; The analysis report generation module is used to annotate the dynamic mechanical parameter set to the corresponding high-speed camera video frame sequence and output a full-process visual analysis report of the fracture.

[0008] In summary, this application includes at least one of the following beneficial technical effects: 1. This application constructs a multi-source synchronous data stream of load, displacement, and strain. It also designs a preprocessing mechanism combining low-pass filtering and wavelet decomposition to address the fracture transient signal, which is traditionally considered noise filtering. This effectively solves the technical gaps in the prior art, such as the misfiltering of high-frequency fluctuations during fracture and the lack of an automatic millisecond-level strain fluctuation pattern identification algorithm. Within the fracture window, this application can extract the true strain transient sequence from a strong noise background and automatically identify the strain rate surge point, strain rebound initiation point, and strain zeroing point, objectively dividing the fracture process into the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage. This automated feature extraction scheme enables the quantification of key dynamic indicators such as peak strain rate, strain rebound amplitude, and critical necking strain rate, which were previously difficult to capture. This fills the long-standing gap in the dynamic response data of rudder stock materials in the last few milliseconds before fracture, providing reliable data support for material ultimate performance evaluation and fracture toughness analysis.

[0009] 2. Addressing the inherent problem mentioned in the background technology—that the data acquisition system and the high-speed camera clock are independent, resulting in a misalignment of several milliseconds to tens of milliseconds—this application uses the fracture moment triggered by real-time monitoring of the load data descent slope as a unified time reference anchor point. Time offset correction is applied to the multi-source synchronous data stream and the high-speed camera video frame sequence, thereby eliminating clock deviation between the two independent hardware systems. Based on this, this application correlates and maps the fracture dynamics feature vector with the load-displacement curve, verifies the mechanical consistency of the stage division, and estimates the crack propagation rate by combining the aligned video frame sequence. Finally, a dynamic mechanical parameter set is generated and labeled to the corresponding image frames. This time synchronization and joint analysis mechanism ensures that the transient characteristics of strain, load fluctuation nodes, and crack morphology evolution achieve precise frame-by-frame correspondence on the time axis, enabling testers to clearly trace the causal relationship between mechanical parameters and physical morphology at each moment. This fundamentally solves the technical problem of the difficulty in accurately verifying mechanical data and image data in traditional testing. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating the data acquisition and analysis method for a rudder lever mechanism tensile test implemented in this application.

[0012] Figure 2 This is a schematic diagram of the data acquisition and fracture event triggering logic flow of this application.

[0013] Figure 3 This is a schematic diagram of the logical flow of fracture boundary identification and dynamic feature vector construction in this application.

[0014] Figure 4 This is a schematic diagram of the overall structure of a rudder lever mechanism breakage test system according to this application.

[0015] Figure 5 This is a schematic diagram of a system module for data acquisition and analysis of a rudder lever mechanism for tensile testing, as described in this application. Detailed Implementation

[0016] The following description, in conjunction with the implementation of the present invention, is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention, and all such modifications and additions should fall within the protection scope of the present invention.

[0017] Example 1

[0018] This application is applicable to the fracture behavior testing and dynamic response analysis of ship rudders, drive shafts, and large metal structural components under axial tensile loads. It is particularly suitable for tensile testing scenarios that require simultaneous acquisition of millisecond-level high-frequency transient signals and load-displacement curves before fracture, and precise time alignment of mechanical data with high-speed camera images. The core idea is to synchronously acquire load, displacement, and strain data throughout the test to form a multi-source synchronous data stream. First, the fracture event is triggered and the fracture moment is locked by real-time dynamic monitoring of the load descent slope. After locking the fracture moment, a fracture window data segment is extracted from the multi-source data stream. The strain transient sequence is extracted through low-pass filtering and wavelet decomposition, and then the points of rapid strain rate increase, strain rebound initiation, and strain zeroing are identified. The fracture process is automatically divided into crack initiation, rapid propagation and penetration, and post-fracture residual stages. Dynamic characteristic parameters for each stage are extracted to construct a fracture dynamics feature vector. Simultaneously, the load-displacement curve and fracture dynamics feature vector are jointly analyzed, and the crack propagation rate is estimated by combining high-speed camera video frame sequences. A dynamic mechanical parameter set is generated and labeled to the corresponding image frames, ultimately outputting a full-process visual analysis report of the fracture. After this processing, even if the fracture transient signal lasts only tens of milliseconds and the data acquisition system is independent of the high-speed camera clock, it can still achieve precise time alignment between the mechanical waveform and the image frame, and automatically extract the high-frequency strain fluctuation mode that is considered noise filtering by traditional methods, effectively solving the problem of long-term lack of data dimensions of material dynamic response at the moment of fracture.

[0019] This application discloses a data acquisition and analysis method for a rudder lever mechanism pull-off test.

[0020] Reference Figure 1-4 A data acquisition and analysis method for a rudder lever mechanism tensile test includes the following steps: Load, displacement, and strain data were collected throughout the tensile test of the rudder lever mechanism, and a multi-source synchronous data stream was formed based on the load, displacement, and strain data.

[0021] Furthermore, continuous signal acquisition was performed throughout the tensile test to obtain the original load simulation signal, original displacement simulation signal, and original strain simulation signal throughout the tensile test. The original load simulation signal, the original displacement simulation signal, and the original strain simulation signal are subjected to signal conditioning and filtering to obtain the conditioned and filtered simulation signals. The analog signals after conditioning and filtering are converted from analog to digital to generate load data, displacement data and strain data; Based on a unified time reference, load data, displacement data, and strain data are time-aligned and synchronously fused to form a multi-source synchronous data stream.

[0022] In one specific implementation, the data acquisition of the rudder stock fracture test involves three types of physical quantities: load, displacement, and strain. These three types of physical quantities reflect the stress state, deformation degree, and local strain response of the rudder stock under axial tensile force, respectively, and together constitute a complete data foundation for describing the mechanical behavior of the entire fracture process.

[0023] Load data is acquired using a load sensor installed between the loading device and the rudder stock specimen. The load sensor works by converting the axial tensile force F applied to the rudder stock into a proportional voltage or current signal output. The amplitude of the output signal changes linearly with the magnitude of the applied load F. During the test, the load sensor continuously outputs a continuous analog electrical signal corresponding to the current axial tensile force, i.e., the original load analog signal. Since the ultimate load of large rudder stocks can reach several meganewtons, the sensor range selection must be based on the estimated peak load of the rudder stock being tested. Matching is used to ensure that the sensor output does not saturate or distort during the ultimate load phase before fracture. The load sensor operates at a sampling frequency... Perform analog-to-digital conversion. With a sampling rate of no less than 10 kHz, which corresponds to a time resolution of 0.1 ms, the accuracy of the location at the moment of fracture meets the requirements of millisecond-level transient analysis. This sampling rate matches the hardware acquisition capability of mainstream dynamic mechanical testing equipment and meets the time resolution requirements of the load drop signal at the moment of fracture.

[0024] Displacement data is acquired using displacement sensors installed between the clamping points at both ends of the tensile testing machine beam or rudder stock specimen. The displacement sensors continuously monitor the change in distance between the clamping points under axial tensile force, converting this change into a continuous, raw displacement analog signal output. The displacement data records the elongation change of the rudder stock throughout its entire lifecycle, from elastic deformation and plastic deformation to fracture separation, providing direct evidence for constructing load-displacement curves and calculating the material's plastic work and elastic strain energy. The displacement sensors operate at a sampling frequency... Perform analog-to-digital conversion. The time resolution is no less than 10 kHz, maintaining the same time resolution as the load sensor, to ensure that the load data and displacement data correspond point by point on the time axis, supporting the accurate construction of subsequent load-displacement curves.

[0025] Strain data were acquired using strain gauges attached to key sections of the rudder stock specimen. The strain gauges work by utilizing the physical effect of the change in resistance of a metal resistance wire under mechanical deformation, which is reflected in the strain. Relative change in resistance The following conditions must be met: ; in The strain gauge's sensitivity coefficient is typically between 1.9 and 2.1. The strain gauge is connected to a Wheatstone bridge circuit, and the bridge output voltage... With strain The relationship is: ; in The excitation voltage is used for the bridge circuit. This converts the minute linear strain ε of the local cross-section of the rudder stock into an acquireable voltage output signal, i.e., the original strain simulation signal. Compared to load and displacement signals, the strain signal is more sensitive to local deformation and can capture transient characteristics such as the millisecond-level increase in strain rate and strain rebound before fracture. The strain gauge is sampled at a frequency... Perform analog-to-digital conversion. Not less than 100 kHz, significantly higher and To ensure that millisecond-level transient features are not affected by the sampling interval Too large to be overlooked.

[0026] The three types of sensors mentioned above output continuously changing analog electrical signals during testing. However, computer data acquisition systems can only process discrete digital signals. Therefore, it is necessary to convert the analog signals into storable and computable digital data through two stages: signal conditioning and analog-to-digital conversion. Signal conditioning is a preprocessing stage before analog-to-digital conversion, and its purpose is to improve the quality of the analog signal so that it meets the input requirements of subsequent analog-to-digital conversion. During the transmission of the raw analog signal output by the sensor to the data acquisition card via wires, it inevitably gets mixed with power frequency interference and high-frequency electrical noise from surrounding electrical equipment. If analog-to-digital conversion is performed directly without processing, these noise components will be digitized in the same way, mixing into the load data, displacement data, and strain data, reducing the signal-to-noise ratio, and introducing spurious fluctuations in subsequent feature extraction. Signal conditioning and filtering processes apply low-pass or band-pass filters to the raw load analog signal, raw displacement analog signal, and raw strain analog signal respectively in the analog domain. This attenuates noise components above the effective signal frequency band before entering the analog-to-digital converter, resulting in conditioned and filtered analog signals, thus ensuring data quality from the source.

[0027] Analog-to-digital conversion (ADC) is the core step in converting a conditioned and filtered continuous analog signal into a discrete digital signal. ADCs operate at fixed sampling intervals. The analog signal is discretely sampled, and at each sampling moment, the amplitude of the current analog signal is read and quantized into a corresponding digital code for storage. This transforms a temporally continuous analog signal into a temporally discrete digital sequence. Sampling frequency. The selection is based on the Nyquist sampling theorem: ; in The highest effective frequency component in the target signal is selected to ensure that the digitized signal can completely reproduce the frequency content of the original analog signal without aliasing distortion. After analog-to-digital conversion, signals are generated at sampling intervals... Discrete load data sequence with time step Displacement data sequence With strain data sequence .

[0028] After completing the analog-to-digital conversion of the three signals respectively, Each of the three sequences forms an independent digital time series, with different sampling frequencies, and each sensor trigger pulse has an inherent delay in arrival time. Without correction, the three digital sequences will deviate on the time axis. In the transient analysis at the millisecond level before fracture, even a time deviation of a few milliseconds will cause a significant misalignment in the correspondence between the load peak and the point of rapid strain increase, directly affecting the accuracy of feature parameter extraction.

[0029] To eliminate the aforementioned time deviation, the master clock of the data acquisition controller is used as the sole time reference, and the original sampling time carried by each data frame is corrected using the following formula: ; in The raw sampling times recorded by each sensor. This represents the delay of the trigger pulse of the i-th sensor relative to the master clock reference, recorded by the master clock reference during synchronous triggering. After time correction, The data are merged and arranged in ascending order according to the corrected timestamps to form a multi-source synchronous data stream in which the three physical quantities are strictly synchronized in time, for continuous reading by the subsequent fracture event detection and data interception modules.

[0030] The load data descent slope is determined based on the load data in the multi-source synchronous data stream. The load data descent slope is compared with a preset trigger threshold. When the load data descent slope exceeds the preset trigger threshold, a fracture event is determined to have occurred and is recorded as the fracture time.

[0031] Furthermore, load data is extracted from the multi-source synchronous data stream, and trend analysis is performed on the load data to obtain information on the load change trend over time. Based on the aforementioned trend information, the rate of change of the load in the downward direction is extracted as an initial estimate of the downward slope of the load data. The initial estimate of the load data descent slope is subjected to stability judgment, and the load data descent slope is confirmed based on the stability judgment result; The load data descent slope is dynamically compared with a preset trigger threshold. When the load data descent slope exceeds the preset trigger threshold and reaches a preset duration, it is determined that a valid fracture event has occurred, and the current time is recorded as the fracture time.

[0032] In one specific implementation, the accurate detection of the fracture event serves as the time reference for all subsequent analytical work; the fracture time... The accuracy of the recording directly determines the quality of the fracture window data extraction. If the fracture time is recorded too early, the fracture window will include too much data from the loading plateau phase, which is irrelevant to the fracture, diluting key transient features. If the fracture time is recorded too late, the early strain rate changes during crack initiation will be truncated, leading to inaccurate identification of stage boundary points. Therefore, it is necessary to design a system that can automatically identify and record the fracture time with the shortest possible delay after fracture occurs. The detection mechanism.

[0033] It is particularly important to note that ship rudder stock is a large, critical load-bearing component, typically with a nominal diameter of 200 mm to 400 mm and an ultimate load capacity reaching the meganewton level. Its cross-sectional dimensions and load magnitude far exceed those of conventional standard tensile specimens. Given this large size, the constraint state, stress triaxiality, and elastic strain energy release magnitude of the crack tip region before fracture exhibit fundamental scale-effect differences compared to conventional material specimens. This difference directly leads to: the structural resonance frequency excited at the moment of fracture in a large rudder stock is much lower than that in a small specimen; high-pass filtering parameters conventionally used for small specimens will misinterpret the actual millisecond-level strain fluctuations as equipment oscillation noise and filter them out directly in this scenario; the strong vibration impact generated by the meganewton-level load drop will severely interfere with the charge amplifiers of strain gauges and extensometers, making the fixed trigger thresholds commonly used in small specimen testing easily fail or trigger falsely on large rudder stock due to fluctuations in the noise floor amplitude. Therefore, the data acquisition parameter settings and transient extraction strategies of conventional material fracture testing cannot be directly applied to large rudder stock mechanisms. A synchronous acquisition and feature identification scheme must be specifically designed for its unique mechanical response mechanism under large size and heavy load.

[0034] Of all observable physical quantities, load data is the most direct signal reflecting the occurrence of a fracture event. In a tensile fracture test, the load values ​​of the rudder stock throughout the entire process from initial loading to fracture are... It exhibits an overall monotonically increasing trend, with only minor fluctuations caused by localized material yielding; however, at the instant of fracture, the load-bearing section loses its load-bearing capacity, and the axial load drops from its peak value in an extremely short time. The load drops sharply, with a descent rate far exceeding any possible load fluctuation during normal loading. This physical characteristic makes the load descent slope the most direct and reliable feature for determining the occurrence of a fracture event. Moreover, compared to strain and displacement data, changes in the load signal respond to fracture events more promptly, making it suitable as the primary basis for triggering fracture event detection.

[0035] Continuously extract payload data sequences from multi-source synchronous data streams. The payload data is scanned in real time using a sliding time window of length W, where W is 5 milliseconds. Within each sliding window, adjacent timestamps are scanned. and Load difference between Divide by the corresponding time interval Calculate the load descent rate at the current moment frame by frame: ; The negative sign indicates that the load is taken as positive in the direction of descent. >0 indicates that the load is decreasing. A larger value indicates a more rapid decrease. The maximum value of the load decrease rate for each frame within the sliding window is taken. This serves as an initial estimate of the descent slope of the load data at the current moment.

[0036] Due to slight jitter in the sensor signal, single point Instantaneous spikes may occur due to signal glitches. If a break is determined solely by a single point slope exceeding a threshold, it carries the risk of false triggering. Therefore, the signal strength calculated frame by frame... To determine trend stability: Requirements In continuous A valid breakage trigger signal can only be confirmed if all sampling points continuously exceed the trigger threshold θ. Take 5 sampling points, corresponding to a duration of .when At this time, the stability discrimination window is 0.5 milliseconds, which is sufficient to filter out single-point anomalies caused by spikes without introducing excessive detection delay.

[0037] Setting a reasonable trigger threshold θ is the core trade-off between fracture detection sensitivity and resistance to false triggering. If the trigger threshold θ is too low, transient load fluctuations caused by localized yielding or stress relaxation during loading will be misjudged as fracture events. If the trigger threshold θ is too high, a longer time will be required to trigger detection after a fracture event, resulting in a later recording of the fracture moment and omitting transient data from the initial stage of fracture. Due to the significant differences in the ultimate load of rudder stock of different sizes and materials—the ultimate load of small rudder stocks may only be tens of kilonewtons, while the ultimate load of large ship rudder stocks with diameters ranging from 200 mm to 400 mm can reach meganewtons—a fixed trigger threshold cannot accommodate such a wide range of load magnitudes. Therefore, before each tensile fracture test, the ultimate load-bearing capacity of the rudder stock under test is estimated by finite element simulation calculations or by referencing historical test data of rudder stocks of the same type, diameter, and material, to obtain the estimated peak load. The trigger threshold θ is determined by the following formula: ; in This is a proportionality coefficient, ranging from 0.30 to 0.50, which is based on the estimated peak load. The load drop per unit sampling interval corresponding to 30% to 50% is used as the trigger threshold, so that θ varies with the load magnitude of the measured rudder stock. Adaptive adjustment. (Take...) The basis for the range of 0.30 to 0.50 is: When the threshold is below 0.30, the load drop corresponding to the trigger threshold is within the normal yield fluctuation range at the end of the loading of a large rudder stock, which is prone to false triggering. When the value is above 0.50, detection can only be triggered after the fracture has become quite severe and the load has dropped significantly, resulting in the loss of critical transient data from the early stages of the fracture.

[0038] When comparing the load data descent slope with the trigger threshold, the calculation of the load data descent slope is also based on the sampling frequency. The denominator is the sampling frequency. After simplifying both sides of the inequality, they cancel each other out. Therefore, the essential criterion for determining the trigger threshold is equivalent to whether the decrease in load amplitude between adjacent sampling points exceeds the estimated peak load. times; sampling frequency It is only used to convert the load amplitude reduction dimension to the load reduction rate dimension, and does not have any additional impact on the physical sensitivity of fracture event detection.

[0039] The slope of the load data decrease Continuously compare with the trigger threshold θ, when Exceeding θ and continuous All sampling points satisfy When the value is greater than θ, it is considered a valid fracture event, and the first sampling point that satisfies the stability criterion is recorded as the fracture time. and with The system sends a capture command to the fracture window data capture module based on the time reference, initiating the backtracking capture and forward extension capture process of the fracture window data.

[0040] Based on the fracture moment, the fracture window data segment is identified from the multi-source synchronous data stream, and fracture window data is generated based on the fracture window data segment. The fracture window data is then subjected to low-pass filtering and wavelet decomposition to generate a strain transient sequence.

[0041] Furthermore, using the break point as the time center, candidate data intervals containing the break point are extracted along the time axis from the multi-source synchronous data stream. Data integrity verification and outlier removal are performed based on candidate data intervals, and the candidate data intervals after data integrity verification and outlier removal are determined as the broken window data segments. The candidate data range after data integrity verification and outlier identification is determined as the broken window data segment, and the multi-source synchronous data streams within the broken window data segment are arranged in the original time order to generate broken window data.

[0042] Furthermore, strain data is extracted from the fracture window data, and low-pass filtering is performed on the strain data to obtain filtered strain data. Based on the wavelet decomposition processing of the filtered strain data, the filtered strain data is decomposed into wavelet coefficients of multiple scales based on the results of the wavelet decomposition processing. Denoising is performed on the wavelet coefficients at each scale to obtain the effective coefficients after denoising. Wavelet reconstruction is performed on the denoised effective coefficients to generate a strain transient sequence. In one specific implementation, extracting the fracture window data is a necessary step to focus massive amounts of full-process data onto key periods of the fracture process. The entire duration of a tensile test typically ranges from several minutes to tens of minutes, while the time period truly containing fracture dynamic characteristics is concentrated within a few hundred milliseconds before and after the fracture moment. If feature extraction is performed directly from the full-process data, not only is the computational load enormous, but a large amount of data from the loading plateau segment, which is unrelated to the fracture, will act as background interference, affecting the accuracy of the identification of stage boundary points. Therefore, it is necessary to extract features from the fracture moment... Using time reference, a data segment containing only complete dynamic information of the fracture process is precisely extracted from the multi-source synchronous data stream as the data basis for all subsequent analyses.

[0043] At the time of fracture Centered on time, each segment traces back in time towards the past. Extending backwards in the direction of the future The truncation range of the broken window data segment is determined as follows: Forward backtracking duration The settings need to fully cover the early characteristics of the crack initiation stage. Crack initiation typically occurs hundreds of milliseconds before the peak load is reached. An excessively short forward window will cause the point of rapid strain rate increase during the initiation stage to fall outside the cutoff range, rendering the identification of boundary points in subsequent stages ineffective; the backward extension time... The settings need to be able to record the release of elastic strain energy and the oscillation of residual stress after penetrating fracture. This process typically completes within tens to two hundred milliseconds after fracture. An excessively short backward window will truncate the strain to zero, resulting in incomplete extraction of characteristic parameters during the penetrating fracture stage. Considering the above factors, Set to 800 milliseconds. The total duration of the data segment in the broken window is set to 200 milliseconds. At the time of fracture It is an asymmetric center. The design basis for the asymmetry before and after is that the dynamic history before fracture is much richer than the residual response after fracture, and a longer time window is needed to fully cover the evolution process from stable loading to crack initiation and rapid propagation, while the data after fracture is only used to capture the strain returning to zero, and the required time is relatively short.

[0044] After determining the truncation range, the time range is retrieved and extracted from the circular buffer of the multi-source synchronized data stream by timestamp. All data frames within the range. After extracting the data frames, the candidate data range needs to be checked for data integrity to see if there are any outliers such as discontinuous timestamps, missing data frames, or amplitudes exceeding the physical reasonable range. The detected outlier data frames are marked and replaced with linear interpolation of adjacent normal frames to ensure the integrity and continuity of the broken window data. Finally, the data segments that have been checked and processed are determined as the broken window data.

[0045] After obtaining the fracture window data, it is necessary to extract the strain transient sequence that can truly reflect the fracture dynamics characteristics. Let the sampling time of the i-th strain sampling point in the fracture window data be denoted as . The corresponding original strain sampling value is The sampling interval is ,in The preset sampling frequency of the strain gauge. The sampling frequency should be no less than 100 kHz. At such a high sampling frequency, high-frequency random noise and residual power frequency interference from the circuit system will be superimposed on the actual strain transient signal, forming the original strain sequence. If directly to When performing first-order differencing to calculate strain rate, high-frequency fluctuations in noise are amplified by the differencing operation, obscuring the true characteristics of rapid strain rate increases and leading to severe inaccuracies in the identification of stage boundary points. Therefore, before performing any feature extraction, the strain data in the fracture window data must first be denoised to generate a transient strain sequence that can truly reflect fracture dynamics.

[0046] The denoising process consists of two sequential steps: low-pass filtering and wavelet decomposition. Each step targets different noise components. Low-pass filtering eliminates broadband random noise with frequencies higher than the effective strain signal frequency band. Specifically, an Nth-order low-pass filter is selected, with the following amplitude-frequency response: ; in Angular frequency, Here, ω is the cutoff angular frequency, and N is the filter order, which is set to N = 4. The filter's amplitude-frequency response within the passband is extremely flat, and it will not distort the amplitude of the strain signal in the effective frequency band, making it suitable for strain data processing where high amplitude accuracy is required. Cutoff frequency Determined based on the highest effective frequency component of the millisecond-level transient characteristics in the strain signal, and set as follows: The scaling factor k is between 10 and 5, meaning the cutoff frequency is 1 / 10 to 1 / 5 of the preset sampling frequency. This preserves the effective frequency components corresponding to millisecond-level transient changes while effectively attenuating noise above this frequency band.

[0047] The filtering method employs bidirectional zero-phase filtering, i.e., filtering the original strain sequence... First, perform forward filtering to obtain the intermediate sequence. And then After time axis reversal, filtering, and reversal again, a smooth strain sequence is finally obtained. The reason for using a bidirectional zero-phase method is that unidirectional filtering introduces phase delay in the filter's transition band. This causes the filtered signal to shift entirely on the time axis, resulting in a shift in the time position of boundary features such as the point of rapid increase in strain rate and the point of strain rebound in the amount of: ; in Is to take The phase value at the specified point is used in the calculation, and the two phase delays of the bidirectional filter are in opposite directions and cancel each other out, ultimately achieving a zero-phase output and ensuring a smooth strain sequence. The time position of each feature point is strictly consistent with the original signal.

[0048] However, while low-pass filters remove high-frequency noise, their cutoff frequency... The above frequency components will be filtered out. If certain millisecond-level transient changes during the fracture process happen to contain... Nearby frequency components may be attenuated by low-pass filtering, causing distortion of characteristic amplitudes. To preserve the transient details of the fracture process as much as possible while denoising, the strain sequence is further smoothed. Perform wavelet decomposition. A db4 wavelet basis was used to perform J = 5-level discrete wavelet decomposition to obtain the detail coefficients of each level. (j = 1, 2, …, 5) and the low-frequency approximation coefficients of the 5th layer , where k is the discrete index of each layer coefficient. The frequency range corresponding to the detail coefficients of the j-th layer is: ; The detail coefficients of layers 1 to 3 correspond to higher frequency scales where millisecond-level transient features reside, and are therefore fully preserved. The detail coefficients of layers 4 and 5 mainly consist of residual low-frequency drift and noise components at their respective frequency scales, and are subject to a soft threshold λ. ; The soft threshold λ is based on the general threshold criterion as follows: , is the noise standard deviation estimate, and M is the total number of coefficients in the current layer.

[0049] The detail coefficients for layers 1 through 3 remain unchanged, using all effective coefficients. Perform wavelet reconstruction to restore the strain transient sequence that removes noise interference while fully preserving the millisecond-level transient change characteristics. This information is provided for the boundary point identification module in subsequent stages to read.

[0050] Stage boundary point identification is performed based on the strain transient sequence to obtain a boundary point set, which includes the strain rate surge point, the strain rebound initiation point, and the strain zeroing point. Based on the boundary point set, the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage are divided, and dynamic feature parameters are extracted for the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage. Based on the extraction results, a fracture dynamic feature vector is constructed.

[0051] Furthermore, based on point-by-point analysis and calculation of the transient strain sequence, a point-by-point strain rate sequence is obtained, and based on the point-by-point strain rate sequence, a local strain rate sequence is obtained. The mean value of the local strain rate sequence within the initial preset time period of the fracture window data is used as the reference strain rate, and the moment when the local strain rate sequence first exceeds the preset multiple of the reference strain rate is marked as the strain rate surge point. The moment when a continuous numerical decrease occurs for the first time in the strain transient sequence after the strain rate surge point and the decrease exceeds a preset range is marked as the strain rebound start point. The moment when the strain rebound in the transient sequence first falls below the order of magnitude of the reference strain rate after the strain rebound initiation point is identified, and this moment is marked as the strain zero point. By combining the point of rapid increase in strain rate, the point of strain rebound onset, and the point of strain returning to zero, we obtain the set of boundary points.

[0052] Furthermore, based on the point of rapid increase in strain rate, the point of strain springback onset, and the point of strain zeroing, the transient strain sequence is segmented to obtain the crack initiation stage, the rapid propagation and penetration stage, and the residual stage after fracture. The maximum value is extracted from the local strain rate sequence during the rapid expansion and penetration phase to obtain the peak strain rate. The amplitude difference between the strain rebound initiation point and the strain zero point is extracted from the strain transient sequence to obtain the strain rebound amplitude. The mean value is extracted from the local strain rate sequence within a preset time period before the point of rapid increase in strain rate to obtain the critical necking strain rate. The plastic power dissipation rate at each stage is obtained by integral calculation based on load data and displacement data in multi-source synchronous data stream. The critical necking strain rate and plastic power dissipation rate corresponding to the crack initiation stage, the peak strain rate and plastic power dissipation rate corresponding to the rapid propagation and penetration stage, and the strain rebound amplitude and plastic power dissipation rate corresponding to the residual stage after fracture are obtained respectively. They are then combined in the order of crack initiation stage, rapid propagation and penetration stage and residual stage after fracture to construct fracture dynamics feature vector.

[0053] The peak strain rate, strain rebound amplitude, critical necking strain rate, and plastic power dissipation rate at each stage are combined in stage order to construct a fracture dynamics characteristic vector.

[0054] In one specific implementation, the entire process of the rudder stock from crack initiation to the residue after fracture is described in the strain transient sequence. The crack exhibits a three-stage evolution with clear physical significance: the crack initiation stage corresponds to the initial moment when the local strain rate changes from stable to rapidly increasing; the rapid propagation and penetration stage corresponds to the interval where the strain value continuously accumulates and reaches its peak value before the crack front penetrates the cross section; and the post-fracture residual stage corresponds to the strain springback and return to zero process caused by the release of elastic strain energy after the fracture surface is completely separated. The boundaries between the three stages are... It has identifiable mathematical characteristics and can automatically determine the stage boundary points through point-by-point analysis of the strain transient sequence, realizing fully automatic stage division without human intervention.

[0055] First, the strain transient sequence Point-by-point calculation of the first-order difference yields the point-by-point strain rate sequence: ; in This represents the strain gauge sampling interval. Since point-by-point differential sampling is highly sensitive to high-frequency residual fluctuations in the signal, it is directly used... Boundary point identification can still be affected by noise interference. Therefore, using a length of... Sliding window By performing root mean square calculations, a local strain rate sequence is obtained that can smooth short-term fluctuations and reflect the overall level of local deformation rate: ; Where the sliding window length Corresponding time width ,Pick =5ms, that is .

[0056] Take the fracture window data starting point = The mean value of the local strain rate sequence within a 100ms time interval is used as the reference strain rate: ; in This represents the total number of sampling points within that time period. Reference strain rate. This represents the background level of the local deformation rate of the rudder stock under stable loading before crack initiation, serving as a reference baseline for subsequent rapid increase judgment.

[0057] In the local strain rate sequence From the start of the fracture window to the fracture time Directional point-by-point scanning will reveal the first occurrence. The moment when the strain rate exceeds the reference strain rate by β times is marked as the strain rate ramp point: ; The scaling factor β is set to 5, meaning that crack initiation is determined when the local strain rate exceeds 5 times the background level. The reason for selecting β=5 is that before entering necking instability, the local strain rate of ductile rudder materials usually increases by 3 to 10 times compared to the steady loading stage. Taking the middle value of 5 can reliably distinguish between normal loading fluctuations and the rate surge caused by crack initiation, and will not miss the early characteristics of the initiation stage due to an excessively high threshold. As the initial boundary of the crack initiation stage.

[0058] exist Subsequently, as the crack rapidly propagates to penetrate the cross-section, the elastic strain energy is released in a concentrated manner at the moment the fracture surface forms, resulting in a transient strain sequence { Local maxima will appear. The initial, sustained, rapid decline is known as strain rebound. Then for { Continuously monitor and record the current local maxima: ; It will first appear in a series The strain at each sampling point continued to decrease, and the cumulative decrease exceeded [amount missing]. The time point that is γ times the value of the strain rebound is marked as the initiation point of the strain rebound: ; When the reduction ratio γ is set to 0.30, the release of elastic strain energy at the moment of fracture in ductile metallic materials exhibits an accelerated rebound characteristic of "slow at first, then rapid." The cumulative strain reduction in the initial rebound stage is typically 20% to 40% of the peak strain. When γ is below 0.20, local yield fluctuations or measurement noise are easily misjudged as the start of rebound. When γ is above 0.40, a significant strain drop is required before the rebound can be triggered, resulting in a significant lag in the initial boundary of the residual stage after fracture. Extensive experimental data from rudder stock fractures have verified that γ=0.30 achieves the optimal balance between noise immunity and timely response.

[0059] Continuous judgment of points Corresponding duration = 2ms. It serves as the boundary between the rapid expansion and penetration phase and the residual phase after fracture.

[0060] exist Subsequently, as the elastic strain energy is fully released, the strain value eventually returns to a stable level close to zero. } exist The moment when the following equation is first satisfied is marked as the strain zeroing point: ; in The zero-return discrimination coefficient is taken as follows: = 3, the signal-to-noise ratio of the signal recovers during the zeroing phase, and 3 times the background fluctuation is sufficient to stably cross the residual noise amplitude, while avoiding excessive hysteresis in zeroing determination. This represents the strain fluctuation amplitude caused by the background deformation rate within a single sampling interval. Using this as a dynamic benchmark for zero-point determination can effectively avoid the influence of measurement noise and signal drift on zero-point identification.

[0061] in accordance with , and right{ The time intervals corresponding to the crack initiation stage are divided into segments. Rapidly expand throughout the corresponding time interval of the phase. The time interval corresponding to the residual stage after fracture Write the three data segments into the stage data cache area respectively.

[0062] After completing the stage division, key parameters that can quantify the fracture dynamics characteristics of each stage are extracted from the stage data cache: Peak strain rate The maximum strain rate level that a material can reach during the rapid propagation and penetration phase is a core indicator for assessing strain rate sensitivity. It is obtained by taking the maximum value from the local strain rate sequence during the rapid propagation and penetration phase. ; And record The moment of appearance The corresponding measuring point numbers are used to locate the spatiotemporal location where the strain rate is most intense during the rapid expansion and penetration phase.

[0063] Strain rebound amplitude The strength reflecting the release of elastic strain energy at the moment of fracture is directly related to the fracture toughness of the material. Extract the amplitude difference between the strain rebound initiation point and the strain zeroing point from the strain rebound point: ; Critical necking strain rate The local deformation rate corresponding to the initiation of necking instability is a critical indicator for judging the transition of a material from stable plastic deformation to localized instability, and the point of rapid increase in strain rate is extracted. forward = Mean value of the local strain rate sequence over 10ms: ; in This represents the total number of sampling points within that time period.

[0064] Plastic power dissipation rate at each stage (s = 1, 2, 3 correspond to the crack initiation stage, rapid propagation and penetration stage, and post-fracture residual stage, respectively) This reflects the material's ability to absorb energy at different fracture stages and is an important basis for evaluating material toughness. The load data corresponding to each stage { } with displacement data The total deformation work at each stage is calculated using the trapezoidal integral method: ; Subtract the elastic strain energy estimated based on the elastic modulus E and cross-sectional area A Then, divide by the corresponding stage duration. The plastic power dissipation rate at each stage is obtained: ; peak strain rate Strain rebound amplitude Critical necking strain rate Dispersion of plastic power consumption at each stage , , Arranged sequentially according to the crack initiation stage, rapid propagation and penetration stage, and post-fracture residual stage, the fracture dynamics feature vector is constructed by combining these stages: ; It fully describes the dynamic characteristics of the rudder stock from crack initiation to complete fracture, which can be used by the subsequent dynamic mechanical parameter set generation module.

[0065] The load and displacement data in the multi-source synchronous data stream are analyzed together to obtain the load-displacement curve. The fracture dynamics feature vector and the load-displacement curve are analyzed together, and the crack propagation rate is estimated by combining the high-speed camera video frame sequence. Based on the crack propagation rate, a dynamic mechanical parameter set is generated.

[0066] Furthermore, load data and displacement data are extracted from the multi-source synchronous data stream, load values ​​and displacement values ​​are paired to obtain data point pairs, and curve fitting is performed on the data point pairs to obtain load-displacement curves. Based on the corresponding intervals of the load-displacement curves mapped from the characteristic vectors of each stage in fracture dynamics, correlation analysis is performed to verify the mechanical consistency of the stage division. Acquire a high-speed camera video frame sequence, extract the position coordinates of the crack tip in adjacent video frames, and calculate the time series of crack propagation velocity; Based on the time index corresponding to each boundary point in the boundary point set, the crack propagation rate time series and the load displacement curve are aligned by time-displacement correlation mapping to form a dynamic mechanical parameter set containing multi-dimensional dynamic mechanical parameters.

[0067] In one specific implementation, the load-displacement curve is the most fundamental macroscopic mechanical characterization in the rudder stock tensile test, completely recording the macroscopic mechanical response from initial loading to fracture completion, and serving as the data foundation for calculating dynamic fracture toughness and dynamic integral. Load data sequences are extracted from multi-source synchronous data streams according to a one-to-one correspondence with timestamps. } and displacement data sequence { }, load values ​​at the same time With displacement value Pairing is for data point pairs Curve fitting is performed on all data point pairs to obtain load-displacement curves with displacement as the abscissa and load as the ordinate. .

[0068] After obtaining the load-displacement curve, the fracture dynamics eigenvectors are... Boundary moments of each stage , and The corresponding abscissa interval of the load-displacement curve F(d) is mapped to perform correlation analysis to verify the mechanical consistency between the stage division results and the macroscopic load variation trend. Specifically, the starting time of the crack initiation stage is determined. The load should be applied at the end of the elastic segment or at the peak of the load on the load-displacement curve. In the vicinity, the rapid expansion and penetration stage corresponds to a period where the load drops sharply, while the residual stage after fracture corresponds to a period where the load falls to zero. If there is a significant deviation between the above correspondence and the shape of the actual load-displacement curve, the boundary point identification results need to be manually reviewed to eliminate misjudgments of stage division caused by noise interference.

[0069] Because the rudder stock material undergoes significant plastic deformation and necking before fracture, this method directly employs dynamic J-integral. As a core indicator for evaluating the dynamic fracture toughness of rudder stock materials, the total deformation work throughout the entire process is calculated using the trapezoidal integral method on the load-displacement curve F(d) in terms of dynamic mechanical parameter calculation. The elastic strain energy estimated by subtracting the slope of the elastic segment Then, the total plastic work was obtained. . Specifically, Engineering estimation is performed based on the ratio of plastic work to effective crack area: ; Its physical meaning is the plastic work consumed per unit area of ​​crack propagation. The larger the value, the stronger the material's ability to resist crack initiation and stable propagation.

[0070] Stress values ​​were calculated from load data within the first 10 ms of the post-fracture residual phase. , and the corresponding local strain rate sequence { Linear least squares fitting is performed in a double logarithmic coordinate system, and the slope of the fitted line is extracted and denoted as the strain rate sensitivity index m. ; The physical meaning of m is the partial derivative of logarithmic stress with respect to logarithmic strain rate, which quantifies the sensitivity of the rudder stock material strength to the increase of strain rate.

[0071] For estimating crack propagation velocity, high-speed video frame sequences provide the only data source that can directly observe the spatial location of the crack front. After acquiring the high-speed video frame sequence, crack initiation frames and penetration frames are located from the video frames. The crack initiation frames correspond to the points of rapid strain rate increase. In the most recent frame of the time-shifted video frame sequence, the strain zeroing point corresponds to the frame that passes through the frame. Nearby video frames. Extract the pixel coordinates of the crack front edge in the crack initiation frame and the crack penetration frame, convert them to physical coordinates using the scale obtained from the pre-test calibration, and calculate the physical displacement of the crack front edge between the two frames. Inter-frame time difference The average crack propagation rate was obtained: ; Dynamic integral Strain rate sensitivity index m and crack propagation rate Merge to generate a dynamic mechanical parameter set This is for use in generating subsequent visualization analysis reports.

[0072] Furthermore, load data and displacement data are extracted from the multi-source synchronous data stream, and the load values ​​and displacement values ​​are paired according to the timestamp to obtain data point pairs. Curve fitting is then performed on the data point pairs to obtain the load-displacement curve. The boundary moments of each stage in the fracture dynamics eigenvector are mapped to the corresponding intervals of the load-displacement curves for correlation analysis to verify the mechanical consistency of the stage division. Acquire a high-speed camera video frame sequence, locate the crack initiation frame and the penetration frame, and calculate the crack propagation speed based on the ratio between the time difference between the two frames and the corresponding physical distance of crack propagation. The dynamic integral, strain rate sensitivity index, and crack propagation rate are combined to generate a set of dynamic mechanical parameters.

[0073] Based on the fracture moment, time offset correction is performed on the high-speed camera video frame sequence and multi-source synchronous data stream, the dynamic mechanical parameter set is labeled to the corresponding high-speed camera video frame sequence, and a full-process visualization analysis report of the fracture is output.

[0074] In one specific implementation, the high-speed camera video frame sequence and the multi-source synchronous data stream are recorded separately by independent hardware systems, each maintaining its own internal clock. During testing, the startup times of the camera system and the data acquisition system differ, and the timing accuracy and drift characteristics of the two systems' internal clocks vary, resulting in a time deviation ranging from several milliseconds to tens of milliseconds. Without correction, directly labeling the dynamic mechanical parameter set onto the video frame sequence will misalign the correspondence between the load curve and the video image at the same physical moment. This will prevent a precise correspondence between fracture morphology and transient strain data, severely impacting the reliability of the analysis conclusions.

[0075] To resolve the aforementioned clock synchronization issue, the break point recorded in step two is used as the reference. As a unified time reference anchor point, the fracture moment corresponds to the starting point of the rapid load drop on the load curve and the first video frame in the high-speed camera video frame sequence where the fracture surface appears. Both have clear and identifiable characteristics, so using the fracture moment as the anchor point can reliably align the two independent time axes.

[0076] Specifically, the inter-frame difference and adaptive threshold detection method is used to identify the video frame in which the fracture surface first appears: After preprocessing the high-speed camera video frame sequence by converting it to grayscale and Gaussian filtering frame by frame, the pixel difference map of adjacent frames is calculated, and the proportion of pixels in the difference map whose grayscale change exceeds a preset threshold τ is counted as the inter-frame difference index. Take the average difference index of the first 50 frames before the start of the fracture window. with standard deviation Establish a background difference benchmark, in order to As a dynamic judgment threshold, Taking a value of 5 to 8, the inter-frame difference in high-speed camera background typically does not exceed a certain range when there are no breaks in the frame. , Setting the value to 5 will reduce the false alarm probability to below one in a million, and The threshold should not exceed 8, to avoid detection delays of more than 1-2 frames, which would affect the time synchronization accuracy. Frames whose difference index first exceeds the threshold and then continues to rise for the next 3 consecutive frames are identified as candidate frames for the first appearance of the fracture surface. Then, the difference map of the frame is binarized and morphologically opened to extract the largest connected region. When its area exceeds the preset minimum crack area threshold, it is finally confirmed as the video frame for the first appearance of the fracture surface.

[0077] Read the internal timestamp of the camera system for this frame. Calculate the fixed offset between the time axis inside the camera system and the time axis of the multi-source synchronous data stream. .Will Apply time offset correction to all internal timestamps of the video frame sequence to ensure that the video frame sequence is precisely aligned with the sensor data frame at the moment of breakage.

[0078] After completing the time offset correction, the dynamic mechanical parameter set will be... Each parameter in the calculation is labeled to the corresponding high-speed camera video frame sequence according to its respective calculation time interval and the corrected timestamp index. Specifically, the dynamic J integral... Marked to peak load The corresponding video frames, the strain rate sensitivity index m, are marked to the range of video frames corresponding to the first 10ms of the residual stage after fracture, and the crack propagation rate. The video frame interval from the crack initiation frame to the penetration frame is marked, so that the fracture evolution stage corresponding to each dynamic mechanical parameter and its physical meaning can be accurately located on the video image.

[0079] Finally, using a unified time axis as the horizontal axis, the load-displacement curves were plotted. Strain transient sequence { }、Local strain rate sequence{ } and crack propagation rate Aligned and superimposed in the same time coordinate system, synchronously embedding crack initiation frames, penetration frames, and boundary moments of each stage. , and The corresponding high-speed camera keyframe images output a full-process fracture visualization analysis report containing all dynamic mechanical parameter values, time ranges for each stage, parameter calculation basis, and corresponding image screenshots.

[0080] The report provides testers with a complete quantitative answer to what exactly happened to the material in the last few hundred milliseconds before the rudder stock broke, in the form of numerical curves and video images that correspond precisely, directly supporting a refined assessment of the ultimate performance of the rudder stock material.

[0081] Example 2

[0082] Reference Figure 5 A data acquisition and analysis system for rudder lever mechanism tensile test, comprising: The data acquisition module is used to collect load data, displacement data and strain data throughout the entire tensile test of the rudder mechanism, and to form a multi-source synchronous data stream based on the load data, displacement data and strain data; The fracture event triggering module is used to identify the load data descent slope from the load data in the multi-source synchronous data stream, compare the load data descent slope with a preset trigger threshold, and determine that a fracture event has occurred when the load data descent slope exceeds the preset trigger threshold, and record it as the fracture time. The data processing and transformation module is used to identify the fracture window data segment from the multi-source synchronous data stream, generate fracture window data based on the fracture window data segment, and perform low-pass filtering and wavelet decomposition on the fracture window data to generate strain transient sequence. The feature extraction module is used to identify stage boundary points in the strain transient sequence to obtain a set of boundary points, which includes the strain rate surge point, the strain rebound initiation point, and the strain zero point. Based on the set of boundary points, the module divides the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage, and extracts dynamic feature parameters for the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage. Based on the extraction results, a fracture dynamic feature vector is constructed. The joint analysis module is used to analyze the joint relationship between load data and displacement data in multi-source synchronous data streams to obtain load-displacement curves; it performs joint analysis on fracture dynamics feature vectors and load-displacement curves, and estimates crack propagation velocity by combining high-speed camera video frame sequences, and generates a dynamic mechanical parameter set based on crack propagation velocity; The analysis report generation module is used to annotate the dynamic mechanical parameter set to the corresponding high-speed camera video frame sequence and output a full-process visual analysis report of the fracture.

[0083] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention, they should all fall within the protection scope of the present invention.

[0084] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0085] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A data acquisition and analysis method for a rudder stick mechanism tensile test, characterized in that, include: The load, displacement, and strain data of the rudder lever mechanism were collected throughout the tensile test, and a multi-source synchronous data stream was formed based on the load, displacement, and strain data. The load data descent slope is determined based on the load data in the multi-source synchronous data stream. The load data descent slope is compared with a preset trigger threshold. When the load data descent slope exceeds the preset trigger threshold, a fracture event is determined to have occurred and is recorded as the fracture time. Based on the fracture moment, the fracture window data segment is identified from the multi-source synchronous data stream, and fracture window data is generated based on the fracture window data segment. Low-pass filtering and wavelet decomposition are performed on the fracture window data to generate a strain transient sequence. Stage boundary point identification is performed based on the strain transient sequence to obtain a boundary point set, which includes the strain rate rapid increase point, the strain rebound initiation point and the strain zero return point. Based on the set of boundary points, the crack initiation stage, the rapid propagation and penetration stage, and the residual stage after fracture are divided. Dynamic feature parameters are extracted for the crack initiation stage, the rapid propagation and penetration stage, and the residual stage after fracture. Based on the extraction results, a fracture dynamic feature vector is constructed. The load and displacement data in the multi-source synchronous data stream are analyzed together to obtain the load-displacement curve; the fracture dynamics feature vector and the load-displacement curve are analyzed together, and the crack propagation rate is estimated by combining the high-speed camera video frame sequence; and a dynamic mechanical parameter set is generated based on the crack propagation rate. The dynamic mechanical parameter set is labeled to the corresponding high-speed camera video frame sequence, and a full-process visual analysis report of the fracture is output.

2. The data acquisition and analysis method for a rudder post mechanism tensile test according to claim 1, characterized in that, Load, displacement, and strain data were collected throughout the tensile test of the rudder post mechanism. Based on this data, a multi-source synchronous data stream was generated, including: Continuous signal acquisition was performed throughout the tensile test to obtain the original load simulation signal, original displacement simulation signal and original strain simulation signal throughout the tensile test. The original load simulation signal, the original displacement simulation signal, and the original strain simulation signal are subjected to signal conditioning and filtering to obtain each simulation signal after conditioning and filtering. The analog signals after conditioning and filtering are converted from analog to digital to generate load data, displacement data and strain data; Based on a unified time reference, load data, displacement data, and strain data are time-aligned and synchronously fused to form a multi-source synchronous data stream.

3. The data acquisition and analysis method for a rudder post mechanism tensile test according to claim 2, characterized in that, The load data descent slope is determined based on the load data in the multi-source synchronous data stream. This slope is compared to a preset trigger threshold. When the load data descent slope exceeds the preset trigger threshold, a fracture event is determined to have occurred and recorded as the fracture moment, including: Load data is extracted from the multi-source synchronous data stream, and trend analysis is performed on the load data to obtain information on the load change trend over time. Based on the aforementioned trend information, the rate of change of the load in the downward direction is extracted as an initial estimate of the downward slope of the load data. The initial estimate of the load data descent slope is subjected to stability judgment, and the load data descent slope is confirmed based on the stability judgment result; The load data descent slope is dynamically compared with a preset trigger threshold. When the load data descent slope exceeds the preset trigger threshold and reaches a preset duration, it is determined that a valid fracture event has occurred, and the current time is recorded as the fracture time.

4. The data acquisition and analysis method for a rudder post mechanism tensile test according to claim 1, characterized in that, Based on the fracture time, the fracture window data segment is identified from the multi-source synchronous data stream, and fracture window data is generated based on the fracture window data segment, including: Using the break point as the time center, candidate data intervals containing the break point are extracted along the time axis from the multi-source synchronous data stream. Data integrity verification and outlier removal are performed based on candidate data intervals, and the candidate data intervals after data integrity verification and outlier removal are determined as the broken window data segments. The candidate data range after data integrity verification and outlier identification is determined as the broken window data segment, and the multi-source synchronous data streams within the broken window data segment are arranged in the original time order to generate broken window data.

5. The data acquisition and analysis method for a rudder post mechanism tensile test according to claim 4, characterized in that, The fracture window data is subjected to low-pass filtering and wavelet decomposition to generate a strain transient sequence, including: Strain data is extracted from the fracture window data, and low-pass filtering is applied to the strain data to obtain filtered strain data. Based on the wavelet decomposition processing of the filtered strain data, the filtered strain data is decomposed into wavelet coefficients of multiple scales based on the results of the wavelet decomposition processing. Denoising is performed on the wavelet coefficients at each scale to obtain the effective coefficients after denoising. Wavelet reconstruction is performed on the effective coefficients after denoising to generate a strain transient sequence.

6. The data acquisition and analysis method for a rudder post mechanism tensile test according to claim 5, characterized in that, Stage boundary point identification is performed based on the strain transient sequence to obtain a boundary point set, which includes the strain rate surge point, the strain rebound initiation point, and the strain zeroing point, including: Based on point-by-point analysis and calculation of the transient strain sequence, a point-by-point strain rate sequence is obtained, and based on the point-by-point strain rate sequence, a local strain rate sequence is obtained. The mean value of the local strain rate sequence within the initial preset time period of the fracture window data is used as the reference strain rate, and the moment when the local strain rate sequence first exceeds the preset multiple of the reference strain rate is marked as the strain rate surge point. The moment when a continuous numerical decrease occurs for the first time in the strain transient sequence after the strain rate surge point and the decrease exceeds a preset range is marked as the strain rebound start point. The moment when the strain rebound in the transient sequence first falls below the order of magnitude of the reference strain rate after the strain rebound initiation point is identified, and this moment is marked as the strain zero point. By combining the point of rapid increase in strain rate, the point of strain rebound onset, and the point of strain returning to zero, we obtain the set of boundary points.

7. The data acquisition and analysis method for a rudder post mechanism tensile test according to claim 6, characterized in that, Based on the boundary point set, the crack initiation stage, rapid propagation and penetration stage, and post-fracture residual stage are divided. Dynamic feature parameters are extracted for each of these stages, and a fracture dynamic feature vector is constructed based on the extraction results, including: Based on the point of rapid increase in strain rate, the point of strain springback onset and the point of strain zeroing, the transient strain sequence is segmented to obtain the crack initiation stage, the rapid propagation and penetration stage and the residual stage after fracture. The peak strain rate is obtained by extracting the maximum value from the local strain rate sequence during the rapid expansion phase. The amplitude difference between the strain rebound initiation point and the strain zero point is extracted from the strain transient sequence to obtain the strain rebound amplitude. The mean value is extracted from the local strain rate sequence within a preset time period before the point of rapid increase in strain rate to obtain the critical necking strain rate. The plastic power dissipation rate at each stage is obtained by integral calculation based on load data and displacement data in multi-source synchronous data stream. The critical necking strain rate and plastic power dissipation rate corresponding to the crack initiation stage, the peak strain rate and plastic power dissipation rate corresponding to the rapid propagation and penetration stage, and the strain rebound amplitude and plastic power dissipation rate corresponding to the residual stage after fracture are obtained respectively. They are then combined in the order of crack initiation stage, rapid propagation and penetration stage and residual stage after fracture to construct fracture dynamics feature vector. The peak strain rate, strain rebound amplitude, critical necking strain rate, and plastic power dissipation rate at each stage are combined in stage order to construct a fracture dynamics characteristic vector.

8. The data acquisition and analysis method for a rudder post mechanism tensile test according to claim 7, characterized in that, Joint relationship analysis is performed on load and displacement data in multi-source synchronous data streams to obtain load-displacement curves. Fracture dynamics feature vectors are jointly analyzed with the load-displacement curves, and crack propagation velocity is estimated using high-speed camera video frame sequences. Based on the crack propagation velocity, a dynamic mechanical parameter set is generated, including: Load and displacement data are extracted from multi-source synchronous data streams, load and displacement values ​​are paired to obtain data point pairs, and curve fitting is performed on the data point pairs to obtain load-displacement curves. Based on the corresponding intervals of the load-displacement curves mapped from the characteristic vectors of each stage in fracture dynamics, correlation analysis is performed to verify the mechanical consistency of the stage division. Acquire a high-speed camera video frame sequence, extract the position coordinates of the crack tip in adjacent video frames, and calculate the time series of crack propagation velocity; The crack propagation rate time series is matched and aligned with the synchronization timestamps corresponding to each boundary point in the boundary point set to form a dynamic mechanical parameter set containing multi-dimensional dynamic mechanical parameters.

9. A data acquisition and analysis system for a rudder stick mechanism tensile test, applied to the data acquisition and analysis method for a rudder stick mechanism tensile test as described in any one of claims 1-8, characterized in that, include: The data acquisition module is used to collect load data, displacement data and strain data throughout the entire tensile test of the rudder mechanism, and to form a multi-source synchronous data stream based on the load data, displacement data and strain data; The fracture event triggering module is used to identify the load data descent slope from the load data in the multi-source synchronous data stream, compare the load data descent slope with a preset trigger threshold, and determine that a fracture event has occurred when the load data descent slope exceeds the preset trigger threshold, and record it as the fracture time. The data processing and transformation module is used to identify the fracture window data segment from the multi-source synchronous data stream and generate fracture window data based on the fracture window data segment. Low-pass filtering and wavelet decomposition are performed on the fracture window data to generate a strain transient sequence. The feature extraction module is used to identify stage boundary points in the strain transient sequence to obtain a set of boundary points, which includes the strain rate surge point, the strain rebound initiation point, and the strain zero point. Based on the set of boundary points, the module divides the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage, and extracts dynamic feature parameters for the crack initiation stage, the rapid propagation and penetration stage, and the post-fracture residual stage. Based on the extraction results, a fracture dynamic feature vector is constructed. The joint analysis module is used to analyze the joint relationship between load data and displacement data in multi-source synchronous data streams to obtain load-displacement curves; it performs joint analysis on fracture dynamics feature vectors and load-displacement curves, and estimates crack propagation velocity by combining high-speed camera video frame sequences, and generates a dynamic mechanical parameter set based on crack propagation velocity; The analysis report generation module is used to annotate the dynamic mechanical parameter set to the corresponding high-speed camera video frame sequence and output a full-process visual analysis report of the fracture.

Images