A multivariate monitoring data fusion method and system for slope tunnel model tests

Abstract

The application discloses a kind of multivariate monitoring data fusion method and system of side slope tunnel model test, is specifically related to geotechnical engineering physical model test technical field, for solving the problem that dynamic stress wave in the model test of existing integrated static and dynamic measurement causes static monitoring signal distortion, and then influence model state assessment accuracy;By synchronously collecting dynamic excitation and static response signal, identify the coupling interference in static signal;Frequency domain comparative analysis is carried out using signal sample before and after model state mutation event acquisition, identify the characteristic frequency component related to mutation, and evaluate the dominant frequency band of coupling interference;Based on the frequency band, separate the effective static response component from the original static signal;Finally, this effective component and the soil dynamic parameter derived from dynamic excitation are fused, to generate comprehensive stability evaluation data to realize;Effectively suppress dynamic interference, improve the authenticity of static signal, improve the reliability of model stability state assessment.

Description

Technical Field

[0001] This invention relates to the field of geotechnical engineering physical model testing technology, and more specifically, to a multi-source monitoring data fusion method and system for slope tunnel model testing. Background Technology

[0002] In indoor physical model tests for slope tunnel engineering, multi-source monitoring methods are typically employed to comprehensively evaluate the mechanical response and stability of the model under complex loading conditions. Existing technologies generally involve embedding commonly used geotechnical instruments such as earth pressure cells and strain gauges within the model to monitor static mechanical parameters. Simultaneously, to obtain dynamic soil properties, a resonant column testing system is integrated to apply controllable dynamic excitation to the model. This test mode, combining static monitoring and dynamic testing, aims to reveal the mechanical behavior of slope tunnel structures more deeply by integrating multi-source monitoring data.

[0003] However, an inherent technical problem exists in such experiments integrating static and dynamic measurements: the dynamic stress waves generated by the resonant column system propagate in the model medium and exert direct mechanical effects on the sensitive elements of the static parameter monitoring sensors embedded within it. This dynamic excitation environment causes significant vibration interference components, unrelated to changes in the overall mechanical state of the model, to be coupled into the output signal of the static sensors, resulting in distortion of the original monitoring data. Since different types of sensors exhibit varying physical response characteristics to dynamic excitation, this signal distortion further complicates the reliable extraction of effective information directly related to model stability from multi-source data, thus limiting the accuracy and reliability of model state assessment results based on multi-source data fusion. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a multi-source monitoring data fusion method and system for slope tunnel model tests to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A multivariate monitoring data fusion method for slope tunnel model tests includes the following steps:

[0007] S1. Collect multivariate monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonant column system and static response signals collected by static parameter monitoring sensors;

[0008] S2. Determine whether there is a coupling interference component caused by dynamic stress wave in the static response signal based on the frequency characteristics of the dynamic excitation signal;

[0009] S3. When it is determined that there is a coupling interference component, after detecting the preset model state change event, collect static response signal segments before and after the event as comparison samples.

[0010] S4. Analyze and compare the frequency domain characteristics of the samples, identify the characteristic frequency components related to mutations, and evaluate the non-characteristic frequency components that overlap with the dynamic excitation frequency band as the dominant frequency band of coupling interference.

[0011] S5. Based on the dominant frequency band of coupling interference, separate the effective static response components related to the mechanical state changes of the slope tunnel model from the static response signal;

[0012] S6. The effective static response components are fused with the soil dynamic parameters derived from the dynamic excitation signal to generate stability assessment data for the slope tunnel model.

[0013] Furthermore, multivariate monitoring data were collected from the slope tunnel model test, including dynamic excitation signals generated by the resonant column system and static response signals collected by static parameter monitoring sensors, including:

[0014] The system synchronously acquires the dynamic excitation signal generated by the resonant column system and the static response signal acquired by the static parameter monitoring sensor, and assigns a unified timestamp to the synchronously acquired dynamic excitation signal and static response signal.

[0015] Furthermore, based on the frequency characteristics of the dynamic excitation signal, the presence of coupling interference components caused by dynamic stress waves in the static response signal is determined, including:

[0016] Analyze the dynamic excitation signal to determine its dominant frequency components;

[0017] The static response signal is converted to the frequency domain to obtain the static response spectrum;

[0018] Determine whether there are significant frequency amplitudes in the frequency band corresponding to the dominant frequency component in the static response spectrum;

[0019] If present, it is determined that there is a coupling interference component caused by dynamic stress wave in the static response signal.

[0020] Furthermore, the determination of whether there is a significant frequency amplitude in the static response spectrum within the frequency band corresponding to the dominant frequency component is achieved by: calculating the average amplitude of the static response spectrum within the frequency band corresponding to the dominant frequency component, and calculating the full-band average amplitude of the entire static response spectrum; if the average amplitude within the frequency band is higher than the full-band average amplitude by a preset multiple, it is determined that there is a significant frequency amplitude.

[0021] Furthermore, when coupling interference is detected, after a pre-defined abrupt change in the model state is detected, static response signal segments before and after the event are collected as comparison samples, including:

[0022] Based on the acoustic emission monitoring device, the acoustic emission signal of the slope tunnel model is monitored. When the energy rate of the acoustic emission signal exceeds the preset threshold, it is determined that a preset model state change event has occurred.

[0023] Based on the unified timestamp assigned to the static response signal, static response signal segments of the same length are extracted before and after the occurrence of the model state mutation event, respectively, and used as static response signal segments before and after the event.

[0024] Furthermore, by analyzing and comparing the frequency domain characteristics of the samples, characteristic frequency components related to abrupt changes were identified, and non-characteristic frequency components overlapping with the dynamic excitation frequency band were assessed as the dominant frequency bands for coupling interference, including:

[0025] Frequency domain analysis was performed on the static response signal segments before and after the event to obtain the pre-event spectrum and the post-event spectrum.

[0026] Calculate the amplitude difference between the spectrum after the event and the spectrum before the event;

[0027] Identify frequency points where the amplitude difference exceeds a preset change threshold, and determine the frequency components corresponding to the frequency points as characteristic frequency components related to mutations;

[0028] The frequency range within the dynamic excitation band that does not belong to the characteristic frequency components is evaluated as the frequency band dominated by coupling interference.

[0029] Furthermore, frequency domain analysis is performed on the static response signal segments before and after the event to obtain the pre-event spectrum and the post-event spectrum. This is achieved by applying a windowing function to the static response signal segments before and after the event to reduce spectral leakage, and then transforming the windowed signal segments to obtain the amplitude distribution of their frequency components, thereby obtaining the pre-event spectrum and the post-event spectrum, respectively.

[0030] Furthermore, based on the dominant frequency band of coupling interference, effective static response components related to the changes in the mechanical state of the slope tunnel model are separated from the static response signal, including:

[0031] Design a band-stop filter based on the evaluated frequency band dominated by coupling interference;

[0032] Apply a band-stop filter to the complete static response signal to filter out frequency components located in the frequency band dominated by coupling interference;

[0033] The remaining frequency components in the filtered signal are reconstructed into effective static response components related to the changes in the mechanical state of the slope tunnel model.

[0034] Furthermore, the effective static response components are fused with the soil dynamic parameters derived from the dynamic excitation signal to generate stability assessment data for the slope tunnel model, including:

[0035] The shear modulus and damping ratio of the soil are calculated from the dynamic excitation signal as dynamic parameters of the soil.

[0036] The time history curve eigenvalues ​​of the effective static response components and the soil dynamic parameters are input into the preset stability evaluation criteria for weighted fusion calculation, generating a comprehensive slope tunnel model stability evaluation index as the slope tunnel model stability evaluation data.

[0037] On the other hand, the present invention provides a multi-source monitoring data fusion system for slope tunnel model tests.

[0038] Includes the following modules:

[0039] The data acquisition module is used to collect multivariate monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonant column system and static response signals collected by static parameter monitoring sensors.

[0040] The interference detection module is used to determine whether there are coupling interference components caused by dynamic stress waves in the static response signal based on the frequency characteristics of the dynamic excitation signal.

[0041] The sample acquisition module is used to collect static response signal fragments before and after a preset model state change event as comparison samples when it is determined that there are coupling interference components.

[0042] The frequency band evaluation module is used to analyze the frequency domain characteristics of the comparison samples, identify the characteristic frequency components related to mutations, and evaluate the non-characteristic frequency components that overlap with the dynamic excitation frequency band as the dominant frequency band of coupling interference.

[0043] The component separation module is used to separate the effective static response components related to the mechanical state changes of the slope tunnel model from the static response signal based on the dominant frequency band of coupling interference.

[0044] The data fusion module is used to fuse the effective static response components with the soil dynamic parameters derived from the dynamic excitation signal to generate stability assessment data for the slope tunnel model.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. By employing a systematic signal processing and data fusion strategy, the problem of static monitoring signal distortion under dynamic excitation environments is effectively solved. First, through synchronous acquisition and correlation analysis of dynamic excitation signals and static response signals, the existence of coupling interference components introduced by dynamic stress waves in the static signals is accurately identified. Then, using the model's own state change events as a time reference, static response signal segments before and after the events are extracted for comparative analysis, transforming the internal damage evolution process of the model into favorable conditions for signal analysis. Through frequency domain differential processing, characteristic frequency components related to the actual mechanical state change of the model can be clearly identified and separated from the broad dynamic excitation frequency band. Finally, by selectively filtering out frequency band components evaluated as dominated by coupling interference, a pure and effective signal component reflecting the static state change of the model is obtained, fundamentally suppressing the adverse effects of dynamic environmental noise on static parameter monitoring and significantly improving the reliability and authenticity of the static response signals.

[0047] 2. By organically integrating the purified effective static response components with the soil dynamic parameters directly derived from the dynamic excitation signal, a multi-dimensional and in-depth assessment of the stability state of the slope tunnel model was achieved. Based on a profound understanding of the physical origin of the signal, signal purification was performed first, followed by feature fusion. The effective static response components reflect the deformation and stress redistribution trends of the model under static loads, while dynamic parameters such as soil shear modulus and damping ratio reveal the stiffness and energy dissipation characteristics of the soil material under dynamic action. By comprehensively evaluating these two types of parameters, which respectively characterize the macroscopic mechanical response of the model and the microscopic dynamic characteristics of the material, the final stability assessment data contains both static evolution information and dynamic characteristic information, overcoming the limitations of single-type data assessment. This assessment method based on the extraction and fusion of signal source features makes the conclusions more comprehensive and accurate in reflecting the comprehensive stability state of the slope tunnel model under complex load conditions, providing a more scientific and reliable basis for model safety assessment. Attached Figure Description

[0048] Figure 1 This is a flowchart of a multivariate monitoring data fusion method for slope tunnel model tests according to the present invention;

[0049] Figure 2 This is a schematic diagram of the structure of a multi-source monitoring data fusion system for slope tunnel model tests according to the present invention. Detailed Implementation

[0050] 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1: Figure 1 This invention presents a multivariate monitoring data fusion method for slope tunnel model tests, which includes the following steps:

[0052] S1. Collect multivariate monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonant column system and static response signals collected by static parameter monitoring sensors;

[0053] S2. Determine whether there is a coupling interference component caused by dynamic stress wave in the static response signal based on the frequency characteristics of the dynamic excitation signal;

[0054] S3. When it is determined that there is a coupling interference component, after detecting the preset model state change event, collect static response signal segments before and after the event as comparison samples.

[0055] S4. Analyze and compare the frequency domain characteristics of the samples, identify the characteristic frequency components related to mutations, and evaluate the non-characteristic frequency components that overlap with the dynamic excitation frequency band as the dominant frequency band of coupling interference.

[0056] S5. Based on the dominant frequency band of coupling interference, separate the effective static response components related to the mechanical state changes of the slope tunnel model from the static response signal;

[0057] S6. The effective static response components are fused with the soil dynamic parameters derived from the dynamic excitation signal to generate stability assessment data for the slope tunnel model.

[0058] To achieve synchronous acquisition and time alignment of multi-source monitoring data in slope tunnel model tests, the data acquisition steps are described in detail. The slope tunnel model tests are conducted in a model box filled with similar materials to simulate the actual slope soil and rock mass. The front of the model box is made of a high-strength transparent material to facilitate observation of the model's deformation and failure process. Static parameter monitoring sensors, including miniature earth pressure cells and strain gauges, are embedded at specific locations inside the model. The miniature earth pressure cells measure the earth pressure distribution inside the model, and the strain gauges measure the strain response of the support structure. The resonant column system consists of a vibrator and a control system. The vibrator is in close contact with the top center of the model via a force transmission rod. The control system generates sinusoidal electrical signals of specific frequency and amplitude to drive the vibrator, thereby applying dynamic excitation to the slope tunnel model.

[0059] The acquisition of dynamic excitation signals is accomplished through the data acquisition unit built into the resonant column system. This data acquisition unit is directly connected to the exciter's control circuit and feedback sensor, and can continuously record the voltage value sequence of the drive electrical signal output by the control system at a first sampling frequency. This voltage value sequence directly reflects the time history change of the dynamic excitation signal. The first sampling frequency is set to 1000 Hz. The acquisition of static response signals is accomplished through an independent multi-channel static data acquisition instrument. This multi-channel static data acquisition instrument is connected to all the miniature earth pressure cells and strain gauges embedded inside the model via shielded cables. The miniature earth pressure cells output voltage signals proportional to the applied earth pressure, and the strain gauges output resistance change signals proportional to the strain value. The multi-channel static data acquisition instrument synchronously scans all channels at a second sampling frequency, converting the received analog signals into digital quantities to form a discrete-time sequence of the static response signal. The second sampling frequency is set to 500 Hz. The first sampling frequency and the second sampling frequency are different.

[0060] To achieve synchronous data acquisition, a scheme combining hardware synchronous triggering and software timestamp alignment is adopted. Specifically, a synchronous controller is used, which generates a high-precision master clock signal with its time base derived from a built-in temperature-compensated crystal oscillator. The synchronous controller is connected to the control system of the resonant column system via a first trigger signal line and to the external trigger port of the multi-channel static data acquisition instrument via a second trigger signal line. At the start of the experiment, the synchronous controller is manually activated, simultaneously sending a rising-edge trigger pulse signal to both the control system of the resonant column system and the multi-channel static data acquisition instrument. Upon receiving the trigger pulse signal, the control system of the resonant column system immediately begins recording the dynamic excitation signal and records the absolute time of receiving the trigger pulse as the starting reference point for dynamic signal acquisition. Upon receiving the same trigger pulse signal, the multi-channel static data acquisition instrument also immediately begins synchronous sampling of all channels and similarly records the arrival time of the trigger pulse as the starting reference point for static signal acquisition.

[0061] Because the data acquisition unit of the resonant column system and the multi-channel static data acquisition instrument use their own independent internal clocks, even though the acquisition start time is synchronized, slight frequency drift between the two clocks during long-term acquisition can lead to accumulated time errors. To assign a unified timestamp, time alignment post-processing is performed after data acquisition. First, dynamic excitation signal data is exported from the resonant column system. This data includes a time series starting from the trigger moment with a first sampling interval as the time step and the corresponding signal amplitude sequence. Then, static response signal data is exported from the multi-channel static data acquisition instrument. This data includes a time series for each channel starting from its own trigger moment with a second sampling interval as the time step and the corresponding signal amplitude sequence. Since the trigger signals originate from the same synchronization controller, the starting reference points of the two systems are at the same physical time.

[0062] Next, a unified time axis is constructed. The zero point of this unified time axis corresponds to the absolute moment when the synchronous controller issues the trigger pulse. The sampling interval of the unified time axis is set to the period corresponding to the least common multiple of the first sampling frequency and the second sampling frequency, or a suitable common sampling interval, such as 1 millisecond, depending on the needs of subsequent analysis. Then, the dynamic excitation signal data is processed. Since its time series is built based on the starting point of the unified time axis, only its clock drift needs to be checked. If the drift is within the allowable range, it is directly used. For the static response signal data, since the second sampling frequency may be different from the sampling frequency of the unified time axis, resampling is required. A linear interpolation algorithm is used to calculate the estimated amplitude at each time point of the unified time axis based on the original discrete time points and amplitudes of the static response signal, thereby generating static response signal data aligned with the unified time axis. The specific process of the linear interpolation algorithm is as follows: For any target time point on a unified time axis, find two time points in the original time series of the static response signal that are immediately before and after the target time point and their corresponding signal amplitudes. Then, assuming that the signal changes linearly between these two original time points, calculate the signal amplitude in a linear proportion according to the position of the target time point relative to these two original time points.

[0063] After the aforementioned synchronization triggering and resampling alignment processes, both the dynamic excitation signal and the static response signal of each channel possess discrete time series based on the same zero point and with the same time interval. This completes the assignment of a unified timestamp to the synchronously acquired dynamic excitation signal and static response signal. This data with a unified timestamp is stored in a specific format data file; for example, each line of data includes a timestamp, the amplitude of the dynamic excitation signal, and the signal amplitude of each static parameter monitoring sensor channel, facilitating direct retrieval and analysis in subsequent steps. This unified timestamp forms the basis for all subsequent time correlation analyses, especially for accurately extracting signal segments before and after events, ensuring the accuracy of time correlations between data from different sources.

[0064] This paper details the process of determining whether coupling interference caused by dynamic stress waves exists in the static response signal based on the frequency characteristics of the dynamic excitation signal. The input data for this process consists of dynamic excitation signals and static response signals acquired according to the methods described in the previous embodiments and assigned a unified timestamp. First, the dynamic excitation signal is analyzed to determine its dominant frequency component. From the dynamic excitation signal data with a unified timestamp, a segment of the dynamic excitation signal that is in the stable loading stage of the slope tunnel model test and whose signal amplitude is stable is selected. The duration of this segment should include a sufficient number of excitation cycles to ensure the accuracy of the frequency analysis; for example, a 10-second segment is selected. The selected dynamic excitation signal segment undergoes frequency domain transformation processing, specifically using the Fast Fourier Transform (FFT) algorithm to transform it from the time domain to the frequency domain, obtaining the frequency-amplitude spectrum of the dynamic excitation signal, called the dynamic excitation spectrum. The specific process of the FFT is as follows: the discrete-time signal sequence is grouped and recursively calculated to finally obtain the complex representation of the signal at discrete frequency points, where the amplitude represents the intensity of the frequency component. Observe the dynamic excitation spectrum and find the frequency component with the largest amplitude. This frequency component is the dominant frequency component of the dynamic excitation signal. For example, if the resonant column system is set to apply a sine wave excitation with a frequency of 5 Hz, a significant peak will appear at 5 Hz in the dynamic excitation spectrum. This 5 Hz frequency is identified as the dominant frequency component. Record the specific value of this dominant frequency component, for example, 5 Hz.

[0065] Next, the static response signal is converted to the frequency domain to obtain the static response spectrum. A static response signal segment that is completely synchronized in time with the aforementioned dynamic excitation signal segment is selected; that is, static response signal data from the same static parameter monitoring sensor (e.g., a specific earth pressure cell) within the same time period are extracted based on a unified timestamp. This static response signal segment is also subjected to a Fast Fourier Transform to convert it into a frequency-amplitude spectrum, thus obtaining the static response spectrum of that channel within this time period. The static response spectrum shows the various frequency components contained in the signal at this monitoring point and their respective intensities.

[0066] Next, it is determined whether there are significant frequency amplitudes in the static response spectrum within the frequency band corresponding to the dominant frequency component. This determination is achieved through quantitative comparison. First, the frequency band corresponding to the dominant frequency component is defined. This frequency band is a narrow frequency range extending to both sides of the definite dominant frequency component (e.g., 5 Hz), for example, set as [dominant frequency component -0.5 Hz, dominant frequency component +0.5 Hz], i.e., [4.5 Hz, 5.5 Hz]. The average amplitude of the static response spectrum within this specific frequency band is calculated and denoted as the band average amplitude. The calculation method is to add the amplitudes corresponding to all discrete frequency points within the band and then divide by the total number of frequency points contained in the band. Next, the full-band average amplitude of the entire static response spectrum is calculated. The full band refers to the effective frequency range that can be represented by the Fast Fourier Transform, i.e., from 0 Hz to the Nyquist frequency (equal to half the sampling frequency; for a sampling frequency of 500 Hz, the Nyquist frequency is 250 Hz). The method for calculating the average amplitude across the entire frequency band is to add up the amplitudes corresponding to all discrete frequency points within the entire frequency band, and then divide by the total number of frequency points within the entire frequency band.

[0067] Next, a significance assessment is performed. The calculated average amplitude of the frequency band is compared with the average amplitude of the entire frequency band. A preset multiple is set as the judgment threshold, for example, a preset multiple of 3. This preset multiple is based on common requirements for signal-to-noise ratio in engineering experience, that is, when the energy of a specific frequency component is significantly higher than the average level of background noise (approximately represented by the average amplitude of the entire frequency band) by more than 3 times, the component is considered significant and non-random. If the average amplitude of the frequency band is higher than the average amplitude of the entire frequency band by this preset multiple (i.e., the average amplitude of the frequency band divided by the average amplitude of the entire frequency band is greater than or equal to 3), then it is determined that there is a significant frequency amplitude in the frequency band corresponding to the dominant frequency component in the static response spectrum. Conversely, if the ratio is lower than the preset multiple, then it is determined that there is no significant frequency amplitude.

[0068] Finally, a final judgment is made based on the significance assessment results. If a significant frequency amplitude is determined, it means that a strong frequency component highly consistent with the dominant frequency component of the dynamic excitation signal has been detected in the static response signal. This strongly suggests that the dynamic stress wave generated by the resonant column system has propagated to the location of the static parameter monitoring sensor and has produced a non-negligible modulation effect on the sensor output, i.e., there is a coupling interference component caused by the dynamic stress wave in the static response signal. If no significant frequency amplitude is determined, it means that the impact of the dynamic excitation on this specific static monitoring point is weak and negligible during the current monitoring period. This judgment will directly determine whether the subsequent step S3 is initiated.

[0069] This paper details how, after determining the presence of coupling interference components caused by dynamic stress waves in the static response signal, the model's state abrupt change events are captured, and comparative samples are obtained for subsequent analysis. The initiation condition for this process is that step S2 confirms the presence of significant coupling interference components in the signal of at least one static parameter monitoring sensor channel. To achieve sensitive capture of damage initiation and propagation within the model, an acoustic emission monitoring device is used as the primary means of identifying model state abrupt change events. The core component of the acoustic emission monitoring device is a piezoelectric acoustic emission sensor, which works by converting elastic stress waves released from material deformation or cracking into electrical signals based on the piezoelectric effect. Multiple acoustic emission sensors are deployed inside the slope tunnel model and at key locations on its surface (such as near potential slip surfaces, tunnel arches, and sidewalls). The sensors are in close contact with the model surface via a coupling agent to ensure good acoustic transmission. The output signals of the acoustic emission sensors are amplified by a preamplifier and continuously recorded by a dedicated acoustic emission acquisition system. This system is also connected to a synchronization controller, ensuring that the recorded data stream has a unified time reference with the dynamic excitation signal and the static response signal.

[0070] Continuously recorded acoustic emission signals are a collection of numerous discrete events, each corresponding to a minor damage activity. To identify critical points representing abrupt changes in macroscopic structural states from these continuous flows, the energy rate parameter of the acoustic emission signals is monitored. The energy rate of the acoustic emission signal is defined as the cumulative sum of the energies of all acoustic emission events recorded per unit time, and its calculation is performed internally within the acoustic emission acquisition system. Specifically, the system calculates the energy rate in a sliding window (e.g., 1 second), that is, every second, the sum of the energies of all acoustic emission events acquired in the past second is calculated as the energy rate value at the current moment. The energy of an acoustic emission event is obtained by squaring the waveform signal of that event and then integrating it over time. Thus, an acoustic emission energy rate time history curve aligned with a uniform timestamp can be obtained.

[0071] The preset model state mutation event is determined by whether the energy rate time history curve exceeds a preset threshold. This preset threshold is set based on the initial background noise level and material properties of the model experiment. Before formal loading begins, a period of acoustic emission background signal is collected when the model is in a stable state, and the average and standard deviation of the acoustic emission energy rate during this background period are calculated. The preset threshold is set to the average background energy rate plus three times the standard deviation. For example, if the average background energy rate is 10 mV² / ms and the standard deviation is 2 mV² / ms, then the preset threshold is set to 16 mV² / ms. This setting is based on the three sigma criterion in statistics, which considers events exceeding three times the standard deviation of the average to be low-probability anomalies, likely corresponding to significant damage mutations within the model, such as the generation or propagation of macroscopic cracks. Therefore, when the real-time monitored acoustic emission energy rate value first exceeds this preset threshold (16 mV² / ms), it is determined that a preset model state mutation event has occurred. The precise moment of this event, i.e., the point in time when the energy rate curve first breaks through the threshold, is recorded, and this moment is determined based on a unified timestamp.

[0072] After determining that a model state abrupt change event has occurred, static response signal segments before and after the event are immediately collected as comparison samples. This operation is strictly based on the uniform timestamp assigned to the static response signals. From the stored complete static response signal data with uniform timestamps, signal segments are extracted for the specific sensor channels to be analyzed. First, the extraction time length is determined. This length should be sufficient to include the characteristic changes of the signal but not too long to avoid introducing unnecessary information; it is usually set based on experience, for example, a length of 10 seconds. Next, the recorded moment of the model state abrupt change event is used as the reference point. The static response signal segment before the event is extracted from the signal data within the time interval starting 10 seconds before the reference point and ending immediately after the reference point. The static response signal segment after the event is extracted from the signal data within the time interval starting from the reference point and ending 10 seconds after the reference point. This results in two static response signal segments with the same time length (both 10 seconds) that are closely connected before and after the moment of the abrupt change event. These two signal segments are labeled as the static response signal segment before the event and the static response signal segment after the event, respectively. Together, they constitute the comparison sample for the next step of frequency domain feature analysis.

[0073] Throughout the process, acoustic emission monitoring provides precise localization of internal damage within the model, while a unified timestamp ensures a precise temporal correlation between acoustic emission events and static response signals. This enables accurate capture of critical time points, such as abrupt changes in the model's state, and allows for the acquisition of comparable before-and-after signal samples. This lays a solid foundation for subsequent identification of signal features related to changes in the model's actual mechanical state. This method effectively transforms the model's own destructive evolution process into a valid time stamp that can be used for signal tracing.

[0074] This paper details how to analyze and compare the frequency domain characteristics of samples to identify signal components related to abrupt changes in the model state and accurately assess the dominant frequency band of coupling interference. The input to this process consists of two key samples obtained in step S3: a static response signal segment before the event and a static response signal segment after the event. First, a detailed frequency domain analysis is required for each of these signal segments to obtain their spectra. Directly transforming a signal of finite length will result in spectral leakage, where energy is dispersed to non-realistic frequency components. To reduce spectral leakage, a windowing function is applied to both the static response signal segments before and after the event before transformation. The windowing function is a weighting function that gradually decays to zero at both ends of the signal segment. A Hanning window is chosen as the windowing function, with its coefficient sequence generated by a cosine function, and specific weight values ​​smoothly decreasing from the center of the window function towards both ends. Each data point in the static response signal segment before the event is multiplied by the corresponding point in the Hanning window function sequence to obtain the windowed signal segment before the event. Similarly, multiplying the static response signal segment after the event occurs by the same Hanning window function sequence yields the windowed signal segment after the event occurs. This multiplication operation effectively suppresses discontinuities at the beginning and end of the signal segment.

[0075] Next, the windowed signal segments before and after the event are transformed to obtain the amplitude distribution of their frequency components. This transformation is implemented using the Fast Fourier Transform (FFT) algorithm. The FFT decomposes the time-domain signal into a series of sine and cosine waves of different frequencies and calculates the complex result corresponding to each frequency component. The magnitude of this complex result (i.e., the square root of the sum of squares) is taken as the amplitude of that frequency component. Performing the FFT on the windowed signal segment before the event yields its frequency-amplitude relationship, called the pre-event spectrum. Similarly, performing the FFT on the windowed signal segment after the event yields the post-event spectrum. Both the pre-event and post-event spectra are sequences composed of discrete frequency points and their corresponding amplitudes, reflecting the frequency components and their intensities present in their respective signal segments.

[0076] After obtaining the pre-event and post-event spectra, the amplitude difference between the post-event and pre-event spectra is calculated. This calculation is performed for each identical discrete frequency point. For each frequency point in the spectrum, the amplitude difference is obtained by subtracting the amplitude of the pre-event spectrum at the same frequency point from the amplitude of the post-event spectrum at that frequency. If the post-event amplitude is greater than the pre-event amplitude, the amplitude difference is positive; otherwise, it is negative. Arranging the amplitude differences of all frequency points in frequency order constitutes an amplitude difference spectrum. This amplitude difference spectrum clearly shows the changes in the static response signal at different frequency components after the model's state abrupt change event.

[0077] Next, identify frequency points in the amplitude difference spectrum where the amplitude difference exceeds a preset change threshold. This preset change threshold is set to distinguish significant frequency changes that may be related to abrupt events, while ignoring minor random fluctuations. The threshold is set based on the background fluctuation level of the static response signal segment in a steady state before the event. Specifically, before calculating the amplitude difference, analyze the pre-event spectrum separately, calculate its amplitude standard deviation across the entire frequency band, and multiply this standard deviation by an empirical coefficient (e.g., 3 times) to obtain the preset change threshold. This means that a frequency component is considered significant only if its change before and after the event exceeds three times its background fluctuation level. Iterate through each frequency point in the amplitude difference spectrum, determining whether the absolute value of its amplitude difference is greater than the preset change threshold. Mark all frequency points that meet the condition.

[0078] The frequency components corresponding to these marked frequency points are identified as characteristic frequency components related to abrupt changes. For example, if the amplitude difference at frequency points f1, f2, and f3 exceeds a preset change threshold, then frequencies f1, f2, and f3 are identified as characteristic frequency components. These characteristic frequency components reflect the specific frequency locations where the static response signal is significantly enhanced or weakened due to abrupt changes in the model's internal state (such as crack initiation or stress redistribution), and they are considered to be closely related to the actual mechanical state changes of the model.

[0079] Finally, the frequency range within the dynamic excitation band that does not belong to the characteristic frequency component is assessed as the dominant frequency band for coupling interference. The dynamic excitation band has been determined in step S2, for example, a range centered on the dominant frequency component (e.g., 5 Hz), such as [4.5 Hz, 5.5 Hz]. All frequency points contained within this dynamic excitation band are examined. Those frequency points that do not belong to the identified characteristic frequency component are filtered out. The continuous or discrete frequency intervals formed by these frequency points constitute the dominant frequency band for coupling interference. For example, if the dynamic excitation frequency band is [4.5Hz, 5.5Hz], and the only identified characteristic frequency component is 5.2Hz, then the dominant frequency bands for coupling interference are the intervals [4.5Hz, 5.2Hz) and (5.2Hz, 5.5Hz]. Assessing this frequency band as dominated by coupling interference means that, within this band, apart from the characteristic point (5.2Hz) related to the abrupt change, the energy of the remaining frequency components mainly originates from the dynamic stress wave interference of the resonant column system, rather than changes in the mechanical state of the model itself. This accurate frequency band assessment provides a direct basis for the next step of effectively separating the interference signal. The entire process, through precise frequency domain differential comparison, successfully isolates and identifies the real changes in the response signal related to model damage from the inherent dynamic coupling interference background.

[0080] This section details the process of separating the effective static response components from the interfered static response signal based on the evaluated dominant frequency band of coupling interference. The inputs to this process include the precisely evaluated dominant frequency band of coupling interference in step S4 (e.g., the intervals formed by frequency points fa to fb and fc to fd) and the complete, time-stamped original static response signal acquired according to method S1. The goal of the separation is to maximally suppress the energy within the dominant frequency band of coupling interference while retaining characteristic frequency components outside and within this band that have been identified as being related to abrupt changes (see S4), thereby obtaining a pure signal component related to the changes in the mechanical state of the slope tunnel model.

[0081] First, a digital band-stop filter is designed based on the evaluated dominant frequency band of coupling interference. The function of a band-stop filter is to block signals within a specific frequency range (stopband) while allowing signals at frequencies on either side of the stopband (passband) to pass. A finite-length unit impulse response filter is designed because its strictly linear phase characteristics avoid introducing signal phase distortion during filtering. The core of filter design is determining its filter coefficients. The design process begins by setting the filter's key performance parameters. The filter order is an important parameter, affecting the steepness of the transition band and the computational cost; empirically, it is usually set between 100 and 200, for example, order 128. The sampling frequency is set to the sampling frequency of the original static response signal, i.e., 500 Hz. The stopband range is directly specified as the evaluated dominant frequency band of coupling interference, such as [fa, fb] and [fc, fd]. The passband range is set to the frequency range on both sides of the stopband, such as [0, fa - Δf], [fb + Δf, fc - Δf] and [fd + Δf, Nyquist frequency], where Δf is the transition band width, which is set according to the filter order and the desired attenuation level, such as 5 Hz.

[0082] Next, the window function method is used to design the filter coefficients. An ideal band-stop filter has a rectangular frequency response in the frequency domain, but its corresponding time-domain impulse response is infinitely long and non-causal. To approximate the ideal response with a finite-length coefficient sequence, the ideal infinite-long impulse response needs to be truncated, and a window function is applied to mitigate the Gibbs phenomenon caused by truncation. First, based on the defined stopband and passband, the unit impulse response sequence of an ideal band-stop filter is calculated. This ideal sequence is obtained by performing an inverse Fourier transform on the ideal frequency response; it is a symmetric sequence centered at zero. Since the ideal sequence is infinitely long, it needs to be truncated to a finite length, equal to the defined filter order plus one (e.g., 129 points). The truncation operation itself is equivalent to applying a rectangular window, but this causes fluctuations in the passband and stopband. To suppress these fluctuations, a higher-performance window function, such as the Hamming window, is used. The coefficient sequence of the Hamming window is also symmetric, with its values ​​smoothly decreasing from the center towards both ends. The truncated ideal unit impulse response sequence is multiplied point-by-point with the Hamming window sequence to obtain the final practical filter coefficients. These coefficients constitute a linear phase filter, with its center coefficient corresponding to the time zero point.

[0083] Then, the designed band-stop filter is applied to the complete static response signal. This is a discrete convolution operation performed in the time domain. The complete static response signal is a discrete-time sequence of length N. The filtering operation involves convolving this input sequence with the filter coefficient sequence of length M (M = filter order + 1) obtained above. The specific steps of convolution are: inverting the filter coefficient sequence; then sliding this inverted coefficient sequence along the input signal sequence; at each sliding position, multiplying the overlapping input signal value with the corresponding filter coefficient value, and summing all products to obtain the value of the output sequence at that position. This operation effectively achieves frequency selection, and the frequency components within the dominant frequency band (stopband) of coupling interference are significantly attenuated during the convolution summation process. To ensure that the length of the filtered signal sequence is consistent with the input signal and to avoid transient effects at the beginning and end, the input signal is usually edge-extended before convolution, for example, by using symmetrical extension, adding M / 2 points of mirror data at the beginning and end of the signal, and then removing the extended part after convolution.

[0084] After processing with a band-stop filter, frequency components within the dominant coupling interference band of the output signal have been significantly filtered out. Finally, the remaining frequency components in the filtered signal are reconstructed into effective static response components related to the changes in the mechanical state of the slope tunnel model. It is important to note that "reconstruction" here does not refer to an independent signal generation step, but rather a clear definition of the physical meaning of the filtered output signal. The signal after the above filtering process is the final effective static response component. It retains all frequency components within the passband of the original signal, as well as the energy of those identified as characteristic frequency components within the dominant coupling interference band (whose energy changes significantly before and after abrupt events and is considered related to the model's true response). Since the band-stop filter is a linear system, the filtering process maintains the linearity of the signal. Therefore, this effective static response component can directly reflect the changing trend of the slope tunnel model's mechanical state (such as earth pressure and structural strain) over time in the time domain, while the periodic oscillation interference caused by dynamic stress waves has been effectively suppressed. This effective static response component, as the key data after purification, will be used for the final comprehensive stability assessment. The entire separation process is based on precise frequency band assessment and customized digital filtering technology, which achieves the goal of extracting effective signals from strong background noise.

[0085] This paper details the complete process of fusing the separated effective static response components with soil dynamic parameters derived from dynamic excitation signals to ultimately generate stability assessment data for a slope tunnel model. The inputs to this process include the effective static response components related to the changes in the mechanical state of the slope tunnel model obtained in step S5, and the dynamic excitation signal segments collected in step S1 that correspond to the effective static response components in time. The goal of the fusion is to combine two types of information—reflecting the evolution of the model's static state and the dynamic characteristics of the soil—to form a comprehensive stability criterion.

[0086] First, the shear modulus and damping ratio of the soil are calculated from the dynamic excitation signal as dynamic parameters of the soil. This calculation is based on the fundamental principles of the resonant column test. A segment of the dynamic excitation signal synchronized with the effective static response component time period is selected; this signal is typically a sine wave with known amplitude and frequency. By analyzing this dynamic excitation signal segment and the model's response to the excitation (usually measured by sensors built into the resonant column system), the dynamic characteristics of the soil can be determined. The specific calculation process is as follows: The shear modulus is obtained by calculating the ratio of dynamic shear stress to dynamic shear strain. The dynamic shear stress is calculated by dividing the dynamic excitation force applied to the top of the model by the equivalent shear area of ​​the model; the excitation force can be converted from the exciter's drive current and calibration coefficient. The dynamic shear strain is calculated by measuring the torsional deformation amplitude of the model under dynamic action and combining it with the model's geometric dimensions. The damping ratio is usually determined by measuring the bandwidth of the resonance curve (the curve of response amplitude changing with frequency) or using the free vibration decay method. For example, in the free vibration decay method, after reaching stable resonance, the excitation is suddenly stopped, and the time history curve of the model's free vibration decay is recorded. The damping ratio can be obtained by calculating the natural logarithm of the ratio of the amplitudes of two adjacent vibration cycles and dividing it by 2π. Through these calculations, the shear modulus (e.g., in kPa) representing the dynamic stiffness characteristics of the model soil during that time period and the damping ratio (dimensionless) representing the energy dissipation characteristics of the soil are obtained. These two parameters are the dynamic parameters of the soil derived from the dynamic excitation signal.

[0087] Next, the time-history feature values ​​of the effective static response component are extracted. The effective static response component is a signal that varies over time, such as the time-history curve of the purified pressure readings of an earth pressure cell. Key mathematical features characterizing the model's stability are extracted from this time-history curve. These features include, but are not limited to: the slope of the trend term (obtained by linearly fitting the time-history curve, reflecting the average rate of change of the physical quantity); the amplitude of fluctuations (calculated by the standard deviation or root mean square value of the time-history curve within a certain time window, reflecting the severity of signal fluctuations); and the amplitude at a specific moment, such as the signal reading at a specific time point after a sudden change in the model's state. These feature values ​​quantify the behavioral patterns of the effective static response component from different perspectives.

[0088] Then, the time-history curve eigenvalues ​​of the effective static response components obtained above, along with the soil dynamic parameters (shear modulus and damping ratio), are input into a preset stability evaluation criterion for weighted fusion calculation. The preset stability evaluation criterion is a multi-parameter comprehensive judgment rule, which can take the form of a weighted summation model. This criterion is established based on statistical analysis of historical data from numerous slope and tunnel model tests or numerical simulation results to determine the contribution weight of each parameter to stability. For example, the stability evaluation criterion can be expressed as: Stability index = Trend term slope × Weight coefficient A + Reciprocal of fluctuation amplitude × Weight coefficient B + Shear modulus × Weight coefficient C - Damping ratio × Weight coefficient D. The values ​​of weight coefficients A, B, C, and D are determined through regression analysis to ensure that the calculated stability index has the best correspondence with the known model stability state (stable, critical, unstable). During the weighted fusion calculation, each eigenvalue and dynamic parameter first needs to be normalized to eliminate dimensional influences and ensure they fall within the same numerical range (e.g., between 0 and 1). The normalization method can employ min-maximum scaling, which involves subtracting the minimum value of each parameter's possible range from its value, and then dividing by the length of its range. The normalized parameter values ​​are then multiplied by their corresponding weighting coefficients, and all weighted values ​​are summed to obtain a preliminary composite index.

[0089] Finally, this preliminary comprehensive index is appropriately transformed (e.g., mapped to a score range of 0-100) to generate a comprehensive slope-tunnel model stability assessment index as the final slope-tunnel model stability assessment data. This index is a single quantitative value, whose magnitude directly reflects the relative stability of the slope-tunnel model at the assessment time; a higher value indicates greater stability, and a lower value indicates poorer stability. This comprehensive index integrates information on both static response trends and soil dynamic characteristics. Compared to single-type data, it can more comprehensively reflect the overall mechanical state of the model, providing an intuitive and quantitative basis for judging the model's safety status. The entire fusion process transforms monitoring data with different physical meanings and dimensions into a unified assessment scale through preset, experience-based, or theoretical criteria, achieving in-depth comprehensive utilization of multi-dimensional monitoring information.

[0090] Example 2: Figure 2 A schematic diagram of a multi-source monitoring data fusion system for slope tunnel model tests according to the present invention is provided. The multi-source monitoring data fusion system for slope tunnel model tests includes the following modules:

[0091] The data acquisition module is used to collect multivariate monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonant column system and static response signals collected by static parameter monitoring sensors.

[0092] The interference detection module is used to determine whether there are coupling interference components caused by dynamic stress waves in the static response signal based on the frequency characteristics of the dynamic excitation signal.

[0093] The sample acquisition module is used to collect static response signal fragments before and after a preset model state change event as comparison samples when it is determined that there are coupling interference components.

[0094] The frequency band evaluation module is used to analyze the frequency domain characteristics of the comparison samples, identify the characteristic frequency components related to mutations, and evaluate the non-characteristic frequency components that overlap with the dynamic excitation frequency band as the dominant frequency band of coupling interference.

[0095] The component separation module is used to separate the effective static response components related to the mechanical state changes of the slope tunnel model from the static response signal based on the dominant frequency band of coupling interference.

[0096] The data fusion module is used to fuse the effective static response components with the soil dynamic parameters derived from the dynamic excitation signal to generate stability assessment data for the slope tunnel model.

[0097] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0098] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0099] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0100] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0102] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0103] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-element monitoring data fusion method for a side slope tunnel model test, characterized in that, The method comprises the following steps: S1, collecting multi-element monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonance column system and static response signals collected by the static parameter monitoring sensor; S2, judging whether there is a coupling interference component caused by dynamic stress wave in the static response signal based on the frequency characteristics of the dynamic excitation signal; S3, when it is judged that there is a coupling interference component, after a preset model state mutation event is monitored, static response signal segments before and after the event occur are collected respectively as comparison samples; S4, analyzing the frequency domain characteristics of the comparison samples, identifying the characteristic frequency components related to the mutation, and evaluating the non-characteristic frequency components overlapping with the dynamic excitation frequency band as the coupling interference dominant frequency band; S5, separating the effective static response component related to the mechanical state change of the slope tunnel model from the static response signal based on the coupling interference dominant frequency band; S6, fusing the effective static response component with the soil dynamic parameters derived from the dynamic excitation signal to generate slope tunnel model stability evaluation data. 2.The method according to claim 1, characterized in that, The method comprises the following steps: S1, collecting multi-element monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonance column system and static response signals collected by the static parameter monitoring sensor; 3.The method according to claim 1, characterized in that, S2, judging whether there is a coupling interference component caused by dynamic stress wave in the static response signal based on the frequency characteristics of the dynamic excitation signal; S3, when it is judged that there is a coupling interference component, after a preset model state mutation event is monitored, static response signal segments before and after the event occur are collected respectively as comparison samples; S4, analyzing the frequency domain characteristics of the comparison samples, identifying the characteristic frequency components related to the mutation, and evaluating the non-characteristic frequency components overlapping with the dynamic excitation frequency band as the coupling interference dominant frequency band; S5, separating the effective static response component related to the mechanical state change of the slope tunnel model from the static response signal based on the coupling interference dominant frequency band; S6, fusing the effective static response component with the soil dynamic parameters derived from the dynamic excitation signal to generate slope tunnel model stability evaluation data.

4. The multi-element monitoring data fusion method for a slope tunnel model test according to claim 3, characterized in that, The method comprises the following steps:

5. The multi-element monitoring data fusion method for a slope tunnel model test according to claim 1, characterized in that, S1, collecting multi-element monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonance column system and static response signals collected by the static parameter monitoring sensor; S2, judging whether there is a coupling interference component caused by dynamic stress wave in the static response signal based on the frequency characteristics of the dynamic excitation signal; S3, when it is judged that there is a coupling interference component, after a preset model state mutation event is monitored, static response signal segments before and after the event occur are collected respectively as comparison samples; S4, analyzing the frequency domain characteristics of the comparison samples, identifying the characteristic frequency components related to the mutation, and evaluating the non-characteristic frequency components overlapping with the dynamic excitation frequency band as the coupling interference dominant frequency band; S5, separating the effective static response component related to the mechanical state change of the slope tunnel model from the static response signal based on the coupling interference dominant frequency band; S6, fusing the effective static response component with the soil dynamic parameters derived from the dynamic excitation signal to generate slope tunnel model stability evaluation data. The method comprises the following steps: S1, collecting multi-element monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonance column system and static response signals collected by the static parameter monitoring sensor; S2, judging whether there is a coupling interference component caused by dynamic stress wave in the static response signal based on the frequency characteristics of the dynamic excitation signal; S3, when it is judged that there is a coupling interference component, after a preset model state mutation event is monitored, static response signal segments before and after the event occur are collected respectively as comparison samples; S4, analyzing the frequency domain characteristics of the comparison samples, identifying the characteristic frequency components related to the mutation, and evaluating the non-characteristic frequency components overlapping with the dynamic excitation frequency band as the coupling interference dominant frequency band; S5, separating the effective static response component related to the mechanical state change of the slope tunnel model from the static response signal based on the coupling interference dominant frequency band; S6, fusing the effective static response component with the soil dynamic parameters derived from the dynamic excitation signal to generate slope tunnel model stability evaluation data.

6. The multi-element monitoring data fusion method for a slope tunnel model test according to claim 1, characterized in that, The frequency domain characteristics of the comparison samples are analyzed, the characteristic frequency components related to the mutation are identified, and the non-characteristic frequency components overlapping with the dynamic excitation frequency band are evaluated as the coupling interference dominant frequency band, including: The static response signal segments before and after the event occur are respectively analyzed in the frequency domain to obtain the pre-event spectrum and the post-event spectrum; The amplitude difference between the post-event spectrum and the pre-event spectrum is calculated; The frequency points with amplitude difference exceeding the preset change threshold are identified, and the frequency components corresponding to the frequency points are determined as the characteristic frequency components related to the mutation; The frequency range within the dynamic excitation frequency band and not belonging to the characteristic frequency components is evaluated as the coupling interference dominant frequency band.

7. The multi-element monitoring data fusion method for a slope tunnel model test according to claim 6, characterized in that, The static response signal segments before and after the event occur are respectively analyzed in the frequency domain to obtain the pre-event spectrum and the post-event spectrum, which is realized by: applying a windowing function to the static response signal segments before and after the event occurs respectively to reduce spectral leakage, and then transforming the windowed signal segments to obtain the amplitude distribution of their frequency components, thereby obtaining the pre-event spectrum and the post-event spectrum respectively. 8.The method according to claim 1, wherein, The effective static response component related to the mechanical state change of the slope tunnel model is separated from the static response signal based on the coupling interference dominant frequency band, including: A band-stop filter is designed according to the evaluated coupling interference dominant frequency band; The band-stop filter is applied to the complete static response signal to filter out the frequency components within the coupling interference dominant frequency band; The remaining frequency components in the filtered signal are reconstructed as the effective static response component related to the mechanical state change of the slope tunnel model. 9.The method of claim 1, wherein, The effective static response component is fused with the soil dynamic parameters derived from the dynamic excitation signal to generate slope tunnel model stability evaluation data, including: The shear modulus and damping ratio of the soil are calculated from the dynamic excitation signal as soil dynamic parameters; The time history curve characteristic values of the effective static response component and the soil dynamic parameters are input into the preset stability evaluation criterion for weighted fusion calculation to generate a comprehensive slope tunnel model stability evaluation index as the slope tunnel model stability evaluation data.

10. A multi-element monitoring data fusion system for a slope tunnel model test, for implementing the multi-element monitoring data fusion method of any one of claims 1-9, characterized in that, It includes the following modules: A data acquisition module for acquiring multi-element monitoring data in the slope tunnel model test, including dynamic excitation signals generated by the resonance column system and static response signals collected by static parameter monitoring sensors; An interference judgment module for judging whether there is a coupling interference component caused by dynamic stress waves in the static response signal based on the frequency characteristics of the dynamic excitation signal; A sample collection module for collecting static response signal segments before and after the occurrence of a preset model state mutation event as comparison samples when it is judged that there is a coupling interference component; A frequency band evaluation module for analyzing the frequency domain characteristics of the comparison samples, identifying characteristic frequency components related to the mutation, and evaluating non-characteristic frequency components overlapping with the dynamic excitation frequency band as the coupling interference dominant frequency band; A component separation module for separating the effective static response component related to the mechanical state change of the slope tunnel model from the static response signal based on the coupling interference dominant frequency band; A data fusion module is configured to fuse the effective static response component with the dynamic parameters of the soil derived from the dynamic excitation signal to generate the slope tunnel model stability evaluation data.

Images