Building engineering quality diagnosis method and device based on multi-sensor fusion and storage medium

Abstract

The invention belongs to the technical field of engineering detection, and particularly provides a building engineering quality diagnosis method based on multi-sensor fusion, comprising the following steps: deploying vibration, strain, temperature and acoustic emission sensor networks at key nodes of a building structure, and performing spatial distribution according to a preset rule; performing noise filtering on the multi-source sensing data; carrying out multi-source data fusion based on an improved evidence theory, and generating quality state credibility evaluation; calculating a quality risk grade through a fuzzy comprehensive evaluation model; and triggering grading alarm and generating a visual diagnosis report. According to the method, the health state of the building structure can be more comprehensively reflected through collaborative collection of multi-modal data including vibration, strain, temperature, acoustic emission and other multi-dimensional data, an improved evidence theory fusion method is adopted, a sensor historical data variance and a real-time consistency index are introduced to serve as reliability weights, the evidence weights are dynamically adjusted, and the reliability of the building structure is improved. And the data fusion precision is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of engineering detection, and specifically relates to a construction engineering quality diagnosis method, device and storage medium based on multi-sensor fusion. Background Art

[0002] Traditional construction project quality inspection mainly relies on manual inspections and single sensor monitoring, which has the following defects: the data dimension is single and only relies on single-modal data such as vibration or strain, which makes it difficult to fully reflect the health status of the structure; the noise interference is serious and the construction site environment is complex, and the sensor signal is easily affected by factors such as mechanical vibration and electromagnetic interference, resulting in a low signal-to-noise ratio; the fusion accuracy is insufficient, and existing evidence theory, fuzzy evaluation and other methods are prone to failure in high-conflict data scenarios, leading to misjudgment; there is a lack of predictive ability and the evolution trend of structural damage cannot be dynamically predicted, making it difficult to support preventive maintenance decisions. On this basis, the present invention proposes an intelligent quality diagnosis method based on multi-sensor fusion. Summary of the Invention

[0003] The present invention mainly provides a construction engineering quality diagnosis method based on multi-sensor fusion to solve the technical problems raised in the above background technology.

[0004] The technical solution adopted by the present invention to solve the above-mentioned technical problems is:

[0005] A construction engineering quality diagnosis method based on multi-sensor fusion includes the following steps:

[0006] S1. Deploy vibration, strain, temperature and acoustic emission sensor networks at key nodes of the building structure and distribute them spatially according to preset rules;

[0007] S2, noise filtering of multi-source sensor data;

[0008] S3, based on the improved evidence theory, multi-source data fusion is performed to generate the quality status credibility assessment;

[0009] S4. Calculate the quality risk level through the fuzzy comprehensive evaluation model;

[0010] S5. Trigger a hierarchical alarm and generate a visual diagnostic report.

[0011] In a further implementation scheme, in step S1, vibration sensors are arranged according to a preset wavelength ratio rule, strain sensors cover stress concentration areas, temperature sensors are distributed based on a gradient of thermal conductivity characteristics, and acoustic emission sensors cover structural joints.

[0012] In a further implementation scheme, in step S1, vibration sensors are deployed at beam-column nodes and mid-span positions of floor slabs at a spacing of no more than one-quarter of the minimum monitoring wavelength. Strain sensors are installed in the form of strain rosettes on cantilever structures, shear wall corners and surfaces of large-span components. Temperature sensors are arranged according to the temperature gradient distribution law based on the calculation results of the structural heat conduction equation. Acoustic emission sensors use broadband resonant probes and cover concrete joints, steel structure welds and prestressed anchor areas.

[0013] In a further embodiment, in step S2, the noise filtering process includes frequency band segmentation of the acoustic emission signal and baseline drift correction of the vibration signal, wherein the wavelet threshold denoising selects the Symlet series wavelet basis function to perform soft threshold processing on the high-frequency coefficients of each layer after decomposition, and the threshold value is dynamically adjusted according to the signal noise level; for the temperature drift interference in the strain signal, a polynomial fitting method is used to establish a temperature-strain compensation model to correct the measurement value in real time; the wavelet basis function formula is as follows:

[0014]

[0015] Where Tj represents the adaptive threshold in the j-th layer wavelet decomposition;

[0016] σj, the noise standard deviation of the j-th layer wavelet coefficients;

[0017] Nj, the number of wavelet coefficients in the jth layer;

[0018] γ, energy ratio adjustment factor, ranges from 0.3 to 0.7, according to γ ​​= 0.5-0.2·e -SNR / 5 Dynamic adjustment;

[0019] The energy of the detail coefficients of the jth layer;

[0020] The energy of the approximation coefficients of the jth layer;

[0021] η, nonlinear index, according to the signal sparsity Adaptive calculation, η = 1.3 + 0.2S;

[0022] tanh(·), the hyperbolic tangent function, compresses the SNR0 / 10 to the (-1, 1) interval.

[0023] In a further embodiment, in step 2, a polynomial fitting method is used to establish a temperature-strain compensation model for the temperature drift interference in the strain signal, and the measured value is corrected in real time; the temperature-strain compensation model is as follows:

[0024]

[0025] Among them, εreal refers to the true strain of the structure after eliminating temperature interference;

[0026] α1, linear thermal expansion coefficient (determined by constant temperature chamber test);

[0027] α2, nonlinear thermal expansion coefficient;

[0028] α3, creep effect coefficient;

[0029] ΔT, the difference between the current temperature and the reference temperature.

[0030] In a further implementation scheme, in step S3, the improved evidence theory fusion method includes the following steps: defining a quality state set as an identification framework and calculating the basic probability distribution of each sensor data; introducing the variance of sensor historical data and real-time consistency index as reliability weights and dynamically adjusting the evidence weights; quantifying the degree of contradiction between different sensor data through conflict factors and correcting the fusion results using a weighted conflict redistribution strategy; and finally outputting the credibility probability distribution of each quality state and screening the state exceeding the preset confidence threshold as the diagnosis conclusion, wherein the dynamic evidence weight function is:

[0031]

[0032] Where, wi is the real-time reliability weight of sensor i;

[0033] μi,σi: mean and standard deviation of historical data of sensor i;

[0034] δi, the real-time deviation, is calculated as

[0035] xi, the real-time measurement value of sensor i;

[0036] μi, the mean vector of historical data of sensor i, calculated by sliding window;

[0037] Σ, historical data covariance matrix.

[0038] In a further implementation scheme, in step S4, the factor set of the fuzzy comprehensive evaluation model includes crack width, vibration amplitude, strain gradient, temperature change rate and acoustic emission energy parameters, and the comment set is divided into three levels: safety, warning and danger; the degree of membership of each factor to the comment set is quantified by a Gaussian membership function, and a fuzzy relationship matrix is ​​constructed; the entropy weight method is combined with the analytic hierarchy process (AHP) to dynamically calculate the weight of each factor, and the comprehensive evaluation results are synthesized by a weighted average operator, and the final risk level is determined according to the maximum membership principle; wherein the fuzzy entropy weight model is:

[0039] in

[0040] Among them, ωj is the composite weight of the jth evaluation factor;

[0041] Hj, the information entropy of the jth factor, based on the statistical frequency of each comment pjk based on historical data;

[0042] rj, the analytic hierarchy process scale value, through the expert scoring matrix A = [a ij ] calculation, satisfying a ij =1 / a ji .

[0043] In a further implementation scheme, in step S5, the risk level output by the fuzzy comprehensive evaluation model is compared with a preset threshold, and when the risk membership exceeds the set threshold, an alarm signal of a corresponding level is triggered.

[0044] The present invention also discloses a construction engineering quality diagnosis device based on multi-sensor fusion, comprising a processor;

[0045] a memory for storing processor-executable instructions;

[0046] Wherein, the processor is configured to execute the above-mentioned quality diagnosis method.

[0047] The present invention also discloses a computer storage medium, wherein the computer storage medium stores computer executable instructions; the computer executable instructions can implement the quality diagnosis method provided above.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention can more comprehensively reflect the health status of building structures through the collaborative collection of multimodal data, including multi-dimensional data such as vibration, strain, temperature and acoustic emission;

[0050] The present invention adopts an improved evidence theory fusion method, introduces the sensor historical data variance and real-time consistency index as reliability weight, dynamically adjusts the evidence weight, and effectively improves the data fusion accuracy;

[0051] The present invention realizes real-time processing and storage of data through noise filtering in the data preprocessing step and the integrated processor and memory in the device, thereby improving the real-time performance and reliability of the diagnosis results.

[0052] The present invention calculates the quality risk level through a fuzzy comprehensive evaluation model and generates a visual diagnosis report, which provides a scientific basis for engineering maintenance and supports the formulation and implementation of preventive maintenance decisions.

[0053] The present invention will be explained in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flow chart of a construction engineering quality diagnosis method based on multi-sensor fusion provided by the present invention. DETAILED DESCRIPTION

[0055] To facilitate understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in different forms and is not limited to the embodiments described in the text. On the contrary, these embodiments are provided to make the content disclosed in the present invention more thorough and comprehensive.

[0056] The present invention provides a construction engineering quality diagnosis method based on multi-sensor fusion, comprising the following steps:

[0057] S1. Sensor network deployment

[0058] Vibration, strain, temperature and acoustic emission sensors are deployed at key nodes of the building structure, such as beam-column joints, mid-span of floor slabs, cantilever structures, shear wall corners and surfaces of large-span components.

[0059] Vibration sensors are arranged at intervals no greater than one-quarter of the minimum monitoring wavelength to ensure that subtle changes in structural vibration can be captured. Strain sensors are installed using rosettes, covering areas of stress concentration, to accurately measure structural strain.

[0060] Temperature sensors are distributed based on the gradient of heat conduction characteristics, with their placement determined by the calculation results of the structural heat conduction equation. Acoustic emission sensors use broadband resonant probes, covering concrete joints, steel welds, and prestressed anchorage areas to monitor for minor damage within the structure.

[0061] S2. Data Preprocessing

[0062] The wavelet threshold denoising method is used to filter out noise. The Symlet series wavelet basis functions are selected to perform soft threshold processing on the high-frequency coefficients of each layer after decomposition. The threshold value is dynamically adjusted according to the signal noise level.

[0063] At the same time, for the temperature drift interference in the strain signal, a polynomial fitting method is used to establish a temperature-strain compensation model to correct the measurement value in real time.

[0064] S3, multi-source data fusion

[0065] Multi-source data fusion is performed based on the improved evidence theory.First, the quality state set is defined as the identification framework, and the basic probability distribution of each sensor data is calculated.

[0066] The sensor historical data variance and real-time consistency index are introduced as reliability weights to dynamically adjust the evidence weight.

[0067] The conflict degree between different sensor data is quantified by the conflict factor, and the weighted conflict redistribution strategy is used to correct the fusion result.

[0068] Finally, the credibility probability distribution of each quality state is output, and the states exceeding the preset credibility threshold are selected as the diagnosis conclusion.

[0069] S4. Quality risk level assessment

[0070] The quality risk level is calculated using a fuzzy comprehensive evaluation model. The factor set includes crack width, vibration amplitude, strain gradient, temperature change rate, and acoustic emission energy parameters.

[0071] Gaussian membership function is used to quantify the degree of membership of each factor to the review set and construct a fuzzy relationship matrix.

[0072] The entropy weight method combined with the analytic hierarchy process (AHP) is used to dynamically calculate the weight of each factor, the comprehensive evaluation results are synthesized through the weighted average operator, and the final risk level is determined according to the maximum membership principle.

[0073] S5. Alarm and report generation

[0074] The risk level (such as low, medium, and high) output by the fuzzy comprehensive evaluation model is compared with the preset threshold. When the risk membership exceeds the set threshold (such as the membership of high risk ≥ 0.8), the corresponding level of alarm signal is triggered, that is, the graded alarm is triggered according to the quality risk level, and a visual diagnostic report is generated to provide decision support for engineering maintenance.

[0075] In step S2, the noise filtering process includes frequency band segmentation of the acoustic emission signal and baseline drift correction of the vibration signal. The wavelet threshold denoising selects the Symlet series wavelet basis function and performs soft threshold processing on the high-frequency coefficients of each layer after decomposition. The threshold value is dynamically adjusted according to the signal noise level. For the temperature drift interference in the strain signal, a polynomial fitting method is used to establish a temperature-strain compensation model to correct the measurement value in real time. The wavelet basis function formula is as follows:

[0076]

[0077] Where Tj represents the adaptive threshold in the j-th layer wavelet decomposition;

[0078] σj, the noise standard deviation of the j-th layer wavelet coefficients;

[0079] Nj, the number of wavelet coefficients in the jth layer;

[0080] γ, energy ratio adjustment factor, ranges from 0.3 to 0.7, according to γ ​​= 0.5-0.2·e -SNR / 5 Dynamic adjustment;

[0081] The energy of the detail coefficients of the jth layer;

[0082] The energy of the approximation coefficients of the jth layer;

[0083] η, nonlinear index, ranges from 1.2 to 1.5, depending on the signal sparsity Adaptive calculation, η = 1.3 + 0.2S;

[0084] tanh(), hyperbolic tangent function, compresses SNR0 / 10 to the (-1,1) interval.

[0085] For example, when applied to concrete crack detection, the decomposition level j = 5, and the wavelet basis is Symlet8;

[0086] SNR0=15dB; σ3=0.12; N3=128; γ=0.5-0.2e -15 / 5 =0.46;

[0087] η=1.3+0.2×0.15=1.33;

[0088] tanh(15 / 10)=tanh(1.5)=0.905;

[0089]

[0090] Noise ratio improvement: 15dB→32dB; crack pulse signal retention rate: 92% (traditional method is only 67%).

[0091] In a further embodiment, in step 2, a polynomial fitting method is used to establish a temperature-strain compensation model for the temperature drift interference in the strain signal, and the measured value is corrected in real time; the temperature-strain compensation model is as follows:

[0092]

[0093] Among them, εreal refers to the true strain of the structure after eliminating temperature interference;

[0094] α1, linear thermal expansion coefficient (determined by constant temperature chamber calibration test, unit: με / ℃);

[0095] α2, nonlinear thermal expansion coefficient (quadratic term coefficient fitted by cyclic temperature loading test);

[0096] α3, creep effect coefficient (calculated according to the material time-temperature equivalence principle, unit: με·h / ℃);

[0097] ΔT, the difference between the current temperature and the reference temperature (the reference temperature is the average temperature).

[0098] If applied to concrete crack detection, the measured strain ε measurement = 250με;

[0099] Temperature difference ΔT = 40°C, temperature change rate = 2°C per hour;

[0100] Material parameters: α1 = 12.5 με / °C, α2 = 0.025 με / °C2, α3 = 0.1 με / (°C·hour);

[0101]

[0102] =-330με; negative value means the weld is under compression, and the actual mechanical strain is 330με compressive strain;

[0103] Temperature compensation of this solution: deduct all 580με → residual <1με, while the residual of the traditional method is 40με.

[0104] In step S3, the improved evidence theory fusion method includes the following steps: defining a set of quality states as an identification framework and calculating the basic probability distribution of each sensor data; introducing the variance of sensor historical data and real-time consistency indicators as reliability weights and dynamically adjusting the evidence weights; quantifying the degree of contradiction between different sensor data through conflict factors and correcting the fusion results using a weighted conflict redistribution strategy; and finally outputting the credibility probability distribution of each quality state and screening the state exceeding the preset confidence threshold as the diagnosis conclusion. The dynamic evidence weight function is:

[0105]

[0106] Where wi is the real-time reliability weight of sensor i; the range is [0,1], and the larger the value, the more reliable the sensor data is;

[0107] μi, σi: the mean and standard deviation of the historical data of sensor i, updated in real time through a sliding window (window size W is 100-500);

[0108] δi, the real-time deviation, is calculated as

[0109] xi, the real-time measurement value of sensor i;

[0110] μi, the mean vector of historical data of sensor i, calculated by sliding window (e.g., window W = 200);

[0111] Σ, historical data covariance matrix, recalculated every 24 hours;

[0112] For example, when applied to steel structure vibration monitoring, the sensor historical data μi=2.1mm; σi=0.3mm;

[0113] Real-time data xi = 3.8 mm; Mahalanobis distance δi = 4.2;

[0114]

[0115] The results show that the weight of abnormal sensors is reduced to 0.07, which avoids the contamination of fusion results.

[0116] In step S4, the factor set of the fuzzy comprehensive evaluation model includes crack width, vibration amplitude, strain gradient, temperature change rate, and acoustic emission energy parameters. The evaluation set is divided into three levels: safety, warning, and danger. The membership degree of each factor to the evaluation set is quantified by a Gaussian membership function, and a fuzzy relationship matrix is ​​constructed. The entropy weight method is combined with the analytic hierarchy process (AHP) to dynamically calculate the weight of each factor. The comprehensive evaluation results are synthesized by the weighted average operator, and the final risk level is determined according to the maximum membership principle. The fuzzy entropy weight model is as follows:

[0117] in

[0118] Among them, ωj is the composite weight of the jth evaluation factor;

[0119] Hj, the information entropy of the jth factor, based on the statistical frequency of each comment pjk based on historical data;

[0120] rj, the analytic hierarchy process scale value, through the expert scoring matrix A = [a ij ] calculation, satisfying a ij =1 / a ji .

[0121] For example, when applied to vibration monitoring of steel structures, the factor set is: crack width (C), vibration amplitude (V), and strain gradient (S);

[0122] Entropy weight calculation: HC = 0.35, HV = 0.28, HS = 0.41, and we get

[0123]

[0124] The expert scoring matrix is ​​rj = [0.4, 0.3, 0.3], and the composite weight

[0125]

[0126] Decision result: Crack risk membership is 0.82 (hazard level), triggering a red alarm. While the embodiments of the present invention have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention. The scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A construction engineering quality diagnosis method based on multi-sensor fusion, characterized in that: The following steps are involved: S1. Deploy vibration, strain, temperature and acoustic emission sensor networks at key nodes of the building structure and distribute them spatially according to preset rules; S2, noise filtering of multi-source sensor data; S3, based on the improved evidence theory, multi-source data fusion is performed to generate the quality status credibility assessment; S4. Calculate the quality risk level through the fuzzy comprehensive evaluation model; S5. Trigger a hierarchical alarm and generate a visual diagnostic report.

2. The method according to claim 1, characterized in that In step S1, vibration sensors are arranged according to a preset wavelength ratio rule, strain sensors cover stress concentration areas, temperature sensors are distributed based on a gradient of heat conduction characteristics, and acoustic emission sensors cover structural joints.

3. The method according to claim 1, characterized in that In step S1, vibration sensors are deployed at beam-column joints and mid-span locations of floor slabs at intervals no greater than one-quarter of the minimum monitoring wavelength. Strain sensors are installed in the form of rosettes on cantilever structures, shear wall corners, and the surfaces of large-span components. Temperature sensors are arranged according to the temperature gradient distribution law based on the calculation results of the structural heat conduction equation. Acoustic emission sensors use broadband resonant probes and cover concrete joints, steel structure welds, and prestressed anchorage areas.

4. The method according to claim 1, wherein In step S2, the noise filtering process includes frequency band segmentation of the acoustic emission signal and baseline drift correction of the vibration signal. The wavelet threshold denoising selects the Symlet series wavelet basis function and performs soft threshold processing on the high-frequency coefficients of each layer after decomposition. The threshold value is dynamically adjusted according to the signal noise level. For the temperature drift interference in the strain signal, a polynomial fitting method is used to establish a temperature-strain compensation model to correct the measurement value in real time. The wavelet basis function formula is as follows: Where Tj represents the adaptive threshold in the j-th layer wavelet decomposition; σj, the noise standard deviation of the j-th layer wavelet coefficients; Nj, the number of wavelet coefficients in the jth layer; γ, energy ratio adjustment factor, according to γ ​​= 0.5-0.2·e -SNR / 5 Dynamic adjustment; The energy of the detail coefficients of the jth layer; The energy of the approximation coefficients of the jth layer; η, nonlinear index, according to the signal sparsity Adaptive calculation, η = 1.3 + 0.2S; tanh(), hyperbolic tangent function, compresses SNR0 / 10 to the (-1,1) interval.

5. The method according to claim 4, characterized in that: In step 2, a polynomial fitting method is used to establish a temperature-strain compensation model for the temperature drift interference in the strain signal, and the measured value is corrected in real time. The temperature-strain compensation model is as follows: Among them, εreal refers to the true strain of the structure after eliminating temperature interference; α1, linear thermal expansion coefficient; α2, nonlinear thermal expansion coefficient; α3, creep effect coefficient; ΔT, the difference between the current temperature and the reference temperature.

6. The method according to claim 1, characterized in that In step S3, the improved evidence theory fusion method includes the following steps: defining a set of quality states as an identification framework and calculating the basic probability distribution of each sensor data; introducing the variance of sensor historical data and real-time consistency indicators as reliability weights and dynamically adjusting the evidence weights; quantifying the degree of contradiction between different sensor data through conflict factors and correcting the fusion results using a weighted conflict redistribution strategy; and finally outputting the credibility probability distribution of each quality state and screening the state exceeding the preset confidence threshold as the diagnosis conclusion. The dynamic evidence weight function is: Where, wi is the real-time reliability weight of sensor i; μi,σi: mean and standard deviation of historical data of sensor i; δi, the real-time deviation, is calculated as xi, the real-time measurement value of sensor i; μi, the mean vector of historical data of sensor i, calculated by sliding window; Σ, historical data covariance matrix.

7. The method according to claim 1, characterized in that In step S4, The fuzzy comprehensive evaluation model's factor set includes crack width, vibration amplitude, strain gradient, temperature change rate, and acoustic emission energy parameters. The evaluation set is divided into three levels: safe, warning, and dangerous. A Gaussian membership function is used to quantify the degree of membership of each factor to the evaluation set, and a fuzzy relationship matrix is ​​constructed. The entropy weight method is combined with the analytic hierarchy process (AHP) to dynamically calculate the weight of each factor. The comprehensive evaluation results are synthesized using a weighted average operator, and the final risk level is determined based on the maximum membership principle. The fuzzy entropy weight model is: in Among them, ωj is the composite weight of the jth evaluation factor; Hj, the information entropy of the jth factor, based on the statistical frequency of each comment pjk based on historical data; rj, the analytic hierarchy process scale value, through the expert scoring matrix A = [a ij ] calculation, satisfying a ij =1 / a ji .

8. The method according to claim 1, characterized in that In step S5, the risk level output by the fuzzy comprehensive evaluation model is compared with a preset threshold value, and when the risk membership exceeds the set threshold value, an alarm signal of the corresponding level is triggered.

9. A construction engineering quality diagnosis device based on multi-sensor fusion, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to execute the quality diagnosis method according to any one of claims 1 to 8.

10. A computer storage medium, characterized in that The computer storage medium stores computer-executable instructions; the computer-executable instructions can implement the quality diagnosis method provided in any one of claims 1 to 8.

Images