Building structure health monitoring method and system based on big data

Abstract

The invention discloses a building structure health monitoring method and system based on big data, and the method comprises the steps: arranging sensors, obtaining the data of the sensors in real time, calculating the health correlation degree between the sensors according to the obtained data of the sensors, and building a health graph model; calculating a harmonic path weight between the sensors based on the health correlation degree; the entropy weight of each sensor is calculated based on the harmonic path weight between the sensors, and the connectivity index of the health graph model is calculated according to the entropy weight; and calculating a building structure health index by using a building structure health index evaluation algorithm based on the harmonic path weight between the sensors and the connectivity index of the health graph model, and comparing the building structure health index with a health threshold to realize building structure health monitoring. According to the method, the health state of the building structure is comprehensively evaluated, the method is sensitive to subtle changes, the stability of abnormal values is enhanced, and the evaluation precision of the health state of the building structure is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent buildings, and specifically relates to a building structure health monitoring method and system based on big data. Background Art

[0002] With the acceleration of urbanization and the continuous development of construction technology, the number of infrastructure is increasing. The health status of building structures is directly related to public safety, economic interests and social stability. It is necessary to monitor the health of building structures, discover potential problems in time, and take preventive and repair measures. However, traditional building structure health monitoring methods rely on manual inspections and regular inspections, which have limitations. With the rapid development of information technology and sensing technology, building structure health monitoring methods based on big data have emerged. By taking advantage of big data technology, real-time, comprehensive and accurate monitoring of building structures can be achieved, providing strong support for building structure health assessment, which has important theoretical significance and practical application value.

[0003] my country's patent application number: CN201910227557.1, publication date: 2019.07.12, discloses a building performance evaluation method and system based on structural health monitoring data, including obtaining building structure response parameters; obtaining the numerical range of the building structure response parameters corresponding to the building performance; comparing the building structure response parameters with the numerical range of the building structure response parameters corresponding to the building performance to calibrate the building performance. The present invention utilizes the building structure response parameters detected by the existing building health monitoring system in the prior art, and obtains the numerical range information of the building structure response parameters corresponding to the building performance through existing specifications and guidance documents. By judging the numerical range of the size of the acquired building structure response parameters, the building performance, i.e., the building damage level, can be calibrated, thereby achieving a comprehensive evaluation of the health status of the building structure.

[0004] However, the above technologies have at least the following technical problems: traditional building structure health monitoring methods are periodic manual inspections or simple static monitoring, which cannot achieve real-time monitoring and dynamic analysis of building structures; data processing capabilities are limited, making it difficult to fully mine and utilize complex information in sensor data; existing technologies lack sensitivity to subtle changes when assessing the health status of building structures, and are unable to capture potential problems in the structure in a timely manner, resulting in low accuracy in health status assessments. Summary of the invention

[0005] The purpose of the present invention is to provide a building structure health monitoring method based on big data in response to the shortcomings of the prior art. The method can solve the problems that the prior art is difficult to fully mine and utilize the complex information in the sensor data, and the prior art lacks sensitivity to subtle changes when evaluating the health status of the building structure, and is unable to capture potential problems in the structure in a timely manner, resulting in low accuracy in health status assessment.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] A building structure health monitoring method based on big data includes the following steps:

[0008] Step 1: Arrange sensors on the building structure, obtain sensor data in real time, calculate the health correlation between sensors based on the obtained sensor data, and establish a health graph model based on the health correlation;

[0009] Step 2: Based on the health association between sensors, the harmonic path weights between sensors are calculated using an adaptive harmonic cycle analysis algorithm;

[0010] Step 3: Based on the harmonic path weights between sensors, the entropy weight connectivity analysis algorithm is used to calculate the entropy weight of each sensor, and the connectivity index of the health graph model is obtained according to the entropy weight calculation;

[0011] Step 4: Based on the harmonic path weights between sensors and the connectivity index of the health graph model, the building structure health index is calculated using the building structure health index assessment algorithm, and the building structure health index is compared with the health threshold to achieve building structure health monitoring.

[0012] Furthermore, the method for calculating the health association degree between sensors in step 1 is specifically as follows:

[0013] Nonlinear transformation and dynamic adjustment mechanism are used for sensor data to capture the complex relationship of sensor data. The hyperbolic tangent function is introduced to calculate the difference between different sensor data to deal with the difference between different sensor data. Then, the dynamic adjustment mechanism of standard deviation and standard deviation square is introduced to adapt to the volatility of sensor data. Finally, the exponential function is used to transform the health correlation between sensors into the range of (0,1).

[0014] The specific calculation formula for the health correlation between sensors is:

[0015]

[0016] Among them, w ij(t) represents the health correlation between the i-th sensor and the j-th sensor at time t; exp represents the exponential function; |tanh(·)| represents the absolute value of the hyperbolic tangent function, and the processing data difference D i (t)-D j (t); D i (t) and D j (t) represents the i-th sensor data and the j-th sensor data at time t; σ ij (t) represents the standard deviation of the i-th sensor data and the j-th sensor data at time t; κ represents the adjustment parameter.

[0017] Furthermore, the method for constructing the health graph model is: taking sensors as nodes and the health correlation between sensors as edges, a health graph model is established. The health graph model is expressed as: G(y) = (V(t), E(t)), where V(t) represents the node set at time t, and E(t) represents the edge set at time t.

[0018] Furthermore, the calculation method of calculating the harmonic path weight in step 2 includes:

[0019] Phase offset is used to capture the periodic characteristics of the sensor, and the sliding window mechanism is used to achieve adaptive period adjustment, and the harmonic path weight is calculated in combination with the health association degree;

[0020] Phase offset is to calculate the error between the harmonic component and the actual sensor data for the current phase offset, and iteratively update the value of the phase offset through the gradient descent method until the error is minimized, thereby matching the actual fluctuation of the sensor data;

[0021] The method of using the sliding window mechanism to achieve adaptive period adjustment is to use fast Fourier transform to perform spectral analysis on sensor data, preliminarily calculate the period of sensor data, and determine the size and step size of the sliding window according to the period. The least squares fitting method is applied to the sensor data in each sliding window to fit a sine or cosine function, and the local period is determined according to the frequency of the sine or cosine function. Finally, the weighted average method is used to integrate the local period into an adaptive period.

[0022] Furthermore, the specific calculation formula of the harmonic path weight between sensors is:

[0023]

[0024] Among them, h ij (t) represents the harmonic path weight between the u-th sensor and the j-th sensor at time t; Represents the summation symbol for n from 1 to N; α n Represents the sinusoidal coefficient of the nth harmonic component; It represents the period T(t) and phase offset φ n sine function; 2π is a constant, representing one cycle; T(t) represents the adaptive cycle at time t; φ n represents the phase shift of the nth harmonic component; γ n Represents the cosine coefficient of the nth harmonic component; represents a cosine function with period T(t); w ij (t) represents the health correlation between the i-th sensor and the j-th sensor at time t.

[0025] Furthermore, in step 3, the entropy weight of each sensor is calculated by combining the harmonic path weights between sensors and the influence of the sensor in the sensor network. The calculation formula of the entropy weight is:

[0026]

[0027] Among them, e i (t) represents the entropy weight of the i-th sensor at time t, which measures the connectivity between sensors; represents the cumulative symbol, |V(t)|-1 represents the number of remaining nodes, that is, the number of sensors remaining except the i-th sensor; represents the modified probability weight, which calculates the connectivity weight between the i-th sensor and the j-th sensor at time t; represents the normalized modified probability weight; h ij (t) represents the harmonic path weight between the i-th sensor and the j-th sensor at time t, reflecting the periodic weight; W represents the logarithm of the modified probability weight, which is used to calculate the information entropy and reflects the uncertainty between the i-th sensor and the j-th sensor; j (t) represents the influence weight of the jth sensor at time t, and the specific calculation formula is:

[0028]

[0029] Among them, d ij (t) represents the actual path distance between the i-th sensor and the j-th sensor at time t.

[0030] Furthermore, in step 3, based on the entropy weight of the sensor, a mathematical statistical method is used to first sum and then average the data to obtain the connectivity index of the health graph model. The specific calculation formula is:

[0031]

[0032] Among them, C G (t) represents the connectivity index of the health graph model at time t, reflecting the health status of the overall structure; represents the average entropy weight of all sensors, and V(t) represents the node set at time t.

[0033] Furthermore, the connectivity index and harmonic path weight of the health graph model are combined to consider the impact of the harmonic path on the fatigue life of the building structure, and the fatigue sensitivity factor and frequency sensitivity factor are used to quantify the impact to calculate the building structure health index. The calculation formula of the building structure health index is:

[0034]

[0035] Among them, H G (t) represents the building structure health index at time t; C G (t) represents the connectivity index of the health graph model at time t; log(1+C G (t)) represents the nonlinear transformation of connectivity index; (C G (t)) m represents the high power of the connectivity index; S(t) represents the stiffness coefficient; q represents the index of the influence of the stiffness on the strength; p represents the index of the stiffness suppression term; exp represents the exponential function; L f (t) represents the predicted value of fatigue life of the building structure at time t.

[0036] Furthermore, in step 4, the health threshold θ(t) is set to represent the building structure health index H G (t) is compared with the health threshold θ(t), when H G When (t)≥θ(t), an alarm is issued, indicating that there may be an abnormality in the building structure; when H G When (t)<θ(t), it indicates that the monitoring status of the building structure is normal.

[0037] Another object of the present invention is to provide a system for implementing the above-mentioned building structure health monitoring method based on big data, comprising:

[0038] A data acquisition module, used to acquire real-time data from sensors arranged on the building structure;

[0039] A health correlation establishment module, used to calculate the health correlation between sensors based on the sensor data acquired by the data acquisition module and establish a health graph model based on the health correlation;

[0040] A harmonic path weight calculation module is used to calculate the harmonic path weights between sensors based on the health association between sensors using an adaptive harmonic period analysis algorithm;

[0041] The entropy weight and connectivity index calculation module is used to calculate the entropy weight of each sensor based on the harmonic path weight between sensors using the entropy weight connectivity analysis algorithm, and obtain the connectivity index of the health graph model according to the entropy weight calculation;

[0042] A building structure health index calculation module is used to calculate the building structure health index using a building structure health index evaluation algorithm based on the harmonic path weights between sensors and the connectivity index of the health graph model;

[0043] The building structure health judgment module is used to judge the health status of the building structure by comparing the building structure health index with the health threshold.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. The present invention converts the health association between sensors into the form of nodes and edges by establishing a dynamic health graph model, so that the complex sensor data relationship can be presented intuitively. The construction of the health graph model helps to fully capture the health status of the building structure and its changing trend, and provides an effective graphical analysis tool to make structural health monitoring more intuitive and easy to understand;

[0046] 2. The dynamic health correlation calculation algorithm adopted by the present invention adopts nonlinear transformation and dynamic adjustment mechanism, processes sensor data differences through hyperbolic tangent function, and combines standard deviation and standard deviation square dynamic adjustment mechanism to enhance the nonlinear characteristics of data differences, adapt to data volatility, and improve the accuracy of health correlation between sensor data, so that the health status assessment of building structures is accurate and reliable;

[0047] 3. The adaptive harmonic period analysis algorithm used in the present invention calculates the harmonic path weights between sensors by capturing the periodic characteristics of sensor data. The phase offset and sliding window mechanism are introduced to adaptively adjust the periodic parameters, which improves the accuracy of the path weights in the sensor network, can reflect the periodic relationship between sensor data, and helps to identify periodic changes and potential hazards in building structures;

[0048] 4. The entropy weighted connectivity analysis algorithm used in the present invention combines the harmonic path weight and the influence of the sensor in the network to calculate the entropy weight of each sensor, reflecting the connectivity of the sensor network. By calculating the entropy weight, the connectivity and influence of the sensor in the network are quantified, and the overall health status of the building structure is effectively evaluated;

[0049] 5. The building structure health index assessment algorithm calculated by the present invention combines the connectivity index, stiffness coefficient and fatigue life prediction value of the building structure to comprehensively assess the health status of the building structure. It is also sensitive to subtle changes, enhances the stability to outliers, and improves the accuracy of the assessment of the health status of the building structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The figure is a flow chart of a building structure health monitoring method based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present invention will be described clearly and completely below in combination with the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0053] The present invention will be further described below in conjunction with specific embodiments, but the present invention is not limited thereto.

[0054] See attached Figure 1 , which shows a flow chart of a building structure health monitoring method based on big data provided by an embodiment of the present invention, the method comprising the following steps:

[0055] Step 1: Arrange sensors on the building structure, obtain sensor data in real time, calculate the health correlation between sensors based on the obtained sensor data, and establish a health graph model based on the health correlation;

[0056] A variety of sensors are arranged on the building structure to acquire sensor data in real time. The sensor data is used to evaluate the real-time status of the building structure, identify potential anomalies or damage, and ensure the safety and stability of the building. In this embodiment, the arranged sensors include acceleration sensors, displacement sensors, and strain sensors, which are distributed to acquire acceleration data, displacement data, and strain data.

[0057] Since different sensor data have different physical quantities, units and magnitudes, in order to make the sensor data comparable in subsequent calculations, it is necessary to normalize the sensor data to obtain normalized sensor data; the normalization method is a technical means well known to those skilled in the art and will not be described in detail here;

[0058] Based on the normalized sensor data, the dynamic health correlation calculation algorithm is used to calculate the health correlation between sensors, identify the complex relationship between sensor data, and provide an accurate description of the building structure status. Specifically, the dynamic health correlation calculation algorithm uses nonlinear transformation and dynamic adjustment mechanism on the normalized sensor data to capture the complex relationship of sensor data, and then calculates the difference between different sensor data to reflect the different responses of different sensors at a specific time point. The hyperbolic tangent function is introduced to convert the input value into an output between -1 and 1 to effectively capture the small difference to process the difference between different sensor data, which can enhance the nonlinear characteristics of sensor data differences, thereby accurately reflecting the health correlation between different sensors. Finally, a dynamic adjustment mechanism for standard deviation and standard deviation square is introduced. For sensor data, the standard deviation can reflect the volatility or instability of sensor data. When the volatility of sensor data is particularly large, the contribution of the standard deviation square to the denominator will increase significantly. The standard deviation and its square ensure the sensitivity to the volatility of sensor data, making the adjustment of health association more obvious in high volatility or abnormal situations. By combining the standard deviation and the square of the standard deviation, the volatility of sensor data can be adapted. The exponential function is used to transform the range of health association between sensors to (0,1).

[0059] The specific calculation formula for the health correlation between sensors is:

[0060]

[0061] Among them, w ij (t) represents the health correlation between the i-th sensor and the j-th sensor at time t; exp represents the exponential function, which converts the calculated difference into the health correlation to ensure that the health correlation result is in the range of (0,1); |tanh(·)| represents the absolute value of the hyperbolic tangent function, which processes the data difference D i (t)-D j (t), enhance the nonlinear characteristics of sensor data through nonlinear transformation; D i (t) and D j (t) represents the i-th sensor data and the j-th sensor data at time t; σ ij (t) represents the standard deviation of the i-th sensor data and the j-th sensor data at time t; κ represents the adjustment parameter, which adjusts the square term of the standard deviation to adapt to the dynamic changes of the sensor data.

[0062] After calculating the health association between sensors, the health graph model of the building structure is established with sensors as nodes and the health association between sensors as edges. The health graph model is expressed as: G(t) = (V(t), E(t)), where V(t) represents the node set at time t and E(t) represents the edge set at time t.

[0063] Step 2: Based on the health association between sensors, the harmonic path weights between sensors are calculated using an adaptive harmonic cycle analysis algorithm;

[0064] Based on the health correlation between sensors, the harmonic path weights between sensors are calculated using the adaptive harmonic periodic analysis algorithm. Sensor data is the dynamic response data of the building structure at different time points, such as acceleration, displacement, and strain. By performing frequency domain analysis (such as Fourier transform) on the sensor data, the time domain signal can be converted into a frequency domain signal and decomposed into multiple frequency components, each of which corresponds to a harmonic component. The amplitude and phase of the harmonic component can describe the periodic characteristics of the signal and reflect the vibration and stress distribution of the building structure. The adaptive harmonic periodic analysis algorithm improves the accuracy of path calculation by capturing the periodic characteristics of the data and combining the health correlation, so that the harmonic path weight not only considers the periodic changes, but also the health status correlation between sensors, which can more accurately reflect the periodic changes and potential fatigue problems of the building structure. The adaptive harmonic periodic analysis algorithm introduces phase offset to capture the periodic characteristics, and uses the sliding window mechanism to achieve adaptive periodic adjustment, thereby reflecting the periodicity and mutual relationship of sensor data.

[0065] The phase offset is adjusted for each harmonic component, that is, for the current phase offset, the error between the harmonic component and the actual sensor data is calculated, and the value of the phase offset can be iteratively updated by the gradient descent method until the error is minimized. By adjusting the phase offset, the starting point or peak position of the sine wave can be changed to match the actual fluctuation of the sensor data, so that each harmonic component can accurately reflect the periodic correlation between the sensor data;

[0066] The use of sliding window mechanism to achieve adaptive period adjustment is to use fast Fourier transform to perform spectrum analysis on sensor data, preliminarily calculate the period of sensor data, and determine the size and step size of sliding window according to the period. The sliding window mechanism is used to perform segmented analysis on sensor data in order to capture the dynamic changes of the period. Specifically, the least squares fitting method is applied to the sensor data in each sliding window to fit the sine or cosine function, and the local period is determined according to the frequency of the sine or cosine function. The weighted average method is further used to integrate the local period into an adaptive period.

[0067] The specific calculation formula for the harmonic path weight between sensors is:

[0068]

[0069] Among them, h ij (t) represents the harmonic path weight between the i-th sensor and the j-th sensor at time t, reflecting the periodic weight; Represents the summation symbol for n from 1 to N, which is used to accumulate the contribution of all harmonic components; α n Represents the sine coefficient of the nth harmonic component, which is used to adjust the amplitude of the sine wave; It represents the period T(t) and phase offset φ n sine function; 2π is a constant, representing one cycle; T(t) represents the adaptive cycle at time t; φ n Represents the phase shift of the nth harmonic component, which is used to capture the periodic characteristics; γ n Represents the cosine coefficient of the nth harmonic component, which is used to adjust the amplitude of the cosine wave; represents a cosine function with period T(t); w ij (t) represents the health correlation between the i-th sensor and the j-th sensor at time t;

[0070] The harmonic path weights reflect the periodic correlations between sensor data and help identify periodic changes and potential fatigue issues in building structures.

[0071] Step 3: Based on the harmonic path weights between sensors, the entropy weight connectivity analysis algorithm is used to calculate the entropy weight of each sensor, and the connectivity index of the health graph model is obtained according to the entropy weight calculation;

[0072] Based on the harmonic path weights between sensors, the entropy weight connectivity analysis algorithm is used to calculate the entropy weight of each sensor. The entropy weight connectivity analysis algorithm combines the harmonic path weights between sensors and the influence of sensors in the sensor network to calculate the entropy weight of each sensor, thereby reflecting the connectivity of the entire sensor network, helping to identify potential building structural weaknesses and failure points, and improving the ability to comprehensively assess the health of building structures.

[0073] The calculation formula of entropy weight is:

[0074]

[0075] Among them, e i (t) represents the entropy weight of the i-th sensor at time t, which measures the connectivity between sensors; represents the cumulative symbol, |V(t)|-1 represents the number of remaining nodes, that is, the number of sensors remaining except the i-th sensor; represents the modified probability weight, which calculates the connectivity weight between the i-th sensor and the j-th sensor at time t; represents the normalized modified probability weight; h ij (t) represents the harmonic path weight between the i-th sensor and the j-th sensor at time t, reflecting the periodic weight; W represents the logarithm of the modified probability weight, which is used to calculate the information entropy and reflects the uncertainty between the i-th sensor and the j-th sensor; j (t) represents the influence weight of the jth sensor at time t, and its specific calculation formula is:

[0076]

[0077] Among them, d ij (t) represents the actual path distance between the i-th sensor and the j-th sensor at time t;

[0078] Based on the entropy weight of the sensor, the mathematical statistical method is used to first sum and then average the data to obtain the connectivity index of the health graph model. The specific calculation formula is:

[0079]

[0080] Among them, C G (t) represents the connectivity index of the health graph model at time t, reflecting the health status of the overall structure; Represents the average value of the entropy weights of all sensors;

[0081] Step 4: Based on the harmonic path weights between sensors and the connectivity index of the health graph model, the building structure health index is calculated using the building structure health index assessment algorithm, and the building structure health index is compared with the health threshold to achieve building structure health monitoring;

[0082] Based on the harmonic path weights between sensors and the connectivity index of the health graph model, the building structure health index is calculated using the building structure health index assessment algorithm. The building structure health index assessment algorithm combines the connectivity index, stiffness coefficient and fatigue life prediction value of the building structure to comprehensively assess the health status of the building structure. It is also sensitive to subtle changes, improving the ability to accurately assess the health status of the building structure.

[0083] The connectivity index of the health graph model is transformed nonlinearly by using logarithmic functions and high-order power functions to further smooth the connectivity index. The ability of the building structure to resist deformation is reflected by the nonlinear processing of the stiffness coefficient. The fatigue life prediction value of the building structure is processed by exponential functions and inverses. The fatigue life prediction value is indexed and reversed to significantly distinguish the fatigue life situation. Among them, the nonlinear processing of the stiffness coefficient requires balancing the numerator and denominator. The key lies in the selection of the exponent. The exponent of the stiffness coefficient of the denominator is selected to be greater than the exponent of the stiffness coefficient of the numerator to ensure that the denominator can effectively suppress the amplification effect of the numerator.

[0084] The fatigue life prediction value of the building structure combines the connectivity index and harmonic path weight of the health graph model, taking into account the impact of the harmonic path on the fatigue life of the building structure, and quantifying the impact through the fatigue sensitivity factor and frequency sensitivity factor. The fatigue sensitivity factor reflects the fatigue characteristics of building materials and structures, while the frequency sensitivity factor measures the impact of vibration frequency on the fatigue life of the building structure. Therefore, the calculation formula of the building structure health index is:

[0085]

[0086] Among them, H G (t) represents the building structure health index at time t, reflecting the overall health status of the building structure; C G (t) represents the connectivity index of the health graph model at time t, reflecting the health status of the overall structure; log(1+C G (t)) represents the nonlinear transformation of the connectivity index, and uses a logarithmic function to process the connectivity index to enhance the stability against outliers; (C G (t)) m Represents the high power of the connectivity index, which enhances the smoothing effect on outliers; S(t) represents the stiffness coefficient, which reflects the ability of the building structure to resist deformation. The calculation method of the stiffness coefficient is a technical means well known to those skilled in the art and will not be described in detail here; q represents the index of adjusting the stiffness to affect the strength, and the value is required to be greater than 1 to enhance the effect of stiffness on the health index; p represents the index of controlling the stiffness inhibition term to ensure that the growth of the health index tends to be gentle; exp represents an exponential function, which is used to enhance the response capability to extreme situations and helps to accurately evaluate the health status of the building structure when it is close to the fatigue limit; L f (t) represents the predicted value of fatigue life of building structure at time t, reflecting the remaining fatigue life of building structure under existing conditions. The specific calculation formula is:

[0087]

[0088] Among them, h l (t) represents the weight of the lth harmonic path at time t; Indicates the accumulation of contributions of M harmonic paths; I l (t) represents the fatigue sensitivity factor, which reflects the sensitivity of the lth harmonic path to the fatigue life of the building structure. It is obtained from experimental data considering the fatigue characteristics of building materials and structures. represents the frequency sensitivity factor, considering the influence of the vibration frequency of the lth harmonic path on the fatigue life of the building structure; f l (t) represents the vibration frequency of the lth harmonic path at time t, which is collected by the sensor; f0 represents the reference frequency, which represents the frequency of the building structure under normal working conditions, which can be set according to the specific implementation scenario and is not limited here; β represents the frequency adjustment factor, which reflects the influence of vibration on the fatigue life of the building structure; B f (l) represents the allowable number of cycles of the lth harmonic path at the current stress level, which is the basis for calculating the fatigue life of building structures and is obtained from experimental data.

[0089] The building structure health index can intuitively reflect the overall health status of the building structure, help management and maintenance personnel to timely understand the health status of the building structure, make scientific maintenance decisions, and ensure the long-term safety and stability of the building structure.

[0090] When judging the health status of a building, a health threshold θ(t) is set in advance. It can be set according to the specific implementation scenario and is not limited here. The building structure health index is compared with the health threshold θ(t). When H G When (t)≥θ(t), an alarm is issued, indicating that there may be an abnormality in the building structure; when H G When (t)<θ(t), it means that the monitoring status of the building structure is normal. By setting the health threshold, an alarm can be issued in time to prevent structural damage or accidents.

[0091] Examples

[0092] 1. Building Description

[0093] The experimental object is a 12-story reinforced concrete frame structure building with a construction area of ​​12,000 square meters. The building is 15 years old and located in a busy traffic area with significant external vibration. This experiment focuses on evaluating the health of the building under dynamic loads and natural environment.

[0094] 2. Sensor equipment

[0095] Accelerometer (MPU6050): measurement range ±16g, accuracy ±0.001g, arranged at the connection point between the main beam and the column.

[0096] Displacement sensor (LVDT type): measurement accuracy 0.01mm, arranged at the floor support point.

[0097] Strain sensor (HBM strain gauge): measuring range ±5000με, accuracy ±1με, arranged on the building foundation column.

[0098] 3. Sensor layout and monitoring points

[0099] Position 1: Foundation column (strain sensor), monitoring the stress and deformation of the column.

[0100] Position 2: Third-layer main beam (acceleration sensor), monitoring the vibration response of the main beam.

[0101] Position 3: Sixth-floor slab support node (displacement sensor), monitoring slab deformation and displacement.

[0102] Data sampling frequency: 100Hz (100 data points per second), 24 hours of continuous monitoring.

[0103] Steps and data analysis

[0104] 1. Sensor data collection

[0105] Position 1 (base cylinder):

[0106] Strain data: [100,101,99,...] (unit: με).

[0107] Data description: The column data fluctuates slightly, indicating that the column is relatively stable in terms of force.

[0108] Position 2 (main beam):

[0109] Acceleration data: [0.01, 0.02, 0.015, ...] (unit: m / s 2 ).

[0110] Data explanation: The acceleration of the main beam is small, which reflects that the vibration amplitude of the building structure is low and the overall stability is good.

[0111] Position 3 (slab support node):

[0112] Displacement data: [0.1, 0.12, 0.09, ...] (unit: mm).

[0113] Data description: The displacement of the floor support is within the normal range, with no obvious abnormal deformation.

[0114] 2. Calculation of health relevance

[0115] Through the dynamic health correlation calculation algorithm, the correlation between different sensors is quantified, revealing the linkage relationship between different parts of the building structure.

[0116] Calculation formula:

[0117]

[0118] Among them, D i (t) and D j (t) represents the i-th sensor data and the j-th sensor data at time t; σ ij (t) represents the standard deviation between the i-th sensor data and the j-th sensor data at time t; κ represents the adjustment parameter, which is set to 0.1.

[0119] Experimental data analysis results:

[0120] 1. Health correlation between position 1 and position 2:

[0121] Strain data vs acceleration data: vs.

[0122] Standard deviation 12 (t) = 0.005;

[0123] Calculation result w 12 (t) = 0.85.

[0124] 2. Health correlation between position 2 and position 3:

[0125] Acceleration data vs displacement data: vs.

[0126] Standard deviation 23 (t) = 0.004;

[0127] Calculation result w 23 (t) = 0.90.

[0128] 3. Harmonic path weight calculation

[0129] An adaptive harmonic period analysis algorithm is used to calculate the harmonic path weights between sensors to capture the periodic vibration characteristics of the building structure and reflect the impact of periodic loads in the building.

[0130] Calculation formula:

[0131]

[0132] Parameter settings:

[0133] N = 3 (number of harmonic components);

[0134] Adaptive period T(t) = 2 seconds;

[0135] Harmonic adjustment coefficient α n =0.7,γ n =0.3, phase shift φ n =0.1.

[0136] Calculation results:

[0137] 1. Harmonic path weights for position 1 and position 2:

[0138] Calculation result h 12 (t) = 0.72.

[0139] 2. Harmonic path weights for positions 2 and 3:

[0140] Calculation result h 23 (t) = 0.75.

[0141] The harmonic path weights show that the periodic response of the building structure is more obvious, especially between the main beams and floor supports, where there may be a certain amount of fatigue accumulation.

[0142] 4. Calculation of entropy weight and connectivity index

[0143] Entropy weights are used to assess the relative influence of each sensor in the network. By calculating entropy weights, the overall connectivity of the building structure can be measured and possible weaknesses of the building structure can be revealed.

[0144] Entropy weight calculation formula:

[0145]

[0146] Among them, W j (t) represents the influence weight of the jth sensor at time t, and the specific calculation formula is:

[0147]

[0148] Connectivity index calculation formula:

[0149]

[0150] C G (t): The overall connectivity index of the building structure, reflecting the health status of the building.

[0151] result:

[0152] 1. Entropy weight of position 1 (base cylinder): e1(t) = 0.65.

[0153] 2. Entropy weight of position 2 (main beam): e2(t)=0.70.

[0154] 3. Entropy weight of position 3 (floor): e3(t)=0.68.

[0155] Overall connectivity index C G (t) = 0.68.

[0156] The connectivity of the building structure is good, the network correlation of each sensor is strong, the building as a whole is in a healthy state, and no obvious structural weaknesses are found.

[0157] 5. Building structure health index assessment

[0158] The building structural health index provides a quantitative health assessment tool by comprehensively considering connectivity indicators, stiffness and fatigue life of building structures.

[0159] Health index calculation formula:

[0160]

[0161] Parameter settings:

[0162] m = 2, q = 1.5, p = 1.2;

[0163] Stiffness coefficient S(t) = 0.85;

[0164] Fatigue life prediction formula:

[0165]

[0166] Parameter settings:

[0167] I l (t)=0.8,f l (t)=200,f0=1000,B f (l) = 100, β = 0.5, M = 3;

[0168] L f (t) = 10000 hours

[0169] Calculation results:

[0170] Building structure health index H G (t) = 1.25

[0171] Threshold settings:

[0172] To evaluate the health status of the building, the threshold θ(t) = 1 is set, and the building structure health index is compared with the threshold. G When (t)≥θ(t), an alarm is issued, indicating that there may be an abnormality in the building structure; when H G When (t)<θ(t), it indicates that the monitoring status of the building structure is normal.

[0173] If the health index is higher than the threshold, it indicates that the building structure is in an unsafe state and requires maintenance or repair. The fatigue condition of the main beams and floor support points should be monitored regularly to prevent the accumulation of potential problems.

[0174] Through the verification of this experiment, the building structure health monitoring method based on big data can accurately reflect the dynamic health status of the building. The experimental results show that the health index of the building under the current working conditions is 1.25, which is higher than the threshold of 1, indicating that the overall structure of the building is abnormal. Maintenance or repair is recommended, and more detailed monitoring of the fatigue status of the main beam and floor support points is recommended to ensure the long-term safety of the building.

[0175] The embodiment of the present invention further provides a system for implementing the above-mentioned building structure health monitoring method based on big data, comprising:

[0176] A data acquisition module, used to acquire real-time data from sensors arranged on the building structure;

[0177] A health correlation establishment module, used to calculate the health correlation between sensors based on the sensor data acquired by the data acquisition module and establish a health graph model based on the health correlation;

[0178] A harmonic path weight calculation module is used to calculate the harmonic path weights between sensors based on the health association between sensors using an adaptive harmonic period analysis algorithm;

[0179] The entropy weight and connectivity index calculation module is used to calculate the entropy weight of each sensor based on the harmonic path weight between sensors using the entropy weight connectivity analysis algorithm, and obtain the connectivity index of the health graph model according to the entropy weight calculation;

[0180] A building structure health index calculation module is used to calculate the building structure health index using a building structure health index evaluation algorithm based on the harmonic path weights between sensors and the connectivity index of the health graph model;

[0181] The building structure health judgment module is used to judge the health status of the building structure by comparing the building structure health index with the health threshold.

[0182] The above are only preferred embodiments of the present invention, and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the contents of the specification of the present invention should be included in the protection scope of the present invention.

Claims

1. A building structure health monitoring method based on big data, characterized in that: The following steps are involved: Step 1: Arrange sensors on the building structure, obtain sensor data in real time, calculate the health correlation between sensors based on the obtained sensor data, and establish a health graph model based on the health correlation; Step 2: Based on the health association between sensors, the harmonic path weights between sensors are calculated using an adaptive harmonic cycle analysis algorithm; Step 3: Based on the harmonic path weights between sensors, the entropy weight connectivity analysis algorithm is used to calculate the entropy weight of each sensor, and the connectivity index of the health graph model is obtained according to the entropy weight calculation; Step 4: Based on the harmonic path weights between sensors and the connectivity index of the health graph model, the building structure health index is calculated using the building structure health index assessment algorithm, and the building structure health index is compared with the health threshold to achieve building structure health monitoring.

2. The building structure health monitoring method based on big data according to claim 1 is characterized in that: The specific method for calculating the health association between sensors in step 1 is: Nonlinear transformation and dynamic adjustment mechanism are used for sensor data to capture the complex relationship of sensor data. The hyperbolic tangent function is introduced to calculate the difference between different sensor data to deal with the difference between different sensor data. Then, the dynamic adjustment mechanism of standard deviation and standard deviation square is introduced to adapt to the volatility of sensor data. Finally, the exponential function is used to transform the health correlation between sensors into the range of (0,1). The specific calculation formula for the health correlation between sensors is: Among them, w ij (t) represents the health correlation between the i-th sensor and the j-th sensor at time t; exp represents the exponential function; |tanh(·)| represents the absolute value of the hyperbolic tangent function, and the processing data difference D i (t)-D j (t); D i (t) and D j (t) represents the i-th sensor data and the j-th sensor data at time t; σ ij (t) represents the standard deviation between the i-th sensor data and the j-th sensor data at time t; k represents the adjustment parameter.

3. The building structure health monitoring method based on big data according to claim 1 is characterized in that: The construction method of the health graph model is: taking sensors as nodes and the health correlation between sensors as edges, a health graph model is established. The health graph model is expressed as: G(t) = (V(t), E(t)), where V(t) represents the node set at time t, and E(t) represents the edge set at time t.

4. The building structure health monitoring method based on big data according to claim 1 is characterized in that: The calculation method for calculating the harmonic path weight in step 2 includes: Phase offset is used to capture the periodic characteristics of the sensor, and the sliding window mechanism is used to achieve adaptive period adjustment, and the harmonic path weight is calculated in combination with the health association degree; Phase offset is to calculate the error between the harmonic component and the actual sensor data for the current phase offset, and iteratively update the value of the phase offset through the gradient descent method until the error is minimized, thereby matching the actual fluctuation of the sensor data; The method of using the sliding window mechanism to achieve adaptive period adjustment is to use fast Fourier transform to perform spectral analysis on sensor data, preliminarily calculate the period of sensor data, and determine the size and step size of the sliding window according to the period. The least squares fitting method is applied to the sensor data in each sliding window to fit a sine or cosine function, and the local period is determined according to the frequency of the sine or cosine function. Finally, the weighted average method is used to integrate the local period into an adaptive period.

5. The building structure health monitoring method based on big data according to claim 4 is characterized in that: The specific calculation formula for the harmonic path weight between sensors is: Among them, h ij (t) represents the harmonic path weight between the i-th sensor and the j-th sensor at time t; Represents the summation symbol for n from 1 to N; α n Represents the sinusoidal coefficient of the nth harmonic component; It represents the period T(t) and phase offset φ n sine function; 2π is a constant, representing one cycle; T(t) represents the adaptive cycle at time t; φ n represents the phase shift of the nth harmonic component; γ n Represents the cosine coefficient of the nth harmonic component; represents a cosine function with period T(t); w ij (t) represents the health correlation between the i-th sensor and the j-th sensor at time t.

6. The building structure health monitoring method based on big data according to claim 1 is characterized in that: In step 3, the entropy weight of each sensor is calculated by combining the harmonic path weights between sensors and the influence of the sensor in the sensor network. The calculation formula of the entropy weight is: Among them, e i (t) represents the entropy weight of the i-th sensor at time t, which measures the connectivity between sensors; represents the cumulative symbol, |V(t)|-1 represents the number of remaining nodes, that is, the number of sensors remaining except the i-th sensor; represents the modified probability weight, which calculates the connectivity weight between the i-th sensor and the j-th sensor at time t; represents the normalized modified probability weight; h ij (t) represents the harmonic path weight between the i-th sensor and the j-th sensor at time t, reflecting the periodic weight; W represents the logarithm of the modified probability weight, which is used to calculate the information entropy and reflects the uncertainty between the i-th sensor and the j-th sensor; j (t) represents the influence weight of the jth sensor at time t, and the specific calculation formula is: Among them, d ij (t) represents the actual path distance between the i-th sensor and the j-th sensor at time t.

7. The building structure health monitoring method based on big data according to claim 6 is characterized in that: In step 3, based on the entropy weight of the sensor, a mathematical statistical method is used to first sum and then average the data to obtain the connectivity index of the health graph model. The specific calculation formula is: Among them, C G (t) represents the connectivity index of the health graph model at time t, reflecting the health status of the overall structure; represents the average entropy weight of all sensors, and V(t) represents the node set at time t.

8. The building structure health monitoring method based on big data according to claim 1 is characterized in that: Combining the connectivity index and harmonic path weight of the health graph model, the influence of the harmonic path on the fatigue life of the building structure is considered, and the influence is quantified by the fatigue sensitivity factor and the frequency sensitivity factor to calculate the building structure health index; the calculation formula of the building structure health index is: Among them, H G (t) represents the building structure health index at time t; C G (t) represents the connectivity index of the health graph model at time t; log(1+C G (t)) represents the nonlinear transformation of connectivity index; (C G (t)) m represents the high power of the connectivity index; S(t) represents the stiffness coefficient; q represents the index of the influence of the stiffness on the strength; p represents the index of the stiffness suppression term; exp represents the exponential function; l f (t) represents the predicted value of fatigue life of the building structure at time t.

9. The building structure health monitoring method based on big data according to claim 1 is characterized in that: In step 4, the health threshold θ(y) is set and the building structure health index H G (t) is compared with the health threshold θ(t), when H G When (t)≥θ(t), an alarm is issued, indicating that there may be an abnormality in the building structure; when H G When (t)<θ(t), it indicates that the monitoring status of the building structure is normal.

10. A system for implementing the building structure health monitoring method based on big data according to any one of claims 1 to 9, characterized in that: include: A data acquisition module, used to acquire real-time data from sensors arranged on the building structure; A health correlation establishment module, used to calculate the health correlation between sensors based on the sensor data acquired by the data acquisition module and establish a health graph model based on the health correlation; A harmonic path weight calculation module is used to calculate the harmonic path weights between sensors based on the health association between sensors using an adaptive harmonic period analysis algorithm; The entropy weight and connectivity index calculation module is used to calculate the entropy weight of each sensor based on the harmonic path weight between sensors using the entropy weight connectivity analysis algorithm, and obtain the connectivity index of the health graph model according to the entropy weight calculation; A building structure health index calculation module is used to calculate the building structure health index using a building structure health index evaluation algorithm based on the harmonic path weights between sensors and the connectivity index of the health graph model; The building structure health judgment module is used to judge the health status of the building structure by comparing the building structure health index with the health threshold.

Images