Engineering skin-based engineering structure performance state inversion method, safety early warning method and system

Abstract

The invention discloses an engineering structure performance state inversion method, a safety early warning method and a safety early warning system based on engineering skin. The inversion method comprises the following steps: acquiring multi-source data acquired by multiple types of sensors arranged on a to-be-monitored engineering structure; preprocessing the multi-source data; fusing the multi-source data by adopting multi-rate Kalman filtering based on a neural network to obtain multi-source data with a unified time step; taking the data of the plurality of flexible piezoelectric sensors after the unified time step as model input, inputting the data into a trained neural network-based key point engineering demand parameter inversion model, and performing inversion to obtain engineering demand parameters of a plurality of key points of the engineering structure; and correcting an inversion result by using a Bayesian algorithm in combination with data of corresponding time steps of the external source sensor. According to the method, the key point engineering demand parameters can be quickly inverted, the inversion efficiency is high, and the precision is high.

Description

Technical Field

[0001] The present invention relates to the field of engineering health monitoring and early warning, and in particular to an engineering structure performance state inversion method based on engineering skin, a safety early warning method and a system. Background Art

[0002] Structural health monitoring (SHM) is a comprehensive damage monitoring of engineering structures, whether whole or in part. By installing sensors at key locations on the structure, damage is monitored and safety warnings are issued. Data analysis can also be used to determine the degree of structural aging, providing data support for repair and assessment.

[0003] Traditional engineering structure monitoring relies primarily on deploying sensors throughout the structure. However, the response behavior of the structure, as measured through field testing, is often limited to a limited number of locations and cannot fully reflect the overall state of the structure. High-density global monitoring with sensors is expensive and even limited by the spatial constraints of sensor deployment, making sensor monitoring difficult in many critical locations (such as within the structure). While traditional numerical methods such as finite element simulation can provide highly accurate global stress, strain, and other response information, processing this global data requires significant computational effort, consumes significant computing resources, and lacks real-time performance, making it difficult to ensure timely safety warnings. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide an engineering structure performance state inversion method, a safety warning method and a system based on engineering skin, which can realize the efficient inversion of key engineering demand parameters, so as to realize global response monitoring of engineering structures and ensure the timeliness of safety warnings.

[0005] In a first aspect, a method for inverting the performance state of an engineering structure based on an engineering skin is provided, comprising the following steps:

[0006] Acquire multi-source data collected by an engineering skin deployed on the engineering structure to be monitored and multiple external sensors; wherein the engineering skin has multiple built-in flexible piezoelectric sensors;

[0007] Preprocess multi-source data by filtering and removing outliers;

[0008] Multi-source data are fused using a multi-rate Kalman filter based on a neural network to obtain multi-source data with a unified time step.

[0009] The data of multiple flexible piezoelectric sensors after a unified time step are used as model input and input into the trained neural network-based key point engineering requirement parameter inversion model to obtain the engineering requirement parameters of multiple key points of the engineering structure (these parameters can reflect the performance status of the engineering structure). The inversion results are corrected using the Bayesian algorithm combined with the data of the corresponding time step from the exogenous sensors.

[0010] According to the first aspect, in some possible implementations, the exogenous sensor includes one or more of a visual sensor, a piezoresistive strain gauge, and a laser displacement sensor.

[0011] According to the first aspect, in some possible implementations, the neural network on which the key point engineering requirement parameter inversion model is based is a physical information neural network embedded with elastic mechanics equilibrium equations and / or coordination equations.

[0012] According to the first aspect, in some possible implementations, the key point engineering requirement parameter inversion model is trained by the following method:

[0013] Using data from multiple flexible piezoelectric sensors as model input and engineering requirement parameters of multiple key points as model output, a sample dataset is constructed; the sample dataset is divided into a training set and a validation set.

[0014] Constructing a physical information neural network embedded with elastic equilibrium equations and / or coordination equations;

[0015] The physical information neural network is trained and verified using the test set and validation set to obtain the inversion model of key point engineering requirement parameters.

[0016] According to the first aspect, in some possible implementations, the model input includes the position coordinates of each flexible piezoelectric sensor, the collected data, and the distance from the key point.

[0017] According to the first aspect, in some possible implementations, the inversion result is corrected using a Bayesian algorithm in combination with data corresponding to a time step of an exogenous sensor, specifically including:

[0018] The network model parameters of the trained key point engineering requirement parameter inversion model are taken as the mean, the coefficient of variation and distribution type are selected, the prior probability distribution of the network model parameters is set, and the initial uncertainty of the network model parameters is characterized;

[0019] Based on the data collected by exogenous sensors, considering the weight differences caused by the accuracy of different exogenous sensors, a multi-source fusion likelihood function is established between the inversion results of the cross-section and the measured data to quantify the consistency between the measured data and the model prediction; the cross-section refers to the part where the monitoring points of the exogenous sensor overlap with multiple key points;

[0020] Applying Bayesian theorem, the prior probability distribution is integrated with the multi-source fusion likelihood function. Through Markov chain Monte Carlo sampling and subset simulation, the posterior distribution samples of the network model parameters are obtained. The posterior probability distribution of the network model parameters is obtained by statistical fitting.

[0021] According to the posterior distribution of the network model parameters, the probability distribution of the engineering requirement parameters of the key points is recalculated, and the highest probability density point estimation method is used to correct the inversion results.

[0022] In a second aspect, a safety early warning method for an engineering structure based on an engineering skin is provided, comprising the following steps:

[0023] The engineering structural performance state inversion method based on engineering skin as described above is used to invert the engineering demand parameters of multiple key points;

[0024] Engineering risk level assessment and safety warning are carried out based on the engineering requirement parameters and reliability analysis of multiple key points obtained by inversion.

[0025] According to the second aspect, in some possible implementations, the engineering risk level assessment and safety warning are performed based on the engineering requirement parameters and reliability analysis of multiple key points obtained based on the inversion, specifically including:

[0026] Based on the probability distribution of engineering requirement parameters at multiple key points corrected by the Bayesian algorithm, combined with the preset dynamic model and industry specifications, the reliability is calculated and the probability distribution of all required engineering requirement parameters is inverted. The probability distribution of local damage indicators and global stability indicators is also calculated.

[0027] Based on the calculated probability distribution of each indicator, the risk level is divided according to the following rules:

[0028] Low risk: The mean of all indicators is less than 70% of the normative threshold, and the reliability is ≥ 90%;

[0029] Medium risk: the mean of at least one indicator is between 70% and 90% of the standard threshold, or the reliability is between 70% and 90%;

[0030] High risk: The mean of any indicator exceeds 90% of the standard threshold, or the reliability is less than 70%;

[0031] Provide security warnings based on risk levels.

[0032] In a third aspect, an engineering structure performance state inversion system based on engineering skin is provided, comprising:

[0033] An engineering skin and multiple exogenous sensors are arranged on the engineering structure to be monitored; wherein the engineering skin has multiple flexible piezoelectric sensors built in;

[0034] Data communication module, used to obtain data collected by each engineered skin and exogenous sensors and transmit it to the data processing module;

[0035] The data processing module is configured to execute the engineering structure performance state inversion method based on engineering skin as described above.

[0036] Fourthly, an engineering structure safety early warning system based on engineering skin is provided, comprising:

[0037] An engineering skin and multiple exogenous sensors are arranged on the engineering structure to be monitored; wherein the engineering skin has multiple flexible piezoelectric sensors built in;

[0038] Data communication module, used to obtain data collected by each engineered skin and exogenous sensors and transmit it to the data processing module;

[0039] The data processing module is configured to execute the engineering structure safety early warning method based on engineering skin as described above.

[0040] The present invention proposes an engineering structure performance state inversion method, safety early warning method and system based on engineering skin, which has the following beneficial effects compared with the existing technology:

[0041] (1) By deploying sensing equipment at some measuring points on the engineering structure to be monitored, the engineering requirement parameters of key points can be inverted, thereby realizing the global response monitoring of the engineering structure, which greatly reduces the installation and maintenance costs;

[0042] (2) Using data from some measurement points and combining it with deep learning technology, we can achieve the inversion of key engineering requirement parameters. Compared with the traditional finite element method, this method greatly reduces the amount of calculation and the consumption of computing resources, greatly improves the real-time performance and efficiency of the inversion of key engineering requirement parameters, and ensures the timeliness of safety warnings.

[0043] (3) Multi-source data was collected, and the inversion results were corrected using the Bayesian algorithm based on the data collected by external sensors, thereby improving the inversion accuracy of the engineering requirement parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1This is a flow chart of an engineering structure performance state inversion method based on engineering skin provided by an embodiment of the present invention;

[0046] Figure 2 This is a flow chart of an engineering structure safety early warning method based on engineering skin provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] To make the objectives, technical solutions, and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other implementations obtained by those of ordinary skill in the art without inventive effort are within the scope of protection of the present invention.

[0048] like Figure 1 As shown, an embodiment of the present invention provides an engineering structure performance state inversion method based on engineering skin, comprising the following steps:

[0049] S1: Acquire the engineering skin deployed on the engineering structure to be monitored and multi-source data collected by multiple external sensors; the engineering skin has multiple built-in flexible piezoelectric sensors.

[0050] The exogenous sensor includes one or more of a visual sensor, a piezoresistive strain gauge, and a laser displacement sensor. The visual sensor or laser displacement sensor can be used to monitor the displacement of a measuring point and convert the displacement into strain or stress; the piezoresistive strain gauge can be built-in or externally attached and can be used to monitor the strain of a measuring point (generally different from the flexible piezoelectric sensor measuring point) and convert the strain into stress or displacement. In this embodiment, the engineered skin includes a flexible piezoelectric sensor, a flexible conductor, and a signal conversion module, and the engineered skin can adapt to a variety of complex and extreme environments. In other embodiments, the engineered skin can also be replaced by a flexible piezoelectric integrated sensor. The flexible piezoelectric integrated sensor consists of piezoelectric ceramics, piezoresistive devices, and a negative Poisson's ratio displacement conversion device. It has the advantages of high flexibility, strong surface adaptability, strong environmental adaptability, the ability to measure large deformations, and a wide-band response. The piezoresistive device is mainly used to monitor low-frequency strain responses and provide an indication for the directional measurement of displacement; the piezoelectric device is mainly used to monitor high-frequency strain responses.

[0051] S2: Preprocessing of multi-source data, including filtering and outlier removal.

[0052] S3: A multi-rate Kalman filter based on a neural network is used to fuse multi-source data to obtain multi-source data with a unified time step.

[0053] Taking the data with higher sampling frequency collected by flexible piezoelectric sensors as input and the data with lower sampling frequency such as visual sensors and piezoresistive strain gauges as target, multi-source data fusion is performed on the low sampling frequency data and the high sampling frequency data to obtain multi-type data results with unified and coordinated sampling frequency and noise distribution, which can be compared and verified at the same time step.

[0054] A multi-rate Kalman filter based on a neural network is used to fuse multi-source data. It aims to process asynchronously sampled multi-source data, facilitate subsequent processing, broaden the signal frequency domain, suppress high-frequency noise and low-frequency drift, and have good robustness to noise, thereby improving the accuracy of multi-source data fusion and providing high-reliability key parameters for engineering demand parameter inversion and engineering structure safety warning.

[0055] In this embodiment, the specific process of fusing multi-source data is as follows:

[0056] S31: Multi-rate data synchronization: Based on interpolation alignment, Kalman prediction, and timestamp unification, it unifies the data sampling frequency of flexible piezoelectric sensors (or flexible piezoelectric-piezoresistive integrated sensors), visual sensors, piezoresistive strain gauges, etc., and broadens the signal frequency domain.

[0057] S32: Multi-rate Kalman filter fusion: Through the dynamic model prediction and update steps of the Kalman filter, measurement noise and sensor drift are suppressed to improve signal accuracy. Using low-frequency measurement data as input (from visual sensors or piezoresistive strain gauges, etc.), the Kalman filter algorithm can estimate the system state at the current step based on the state estimate of the system in the previous step and the observation of the current step, realizing "nonlinear interpolation" of low-frequency data to obtain high-frequency data. During the prediction process, the Kalman gain is estimated through the recurrent neural network GRU to avoid dependence on given noise statistical parameters (such as noise variance and noise distribution).

[0058] S33: Result verification and optimization: Based on residual calculation, parameter tuning, and robustness testing, the fused state estimation is optimized to meet the requirements of accuracy and robustness; the time series backpropagation algorithm is used to continuously optimize the network parameters and improve the performance of the output high-frequency signal; when training the network, the training set can be obtained based on downsampling of the high-frequency data, and the original high-frequency data can be used as the true value to verify the combined performance of the trained neural network-based multi-rate Kalman filter.

[0059] It should be noted that the basic implementation principle of the multi-rate Kalman filter based on the neural network is an existing technology, which is not the focus of the present invention and will not be described in detail here.

[0060] S4: The data of multiple flexible piezoelectric sensors after a unified time step are used as model input, and are input into the trained neural network-based key point engineering requirement parameter inversion model to obtain the engineering requirement parameters of multiple key points of the engineering structure through inversion; the inversion results are corrected using the Bayesian algorithm in combination with the data of the corresponding time step of the exogenous sensor.

[0061] Among them, the engineering requirement parameters can be stress, strain, displacement, etc. In this embodiment, strain data collected based on multiple flexible piezoelectric sensors can be selected as model input to predict the strain data of multiple key points, and then the predicted strain data can be converted into stress or displacement data.

[0062] In this embodiment, the key point engineering requirement parameter inversion model is trained by the following method:

[0063] S41: Use the data of multiple flexible piezoelectric sensors as model input and the engineering requirement parameters of multiple key points as model output to construct a sample data set; divide the sample data set into a training set and a validation set.

[0064] The model input of each sample is the real data monitored by the deployed flexible piezoelectric sensors, and the model output of the sample is obtained through finite element simulation or through finite element simulation combined with real data monitored by external sensors. In some embodiments, the model input can optionally include only the strain data collected by each flexible piezoelectric sensor; in some preferred embodiments, the model input includes the position coordinates of each flexible piezoelectric sensor, the collected data, and the distance from the key point; by incorporating the position coordinates of each flexible piezoelectric sensor and the distance from the key point, and incorporating physical space characteristics, the inversion accuracy of the model can be improved; the distance from the key point can be L1 distance or L2 distance, etc.

[0065] S42: Construct a physical information neural network embedded with elastic equilibrium equations and / or coordination equations.

[0066] Specifically, the main architecture of the physical information neural network can use a neural network based on the spatiotemporal attention mechanism, such as Transformer, FredFormer network, etc., and embed the elastic mechanics equilibrium equation and / or coordination equation in the loss function.

[0067] Damage function L of physical information neural network total It is expressed as follows:

[0068] L total =L data +λ balance L balance +λ coordination L coordination

[0069] Among them, L data Represents the regression loss, which is composed of the MSE, MAE or RMSE loss between the model prediction value and the true value; L balance and λ balance Respectively represent the residual loss of the elastic mechanics equilibrium equation and the corresponding weight; when λ balance When it is 0, it means that the physical information neural network only embeds the coordination equation; L coordination and λ coordination Represent the residual loss of the coordination equation and the corresponding weight respectively; when λ coordination When it is 0, it means that the physical information neural network only embeds the elastic mechanics equilibrium equation.

[0070] The coordination equations include normal strain coordination equation and shear strain coordination equation;

[0071] The normal strain compatibility equation is expressed as follows:

[0072]

[0073] The shear strain compatibility equation is expressed as follows:

[0074]

[0075] Where, ε 11 , ε 22 , ε 33 are all positive strain components; ε 12 , ε 23 , ε 13 are all shear strain components; x1, x2, and x3 represent the directions of the three orthogonal coordinate axes in the three-dimensional Cartesian coordinate system;

[0076] The equilibrium equation of elasticity is expressed as follows:

[0077]

[0078] Where σ xx , σ yy , σ zz are all normal stress components; τ xy =τ yx , τ yz =τ zy , τ zx =τ xz are all shear stress components; F x 、F y 、F z are all body force components; x, y, and z represent the three orthogonal coordinate axis directions in the three-dimensional Cartesian coordinate system, which are equivalent to the aforementioned x1, x2, and x3.

[0079] Residual loss L of the elastic equilibrium equationbalance It is expressed as follows:

[0080]

[0081] Where N is the number of samples; R balance (i) represents the residual of the elastic mechanics equilibrium equation corresponding to the i-th sample, which is the sum of the residuals of the elastic mechanics equilibrium equation in all directions;

[0082] Coordination equation residual loss L coordination It is expressed as follows:

[0083]

[0084] Where R coordination (i) represents the residual of the compatibility equation corresponding to the i-th sample, which is the sum of the residuals of all normal strain compatibility equations and all shear strain compatibility equations.

[0085] S43: Use the test set and validation set to train and validate the physical information neural network to obtain the inversion model of key point engineering requirement parameters.

[0086] After training the inversion model of key point engineering requirement parameters, the data collected in real time by the flexible piezoelectric sensor can be used to construct the model input and quickly invert the engineering requirement parameters of the key points.

[0087] In some preferred embodiments, the overlap between the monitoring points of the exogenous sensor and multiple key points is taken into account, i.e., the intersection. Therefore, the real engineering requirement parameters of the key points can be obtained through the exogenous sensor in the intersection. Therefore, this data can be used to optimize the inversion model of the key point engineering requirement parameters to improve the global (including non-intersection) inversion accuracy of the key point engineering requirement parameter inversion model. Specifically, the inversion results are corrected using the Bayesian algorithm in combination with the data of the corresponding time step of the exogenous sensor, which specifically includes:

[0088] S401: The network model parameters of the trained key point engineering requirement parameter inversion model are taken as the mean, the coefficient of variation and distribution type are selected, and the prior probability distribution π(θ) of the network model parameters is set to represent the initial uncertainty of the network model parameters. The prior probability distribution π(θ) of the network model parameters is expressed as follows:

[0089]

[0090] Where, To train the network model parameter values for the key point engineering requirement parameter inversion model; σ θ and f specifies the standard deviation and distribution type (uniform, normal, lognormal, etc.);

[0091] S402: Based on the data collected by the external sensors, considering the weight differences caused by the accuracy of different external sensors, a multi-source fusion likelihood function L(y|θ) is established between the cross-part inversion results and the measured data to quantify the consistency between the measured data and the model prediction. The multi-source fusion likelihood function L(y|θ) is expressed as follows:

[0092]

[0093] In the formula, y=(y1,y2,...,y n ) is the engineering requirement parameter data measured by external sensors such as visual sensors and piezoresistive strain gauges, and n is the number of external sensors; is the weight coefficient of the i-th external sensor data; h(·) is the inversion model of the key point engineering demand parameters; is the model input based on the data collected by the flexible piezoelectric sensor; θ is the network model parameter; R is the coordinate of the key point

[0094] S403: Apply Bayesian theorem to fuse the prior probability distribution with the multi-source fusion likelihood function. Through Markov chain Monte Carlo sampling and subset simulation, we obtain the posterior distribution samples of the network model parameters. Statistical fitting is used to obtain the posterior probability distribution p(θ) of the network model parameters, which is expressed as follows:

[0095]

[0096] S404: Based on the posterior distribution of the network model parameters, the probability distribution of the engineering requirement parameters of the key points is recalculated, and the inversion results are corrected using the highest probability density point estimation method. The process is shown as follows:

[0097]

[0098] Where h R The revised value of the engineering requirement parameter of the key point with coordinate R.

[0099] The above embodiment provides an engineering structure performance state inversion method based on engineering skin, which has the following advantages: by deploying sensing equipment at some measuring points on the engineering structure to be monitored, the engineering demand parameters of key points can be inverted, thereby realizing global response monitoring of the engineering structure, greatly reducing the installation and maintenance costs; using the data of some measuring points in combination with deep learning technology to realize the inversion of engineering demand parameters of key points, compared with the traditional finite element method, the amount of calculation is greatly reduced, the consumption of computing resources is reduced, and the real-time and efficiency of the inversion of engineering demand parameters of key points are greatly improved, which can ensure the timeliness of safety warnings; multi-source data is collected, and the inversion results are corrected using the Bayesian algorithm based on the data collected by external sensors, thereby improving the inversion accuracy of the engineering demand parameters.

[0100] Based on the above embodiment, a method for inverting the performance state of an engineering structure based on an engineering skin is provided. The embodiment of the present invention further provides a method for early warning of the safety of an engineering structure based on an engineering skin. Figure 2 As shown, the following steps are included:

[0101] S1: Acquire multi-source data collected by an engineering skin deployed on the engineering structure to be monitored and multiple external sensors; the engineering skin has multiple built-in flexible piezoelectric sensors;

[0102] S2: Preprocess multi-source data by filtering and removing outliers;

[0103] S3: Multi-source data are fused using a multi-rate Kalman filter based on a neural network to obtain multi-source data with a unified time step;

[0104] S4: Using the data of multiple flexible piezoelectric sensors after a unified time step as model input, input it into the trained neural network-based key point engineering requirement parameter inversion model to obtain the engineering requirement parameters of multiple key points of the engineering structure; combining the data of the corresponding time step of the external sensor, the Bayesian algorithm is used to correct the inversion results;

[0105] S5: Based on the engineering requirement parameters and reliability analysis of multiple key points obtained by inversion, engineering risk level assessment and safety warning are carried out.

[0106] In this embodiment, the engineering risk level assessment and safety warning are performed based on the engineering requirement parameters and reliability analysis of multiple key points obtained based on the inversion, specifically including:

[0107] S51: Based on the probability distribution of engineering requirement parameters at multiple key points corrected by the Bayesian algorithm, combined with the preset dynamic model and industry specifications, the reliability is calculated and the probability distribution of all required engineering requirement parameters is inverted. The probability distribution of local damage indicators and global stability indicators is also calculated.

[0108] Among them, reliability is the probability that the engineering requirement parameter is lower than the specification threshold. The probability (reliability) of the engineering requirement parameter being lower than the specification threshold can be calculated based on the posterior probability distribution:

[0109]

[0110] Where p s is the reliability, h thr Indicates the specification threshold of the engineering requirement parameter;

[0111] Among them, local damage indicators include stress, strain, displacement at specific points (and specific components or weighted quantities of the above physical quantities, such as the first stress component, von Mises stress, etc.); global stability indicators include inter-story displacement angle, residual bearing capacity, damage index (generally defined by the ratio of elastic moduli), etc. The calculation methods of local damage indicators and global stability indicators are all existing technologies and are not the focus of the present invention, so they will not be elaborated here.

[0112] S52: Based on the calculated probability distribution of each indicator, the risk level is divided according to the following rules:

[0113] Low risk: The mean of all indicators is less than 70% of the normative threshold, and the reliability is ≥ 90%;

[0114] Medium risk: the mean of at least one indicator is between 70% and 90% of the standard threshold, or the reliability is between 70% and 90%;

[0115] High risk: The mean of any indicator exceeds 90% of the standard threshold, or the reliability is less than 70%;

[0116] S53: Provide safety warnings based on risk levels, including real-time warning signal generation and warning information visualization.

[0117] In some embodiments, the accuracy of the early warning can also be verified based on subsequent monitoring data, and the above-mentioned risk level classification thresholds can be updated.

[0118] In addition, an embodiment of the present invention further provides an engineering structure performance state inversion system based on engineering skin, comprising:

[0119] An engineering skin and multiple exogenous sensors are arranged on the engineering structure to be monitored; wherein the engineering skin has multiple flexible piezoelectric sensors built in;

[0120] Data communication module, used to obtain data collected by each engineered skin and exogenous sensors and transmit it to the data processing module;

[0121] The data processing module is configured to execute the engineering structure performance state inversion method based on engineering skin as described above.

[0122] In addition, an embodiment of the present invention further provides an engineering structure safety early warning system based on engineered skin, comprising:

[0123] An engineering skin and multiple exogenous sensors are arranged on the engineering structure to be monitored; wherein the engineering skin has multiple flexible piezoelectric sensors built in;

[0124] Data communication module, used to obtain data collected by each engineered skin and exogenous sensors and transmit it to the data processing module;

[0125] The data processing module is configured to execute the engineering structure safety early warning method based on engineering skin as described above.

[0126] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0127] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0128] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0129] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0130] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0131] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for inverting the performance state of an engineering structure based on an engineering skin, characterized in that: The steps include: Acquire multi-source data collected by the engineering skin deployed on the engineering structure to be monitored and multiple external sensors; The engineered skin has multiple flexible piezoelectric sensors built into it; Preprocess multi-source data by filtering and removing outliers; Multi-source data are fused using a multi-rate Kalman filter based on a neural network to obtain multi-source data with a unified time step. The data of multiple flexible piezoelectric sensors after a unified time step are used as model input and input into the trained neural network-based key point engineering requirement parameter inversion model to obtain the engineering requirement parameters of multiple key points of the engineering structure. The inversion results are corrected using the Bayesian algorithm combined with the data of the corresponding time step of the exogenous sensor.

2. The method for inverting the performance state of an engineering structure based on an engineering skin according to claim 1, characterized in that: The external source sensor includes one or more of a visual sensor, a piezoresistive strain gauge and a laser displacement sensor.

3. The method for inverting the performance state of an engineering structure based on an engineering skin according to claim 1, characterized in that: The neural network on which the key point engineering requirement parameter inversion model is based is a physical information neural network embedded with elastic mechanics equilibrium equations and / or coordination equations.

4. The method for inverting the performance state of an engineering structure based on an engineering skin according to claim 1, characterized in that: The key point engineering requirement parameter inversion model is trained by the following method: Using data from multiple flexible piezoelectric sensors as model input and engineering requirement parameters of multiple key points as model output, a sample dataset is constructed; the sample dataset is divided into a training set and a validation set. Constructing a physical information neural network embedded with elastic equilibrium equations and / or coordination equations; The physical information neural network is trained and verified using the test set and validation set to obtain the inversion model of key point engineering requirement parameters.

5. The method for inversion of engineering structure performance state based on engineering skin according to claim 4 is characterized in that: The model input includes the position coordinates of each flexible piezoelectric sensor, the collected data and the distance from the key point.

6. The method for inversion of engineering structure performance state based on engineering skin according to claim 1, characterized in that: The inversion result is corrected by using the Bayesian algorithm in combination with the data of the corresponding time step of the exogenous sensor, specifically including: The network model parameters of the trained key point engineering requirement parameter inversion model are taken as the mean, the coefficient of variation and distribution type are selected, the prior probability distribution of the network model parameters is set, and the initial uncertainty of the network model parameters is characterized; Based on the data collected by exogenous sensors, considering the weight differences caused by the accuracy of different exogenous sensors, a multi-source fusion likelihood function is established between the inversion results of the cross-section and the measured data to quantify the consistency between the measured data and the model prediction; the cross-section refers to the part where the monitoring points of the exogenous sensor overlap with multiple key points; Applying Bayesian theorem, the prior probability distribution is integrated with the multi-source fusion likelihood function. Through Markov chain Monte Carlo sampling and subset simulation, the posterior distribution samples of the network model parameters are obtained. The posterior probability distribution of the network model parameters is obtained by statistical fitting. According to the posterior distribution of the network model parameters, the probability distribution of the engineering requirement parameters of the key points is recalculated, and the highest probability density point estimation method is used to correct the inversion results.

7. A safety early warning method for engineering structures based on engineering skin, characterized in that: The steps include: The engineering structure performance state inversion method based on the engineering skin as described in any one of claims 1 to 6 is used to invert and obtain the engineering requirement parameters of multiple key points; Engineering risk level assessment and safety warning are carried out based on the engineering requirement parameters and reliability analysis of multiple key points obtained by inversion.

8. The engineering structure safety early warning method based on engineering skin according to claim 6 is characterized in that: The engineering requirement parameters and reliability analysis of multiple key points obtained based on the inversion process are used to perform engineering risk level assessment and safety warning, specifically including: Based on the probability distribution of engineering requirement parameters at multiple key points corrected by the Bayesian algorithm, combined with the preset dynamic model and industry specifications, the reliability is calculated and the probability distribution of all required engineering requirement parameters is inverted. The probability distribution of local damage indicators and global stability indicators is also calculated. Based on the calculated probability distribution of each indicator, the risk level is divided according to the following rules: Low risk: The mean of all indicators is less than 70% of the normative threshold, and the reliability is ≥ 90%; Medium risk: the mean of at least one indicator is between 70% and 90% of the standard threshold, or the reliability is between 70% and 90%; High risk: The mean of any indicator exceeds 90% of the standard threshold, or the reliability is less than 70%; Provide security warnings based on risk levels.

9. An engineering structure performance state inversion system based on engineering skin, characterized in that: include: An engineering skin and multiple exogenous sensors are arranged on the engineering structure to be monitored; The engineered skin has multiple flexible piezoelectric sensors built into it; Data communication module, used to obtain data collected by each engineered skin and exogenous sensors and transmit it to the data processing module; The data processing module is configured to execute the engineering structure performance state inversion method based on engineering skin according to any one of claims 1 to 6.

10. An engineering structure safety early warning system based on engineering skin, characterized in that: include: An engineering skin and multiple exogenous sensors are arranged on the engineering structure to be monitored; The engineered skin has multiple flexible piezoelectric sensors built into it; Data communication module, used to obtain data collected by each engineered skin and exogenous sensors and transmit it to the data processing module; The data processing module is configured to execute the engineering structure safety early warning method based on engineering skin as described in claim 7 or 8.

Images