Global data sensing method, device and equipment based on diffusion model

Abstract

The invention relates to a global data sensing method, device and equipment based on a diffusion model. The method comprises the following steps: obtaining local monitoring data corresponding to a building structure, obtaining a target posterior differential model, then obtaining initial global sensing data corresponding to the building structure, and carrying out iterative solution on the target posterior differential model according to the local monitoring data and the initial global sensing data, so as to obtain the target posterior differential model. And obtaining target global sensing data corresponding to the building structure. By adopting the method, the monitoring accuracy of the global sensing data can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of civil engineering structures, and in particular to a global data sensing method, device and equipment based on a diffusion model. Background Art

[0002] In the field of civil engineering, in order to ensure that the building structure works in a safe state, real-time monitoring of the global sensor data of the building structure is required.

[0003] Traditional global data sensing methods based on diffusion models usually place sensors at different locations of the building structure to obtain the global sensing data of the building structure; or use conventional machine learning models to calculate the global sensing data of the building structure.

[0004] However, the above-mentioned global data sensing method based on diffusion model has the problem of low monitoring accuracy. Summary of the invention

[0005] Based on this, it is necessary to provide a global data sensing method, device and equipment based on a diffusion model that can improve monitoring accuracy in response to the above technical problems.

[0006] In a first aspect, the present application provides a global data sensing method based on a diffusion model, comprising:

[0007] Obtain local monitoring data corresponding to the building structure and obtain a target posterior differential model, where the target posterior differential model is trained based on sample global sensor data corresponding to the building structure;

[0008] Acquire initial global sensor data corresponding to the building structure, where the initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution;

[0009] The target posterior differential model is iteratively solved according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

[0010] In one embodiment, the target posterior differential model is iteratively solved according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure, including:

[0011] For each iterative solution process, the target posterior differential model is solved according to the local monitoring data and the first intermediate global sensor data to obtain the second intermediate global sensor data;

[0012] If the current number of iterative solutions reaches the target number of iterative solutions, determining the second intermediate global sensor data as the target global sensor data;

[0013] Among them, when the current iterative solution process is the first iterative solution process, the first intermediate global sensor data is the initial global sensor data; when the current iterative solution process is not the first iterative solution process, the first intermediate global sensor data is the second intermediate global sensor data obtained in the previous iterative solution process.

[0014] In one embodiment, solving the target posterior differential model according to the local monitoring data and the first intermediate global sensor data to obtain the second intermediate global sensor data includes:

[0015] Determine a priori score corresponding to the current iterative solution process according to the first intermediate global sensor data and a target priori score estimator included in the target posterior differential model, and determine noise-free global sensor data according to the priori score;

[0016] Determine the likelihood score corresponding to the current iterative solution process based on the noise-free global sensor data and the local monitoring data;

[0017] The second intermediate global sensing data is determined according to the prior score, the likelihood score, and the target posterior differential model.

[0018] In one embodiment, the method further comprises:

[0019] Acquire multiple sample global sensor data, and iteratively optimize the initial prior score estimator according to each sample global sensor data to obtain a target prior score estimator; wherein the multiple sample global sensor data include historical global sensor data and / or simulated global sensor data corresponding to the building structure;

[0020] Determine the target posterior differentiation model from the target prior score estimator.

[0021] In one embodiment, an initial prior score estimator is iteratively optimized according to the global sensor data of each sample to obtain a target prior score estimator, including:

[0022] For each iterative optimization process, the sample global sensor data corresponding to the iterative optimization process is subjected to noise processing according to different sampling time steps, so as to obtain the sample global sensor data after noise processing corresponding to each sampling time step;

[0023] The intermediate prior score estimator is optimized according to the global state number of each sample after noise processing to obtain an optimized intermediate prior score estimator;

[0024] Detecting whether a preset convergence condition is satisfied, and if so, determining the optimized intermediate prior score estimator as the target prior score estimator;

[0025] Among them, when the iterative optimization process is the first iterative optimization process, the intermediate prior score estimator is the initial prior score estimator; when the iterative optimization process is not the first iterative optimization process, the intermediate prior score estimator is the optimized intermediate prior score estimator obtained in the previous iterative optimization process.

[0026] In one embodiment, the sample global sensor data corresponding to the iterative optimization process is subjected to noise processing according to different sampling time steps to obtain the sample global sensor data after noise processing corresponding to each sampling time step, including:

[0027] For each sampling time step, obtain the sampling noise corresponding to the sampling time step;

[0028] The sample global sensor data is subjected to noise processing according to the sampling noise to obtain the sample global sensor data after the noise processing corresponding to the sampling time step.

[0029] In a second aspect, the present application also provides a data monitoring device, comprising:

[0030] An acquisition module is used to acquire local monitoring data corresponding to the building structure and to acquire a target posterior differential model, where the target posterior differential model is trained based on sample global sensor data corresponding to the building structure;

[0031] An initial value acquisition module is used to obtain initial global sensor data corresponding to the building structure. The initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution.

[0032] The solution module is used to iteratively solve the target posterior differential model according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

[0033] In a third aspect, an embodiment of the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method of the first aspect when executing the computer program.

[0034] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method of the first aspect described above.

[0035] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the method in the first aspect are implemented.

[0036] The above-mentioned global data sensing method, device and equipment based on the diffusion model obtains local monitoring data corresponding to the building structure and obtains the target posterior differential model, and then obtains the initial global sensing data corresponding to the building structure. The target posterior differential model can be iteratively solved according to the local monitoring data and the initial global sensing data to obtain the target global sensing data corresponding to the building structure, wherein the initial global sensing data is obtained by sampling a standard multivariate Gaussian distribution, and the target posterior differential model is trained according to the sample global sensing data corresponding to the building structure. In this way, the target posterior differential model obtained by training the sample global sensor data based on the initial global sensor data is iteratively solved, and the local monitoring data of the building structure is integrated in the iterative solution process, and finally the target global sensor data corresponding to the target building structure is automatically obtained, thereby avoiding the problem of low monitoring accuracy caused by insufficient deployment of sensors in the related technology when obtaining the global sensor data of the building structure by arranging sensors at different positions of the building structure, and the problem of low monitoring accuracy caused by sensitivity to noise when using conventional machine learning models to calculate the global sensor data of the building structure. The target global sensor data of the building structure obtained by the technical solution provided in the present application is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 is an application environment diagram of a global data sensing method based on a diffusion model in one embodiment;

[0039] Figure 2 is a schematic flow chart of a global data sensing method based on a diffusion model in one embodiment;

[0040] Figure 3 A schematic diagram of a process of determining a target posterior differential model in another embodiment;

[0041] Figure 4 is a schematic flow chart of step 203 in another embodiment;

[0042] Figure 5 is a schematic flow chart of an exemplary global data sensing method based on a diffusion model in an embodiment;

[0043] Figure 6A schematic diagram of a flow chart of a global data sensing method based on a diffusion model in an embodiment in which the building structure is a steel plate shear wall;

[0044] Figure 7 is a structural block diagram of a data monitoring device in one embodiment;

[0045] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0047] In the field of civil engineering, in order to ensure that the building structure works in a safe state, real-time monitoring of the global sensor data of the building structure is required.

[0048] Traditional global data sensing methods based on diffusion models usually deploy sensors at different locations of the building structure to obtain the global sensing data of the building structure. However, due to factors such as cost and location restrictions, these sensors can often only cover part of the structure, making it difficult to directly monitor and obtain the global sensing data of the building structure. For example, on a 1:10 nuclear containment model, although hundreds of sensors are deployed, the coverage is still negligible for a surface area of ​​180m². Therefore, this method has the problem of low monitoring accuracy.

[0049] Alternatively, the related technology also uses conventional machine learning models to calculate the global sensor data of the building structure, such as multi-layer perceptron, convolutional neural network, long short-term memory model, etc. However, conventional machine learning models rely on a large amount of data for training and are extremely sensitive to noise, which will be amplified to the output of the conventional machine learning model, resulting in low monitoring accuracy.

[0050] In view of this, the present application provides a global data sensing method, device and equipment based on a diffusion model, by obtaining local monitoring data corresponding to a building structure and obtaining a target posterior differential model, and then obtaining initial global sensing data corresponding to the building structure, the target posterior differential model can be iteratively solved according to the local monitoring data and the initial global sensing data to obtain the target global sensing data corresponding to the building structure, wherein the target posterior differential model is trained based on sample global sensing data corresponding to the building structure, and the initial global sensing data is obtained by sampling a standard multivariate Gaussian distribution. In this way, the target posterior differential model obtained by training the sample global sensor data based on the initial global sensor data is iteratively solved, and the local monitoring data of the building structure is integrated in the iterative solution process, and finally the target global sensor data corresponding to the target building structure is automatically obtained, thereby avoiding the problem of low monitoring accuracy caused by insufficient deployment of sensors in the related technology when obtaining the global sensor data of the building structure by arranging sensors at different positions of the building structure, and the problem of low monitoring accuracy caused by sensitivity to noise when using conventional machine learning models to calculate the global sensor data of the building structure. The global sensor data of the building structure obtained by the technical solution provided in the present application is more accurate.

[0051] The global data sensing method based on the diffusion model provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the data storage system can store the data that the server 101 needs to process. The data storage system can be integrated on the server 101, or it can be placed on the cloud or other network servers. The server 101 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0052] In an exemplary embodiment, Figure 2 As shown in the figure, a global data sensing method based on diffusion model is provided. Figure 1 The server 101 in the example is used as an example to illustrate, including the following steps 201 to 203. Among them:

[0053] Step 201, obtaining local monitoring data corresponding to the building structure, and obtaining a target posterior differential model.

[0054] The local monitoring data may be the status data of the building structure at a preset position, and the status data may include the current displacement, deformation, etc. of this part of the position; the local monitoring data may also be the relevant environmental data of the environment in which the building structure is located.

[0055] Optionally, the server may obtain local monitoring data through sensors deployed at preset positions of the building structure; optionally, the server may obtain archived local monitoring data from a database.

[0056] After obtaining the local monitoring data, the server can obtain a target a posteriori differential model for processing the local monitoring data. In an embodiment of the present application, the target a posteriori differential model is constructed based on a diffusion model, and is trained based on sample global sensor data corresponding to the building structure. The target a posteriori differential model can be a model constructed based on a diffusion model, and can reversely restore the global sensor data that is forward diffused into data noise to the global sensor data corresponding to the local monitoring data based on the local monitoring data.

[0057] Optionally, the server may directly obtain a pre-trained target posterior differential model from a database; optionally, the server may obtain sample global sensor data corresponding to the building structure, and train the target posterior differential model based on the sample global sensor data.

[0058] In an embodiment of the present application, the global sensor data refers to all current status data of the building structure, and the sample global sensor data can be used to train the target posterior differential model. Optionally, the server can obtain historical global sensor data collected by the sensor through a database as sample global sensor data; optionally, the server can simulate the building structure to obtain simulated global sensor data related to the building structure, so that the server can use the simulated global sensor data as sample global sensor data.

[0059] Step 202: Acquire initial global sensor data corresponding to the building structure.

[0060] In an embodiment of the present application, in order to solve the target a posteriori differential model, it is necessary to determine an initial global sensor data, which may be data noise corresponding to the building structure, and the data noise is pure noise data. The server may use the target a posteriori differential model, based on the local monitoring data, to reversely restore the data noise to the global sensor data corresponding to the local monitoring data.

[0061] In one possible implementation, the initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution. In this embodiment of the present application, the server needs to first determine the target iterative solution times of the target posterior differential model. The server can sample the preset data distribution according to the target iterative solution times to obtain the initial state data corresponding to the building structure. The preset data distribution can be the distribution that the global sensor data obeys after continuous diffusion and noise addition. In this embodiment of the present application, the preset data distribution can be a standard multivariate Gaussian distribution.

[0062] Step 203, iteratively solving the target posterior differential model according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

[0063] After obtaining the local monitoring data and the initial global sensor data, the server can iteratively solve the target posterior differential model based on the local monitoring data and the initial global sensor data.

[0064] Optionally, the server can set a preset number of iterative solutions. When the iterative solution process reaches the preset number of iterative solutions, the server can obtain the target global sensor data. Optionally, the server can detect whether the result of each iterative solution process tends to be stable. The detection process may include: detecting whether the difference between the results of multiple consecutive iterative solution processes is within a preset numerical range. If so, it is determined that the result tends to be stable. The server can determine the target global sensor data based on the result of the iterative solution.

[0065] In this way, in the above embodiment, the target posterior differential model obtained by training the sample global sensor data based on the initial global sensor data is iteratively solved, and the local monitoring data of the building structure is integrated in the iterative solution process, and finally the target global sensor data corresponding to the target building structure is automatically obtained, thereby avoiding the problem of low monitoring accuracy caused by insufficient deployment of sensors in the related technology when obtaining the global sensor data of the building structure by arranging sensors at different positions of the building structure, and the problem of low monitoring accuracy caused by sensitivity to noise when using conventional machine learning models to calculate the global sensor data of the building structure. The global sensor data of the building structure obtained by the technical solution provided in the present application is more accurate.

[0066] In one embodiment, based on the above Figure 2 The embodiment shown, see Figure 3 , this embodiment involves the process of determining the target posterior differential model. Figure 3 As shown, the process may include step 301 and step 302 .

[0067] Step 301 , obtain a plurality of sample global sensor data, and iteratively optimize an initial prior score estimator according to each sample global sensor data to obtain a target prior score estimator.

[0068] In an embodiment of the present application, the target posterior differential model is constructed based on a diffusion model, which includes a forward diffusion process and a reverse diffusion process. The forward diffusion process is to iteratively add noise to conventional global sensor data. After multiple (T) iterative diffusions, the conventional global sensor data is diffused into data noise that obeys a Gaussian distribution with a mean of 0 and a variance of a unit matrix. The reverse diffusion process is to gradually reversely diffuse the pure noise sampled from the Gaussian distribution into the corresponding global sensor data.

[0069] According to the above principles, a target posterior differential model can be constructed.

[0070] In the process of constructing the target posterior differential model, the target prior score estimator is first obtained. In an embodiment of the present application, the server can iteratively optimize the initial prior score estimator through multiple sample global sensor data to obtain the target prior score estimator, wherein the multiple sample global sensor data include historical global sensor data and / or simulated global sensor data corresponding to the building structure.

[0071] Suppose the sample global sensor data of forward iterative diffusion t times is x t , the forward diffusion process can be modeled as a Markov chain, and the transfer kernel function of each iterative diffusion can refer to the following formula:

[0072] (1)

[0073] in, represents the distribution of the global sensor data of the sample after the tth forward iterative diffusion, β t represents the noise scaling factor in the t-th forward iterative diffusion, which is used to control the noise intensity added in each forward iterative diffusion process. In the embodiment of the present application, β t <1, I represents the identity matrix.

[0074] The reparameterized iteration formula in each forward iterative diffusion process can refer to the following formula:

[0075] (2)

[0076] in, is the noise sampled from a standard Gaussian distribution.

[0077] make ,and , substituting this parameter into formula 1 and 2, we can get the following formula:

[0078] , (3)

[0079] In summary, it can be seen that when the number of forward iterative diffusion times t is large enough, the information attached to the global sensor data of the sample is gradually eliminated and becomes pure data noise that obeys the N(0,I) distribution.

[0080] Therefore, by mathematically transforming formulas 1 and 2, we can obtain their corresponding inverse stochastic differential equations:

[0081] (4)

[0082] The only unknown term in Equation 4 is is the prior probability distribution The prior score, then the initial prior score estimator can be set as In the initial a priori score estimator, θ represents the parameter to be optimized in the initial a priori score estimator, x(t) is the global sensor data of the sample after t times of forward iterative diffusion, and t is the number of forward iterative diffusions. Based on the above formula, the relevant objective function of the initial a priori score estimator can be set. For details, refer to the following formula:

[0083] (5)

[0084] In this way, based on the above objective function, the server can iteratively optimize the initial prior score estimator according to the global sensor data of each sample to obtain the target prior score estimator.

[0085] Regarding the specific iterative optimization process, the server may preset a loss function, so that the server may iteratively optimize the target prior score estimator according to the above-mentioned optimization objective function and the preset loss function, thereby obtaining the target prior score estimator.

[0086] In a possible implementation, for each iterative optimization process, the server may perform noise processing on the sample global sensor data corresponding to the iterative optimization process according to different sampling time steps to obtain the sample global sensor data after noise processing corresponding to each sampling time step, and optimize the intermediate prior score estimator according to the number of global states of each sample after noise processing to obtain the optimized intermediate prior score estimator. At this time, the server may detect whether a preset convergence condition is met. If the preset convergence condition is met, the optimized intermediate prior score estimator is determined to be the target prior score estimator; wherein, when the iterative optimization process is the first iterative optimization process, the intermediate prior score estimator is the initial prior score estimator; when the iterative optimization process is not the first iterative optimization process, the intermediate prior score estimator is the optimized intermediate prior score estimator obtained in the previous iterative optimization process.

[0087] In the embodiment of the present application, for each iterative optimization process, multiple sampling time steps may be uniformly sampled first. For example, the number of sampling time steps may be M, and the sampling time step may be set to t j ,in, , j represents the sequence number corresponding to each sampling time step.

[0088] According to different sampling time steps, the server can perform noise processing on the sample global sensor data corresponding to the iterative optimization process, so as to obtain the sample global sensor data after noise processing corresponding to each sampling time step. The process may specifically include: for each sampling time step, the server obtains the sampling noise corresponding to the sampling time step, and performs noise processing on the sample global sensor data according to the sampling noise, so as to obtain the sample global sensor data after noise processing corresponding to the sampling time step.

[0089] In the embodiment of the present application, for each sampling time step, the server can sample from the standard normal distribution according to the sampling time step to obtain sampling noise, and the sampling noise corresponds to the above-mentioned data noise , the server can substitute the sampling noise and the sample global sensor data into the above formula 3 to obtain the sample global sensor data after noise processing corresponding to the sampling time step.

[0090] After obtaining the noise-processed sample global sensor data corresponding to each sampling time step, the server can optimize the intermediate prior score estimator according to the noise-processed sample global sensor data to obtain the optimized intermediate prior score estimator. Therefore, the server can preset the loss function and optimize the intermediate prior score estimator based on the loss function, the above-mentioned objective function and the noise-processed sample global sensor data.

[0091] It can be understood that the purpose of optimizing the prior score estimator is to determine the accurate evaluation of the prior score. Therefore, the server needs to determine the relationship between the data noise and the prior score gradient. According to the above formula 3, the data noise can be determined The relationship between the prior score gradient can be specifically referred to the following formula:

[0092] (6)

[0093] Based on this, the loss function corresponding to each iterative optimization process can be specifically referred to the following formula:

[0094] (7)

[0095] That is, for each iterative optimization process, the average value of the error between the prior score evaluated by the prior score estimator corresponding to each sampling time step and the prior score gradient is calculated. This average value is the loss value. By making it as small as possible, a more optimized prior score estimator can be obtained.

[0096] In an embodiment of the present application, after each round of iterative optimization, the server can detect whether a preset convergence condition is met. Optionally, the preset convergence condition may include whether the number of current iterative optimizations reaches a preset number; optionally, the preset convergence condition may also include whether the loss value reaches a preset threshold; optionally, the preset convergence condition may also include whether the loss value tends to be stable. It can be understood that the server can detect whether one of the above preset convergence conditions is met, thereby determining the target prior score estimator. The server can also detect whether the above preset convergence conditions are met at the same time, thereby determining the target prior score estimator.

[0097] Step 302, determining a target a priori differential model based on the target a priori score estimator.

[0098] In the embodiment of the present application, in order to integrate the local monitoring data y into the sampling process, the prior score in the above formula 4 can be Replaced by the posterior score , and obtain the posterior differential model, so that the local monitoring data y can be used to sample the pure data noise from the Gaussian distribution and gradually diffuse it back to the corresponding global sensor data. The posterior differential model can refer to the following formula:

[0099] (8)

[0100] To facilitate calculation, the server can use a preset method (such as the Bayesian formula) to convert the posterior score Split into posterior scores and likelihood scores. The split results can refer to the following formula:

[0101] (9)

[0102] Among them, the prior score The target prior score estimator obtained above can be used for estimation, thereby obtaining the final target posterior differential equation.

[0103] Regarding the likelihood score, the specific calculation process can refer to the following content.

[0104] In a possible implementation, the likelihood score may be first approximated as follows:

[0105] (10)

[0106] In this way, the ambiguous probability distribution Approximately a relatively well-defined probability distribution ,in, is the estimated value of the global sensor data.

[0107] To sum up, it can be understood that the probability distribution It is characterized that the global sensor data is When , the local monitoring data is the probability distribution of y, and It can be estimated by the following formula:

[0108] (11)

[0109] In the current scenario, the approximate likelihood function The described direct problem has a unique solution and the solution is stable and can be treated in a relatively simple way.

[0110] Since the likelihood function describes the estimated value of a given global sensor data In the case of , the probability distribution of the local monitoring data y is observed, that is, the likelihood function is constructed based on the forward diffusion model. Therefore, an abstract forward diffusion model can be defined, which can be specifically referred to the following formula:

[0111] (12)

[0112] A(·) maps the global sensor data to the local monitoring data y and introduces Gaussian noise n to quantify the error in the local monitoring data. The standard deviation σ can be determined according to the actual acquisition scenario of the local monitoring data.

[0113] When the physical quantities of the global sensor data and local monitoring data of the building structure are consistent, the observation matrix can be used as the forward model to directly extract the local monitoring data from the estimated global sensor data. This method is relatively simple; if there is a difference in the physical quantities of the global sensor data and the local monitoring data, a neural network should be used to construct a differentiable mapping relationship between the two. The specific process is to first obtain training data through numerical simulation and field measurement, and then use deep learning algorithms such as multi-layer perceptrons to learn and approximate the complex mapping relationship between the physical quantities of global sensor data and local monitoring data, so as to train the forward model.

[0114] The training process of the forward model can refer to the following: input the sample global sensor data into the forward model to obtain the predicted local monitoring data output by the forward model, and continuously adjust the relevant parameters of the forward model according to the predicted local monitoring data, the local monitoring data and the preset loss function, so as to obtain a trained forward model.

[0115] After obtaining the trained forward model, the server can model the likelihood function and then derive the likelihood score. Combining the expressions of the likelihood score and the prior score, the stochastic differential equation with the posterior score can be modeled.

[0116] The specific formula of the likelihood function can refer to the following formula:

[0117] (13)

[0118] Taking the logarithmic derivative of it, we can get the final likelihood score:

[0119] (14)

[0120] Among them, the hyperparameters This parameter is used to quantify the credibility of local monitoring data. The specific value is determined according to the actual collection scenario of local monitoring data.

[0121] Substituting the likelihood score and the prior score estimated by the prior score estimator into Formula 8, the target posterior differential model can be obtained.

[0122] In one embodiment, based on the above Figure 2 The embodiment shown, see Figure 4 This embodiment involves a process of iteratively solving the target posterior differential model based on local monitoring data and initial global sensor data to obtain the target global sensor data corresponding to the building structure. Figure 4 As shown, step 203 may include step 401 and step 402 .

[0123] Step 401 , for each iterative solution process, solve the target posterior differential model according to the local monitoring data and the first intermediate global sensor data to obtain the second intermediate global sensor data.

[0124] It can be understood that the target posterior differential model is a continuous back-diffusion process. Therefore, the server can be based on the local monitoring data y and the initial global sensor data x T The target posterior differential model is iteratively solved to obtain the final global sensor data x0.

[0125] For each iterative solution process, in a possible implementation, based on Figure 3 The target a posteriori differential model shown can be solved by the server using a preset solution method, namely the Euler-Maruyama method.

[0126] In a possible implementation, the target iteration number T can be preset. It is understandable that when T is large enough, the initial global sensor data x TIt obeys the standard multivariate Gaussian distribution and can be sampled from the standard multivariate Gaussian distribution according to the number of iterative solutions.

[0127] Based on the above, the first intermediate global sensing data can be set to x t , t represents the value obtained by subtracting the current number of iterations from the target number of iterations T.

[0128] For each iterative solution process, in a possible implementation, the server can determine the prior score corresponding to the current iterative solution process based on the first intermediate global sensor data and the target prior score estimator included in the target posterior differential model, and determine the noise-free global sensor data based on the prior score. Based on the noise-free global sensor data and the local monitoring data, the server can determine the likelihood score corresponding to the current iterative solution process. Based on the prior score, the likelihood score and the target posterior differential model, the server can determine the second intermediate global sensor data.

[0129] In the embodiment of the present application, the server can t and the target prior score estimator Estimate the prior score .

[0130] According to the prior score and the above formula 11, the server can determine the first intermediate global sensing data x t The corresponding noise-free global sensor data .

[0131] Based on noise-free global sensor data And the local monitoring data y, substitute into the above formula 14, and the likelihood score corresponding to the current iterative solution process can be obtained .

[0132] According to the prior score and likelihood score, substitute them into formula 9 to get the posterior score corresponding to the current iterative solution process: .

[0133] According to the posterior score and the target posterior differential model (Formula 8), the server can obtain the second intermediate global sensor data x corresponding to the current iterative solution process t-1 .

[0134] Step 402: If the current iterative solution times reaches the target iterative solution times, the second intermediate global sensor data is determined as the target global sensor data.

[0135] Among them, when the current iterative solution process is the first iterative solution process, the first intermediate global sensor data is the initial global sensor data; when the current iterative solution process is not the first iterative solution process, the first intermediate global sensor data is the second intermediate global sensor data obtained in the previous iterative solution process.

[0136] It can be understood that, after obtaining the second intermediate global sensing data x t-1 After that, the server can monitor whether the number of previous iterations reaches the target number of iterations T. If so, the server can determine that the iteration process is over and the second intermediate global sensor data x t-1 That is, the target global sensor data x0.

[0137] In one embodiment, referring to Figure 5 , provides an exemplary global data sensing method based on diffusion model, which can be applied to Figure 1 Servers in the illustrated implementation.

[0138] Step 501 , obtaining a plurality of sample global sensor data, and obtaining the sampling noise corresponding to the sampling time step for each iterative optimization process and for each sampling time step.

[0139] Step 502 , performing noise processing on the sample global sensor data according to the sampling noise, to obtain the sample global sensor data after the noise processing corresponding to the sampling time step.

[0140] Step 503: Optimize the intermediate a priori score estimator according to the global state number of each sample after the noise processing to obtain an optimized intermediate a priori score estimator.

[0141] Step 504 , detecting whether a preset convergence condition is satisfied. If the preset convergence condition is satisfied, determining the optimized intermediate prior score estimator as the target prior score estimator.

[0142] Wherein, when the iterative optimization process is the first iterative optimization process, the intermediate prior score estimator is the initial prior score estimator. When the iterative optimization process is not the first iterative optimization process, the intermediate prior score estimator is the optimized intermediate prior score estimator obtained in the previous iterative optimization process.

[0143] The multiple sample global sensor data include historical global sensor data and / or simulated global sensor data corresponding to the building structure.

[0144] Step 505, determining a target a priori differential model according to the target a priori score estimator.

[0145] Step 506, obtaining local monitoring data corresponding to the building structure, and obtaining a target posterior differential model.

[0146] Among them, the target posterior differential model is trained based on the sample global sensor data corresponding to the building structure.

[0147] Step 507: Acquire initial global sensor data corresponding to the building structure.

[0148] The initial global sensing data is obtained by sampling the standard multivariate Gaussian distribution.

[0149] Step 508, for each iterative solution process, determine the prior score corresponding to the current iterative solution process according to the first intermediate global sensor data and the target prior score estimator included in the target posterior differential model, and determine the noise-free global sensor data according to the prior score.

[0150] Step 509 : determining a likelihood score corresponding to the current iterative solution process according to the noise-free global sensor data and the local monitoring data.

[0151] Step 510, determining second intermediate global sensor data according to the prior score, the likelihood score, and the target posterior differential model.

[0152] Step 511: If the current iterative solution times reaches the target iterative solution times, the second intermediate global sensor data is determined as the target global sensor data.

[0153] Wherein, when the current iterative solution process is the first iterative solution process, the first intermediate global sensor data is the initial global sensor data. When the current iterative solution process is not the first iterative solution process, the first intermediate global sensor data is the second intermediate global sensor data obtained in the previous iterative solution process.

[0154] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0155] based on Figure 5 The embodiment shown, referring to Figure 6, is the process of obtaining the target global sensor data by taking the steel plate shear wall as an example. It can be seen that the process includes T iterative solution processes. The server first extracts the local monitoring data from the corresponding information collected by the sensor, and samples the initial global sensor data. Based on this, the prior score and the posterior score are estimated, and finally the global sensor data is obtained.

[0156] Based on the same inventive concept, the embodiment of the present application also provides a data monitoring device for implementing the global data sensing method based on the diffusion model involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more data monitoring device embodiments provided below can refer to the limitations of the global data sensing method based on the diffusion model above, and will not be repeated here.

[0157] In an exemplary embodiment, Figure 7 As shown, a data monitoring device is provided, including: an acquisition module 701, an initial value acquisition module 702 and a solution module 703, wherein:

[0158] An acquisition module 701 is used to acquire local monitoring data corresponding to the building structure and acquire a target posterior differential model, wherein the target posterior differential model is trained based on sample global sensor data corresponding to the building structure;

[0159] An initial value acquisition module 702 is used to acquire initial global sensor data corresponding to the building structure, wherein the initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution;

[0160] The solving module 703 is used to iteratively solve the target posterior differential model according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

[0161] In one embodiment, the solution module 703 includes:

[0162] An intermediate solving unit, configured to solve the target posterior differential model according to the local monitoring data and the first intermediate global sensor data for each iterative solving process, to obtain second intermediate global sensor data;

[0163] a target determination unit, configured to determine the second intermediate global sensor data as the target global sensor data if the current number of iterative solutions reaches a target number of iterative solutions;

[0164] Among them, when the current iterative solution process is the first iterative solution process, the first intermediate global sensor data is the initial global sensor data; when the current iterative solution process is not the first iterative solution process, the first intermediate global sensor data is the second intermediate global sensor data obtained in the previous iterative solution process.

[0165] In one embodiment, the intermediate solution unit is specifically used to perform:

[0166] Determine a priori score corresponding to the current iterative solution process according to the first intermediate global sensor data and a target priori score estimator included in the target posterior differential model, and determine noise-free global sensor data according to the priori score;

[0167] Determining a likelihood score corresponding to the current iterative solution process according to the noise-free global sensor data and the local monitoring data;

[0168] The second intermediate global sensor data is determined according to the prior score, the likelihood score, and the target posterior differential model.

[0169] In one embodiment, the apparatus further comprises:

[0170] an iterative optimization module, used for acquiring a plurality of said sample global sensor data, and iteratively optimizing an initial prior score estimator according to each of said sample global sensor data, so as to obtain a target prior score estimator; wherein said plurality of said sample global sensor data include historical global sensor data and / or simulated global sensor data corresponding to said building structure;

[0171] A model determination module is used to determine the target a priori differential model according to the target prior score estimator.

[0172] In one embodiment, the iterative optimization module includes:

[0173] A noise adding unit is used for performing noise adding processing on the sample global sensor data corresponding to the iterative optimization process according to different sampling time steps for each iterative optimization process, so as to obtain the sample global sensor data after noise adding processing corresponding to each sampling time step;

[0174] An optimization unit, used for optimizing the intermediate prior score estimator according to the global state number of each sample after the noise processing, to obtain an optimized intermediate prior score estimator;

[0175] A detection unit, used to detect whether a preset convergence condition is satisfied, and if the preset convergence condition is satisfied, determine the optimized intermediate prior score estimator as the target prior score estimator;

[0176] Among them, when the iterative optimization process is the first iterative optimization process, the intermediate prior score estimator is the initial prior score estimator; when the iterative optimization process is not the first iterative optimization process, the intermediate prior score estimator is the optimized intermediate prior score estimator obtained in the previous iterative optimization process.

[0177] In one embodiment, the noise adding unit is specifically used to perform:

[0178] For each of the sampling time steps, obtaining the sampling noise corresponding to the sampling time step;

[0179] The sample global sensor data is subjected to noise addition processing according to the sampling noise to obtain the sample global sensor data after the noise addition processing corresponding to the sampling time step.

[0180] Each module in the above data monitoring device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0181] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data monitoring data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a global data sensing method based on a diffusion model is implemented.

[0182] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0183] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0184] Acquire local monitoring data corresponding to the building structure, and acquire a target posterior differential model, wherein the target posterior differential model is trained based on sample global sensor data corresponding to the building structure;

[0185] Acquire initial global sensor data corresponding to the building structure, wherein the initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution;

[0186] The target posterior differential model is iteratively solved according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

[0187] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0188] For each iterative solution process, the target posterior differential model is solved according to the local monitoring data and the first intermediate global sensor data to obtain second intermediate global sensor data;

[0189] If the current number of iterative solutions reaches the target number of iterative solutions, determining the second intermediate global sensor data as the target global sensor data;

[0190] Among them, when the current iterative solution process is the first iterative solution process, the first intermediate global sensor data is the initial global sensor data; when the current iterative solution process is not the first iterative solution process, the first intermediate global sensor data is the second intermediate global sensor data obtained in the previous iterative solution process.

[0191] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0192] Determine a priori score corresponding to the current iterative solution process according to the first intermediate global sensor data and a target priori score estimator included in the target posterior differential model, and determine noise-free global sensor data according to the priori score;

[0193] Determining a likelihood score corresponding to the current iterative solution process according to the noise-free global sensor data and the local monitoring data;

[0194] The second intermediate global sensor data is determined according to the prior score, the likelihood score, and the target posterior differential model.

[0195] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0196] Acquire a plurality of the sample global sensor data, and iteratively optimize an initial prior score estimator according to each of the sample global sensor data to obtain a target prior score estimator; wherein the plurality of the sample global sensor data include historical global sensor data and / or simulated global sensor data corresponding to the building structure;

[0197] The target a priori differential model is determined based on the target a priori score estimator.

[0198] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0199] For each iterative optimization process, performing noise processing on the sample global sensor data corresponding to the iterative optimization process according to different sampling time steps, so as to obtain the sample global sensor data after noise processing corresponding to each sampling time step;

[0200] Optimizing the intermediate prior score estimator according to the global state number of each sample after the noise processing to obtain an optimized intermediate prior score estimator;

[0201] Detecting whether a preset convergence condition is satisfied, and if the preset convergence condition is satisfied, determining the optimized intermediate prior score estimator as the target prior score estimator;

[0202] Among them, when the iterative optimization process is the first iterative optimization process, the intermediate prior score estimator is the initial prior score estimator; when the iterative optimization process is not the first iterative optimization process, the intermediate prior score estimator is the optimized intermediate prior score estimator obtained in the previous iterative optimization process.

[0203] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0204] For each of the sampling time steps, obtaining the sampling noise corresponding to the sampling time step;

[0205] The sample global sensor data is subjected to noise addition processing according to the sampling noise to obtain the sample global sensor data after the noise addition processing corresponding to the sampling time step.

[0206] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0207] Acquire local monitoring data corresponding to the building structure, and acquire a target posterior differential model, wherein the target posterior differential model is trained based on sample global sensor data corresponding to the building structure;

[0208] Acquire initial global sensor data corresponding to the building structure, wherein the initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution;

[0209] The target posterior differential model is iteratively solved according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

[0210] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0211] For each iterative solution process, the target posterior differential model is solved according to the local monitoring data and the first intermediate global sensor data to obtain second intermediate global sensor data;

[0212] If the current number of iterative solutions reaches the target number of iterative solutions, determining the second intermediate global sensor data as the target global sensor data;

[0213] Among them, when the current iterative solution process is the first iterative solution process, the first intermediate global sensor data is the initial global sensor data; when the current iterative solution process is not the first iterative solution process, the first intermediate global sensor data is the second intermediate global sensor data obtained in the previous iterative solution process.

[0214] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0215] Determine a priori score corresponding to the current iterative solution process according to the first intermediate global sensor data and a target priori score estimator included in the target posterior differential model, and determine noise-free global sensor data according to the priori score;

[0216] Determining a likelihood score corresponding to the current iterative solution process according to the noise-free global sensor data and the local monitoring data;

[0217] The second intermediate global sensor data is determined according to the prior score, the likelihood score, and the target posterior differential model.

[0218] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0219] Acquire a plurality of the sample global sensor data, and iteratively optimize an initial prior score estimator according to each of the sample global sensor data to obtain a target prior score estimator; wherein the plurality of the sample global sensor data include historical global sensor data and / or simulated global sensor data corresponding to the building structure;

[0220] The target a priori differential model is determined based on the target a priori score estimator.

[0221] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0222] For each iterative optimization process, performing noise processing on the sample global sensor data corresponding to the iterative optimization process according to different sampling time steps, so as to obtain the sample global sensor data after noise processing corresponding to each sampling time step;

[0223] Optimizing the intermediate prior score estimator according to the global state number of each sample after the noise processing to obtain an optimized intermediate prior score estimator;

[0224] Detecting whether a preset convergence condition is satisfied, and if the preset convergence condition is satisfied, determining the optimized intermediate prior score estimator as the target prior score estimator;

[0225] Among them, when the iterative optimization process is the first iterative optimization process, the intermediate prior score estimator is the initial prior score estimator; when the iterative optimization process is not the first iterative optimization process, the intermediate prior score estimator is the optimized intermediate prior score estimator obtained in the previous iterative optimization process.

[0226] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0227] For each of the sampling time steps, obtaining the sampling noise corresponding to the sampling time step;

[0228] The sample global sensor data is subjected to noise addition processing according to the sampling noise to obtain the sample global sensor data after the noise addition processing corresponding to the sampling time step.

[0229] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0230] Acquire local monitoring data corresponding to the building structure, and acquire a target posterior differential model, wherein the target posterior differential model is trained based on sample global sensor data corresponding to the building structure;

[0231] Acquire initial global sensor data corresponding to the building structure, wherein the initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution;

[0232] The target posterior differential model is iteratively solved according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

[0233] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0234] For each iterative solution process, the target posterior differential model is solved according to the local monitoring data and the first intermediate global sensor data to obtain second intermediate global sensor data;

[0235] If the current number of iterative solutions reaches the target number of iterative solutions, determining the second intermediate global sensor data as the target global sensor data;

[0236] Among them, when the current iterative solution process is the first iterative solution process, the first intermediate global sensor data is the initial global sensor data; when the current iterative solution process is not the first iterative solution process, the first intermediate global sensor data is the second intermediate global sensor data obtained in the previous iterative solution process.

[0237] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0238] Determine a priori score corresponding to the current iterative solution process according to the first intermediate global sensor data and a target priori score estimator included in the target posterior differential model, and determine noise-free global sensor data according to the priori score;

[0239] Determining a likelihood score corresponding to the current iterative solution process according to the noise-free global sensor data and the local monitoring data;

[0240] The second intermediate global sensor data is determined according to the prior score, the likelihood score, and the target posterior differential model.

[0241] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0242] Acquire a plurality of the sample global sensor data, and iteratively optimize an initial prior score estimator according to each of the sample global sensor data to obtain a target prior score estimator; wherein the plurality of the sample global sensor data include historical global sensor data and / or simulated global sensor data corresponding to the building structure;

[0243] The target a priori differential model is determined based on the target a priori score estimator.

[0244] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0245] For each iterative optimization process, performing noise processing on the sample global sensor data corresponding to the iterative optimization process according to different sampling time steps, so as to obtain the sample global sensor data after noise processing corresponding to each sampling time step;

[0246] Optimizing the intermediate prior score estimator according to the global state number of each sample after the noise processing to obtain an optimized intermediate prior score estimator;

[0247] Detecting whether a preset convergence condition is satisfied, and if the preset convergence condition is satisfied, determining the optimized intermediate prior score estimator as the target prior score estimator;

[0248] Among them, when the iterative optimization process is the first iterative optimization process, the intermediate prior score estimator is the initial prior score estimator; when the iterative optimization process is not the first iterative optimization process, the intermediate prior score estimator is the optimized intermediate prior score estimator obtained in the previous iterative optimization process.

[0249] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0250] For each of the sampling time steps, obtaining the sampling noise corresponding to the sampling time step;

[0251] The sample global sensor data is subjected to noise addition processing according to the sampling noise to obtain the sample global sensor data after the noise addition processing corresponding to the sampling time step.

[0252] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0253] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0254] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0255] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A global data sensing method based on a diffusion model, characterized in that: The method comprises: Acquire local monitoring data corresponding to the building structure, and acquire a target posterior differential model, wherein the target posterior differential model is trained based on sample global sensor data corresponding to the building structure; Acquire initial global sensor data corresponding to the building structure, wherein the initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution; The target posterior differential model is iteratively solved according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

2. The method according to claim 1, characterized in that The iteratively solving the target posterior differential model according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure includes: For each iterative solution process, the target posterior differential model is solved according to the local monitoring data and the first intermediate global sensor data to obtain second intermediate global sensor data; If the current number of iterative solutions reaches the target number of iterative solutions, determining the second intermediate global sensor data as the target global sensor data; Among them, when the current iterative solution process is the first iterative solution process, the first intermediate global sensor data is the initial global sensor data; when the current iterative solution process is not the first iterative solution process, the first intermediate global sensor data is the second intermediate global sensor data obtained in the previous iterative solution process.

3. The method according to claim 2, characterized in that The step of solving the target posterior differential model according to the local monitoring data and the first intermediate global sensor data to obtain the second intermediate global sensor data includes: Determine a priori score corresponding to the current iterative solution process according to the first intermediate global sensor data and a target priori score estimator included in the target posterior differential model, and determine noise-free global sensor data according to the priori score; Determining a likelihood score corresponding to the current iterative solution process according to the noise-free global sensor data and the local monitoring data; The second intermediate global sensor data is determined according to the prior score, the likelihood score, and the target posterior differential model.

4. The method according to claim 1, characterized in that: The method further comprises: Acquire a plurality of the sample global sensor data, and iteratively optimize an initial prior score estimator according to each of the sample global sensor data to obtain a target prior score estimator; wherein the plurality of the sample global sensor data include historical global sensor data and / or simulated global sensor data corresponding to the building structure; The target a priori differential model is determined based on the target a priori score estimator.

5. The method according to claim 4, characterized in that The iterative optimization of the initial a priori score estimator according to the global sensor data of each sample to obtain the target a priori score estimator includes: For each iterative optimization process, performing noise processing on the sample global sensor data corresponding to the iterative optimization process according to different sampling time steps, so as to obtain the sample global sensor data after noise processing corresponding to each sampling time step; Optimizing the intermediate prior score estimator according to the global state number of each sample after the noise processing to obtain an optimized intermediate prior score estimator; Detecting whether a preset convergence condition is satisfied, and if the preset convergence condition is satisfied, determining the optimized intermediate prior score estimator as the target prior score estimator; Among them, when the iterative optimization process is the first iterative optimization process, the intermediate prior score estimator is the initial prior score estimator; when the iterative optimization process is not the first iterative optimization process, the intermediate prior score estimator is the optimized intermediate prior score estimator obtained in the previous iterative optimization process.

6. The method according to claim 5, characterized in that The step of performing noise processing on the sample global sensor data corresponding to the iterative optimization process according to different sampling time steps to obtain the sample global sensor data after noise processing corresponding to each sampling time step includes: For each of the sampling time steps, obtaining the sampling noise corresponding to the sampling time step; The sample global sensor data is subjected to noise addition processing according to the sampling noise to obtain the sample global sensor data after the noise addition processing corresponding to the sampling time step.

7. A global data sensing device based on a diffusion model, characterized in that: The device comprises: An acquisition module, used to acquire local monitoring data corresponding to the building structure and acquire a target posterior differential model, wherein the target posterior differential model is trained based on sample global sensor data corresponding to the building structure; An initial value acquisition module, used to acquire initial global sensor data corresponding to the building structure, wherein the initial global sensor data is obtained by sampling a standard multivariate Gaussian distribution; A solution module is used to iteratively solve the target posterior differential model according to the local monitoring data and the initial global sensor data to obtain the target global sensor data corresponding to the building structure.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images