A multi-source data bias detection and optimization method based on digital twinning

Abstract

The application discloses a multi-source data bias detection and optimization method based on digital twinning, relates to the technical field of digital twinning, and comprises the following steps: constructing a real and twinned space-time matrix, using a vector splicing method to construct a coupling bias vector, screening high-bias physical quantities, regarding the space points corresponding to the high-bias physical quantities as grid nodes, constructing a space grid graph, using a path planning algorithm to define a heuristic function, generating an optimized sensor path, using a linear interpolation method to update non-path points, obtaining an updated real space-time matrix, using a gradient descent method to optimize digital twinning model parameters, and generating updated twinned data. The application simulates multi-scale perturbations by means of a sinusoidal wave superposition method and screens main frequencies by means of an elbow rule, thereby enhancing the sensitivity to nonlinear perturbation responses, and through nonlinear fusion and path optimization, the efficiency of sensor deployment and the accuracy of model parameter optimization are improved.

Description

Technical Field

[0001] The present invention relates to the field of digital twin technology, and in particular to a multi-source data bias detection and optimization method based on digital twins. Background Art

[0002] With the continuous development of sensor technology, the Internet of Things, and digital twin technology, industrial scenario modeling and analysis based on multi-source heterogeneous data has become a key technology for intelligent manufacturing and smart industry. By constructing virtual mapping models, digital twins can simulate and predict physical entities and implement process control under non-disturbance conditions. They have been widely used in scenarios such as equipment predictive maintenance, energy consumption analysis, and process optimization. In terms of multi-source data fusion, existing research focuses on data alignment, missing value filling, feature extraction, and modeling strategies. Due to the heterogeneity of multi-source data sources, the uneven spatial and temporal distribution, and the accumulation of sensor errors, data bias frequently arises, directly affecting modeling accuracy and predictive robustness.

[0003] Existing technologies still have shortcomings in digital twin model construction and data analysis. Existing technologies usually analyze frequency domain information entropy, energy flow density or spectral energy difference separately, lack a nonlinear fusion mechanism for these features, and cannot fully explore the coupling relationship between multiple physical quantities, which limits the accuracy of the bias factor. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a multi-source data bias detection and optimization method based on digital twins, which solves the problem that the existing technology usually analyzes frequency domain information entropy, energy flow density or spectral energy difference separately, lacks a nonlinear fusion mechanism for these features, cannot fully explore the coupling relationship between multiple physical quantities, and limits the accuracy of the bias factor.

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

[0007] In a first aspect, the present invention provides a multi-source data bias detection and optimization method based on digital twins, which includes:

[0008] Collect multi-source data and pre-process it, construct the real and twin space-time matrices, count the bin probabilities, calculate the frequency domain information entropy, construct the frequency domain information entropy difference vector, treat the real and twin space-time matrices as field strengths, calculate the local energy flux density, construct the flux density difference vector, use the sine wave superposition method to generate the disturbance value, perform frequency domain conversion on the disturbance vector, screen the power spectrum density to calculate the spectral energy sum, and construct the spectral energy difference vector;

[0009] Use the vector concatenation method to construct the coupling bias vector, calculate the bias factor, calculate the spatial average bias factor, screen high-bias physical quantities, treat the spatial points corresponding to the high-bias physical quantities as grid nodes, construct a spatial grid graph, use the path planning algorithm to define the heuristic function, generate the sensor optimization path, use the linear interpolation method to update the non-path points, obtain the updated real time-space matrix, use the gradient descent method to optimize the digital twin model parameters, and generate the updated twin data;

[0010] Build a visual interface to display the updated twin data and store the multi-source data generated by collection and analysis.

[0011] As a preferred solution of the multi-source data bias detection and optimization method based on digital twins of the present invention, the construction of the frequency domain information entropy difference vector and the flow density difference vector includes:

[0012] Based on real and twin multi-source data, the matrix construction method is used to construct the real and twin space-time matrices, and each column of the real and twin space-time matrices is subjected to wavelet transform to obtain the wavelet coefficients and normalize them. The normalized coefficients are divided into equal-width bins, the probability of each bin is counted, the frequency domain information entropy is calculated, and the frequency domain information entropy difference between the real and twin multi-source data is calculated. The frequency domain information entropy difference is combined into a frequency domain information entropy difference vector using the vector construction method;

[0013] The real and twin space-time matrices are regarded as field strengths, the time and space derivatives are calculated using the central difference method, the local energy flux density is calculated using the Euclidean norm, the energy flux density difference between the local energy flux densities is calculated using the absolute difference, and the energy flux density differences are combined into energy flux density difference vectors using the vector construction method.

[0014] As a preferred solution of the multi-source data bias detection and optimization method based on digital twins described in the present invention, the method of generating disturbance values ​​using the sine wave superposition method and constructing the spectral energy difference vector includes:

[0015] Based on the real and twin space-time matrices, the sine wave superposition method is used to generate disturbance values. The vector expansion method is used to combine the disturbance values ​​into a disturbance vector to construct a trajectory vector. The disturbance vector is superimposed on the trajectory vector using vector addition. The disturbance vector is converted into the frequency domain using fast Fourier transform to obtain the power spectrum density. The power spectrum density is sorted from large to small, and the first K power spectrum densities are selected to calculate the spectral energy sum and the spectral energy difference. The spectral energy difference is then combined into a spectral energy difference vector using the vector construction method.

[0016] As a preferred solution of the multi-source data bias detection and optimization method based on digital twins described in the present invention, the method of constructing a coupling bias vector using a vector splicing method and screening high-bias physical quantities includes:

[0017] The coupling bias vector is constructed using a vector splicing method, a bias factor is calculated, and a spatial average bias factor is calculated;

[0018] The detection threshold is set using the percentile method, the spatial average bias factor greater than the detection threshold is screened, the corresponding physical quantity is extracted, and is marked as a high bias physical quantity.

[0019] As a preferred scheme of the multi-source data bias detection and optimization method based on digital twinning, wherein: the non-path point is updated using the linear interpolation method, and the digital twinning model parameters are optimized using the gradient descent method, including:

[0020] The spatial point corresponding to the high bias physical quantity is regarded as a grid node, the distance threshold is set using the statistical analysis method, the Euclidean distance between the grid nodes is calculated, the Euclidean distance less than the distance threshold is defined as an edge, the spatial grid graph is constructed, the starting point is randomly set, the end point is set using the target node selection method, and the path planning algorithm is used to define the heuristic function, generate the sensor optimization path, obtain the path point based on the sensor optimization path, screen the non-path point based on the real-time space matrix, update the non-path point using the linear interpolation method, and obtain the updated real-time space matrix; The path planning algorithm defines the heuristic function, generates the sensor optimization path, obtains the path point based on the sensor optimization path, screens the non-path point based on the real-time space matrix, updates the non-path point using the linear interpolation method, and obtains the updated real-time space matrix;

[0021] Based on the updated real-time space matrix, the digital twinning model parameters including the temperature boundary value, the pressure boundary value, the thermal conductivity and the resistivity are optimized using the gradient descent method, and the updated twinning data is generated.

[0022] As a preferred scheme of the multi-source data bias detection and optimization method based on digital twinning, wherein: the updated twinning data is displayed on the visual interface, including:

[0023] The visual interface is constructed using the front-end framework React.js, and the updated twinning data is visually displayed;

[0024] The user after real-name verification is allowed to review.

[0025] As a preferred scheme of the multi-source data bias detection and optimization method based on digital twinning, wherein: the multi-source data collected and analyzed is stored, including:

[0026] The collected multi-source data and the updated twinning data generated by analysis are stored in the central database, and security access measures are set, the central database stores the data in the cloud backup, and regularly detects the integrity of the stored data and the backup data, generates an integrity detection record after detection, and synchronously stores the record in the central database.

[0027] As a preferred scheme of the multi-source data bias detection and optimization method based on digital twinning provided in the application, wherein the collection of multi-source data and preprocessing comprises:

[0028] Real multi-source data is collected using intelligent sensors, and denoising and normalization processing are performed;

[0029] The intelligent sensors comprise GPS, temperature, current, pressure, and conductivity sensors;

[0030] The real multi-source data comprises spatial position, temperature, current, pressure, and conductivity data;

[0031] A digital twinning model is constructed using ANSYS, the digital twinning model parameters are calibrated based on real multi-source data using a Monte Carlo method, and twin multi-source data is simulated and generated, and denoising and normalization processing are performed;

[0032] The twin multi-source data comprises spatial position, temperature, current, pressure, and conductivity data.

[0033] In a second aspect, the application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the multi-source data bias detection and optimization method based on digital twinning according to the first aspect of the application.

[0034] In a third aspect, the application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the multi-source data bias detection and optimization method based on digital twinning according to the first aspect of the application.

[0035] The application has the following advantages: the application simulates multi-scale perturbations using a sine wave superposition method and filters main frequencies using an elbow rule, thereby enhancing sensitivity to nonlinear perturbation responses; and the application improves sensor deployment efficiency and the accuracy of model parameter optimization through nonlinear fusion and path optimization. BRIEF DESCRIPTION OF DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0037] Figure 1 A flowchart of the multi-source data bias detection and optimization method based on digital twinning in embodiment 1;

[0038] Figure 2This is a flowchart for calculating the spatial average bias factor in the multi-source data bias detection and optimization method based on digital twins in Example 1. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0042] Example 1, with reference to Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a multi-source data bias detection and optimization method based on digital twins, comprising the following steps:

[0043] S1. Collect multi-source data and pre-process them, construct the real and twin space-time matrices, count the bin probabilities, calculate the frequency domain information entropy, construct the frequency domain information entropy difference vector, regard the real and twin space-time matrices as field strengths, calculate the local energy flow density, construct the flow density difference vector, use the sine wave superposition method to generate the disturbance value, perform frequency domain conversion on the disturbance vector, screen the power spectrum density to calculate the spectral energy sum, and construct the spectral energy difference vector;

[0044] Specifically, collect multi-source data and perform preprocessing, including:

[0045] Use smart sensors to collect real multi-source data and perform denoising and normalization processing;

[0046] The smart sensors include GPS, temperature, current, pressure and conductivity sensors;

[0047] The real multi-source data includes spatial position, temperature, current, pressure and conductivity data;

[0048] Use ANSYS to build a digital twin model, calibrate the digital twin model parameters using the Monte Carlo method based on real multi-source data, simulate and generate twin multi-source data, and perform denoising and normalization processing;

[0049] The twin multi-source data includes spatial position, temperature, current, pressure, and conductivity data.

[0050] By deploying multiple types of intelligent sensors, real multi-source data containing multiple physical quantity characteristics are collected to provide a basis for subsequent twin modeling and bias analysis. Through denoising and normalization processing, the stability and comparability of the data are enhanced, the occasional errors introduced by equipment precision fluctuations or environmental interference are reduced, the introduction of the Monte Carlo method increases the robustness and credibility of the modeling results, the uncertainty of multi-source input can be fully quantified, which helps to align the real and simulation data in a statistical sense, and improves the modeling accuracy, enhances the model generalization ability, and constructs a high-dimensional spatiotemporal consistent twin data set of the digital twin system.

[0051] Further, a frequency domain information entropy difference vector and a flow density difference vector are constructed, including:

[0052] Based on the real and twin multi-source data, a matrix construction method is used to construct real and twin spatiotemporal matrices. Wavelet transform is performed on each column of the real and twin spatiotemporal matrices to obtain wavelet coefficients and perform normalization processing. The normalized coefficients are divided into equal-width bins, the probability of each bin is counted, and the formula is:

[0053] ,

[0054] ,

[0055] wherein and are the probabilities of the bth bin of the real and twin multi-source data, respectively, indicating the frequency of the normalized coefficients falling into the bin, and are the i th wavelet coefficient amplitude of the spatial point , scale l of the real and twin multi-source data at time , l is the wavelet decomposition scale, and the maximum decomposition scale method is used to set is the spatial sampling point index, is the time sampling point index, and are the number of coefficients of the real and twin multi-source data falling into the bth bin, respectively, and are the total number of coefficients of the real and twin multi-source data, respectively, and b is the bin index;

[0056] Based on the bin probability, the frequency domain information entropy is calculated, and the formula is:

[0057] ,

[0058] ,

[0059] in and are the physical quantities x and , scale l, spatial point If the information entropy of ,but , x is the physical quantity type, corresponding to the type of multi-source data, B is the number of bins, set using Scott's rule;

[0060] Calculate the frequency domain information entropy difference between real and twin multi-source data. The formula is:

[0061] ,

[0062] in is the physical quantity x, the spatial point The frequency domain information entropy difference, L is the total number of wavelet decomposition scales;

[0063] Use vector construction method to combine frequency domain information entropy difference into frequency domain information entropy difference vector;

[0064] The real and twin space-time matrices are regarded as field strengths, and the time and space derivatives are calculated using the central difference method, as follows:

[0065] ,

[0066] ,

[0067] ,

[0068] ,

[0069] in 、 、 and are the physical quantities x and In time , spatial point The time and space derivatives of and are the time in real and twin multi-source data respectively , spatial point The physical value of

[0070] Based on the time and space derivatives, the local energy flux density is calculated using the Euclidean norm, as follows:

[0071] ,

[0072] ,

[0073] in and are the physical quantities x and In time , spatial point The local energy flux density;

[0074] The absolute difference is used to calculate the energy flux density difference between the local energy flux densities. The formula is:

[0075] ,

[0076] ,

[0077] in is the physical quantity x at time , spatial point The energy flux density difference, is the space-time average energy flux density difference of the physical quantity x, T is the total number of time sampling points, and S is the total number of space sampling points;

[0078] The energy flux density differences are combined into energy flux density difference vectors using a vector construction method.

[0079] By constructing a frequency-domain information entropy difference vector, the structural deviations between the real and twin systems in the frequency domain can be explicitly characterized. This is more sensitive than traditional time-domain comparisons and can accurately identify local anomalies caused by insufficient system modeling, perception delays, or noise amplification, thereby improving the resolution and accuracy of deviation detection. Treating the real and twin data as physical field strengths and extracting their spatiotemporal derivatives using the central difference method, the local energy flux density is then calculated, revealing the patterns of energy transmission in the spatiotemporal dimensions. The local energy flux density reflects the changing trends of physical quantities in the system and is a highly dynamic structural indicator. The energy flux density difference vector obtained by calculating the difference can be used to characterize the deviation trends of system behavioral dynamics, making it particularly suitable for analyzing non-steady-state systems. By integrating entropy statistics with energy flux dynamics, a multidimensional bias representation space is constructed, providing frequency-space-time multidimensional cross-validation and significantly enhancing the accuracy and stability of bias detection.

[0080] Furthermore, the sine wave superposition method is used to generate disturbance values ​​and construct the spectral energy difference vector, including:

[0081] Based on the real and twin space-time matrices, the perturbation value is generated using the sine wave superposition method, and the formula is:

[0082] ,

[0083] in is the physical quantity x at the spatial point The disturbance value of M is the number of sine wave components, which is set using the empirical parameter selection method. is the amplitude of the mth sine wave, m is the sine wave index, is the frequency of the mth sine wave, For spatial points The z-axis coordinate of is the phase of the mth sine wave;

[0084] Combine the perturbation values ​​into a perturbation vector using the vector expansion method;

[0085] For each spatial point and physical quantity, construct the trajectory vector, the formula is:

[0086] ,

[0087] ,

[0088] in and are the physical quantities x and At a point in space The trajectory vector of

[0089] Use vector addition to add the disturbance vector to the trajectory vector, the formula is:

[0090] ,

[0091] ,

[0092] in and are the physical quantities x and At a point in space The perturbation vector, is the physical quantity x at the spatial point The perturbation vector of

[0093] Use fast Fourier transform to transform the disturbance vector into frequency domain and obtain the power spectrum density, which is:

[0094] ,

[0095] ,

[0096] in and are the physical quantities x and At a point in space ,frequency The power spectral density, is the frequency, It is a Fourier transform operation, which returns the complex coefficients in the frequency domain;

[0097] Sort the power spectrum densities from large to small, select the first K power spectrum densities and calculate the spectral energy sum. The formula is:

[0098] ,

[0099] ,

[0100] in and are the physical quantities x and At a point in space The spectral energy of the first K main frequencies of , K is the number of main frequencies, set using the elbow rule, is the set of the first K power spectral densities, sorted by power spectral density;

[0101] Based on the spectral energy sum, the spectral energy difference is calculated as follows:

[0102] ,

[0103] in is the physical quantity x at the spatial point The spectral energy difference;

[0104] The spectral energy differences are combined into a spectral energy difference vector using a vector construction method.

[0105] By combining multiple sinusoidal bands, the disturbance response of the system at different frequencies can be simulated. Compared with random disturbances, it is more controllable and explainable, and can effectively detect the response differences of the system in different frequency dimensions, and then identify spectrum offsets. After superimposing the disturbance on the trajectory vector through vector addition, the fast Fourier transform is used to convert the disturbed signal into the frequency domain space, and then the power spectral density is calculated, which accurately identifies the simulation ability differences of the twin model in the frequency domain structure, and can locate the system response anomalies in a specific frequency band. It is suitable for the detection of structural periodic behavior. The spectral energy difference vector can be used as a discriminant feature and included in the coupling analysis, which is conducive to forming a more comprehensive bias description. The vector splicing method is used to integrate the three into a coupled bias vector, and then combined with the nonlinear function and the exponential attenuation factor to calculate the bias factor of each physical quantity at the spatial point, which can achieve spatial precise positioning and modeling feedback of the bias area. The spatial graph modeling and path planning method supports the optimization of perception deployment under limited sensor resources and improves sampling efficiency.

[0106] S2. Use the vector splicing method to construct the coupling bias vector, calculate the bias factor, calculate the spatial average bias factor, screen the high-bias physical quantities, regard the spatial points corresponding to the high-bias physical quantities as grid nodes, construct a spatial grid graph, use the path planning algorithm to define the heuristic function, generate the sensor optimization path, use the linear interpolation method to update the non-path points, obtain the updated real time-space matrix, use the gradient descent method to optimize the digital twin model parameters, and generate the updated twin data;

[0107] Specifically, the vector splicing method is used to construct the coupling bias vector and screen the high-bias physical quantities, including:

[0108] The coupling bias vector is constructed using the vector concatenation method, and the bias factor is calculated using the formula:

[0109] ,

[0110] ,

[0111] in is the physical quantity x at the spatial point The coupling bias vector, 、 as well as is the physical quantity x at the spatial point Frequency domain information entropy difference vector, energy flow density difference vector and spectrum energy difference vector, is the physical quantity x at the spatial point The bias factor, and are the nonlinear coupling parameter and exponential decay parameter, respectively, which are set using the grid search method;

[0112] Bias factor vector, calculate the spatial average bias factor, the formula is:

[0113] ,

[0114] in is the spatial average bias factor of the physical quantity x;

[0115] The percentile method is used to set the detection threshold, and the spatial average bias factor greater than the detection threshold is screened. The corresponding physical quantity is extracted, which corresponds to the real multi-source data and marked as a high-bias physical quantity.

[0116] By fusing the frequency domain information entropy difference vector, energy flux density difference vector and spectral energy difference vector, a multi-scale mapping relationship of three-dimensional physical information differences is essentially established. The frequency domain information entropy difference vector reveals the differences in the fluctuation complexity of physical quantities, the energy flux density difference vector reflects the inconsistency of energy propagation intensity or energy transfer direction, and the spectral energy difference vector focuses on analyzing the mismatch of energy concentration in the main frequency domain. Through the vector splicing method, this information is unified into the same space for combined modeling, so that the subsequent calculation of the bias factor is no longer one-sided and has high robustness and comprehensive discrimination ability. The introduction of nonlinear coupling parameters and exponential decay parameters is a mathematical modeling of the inhomogeneous disturbance law in complex systems, which makes the bias factor sensitive to local anomalies and smooth to overall changes, avoiding over-response to accidental noise. The spatial average value of the bias factor is used to construct a statistical distribution, reducing the false alarm rate and missed alarm rate of anomaly detection and improving positioning accuracy.

[0117] Furthermore, linear interpolation is used to update non-path points, and gradient descent is used to optimize the parameters of the digital twin model, including:

[0118] The spatial points corresponding to the high-bias physical quantities are regarded as grid nodes, and the distance threshold is set using the statistical analysis method. The Euclidean distance between the grid nodes is calculated, and the Euclidean distances less than the distance threshold are defined as edges. The spatial grid graph is constructed, the starting point is randomly set, and the end point is set using the target node selection method. The path planning algorithm defines a heuristic function to generate the sensor optimized path, which is:

[0119] ,

[0120] ,

[0121] in For grid nodes The total cost estimate, From the starting point to the grid node The actual path cost, From the grid node Estimated cost to the target node, Manhattan distance, 、 as well as For grid nodes The three-dimensional coordinates of 、 as well as is the three-dimensional coordinate of the end point;

[0122] Based on the sensor optimization path, we obtain the path points, filter the non-path points based on the real space-time matrix, and use the linear interpolation method to update the non-path points to obtain the updated real space-time matrix. The formula is:

[0123] ,

[0124] ,

[0125] in is the updated physical quantity x at time , grid nodes The value of is the physical quantity x at time , spatial point The value of represents the real space-time matrix value, is the interpolation weight, is the Euclidean distance between the grid node and the spatial point;

[0126] Based on the updated real-time space matrix, the gradient descent method is used to optimize the digital twin model parameters, including temperature boundary values, pressure boundary values, thermal conductivity, and resistivity, to generate updated twin data. The formula is:

[0127] ,

[0128] in For the optimized digital twin model, P, Z, I, R and C are real multi-source data. is the physical quantity x at time , spatial point The value of represents the twin space-time matrix value.

[0129] A The heuristic function introduced in the algorithm, combined with the Manhattan distance, allows the search path to prioritize approaching the target area, achieving an optimal balance between path length and resource overhead. It is not only suitable for static data environments, but also has good dynamic adaptability. The use of Manhattan distance strengthens the directionality of the path and is more effective in industrial grid structures (such as factory workshops and mine tunnels). The heuristic cost function can embed multi-objective factors, such as considering multi-dimensional costs such as temperature gradients and failure probabilities, to further enrich the path evaluation dimensions. The linear interpolation method based on Euclidean distance weighting maintains the continuity and smoothness of the data field, bridges the data gaps in the blank areas covered by the sensor, and avoids the deviation of non-path point information from model training. Based on the updated real space-time matrix, the twin model parameters are re-inverted and optimized. The model parameters continue to approach the real physical state with the measured data. The optimized boundary conditions and material parameters are closer to reality, giving the digital twin the ability to "learn" and can be gradually refined during operation.

[0130] S3. Build a visual interface to display the updated twin data and store the multi-source data generated by collection and analysis;

[0131] Specifically, the visual interface is constructed to display the updated twin data, including:

[0132] The visual interface is constructed using the front-end framework React.js to visualize the updated twin data.

[0133] The real-name verified user is allowed to review.

[0134] Using the virtual DOM differential update mechanism, React can realize real-time visualization rendering of thousands of data points, and is particularly suitable for processing high-frequency and multi-dimensional twin data, and can smoothly display dynamic views such as device running status, fault heat map, and parameter curve. By combining React with the identity verification system and the front-end permission control mechanism, different users of different roles can be provided with different data views, forming a role-driven data interaction experience.

[0135] Further, the multi-source data collected and analyzed is stored, including:

[0136] The collected multi-source data and the updated twin data generated by analysis are stored in the central database, and security access measures are set. The central database stores the data in the cloud backup, and periodically detects the integrity of the stored data and the backup data. After the detection is completed, an integrity detection record is generated and stored in the central database.

[0137] Through the real-name verification mechanism, each data review operation can be bound to a specific user identity, realizing log recording and permission constraints of data access behavior, which helps to meet the safety compliance requirements of industrial data. The central database serves as a unified data management hub, which can integrate spatiotemporal data, frequency domain data, image data, and text description of heterogeneous information. Through Schema design, unified modeling and relationship mapping are realized, which is convenient for subsequent analysis and calling. Through periodic hash verification, redundant backup comparison, and storage link comparison mechanisms, data tampering, file damage, and version inconsistency can be detected in time, providing a reliable data foundation for the operation of the twin model.

[0138] The embodiment also provides a computer device suitable for the multi-source data bias detection and optimization method based on digital twinning, including a memory and a processor. The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the multi-source data bias detection and optimization method based on digital twinning as proposed in the above embodiment.

[0139] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0140] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-source data bias detection and optimization method based on digital twins as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

[0141] In summary, the present invention simulates multi-scale disturbances through the sine wave superposition method and screens the main frequency through the elbow rule, thereby enhancing the sensitivity to nonlinear disturbance response. Through nonlinear fusion and path optimization, the efficiency of sensor deployment and the accuracy of model parameter optimization are improved.

[0142] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A multi-source data bias detection and optimization method based on digital twins, characterized by: include, Collect multi-source data and pre-process it, construct the real and twin space-time matrices, count the bin probabilities, calculate the frequency domain information entropy, construct the frequency domain information entropy difference vector, treat the real and twin space-time matrices as field strengths, calculate the local energy flux density, construct the flux density difference vector, use the sine wave superposition method to generate the disturbance value, perform frequency domain conversion on the disturbance vector, screen the power spectrum density to calculate the spectral energy sum, and construct the spectral energy difference vector; Use the vector concatenation method to construct the coupling bias vector, calculate the bias factor, calculate the spatial average bias factor, screen high-bias physical quantities, treat the spatial points corresponding to the high-bias physical quantities as grid nodes, construct a spatial grid graph, use the path planning algorithm to define the heuristic function, generate the sensor optimization path, use the linear interpolation method to update the non-path points, obtain the updated real time-space matrix, use the gradient descent method to optimize the digital twin model parameters, and generate the updated twin data; Build a visual interface to display updated twin data and store multi-source data generated by collection and analysis; The linear interpolation method is used to update the non-path points, and the gradient descent method is used to optimize the parameters of the digital twin model, including: The spatial points corresponding to the high-bias physical quantities are regarded as grid nodes, and the distance threshold is set using the statistical analysis method. The Euclidean distance between the grid nodes is calculated, and the Euclidean distances less than the distance threshold are defined as edges. The spatial grid graph is constructed, the starting point is randomly set, and the end point is set using the target node selection method. The path planning algorithm defines a heuristic function to generate a sensor-optimized path, obtains path points based on the sensor-optimized path, filters non-path points based on the real space-time matrix, and updates the non-path points using linear interpolation to obtain an updated real space-time matrix. Based on the updated real space-time matrix, the gradient descent method is used to optimize the digital twin model parameters, including temperature boundary values, pressure boundary values, thermal conductivity, and resistivity, to generate updated twin data.

2. The multi-source data bias detection and optimization method based on digital twins according to claim 1, characterized in that: The constructing of the frequency domain information entropy difference vector and the flow density difference vector includes: Based on real and twin multi-source data, the matrix construction method is used to construct the real and twin space-time matrices, and each column of the real and twin space-time matrices is subjected to wavelet transform to obtain the wavelet coefficients and perform normalization. The normalized coefficients are divided into equal-width bins, the probability of each bin is counted, the frequency domain information entropy is calculated, and the frequency domain information entropy difference between the real and twin multi-source data is calculated. The frequency domain information entropy difference is combined into a frequency domain information entropy difference vector using the vector construction method; The real and twin space-time matrices are regarded as field strengths, the time and space derivatives are calculated using the central difference method, the local energy flux density is calculated using the Euclidean norm, the energy flux density difference between the local energy flux densities is calculated using the absolute difference, and the energy flux density differences are combined into energy flux density difference vectors using the vector construction method.

3. The multi-source data bias detection and optimization method based on digital twins according to claim 2, characterized in that: The method of using the sine wave superposition method to generate a disturbance value and construct a spectral energy difference vector includes: Based on the real and twin space-time matrices, the sine wave superposition method is used to generate disturbance values. The vector expansion method is used to combine the disturbance values ​​into a disturbance vector to construct a trajectory vector. The disturbance vector is superimposed on the trajectory vector using vector addition. The disturbance vector is converted into the frequency domain using fast Fourier transform to obtain the power spectrum density. The power spectrum density is sorted from large to small, and the first K power spectrum densities are selected to calculate the spectral energy sum and the spectral energy difference. The spectral energy difference is then combined into a spectral energy difference vector using the vector construction method.

4. The multi-source data bias detection and optimization method based on digital twins according to claim 3, characterized in that: The vector splicing method is used to construct a coupling bias vector and screen high-bias physical quantities, including: Use the vector concatenation method to construct the coupling bias vector, calculate the bias factor, and calculate the spatial average bias factor; The percentile method is used to set the detection threshold, and the spatial average bias factor greater than the detection threshold is screened. The corresponding physical quantity is extracted and marked as a high-bias physical quantity.

5. The multi-source data bias detection and optimization method based on digital twins according to claim 4, characterized in that: The visualization interface constructed to display the updated twin data includes: Use the front-end framework React.js to build a visualization interface to visualize the updated twin data; Users who have passed real-name verification are allowed to access the information.

6. The multi-source data bias detection and optimization method based on digital twins according to claim 5, characterized in that: The multi-source data generated by the storage, collection and analysis includes: The collected multi-source data and the updated twin data generated by analysis are stored in the central database, and security access measures are set up. The central database will back up the stored data to the cloud and regularly perform integrity checks on the stored data and backup data. After the test is completed, the integrity test record will be generated and stored synchronously in the central database.

7. The multi-source data bias detection and optimization method based on digital twins according to claim 1, characterized in that: The multi-source data collection and pre-processing includes: Use smart sensors to collect real multi-source data and perform denoising and normalization processing; The smart sensors include GPS, temperature, current, pressure and conductivity sensors; The real multi-source data includes spatial position, temperature, current, pressure and conductivity data; Use ANSYS to build a digital twin model, calibrate the digital twin model parameters using the Monte Carlo method based on real multi-source data, simulate and generate twin multi-source data, and perform denoising and normalization processing; The twin multi-source data includes spatial position, temperature, current, pressure and conductivity data.

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 multi-source data bias detection and optimization method based on digital twins 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 multi-source data bias detection and optimization method based on digital twins are implemented.

Images