Business process dynamic optimization decision system based on ai industrial big data processing

Abstract

The present application relates to the technical field of data processing, in particular to a business process dynamic optimization decision system based on AI industrial big data processing, comprising: acquiring stress-strain transmission data and corresponding pipe network operation parameters of a petrochemical pipeline network under different operating states, and constructing a historical decision case library; adaptively determining a core transmission feature set from the stress-strain transmission data, and iteratively optimizing the feature screening weight through historical data; inputting the transmission feature set and the pipe network operation parameters into a dynamic correlation modeling engine to generate a self-updating transmission-operation correlation model, and correcting the correlation coefficients according to real-time data deviation; based on the transmission-operation correlation model and the historical decision case library, performing dynamic adjustment decision of the pipe network operation parameters. Stress-strain and operation parameters are collected through distributed sensing, and the space-time deviation is eliminated to form a space-time correlation data set, the feature screening weight is iteratively optimized through historical feature importance, and the key transmission path is automatically identified.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a business process dynamic optimization decision-making system based on AI industrial big data processing. Background Art

[0002] As the core infrastructure for energy transportation, the safe and stable operation of petrochemical pipeline networks is directly related to industrial production efficiency and safe production. During long-term operation, pipelines are affected by media erosion, environmental corrosion, and material aging, and are prone to risks such as stress concentration and local leakage. Dynamic regulation must be achieved by optimizing pipeline network operating parameters. Traditional decision-making methods have significant limitations: structural data such as stress and strain and operating parameters such as medium flow rate often produce temporal and spatial deviations due to acquisition time differences and sensor position offsets, resulting in distorted correlations between structural state and operating parameters. Relying on manually preset features such as fixed transmission paths and static attenuation coefficients, they cannot adapt to dynamic scenarios such as pipeline aging and topology changes. Key features are easily missed, and correlation models are mostly statically constructed. When the pipeline state changes, the prediction error increases, requiring manual retraining. Parameter adjustment plans are often based on simple analogies with historical cases and lack a quantitative balance between multiple objectives such as safety thresholds, energy consumption, and equipment life. Summary of the Invention

[0003] In response to the technical problems existing in the prior art, the present invention provides a business process dynamic optimization decision-making system based on AI industrial big data processing.

[0004] The technical solution of the present invention to solve the above technical problems is as follows: a business process dynamic optimization decision-making system based on AI industrial big data processing, comprising:

[0005] Data acquisition module: obtains stress and strain transfer data and corresponding network operating parameters of the petrochemical pipeline network under different operating conditions, and builds a historical decision-making case library;

[0006] Feature extraction module: Adaptively determine the core transfer feature set from stress-strain transfer data, and iteratively optimize feature screening weights through historical data;

[0007] Dynamic correlation module: Input the transfer feature set and pipe network operation parameters into the dynamic correlation modeling engine to generate a self-updating transfer-operation correlation model and correct the correlation coefficient according to the real-time data deviation;

[0008] Dynamic adjustment decision module: Based on the transfer-operation association model and the historical decision case library, it executes dynamic adjustment decisions on the pipe network operation parameters.

[0009] In a preferred embodiment, the data acquisition module deploys distributed stress and strain sensors in the pipeline at a preset density to ensure the spatial coverage integrity of each node in the pipeline network. The sampling frequency of the sensors strictly satisfies the Nyquist sampling theorem to fully capture the instantaneous transmission characteristics of stress and strain, including high-frequency vibration signals caused by turbulent impact and low-frequency deformation signals caused by continuous pressure. The sensor data is connected to the data acquisition terminal through wireless transmission, and the stress and strain values ​​of each node in the pipeline network are collected in real time. The stress and strain values ​​together represent the structural bearing state of the pipeline. The collected stress and strain values ​​are filtered and stored in the form of a time-series signal stream.

[0010] At the same deployment node position of the distributed stress and strain sensor, the network operation parameters are synchronously obtained, including: medium flow rate, pipeline pressure and ambient temperature. The sensor node number is used as the unique index to construct a multidimensional parameter matrix. The network operation parameters of the same node at the same timestamp are arranged in time series to form a time-parameter two-dimensional matrix. The network operation parameters collected by the node include: medium flow rate v, pipeline pressure P, ambient temperature T env , suppose the collected timestamp sequence is t1, t2, ..., t n , then the time-parameter two-dimensional matrix M k Defined as:

[0011]

[0012] Its element expression is: M k (i,j)=param j (t i ), where M k Represents the time-parameter two-dimensional matrix of node k, i represents the time index, j represents the parameter index, and j=1 corresponds to the medium flow velocity v, j=2 corresponds to the pipeline pressure P, and j=3 corresponds to the ambient temperature T env .

[0013] Automatically optimize the spatial coordinate association algorithm based on historical matching errors, dynamically align the real-time stress and strain transmission signals with the pipeline network operating parameters, and determine the spatiotemporal correlation data set. Historical matching errors refer to the actual spatial deviation of the theoretically associated pipeline network nodes in past data and the actual timestamp deviation of the theoretical synchronization time. Using historical matching errors as the basis for algorithm optimization, the calibration coefficient of the spatial coordinate association algorithm and the delay compensation value of the time alignment algorithm are automatically adjusted by analyzing the distribution of historical matching errors.

[0014] The steps to implement dynamic alignment are as follows:

[0015] Time dimension: Compare the trend consistency of the two types of signals through a sliding window and correct the timestamp deviation. The specific calculation formula is as follows:

[0016]

[0017] Δt offset =arg max τ ρ(τ)

[0018] Where ρ represents the trend correlation coefficient between stress-strain signal and pipeline network operation parameters, x t Represents the stress-strain signal value at time t, y t represents the network operation parameter value at time t, T represents the sliding window length, τ represents the time offset, Δt offset represents the optimal time offset, Represents x t 、y t The mean value in the time window T;

[0019] Spatial dimension: Combine the pipeline topology to calibrate the coordinates to ensure that the stress and strain signal nodes coincide with the operating parameter nodes in physical space. The specific calculation formula for spatial dimension coordinate calibration is as follows:

[0020] (x',y',z')=(x+Δx k ,y+Δy k ,z+Δz k )

[0021] Among them, x, y, z represent the theoretical installation coordinates of the sensor, x', y', z' represent the actual space coordinates after calibration, Δx k , Δy k , Δz k represents the coordinate correction of node k, where k represents the sensor node number.

[0022] In a preferred embodiment, the feature extraction module loads historical feature importance evaluation results, performs intelligent transfer path analysis on the spatiotemporal correlation dataset, and automatically identifies the associated paths. The historical feature importance evaluation results are quantified values ​​of the contribution strength of the features obtained through historical data analysis to pipeline status assessment. Using the spatiotemporal correlation dataset as input, a graph neural network is used to construct a node-node stress transfer network model. By calculating the stress transfer efficiency and global influence of each path, the critical transfer path is automatically identified.

[0023] The specific calculation formula of the stress transfer efficiency is as follows:

[0024]

[0025] Among them, η k→l represents the stress transfer efficiency from node k to node l, σ k (t), σ l(t) represents the stress value of nodes k and l at time t, represents the average stress value of nodes k and l in the time window T;

[0026] The specific calculation formula of the global influence is as follows:

[0027]

[0028] Among them, δ p represents the global influence of path p, ∑ all represents the set of stress values ​​of all nodes, ∑ all\p represents the set of stress values ​​of all nodes after removing path p, and Var(·) represents the variance function, which measures the degree of discreteness of stress distribution.

[0029] The dynamic attenuation coefficient and diffusion coefficient of each key transmission path are calculated and dynamically corrected as the pipeline ages. The dynamic attenuation coefficient represents the energy attenuation degree of stress along the transmission path with distance. The specific calculation formula is as follows:

[0030] α p =α p0 ·(1+λ·γ k )

[0031] Among them, α p represents the dynamic attenuation coefficient of path p, α p0 represents the initial attenuation coefficient of path p, λ represents the aging influence coefficient, γ k It represents the aging degree of node k on path p. The diffusion coefficient represents the diffusion range of stress from the main path to the surrounding nodes. The specific calculation formula is as follows:

[0032] β p =β p0 ·(1+μ·γ k )

[0033] Among them, β p represents the diffusion coefficient of path p, β p0 represents the initial diffusion coefficient of path p, μ represents the aging diffusion coefficient. Pipeline aging will change its mechanical properties. The stress transfer resistance of the aging node increases, resulting in an increase in the attenuation coefficient. At the same time, the stress distribution uniformity in the aging area decreases, resulting in an increase in the diffusion coefficient. The pipeline aging data is obtained in real time, and a mapping relationship between aging degree and coefficient correction amount is established.

[0034] The top-N feature combinations are screened from the dynamic attenuation coefficient and diffusion coefficient to determine the core transfer feature set. The influence of the quantitative features on the pipeline network status assessment task is ranked. The specific calculation formula is as follows:

[0035]

[0036] Among them, Imp(f) represents the contribution of feature f, f represents the feature to be evaluated, M represents the number of decision trees in random forest, OOB err,t Represents the out-of-bag error of the t-th tree, OOB err,t(f) It represents the out-of-bag error of the tth tree after randomly shuffling the feature f. The sorting criteria include the correlation between the feature and the maximum stress value and failure probability of the pipeline network, as well as the effective frequency of the feature in historical decision-making. After sorting from high to low by contribution, the top N features are selected for combination, where the value of N is determined by the data complexity, which is quantified by the number of nodes and the parameter fluctuation variance index. The core transfer feature set concentrates on the features that are most explanatory of the pipeline network status.

[0037] In a preferred embodiment, the dynamic association module performs intelligent clustering on the pipe network operation parameters to obtain parameter clusters corresponding to the key transfer paths. The intelligent clustering is based on the multidimensional characteristics of the pipe network operation parameters and adopts a density clustering algorithm to calculate the Euclidean distance between the parameters. The specific calculation formula is as follows:

[0038]

[0039] Among them, d i,j represents the Euclidean distance between the operating parameters of the pipe network of group i and group j, v i 、v j Indicates the medium flow rate of group i and j, P i 、P j Indicates the pipeline pressure of group i and j, T i 、T j represents the ambient temperature of the i-th and j-th groups. Parameters with similar change trends are divided into the same cluster. The specific calculation formula is as follows:

[0040]

[0041] Among them, Core(i) indicates whether the i-th group of parameters is a core point, ∈ represents the radius threshold of density clustering, Num(j|·) represents the number of parameter groups that meet the conditions, and MinPts represents the minimum number of neighbors of the core point. A sub-association model is established for each parameter cluster and the corresponding transmission path. The specific calculation formula is as follows:

[0042]

[0043] in, It represents the core transfer feature value of path p predicted by the sub-model, v, P, and T represent the medium flow rate, pipeline pressure, and ambient temperature in the parameter cluster, ω0 represents the bias term, ω1, ω2, and ω3 represent the regression coefficients, corresponding to the weights of flow rate, pressure, and temperature, respectively. The larger the value, the stronger the influence of the parameter on the feature. The confidence threshold is set for the sub-association model. The sub-association model takes the operating parameters of the parameter cluster as input and the core features of the corresponding transfer path as output, and is trained based on the multivariate linear regression algorithm.

[0044] When real-time data is input, if the prediction error of the sub-association model exceeds the confidence threshold, the model reconstruction is automatically triggered, and the updated sub-models are integrated to obtain the transfer-run association model. The specific calculation formula for the prediction error and threshold judgment is as follows:

[0045]

[0046] Among them, e m represents the prediction error of sub-model m, Represents the core feature value predicted by the sub-model, f p represents the core feature value actually collected, θ represents the confidence threshold, and flag represents the reconstruction trigger flag. The specific calculation formula for updating the sub-model parameters after reconstruction is as follows:

[0047]

[0048] Among them, ω' k represents the regression coefficient of the reconstructed sub-model m, L represents the sample size of the latest real-time data, f p,t It represents the predicted eigenvalue and actual eigenvalue of the t-th group of samples. The prediction error refers to the deviation between the result of the sub-model predicting the core transfer feature based on the real-time operation parameters and the actual collected eigenvalue. If the deviation exceeds the confidence threshold, it indicates that the original model cannot adapt to the current association law. The model reconstruction calls the latest real-time data to retrain the corresponding sub-model and adjust the model parameters. After the reconstruction is completed, the engine fuses all sub-models according to the path weight to generate a new transfer-operation association model. The specific calculation formula is as follows:

[0049]

[0050] Where F represents the output of the final transfer-operation correlation model, Z represents the total number of critical transfer paths, and ω p represents the weight of path p, Represents the output of sub-model m.

[0051] In a preferred embodiment, the dynamic adjustment decision module obtains the real-time transmission data of stress and strain of the current pipeline network, inputs the transmission-operation correlation model to obtain the initial adjustment parameter range, and calls the historical decision case library, each case containing the stress and strain state D hist , adjust the parameter v hist 、P hist 、T hist , implementation effect E hist ,Through the case similarity matching algorithm, the effective parameter set verified by practice is screened out from the parameter range. The specific calculation formula is as follows:

[0052]

[0053] Among them, S represents the similarity, ranging from [0, 1], σ hist,k , ε hist,k The specific calculation formulas for stress and strain values ​​of historical cases and effective parameter screening are as follows:

[0054] Valid={(v,P,T)|v∈[v low ,v hist ],P∈[P low ,P hist ],T∈[T low ,T hist ],S

[0055] ≥S0,E hist ≥E0}

[0056] Among them, Valid represents the set of valid parameter pairs, that is, the set of parameter combinations that meet all conditions after screening, | represents the following conditions are met, v low represents the lowest reasonable threshold value of parameter v, v hist represents the highest reasonable threshold of parameter v, P low represents the lowest reasonable threshold of parameter P, P hist represents the highest reasonable threshold of parameter P, T low represents the lowest reasonable threshold of parameter T, T hist Indicates the highest reasonable threshold of parameter T, S≥S0 indicates that the feature contribution of the parameter pair must meet the minimum requirement, E hist ≥E0 indicates that the historical validity of the parameter pair must meet the minimum requirement, and the valid parameter sets are prioritized to obtain the optimal adjustment plan.

[0057] The beneficial effects of the present invention are as follows: the present invention collects stress, strain and operating parameters through distributed sensing, eliminates spatiotemporal deviations, and forms a spatiotemporal correlation data set of spatial coordinates-time-structural parameters-operating parameters. The data can truly reflect the real-time correlation between structural stress and operating status, and iteratively optimizes the screening weights through historical feature importance, automatically identifies key transfer paths, and calculates the attenuation coefficient and diffusion coefficient dynamically corrected with aging. Finally, the Top-N core features are screened, breaking away from the limitations of manual presets, ensuring that the features always correspond to the current core stress characteristics of the pipeline, improving the feature pertinence, and automatically selecting the modeling algorithm based on the feature type to construct a sub-model for the key path and the corresponding parameter cluster. When the real-time data prediction error exceeds the threshold, the sub-model is automatically reconstructed and fused and updated, which can capture the changes in the correlation rules caused by pipeline aging and changes in operating conditions in real time, ensuring that the model always adapts to the actual status. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flow chart of the present invention;

[0059] Figure 2 This is a system block diagram of the present invention. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0061] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the described features. In the description of this application, "plurality" means two or more, unless otherwise specifically specified.

[0062] In the description of this application, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art will recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in this application.

[0063] like Figure 1 This embodiment provides: a business process dynamic optimization decision system based on AI industrial big data processing, including:

[0064] Data acquisition module: obtains stress and strain transfer data and corresponding network operating parameters of the petrochemical pipeline network under different operating conditions, and builds a historical decision-making case library;

[0065] In this embodiment, the data acquisition module requires specific explanation. The data acquisition module deploys distributed stress and strain sensors at a preset density within the pipeline (including stress concentration areas such as elbows and welds) to ensure the spatial coverage integrity of each node in the pipeline network. The sampling frequency of the sensors strictly satisfies the Nyquist sampling theorem to fully capture the instantaneous transmission characteristics of stress and strain, including high-frequency vibration signals caused by turbulent impact and low-frequency deformation signals caused by continuous pressure. The sensor data is connected to the data acquisition terminal via wireless transmission, and the stress value (reflecting the internal force per unit area of ​​the pipeline) and strain value (reflecting the ratio of pipeline deformation to original length) of each node in the pipeline network are collected in real time. The stress and strain values ​​together represent the structural load-bearing state of the pipeline. The collected stress and strain values ​​are filtered and stored in the form of a time-series signal stream. This storage mode completely preserves the transmission law of stress and strain in the pipeline network through continuous recording in the time dimension, providing time domain feature support for subsequent analysis of the relationship between the stress transmission path and the pipeline operation state, and ensuring that the original signal can truly reflect the dynamic changes in the force of the pipeline under different operating conditions.

[0066] At the same deployment node position of the distributed stress and strain sensor, the pipeline network operation parameters are synchronously obtained, including: medium flow rate (reflecting the flow intensity of the fluid in the pipeline), pipeline pressure (reflecting the static pressure of the fluid on the inner wall of the pipeline) and ambient temperature (reflecting the physical impact of the external environment on the pipeline material). The sensor node number is used as the unique index to construct a multidimensional parameter matrix. The pipeline network operation parameters of the same node at the same timestamp are arranged in time series to form a time-parameter two-dimensional matrix. The pipeline network operation parameters collected by the node include: medium flow rate v, pipeline pressure P, ambient temperature T env , suppose the collected timestamp sequence is t1, t2, ..., t n , then the time-parameter two-dimensional matrix M k Defined as:

[0067]

[0068] Its element expression is: M k (i,j)=param j (t i ), where M k Represents the time-parameter two-dimensional matrix of node k, i represents the time index, j represents the parameter index, and j=1 corresponds to the medium flow velocity v, j=2 corresponds to the pipeline pressure P, and j=3 corresponds to the ambient temperature T env ;

[0069] Based on the historical matching error, the spatial coordinate association algorithm is automatically optimized, the real-time transmission signal of stress and strain is dynamically aligned with the pipeline network operation parameters, and the spatiotemporal correlation data set is determined to eliminate the spatiotemporal deviation in the data acquisition process. The spatiotemporal deviation is caused by the non-ideality of the actual acquisition scenario: in the time dimension, there is a difference in the signal transmission delay between the stress and strain sensor and the pipeline network monitoring unit, resulting in a deviation in the signal timestamp at the same moment. In the spatial dimension, the sensor installation position may be slightly offset due to pipeline vibration, resulting in an error between the pipeline network node number association and the actual spatial position. The historical matching error refers to the actual spatial deviation of the theoretically associated pipeline network nodes in the past data, and the actual timestamp deviation of the theoretical synchronization time. The historical matching error is used as the basis for algorithm optimization. By analyzing the distribution of historical matching errors, the calibration coefficient of the spatial coordinate association algorithm and the delay compensation value of the time alignment algorithm are automatically adjusted.

[0070] The steps to implement dynamic alignment are as follows:

[0071] Time dimension: Compare the trend consistency of the two types of signals through a sliding window and correct the timestamp deviation. The specific calculation formula is as follows:

[0072]

[0073] Δt offset=arg max τ ρ(τ)

[0074] Where ρ represents the trend correlation coefficient between stress-strain signal and pipeline network operation parameters, x t Represents the stress-strain signal value at time t, y t represents the network operation parameter value at time t, T represents the sliding window length, τ represents the time offset, Δt offset represents the optimal time offset, Represents x t 、y t The mean value in the time window T;

[0075] It should be noted that the time dimension alignment is achieved by calculating the trend correlation coefficient of the two types of signals. For different time offsets τ, the stress-strain signal and the network operation parameter ρ are calculated, and the τ with the largest ρ is selected as Δt offset , correct the timestamp of the stress and strain signal, and ensure that the signals at the same physical moment strictly correspond in the time dimension by eliminating the time deviation caused by transmission delay.

[0076] Spatial dimension: The coordinates are calibrated based on the pipeline topology (the distance and connection relationship between pipeline network nodes) to ensure that the stress and strain signal nodes coincide with the operating parameter nodes in physical space. The specific calculation formula for spatial dimension coordinate calibration is as follows:

[0077] (x',y',z')=(x+Δx k ,y+Δy k ,z+Δz k )

[0078] Among them, x, y, z represent the theoretical installation coordinates of the sensor, x', y', z' represent the actual space coordinates after calibration, Δx k , Δy k , Δz k represents the coordinate correction of node k, where k represents the sensor node number;

[0079] It should be noted that the spatial dimension calibration is based on the historical matching error, such as the deviation between the actual position of node k in the past and the theoretical coordinates, and the coordinate correction Δx is calculated. k , Δy k , Δz k ,By correcting the theoretical coordinates to obtain the actual space coordinates, it is ensured that the stress and strain sensors and the monitoring nodes of the pipeline operation parameters coincide in the physical space, eliminating the spatial deviation caused by the installation offset.

[0080] The final output spatiotemporal correlation data set includes "spatial coordinates - timestamp - stress value - strain value - medium flow rate - pipeline pressure - ambient temperature" to achieve unbiased correlation between the two types of signals in the spatiotemporal dimension.

[0081] It should be noted that the construction and generation logic of the historical decision case library is as follows:

[0082] The distributed stress and strain sensors are used to collect stress and strain values ​​at each node in the pipeline to form a time-series signal stream. The pipeline network monitoring module is used to simultaneously collect the medium flow rate, pipeline pressure, and ambient temperature at the corresponding node to construct a multi-dimensional parameter matrix.

[0083] Optimize spatial coordinate association algorithm and time alignment rules based on historical matching errors:

[0084] Time dimension: Compare the trend consistency of stress and strain signals with operating parameters through a sliding window and correct timestamp deviation (for example, adjust the stress signal timestamp to be strictly synchronized with the flow rate parameter);

[0085] Spatial dimension: The sensor coordinates are calibrated in combination with the pipeline topology to ensure that the spatial positions of the "stress collection point" and the "operating parameter monitoring point" are completely corresponding.

[0086] After preprocessing, structured data including "node coordinates - timestamp - stress value - strain value - medium flow rate - pipeline pressure - ambient temperature" is formed. This dataset is the basic information for the case library to record the "initial state of the pipeline".

[0087] Feature extraction module: Adaptively determine the core transfer feature set from stress-strain transfer data, and iteratively optimize feature screening weights through historical data;

[0088] In this embodiment, the feature extraction module requires specific explanation. The feature extraction module loads the historical feature importance evaluation results, performs intelligent transfer path analysis on the spatiotemporal correlation dataset, and automatically identifies the associated paths. The historical feature importance evaluation results are quantified values ​​of the contribution strength of the features obtained through historical data analysis to the pipeline status assessment. Using the spatiotemporal correlation dataset as input, a graph neural network is used to construct a node-node stress transfer network model. By calculating the stress transfer efficiency of each path (such as the synchronization rate of stress changes of adjacent nodes) and the global influence (such as the degree of interference of the path terminal on the overall pipeline network stress distribution), the critical transfer paths are automatically identified.

[0089] The specific calculation formula of the stress transfer efficiency is as follows:

[0090]

[0091] Among them, η k→l represents the stress transfer efficiency from node k to node l, σk (t), σ l (t) represents the stress value of nodes k and l at time t, represents the average stress value of nodes k and l in the time window T;

[0092] The specific calculation formula of the global influence is as follows:

[0093]

[0094] Among them, δ p represents the global influence of path p, ∑ all represents the set of stress values ​​of all nodes, ∑ all\p represents the set of stress values ​​of all nodes after removing path p, Var(·) represents the variance function, which measures the degree of discreteness of stress distribution;

[0095] It should be noted that the stress transfer efficiency is calculated by the correlation coefficient of the stress changes of adjacent nodes, reflecting the synchronization of stress transfer from node k to node l. This value is the core indicator for judging whether a path is critical. The higher the efficiency, the more likely the path is to be the main stress transfer channel.

[0096] It should be noted that the global influence is calculated by comparing the variance change of the overall pipeline network stress distribution when the path exists and when the path is interrupted. If the pipeline network stress variance decreases significantly after the path p is interrupted, it means that the path is crucial to maintaining the overall stress distribution and has a high global influence. By combining the stress transfer efficiency and the global influence, the critical transfer path can be automatically identified.

[0097] The dynamic attenuation coefficient and diffusion coefficient of each key transmission path are calculated and dynamically corrected as the pipeline ages. The dynamic attenuation coefficient represents the energy attenuation degree of stress along the transmission path with distance. The specific calculation formula is as follows:

[0098] α p =α p0 ·(1+λ·γ k )

[0099] Among them, α p represents the dynamic attenuation coefficient of path p, α p0 represents the initial attenuation coefficient of path p, λ represents the aging influence coefficient, γ k It represents the aging degree of node k on path p. The diffusion coefficient represents the diffusion range of stress from the main path to the surrounding nodes. The specific calculation formula is as follows:

[0100] β p =β p0 ·(1+μ·γ k )

[0101] Among them, β prepresents the diffusion coefficient of path p, β p0 represents the initial diffusion coefficient of path p, and μ represents the aging diffusion coefficient. Pipeline aging (such as corrosion leading to wall thinning and material fatigue) will change its mechanical properties. The stress transfer resistance of the aging node increases, resulting in an increase in the attenuation coefficient. At the same time, the stress distribution uniformity in the aging area decreases, resulting in an increase in the diffusion coefficient (stress diffuses to non-main paths). Real-time pipeline aging data (such as operating time, corrosion rate, and wall thickness detection value) is obtained to establish a mapping relationship between aging degree and coefficient correction value.

[0102] Based on the feature contribution ranking algorithm, the top-N feature combinations are screened from the dynamic attenuation coefficient and diffusion coefficient to determine the core transfer feature set. The feature contribution ranking algorithm ranks the impact of quantitative features on pipeline network status assessment tasks (such as stress anomaly warning and operation optimization decision-making). The specific calculation formula is as follows:

[0103]

[0104] Among them, Imp(f) represents the contribution of feature f, f represents the feature to be evaluated, M represents the number of decision trees in random forest, OOB err,t Represents the out-of-bag error of the t-th tree, OOB err,t(f) It represents the out-of-bag error of the tth tree after randomly shuffling the feature f, such as the feature importance ranking based on random forest and the contribution calculation of the gradient boosting tree. The ranking criteria include the correlation between the feature and the maximum stress value and failure probability of the pipeline network, as well as the effective frequency of the feature in historical decision-making (such as whether the feature is used many times to accurately predict anomalies). After sorting from high to low by contribution, the top N features are selected for combination, where the value of N is determined by the data complexity, which is quantified by the number of nodes and the parameter fluctuation variance index. For example, when the number of nodes is >50, N=8, and when the number of nodes is <50, N=5. The core transfer feature set concentrates on the features that are most explanatory of the pipeline network status, such as the dynamic attenuation coefficient and high diffusion coefficient of the critical path.

[0105] Dynamic correlation module: Input the transfer feature set and pipe network operation parameters into the dynamic correlation modeling engine to generate a self-updating transfer-operation correlation model and correct the correlation coefficient according to the real-time data deviation;

[0106] In this embodiment, the dynamic association module needs to be specifically explained. The dynamic association module performs intelligent clustering on the pipe network operation parameters to obtain parameter clusters corresponding to the key transfer paths. The intelligent clustering is based on the multidimensional characteristics of the pipe network operation parameters and adopts a density clustering algorithm to calculate the Euclidean distance between the parameters. The specific calculation formula is as follows:

[0107]

[0108] Among them, d i,j represents the Euclidean distance between the operating parameters of the pipe network of group i and group j, v i 、v j Indicates the medium flow rate of group i and j, P i 、P j Indicates the pipeline pressure of group i and j, T i 、T j represents the ambient temperature of the i-th and j-th groups. Parameters with similar change trends are divided into the same cluster. The specific calculation formula is as follows:

[0109]

[0110] Among them, Core(i) indicates whether the i-th group of parameters is a core point, ∈ represents the radius threshold of density clustering, Num(j|·) represents the number of parameter groups that meet the conditions, and MinPts represents the minimum number of neighbors of the core point. A sub-association model is established for each parameter cluster and the corresponding transmission path. The specific calculation formula is as follows:

[0111]

[0112] in, represents the core transfer feature value of path p predicted by the sub-model, v, P, and T represent the medium flow rate, pipeline pressure, and ambient temperature within the parameter cluster, ω0 represents the bias term, and ω1, ω2, and ω3 represent regression coefficients, corresponding to the weights of flow rate, pressure, and temperature, respectively. The larger the value, the stronger the influence of the parameter on the feature. The confidence threshold is set for the sub-association model. The sub-association model uses the operating parameters of the parameter cluster as input and the core features of the corresponding transfer path as output. It is trained based on the multivariate linear regression algorithm.

[0113] When real-time data is input, if the prediction error of the sub-association model exceeds the confidence threshold, the model reconstruction is automatically triggered, and the updated sub-models are integrated to obtain the transfer-run association model. The specific calculation formula for the prediction error and threshold judgment is as follows:

[0114]

[0115] Among them, e m represents the prediction error of sub-model m, Represents the core feature value predicted by the sub-model, f p represents the core feature value actually collected, θ represents the confidence threshold, and flag represents the reconstruction trigger flag. The specific calculation formula for updating the sub-model parameters after reconstruction is as follows:

[0116]

[0117] Among them, ω' krepresents the regression coefficient of the reconstructed sub-model m, L represents the sample size of the latest real-time data, f p,t It represents the predicted eigenvalue and actual eigenvalue of the t-th group of samples. The prediction error refers to the deviation between the result of the sub-model predicting the core transfer feature based on the real-time operation parameters and the actual collected eigenvalue. If the deviation exceeds the confidence threshold, it indicates that the original model cannot adapt to the current association law. The model reconstruction calls the latest real-time data to retrain the corresponding sub-model and adjust the model parameters. After the reconstruction is completed, the engine fuses all sub-models according to the path weight to generate a new transfer-operation association model. The specific calculation formula is as follows:

[0118]

[0119] Where F represents the output of the final transfer-operation correlation model, Z represents the total number of critical transfer paths, and ω p represents the weight of path p, Represents the output of sub-model m.

[0120] It should be noted that the fusion integrates the outputs of all sub-models through weighted summation, and the weight ω p According to the global impact distribution of the path, the final output F comprehensively reflects the core characteristics of all critical paths, forming a self-updating transfer-operation association model, which provides a basis for subsequent parameter adjustment decisions.

[0121] Dynamic adjustment decision module: Based on the transfer-operation correlation model and historical decision case library, it executes dynamic adjustment decisions on pipeline network operation parameters;

[0122] In this embodiment, the dynamic adjustment decision module needs to be specifically explained. The dynamic adjustment decision module obtains the real-time transmission data of stress and strain of the current pipeline network, inputs the transmission-operation correlation model to obtain the initial adjustment parameter range, and calls the historical decision case library. Each case contains the stress and strain state D hist , adjust the parameter v hist 、P hist 、T hist , implementation effect E hist ,Through the case similarity matching algorithm, the effective parameter set verified by practice is screened out from the parameter range. The specific calculation formula is as follows:

[0123]

[0124] Among them, S represents the similarity, ranging from [0, 1], σ hist,k , ε hist,k The specific calculation formulas for stress and strain values ​​of historical cases and effective parameter screening are as follows:

[0125] Valid={(v,P,T)|v∈[vlow ,v hist ],P∈[P low ,P hist ],T∈[T low ,T hist ],S

[0126] ≥S0,E hist ≥E0}

[0127] Among them, Valid represents the set of valid parameter pairs, that is, the set of parameter combinations that meet all conditions after screening, | represents the following conditions are met, v low represents the lowest reasonable threshold value of parameter v, v hist represents the highest reasonable threshold of parameter v, P low represents the lowest reasonable threshold of parameter P, P hist represents the highest reasonable threshold of parameter P, T low represents the lowest reasonable threshold of parameter T, T hist Indicates the highest reasonable threshold of parameter T, S≥S0 indicates that the feature contribution of the parameter pair must meet the minimum requirement, E hist ≥E0 means that the historical validity of the parameter pair must meet the minimum requirement, and the valid parameter sets are prioritized to obtain the optimal adjustment plan;

[0128] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0129] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0130] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0131] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0132] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0133] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0134] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A business process dynamic optimization decision-making system based on AI industrial big data processing, characterized by: include: Data acquisition module: obtains stress and strain transfer data and corresponding network operating parameters of the petrochemical pipeline network under different operating conditions, and builds a historical decision-making case library; Feature extraction module: Adaptively determine the core transfer feature set from stress-strain transfer data, and iteratively optimize feature screening weights through historical data; Dynamic correlation module: Input the transfer feature set and pipe network operation parameters into the dynamic correlation modeling engine to generate a self-updating transfer-operation correlation model and correct the correlation coefficient according to the real-time data deviation; Intelligent clustering is performed on the pipe network operating parameters to obtain parameter clusters that correspond one-to-one to the key transfer paths. Intelligent clustering is based on the multidimensional characteristics of the pipe network operating parameters and adopts a density clustering algorithm. The Euclidean distance between the parameters is calculated. The specific calculation formula is as follows: Among them, d i,j represents the Euclidean distance between the operating parameters of the pipe network of group i and group j, v i 、v j Indicates the medium flow rate of group i and j, P i 、P j Indicates the pipeline pressure of group i and j, T i 、T j represents the ambient temperature of the i-th and j-th groups. Parameters with similar change trends are divided into the same cluster. The specific calculation formula is as follows: Among them, Core(i) indicates whether the i-th group of parameters is a core point, ∈ represents the radius threshold of density clustering, Num(j|·) represents the number of parameter groups that meet the conditions, and MinPts represents the minimum number of neighbors of the core point. A sub-association model is established for each parameter cluster and the corresponding transmission path. The specific calculation formula is as follows: in, represents the core transfer feature value of path p predicted by the sub-model, v, P, and T represent the medium flow rate, pipeline pressure, and ambient temperature within the parameter cluster, ω0 represents the bias term, and ω1, ω2, and ω3 represent regression coefficients, corresponding to the weights of flow rate, pressure, and temperature, respectively. The larger the value, the stronger the influence of the parameter on the feature. The confidence threshold is set for the sub-association model. The sub-association model uses the operating parameters of the parameter cluster as input and the core features of the corresponding transfer path as output. It is trained based on the multivariate linear regression algorithm. When real-time data is input, if the prediction error of the sub-association model exceeds the confidence threshold, the model reconstruction is automatically triggered, and the updated sub-models are integrated to obtain the transfer-run association model. The specific calculation formula for the prediction error and threshold judgment is as follows: Among them, e m represents the prediction error of sub-model m, Represents the core feature value predicted by the sub-model, f p represents the core feature value actually collected, θ represents the confidence threshold, and flag represents the reconstruction trigger flag. The specific calculation formula for updating the sub-model parameters after reconstruction is as follows: Among them, ω' k represents the regression coefficient of the reconstructed sub-model m, L represents the sample size of the latest real-time data, f p,t It represents the predicted eigenvalue and actual eigenvalue of the t-th group of samples. The prediction error refers to the deviation between the result of the sub-model predicting the core transfer feature based on the real-time operation parameters and the actual collected eigenvalue. If the deviation exceeds the confidence threshold, it indicates that the original model cannot adapt to the current association law. The model reconstruction calls the latest real-time data to retrain the corresponding sub-model and adjust the model parameters. After the reconstruction is completed, the engine fuses all sub-models according to the path weight to generate a new transfer-operation association model. The specific calculation formula is as follows: Where F represents the output of the final transfer-operation correlation model, Z represents the total number of critical transfer paths, and ω p represents the weight of path p, represents the output of sub-model m, Dynamic adjustment decision module: Based on the transfer-operation association model and the historical decision case library, it executes dynamic adjustment decisions on the pipe network operation parameters.

2. The business process dynamic optimization decision system based on AI industrial big data processing according to claim 1 is characterized in that: The data acquisition module deploys distributed stress and strain sensors in the pipeline at a preset density to ensure the spatial coverage integrity of each node in the pipeline network. The sampling frequency of the sensors strictly meets the Nyquist sampling theorem to fully capture the instantaneous transmission characteristics of stress and strain, including high-frequency vibration signals caused by turbulent impact and low-frequency deformation signals caused by continuous pressure. The sensor data is connected to the data acquisition terminal via wireless transmission, and the stress and strain values ​​of each node in the pipeline network are collected in real time. The stress and strain values ​​together represent the structural bearing state of the pipeline. The collected stress and strain values ​​are filtered and stored in the form of a time-series signal stream.

3. The business process dynamic optimization decision system based on AI industrial big data processing according to claim 2 is characterized in that: At the same deployment node position of the distributed stress and strain sensor, the network operation parameters are synchronously obtained, including: medium flow rate, pipeline pressure and ambient temperature. The sensor node number is used as the unique index to construct a multidimensional parameter matrix. The network operation parameters of the same node at the same timestamp are arranged in time series to form a time-parameter two-dimensional matrix. The network operation parameters collected by the node include: medium flow rate v, pipeline pressure P, ambient temperature T env , suppose the collected timestamp sequence is t1, t2, ..., t n , then the time-parameter two-dimensional matrix M k Defined as: Its element expression is: M k (i,j)=param j (t i ), where M k Represents the time-parameter two-dimensional matrix of node k, i represents the time index, j represents the parameter index, and j=1 corresponds to the medium flow velocity v, j=2 corresponds to the pipeline pressure P, and j=3 corresponds to the ambient temperature T env .

4. The business process dynamic optimization decision system based on AI industrial big data processing according to claim 3 is characterized in that: Automatically optimize the spatial coordinate association algorithm based on historical matching errors, dynamically align the real-time stress and strain transmission signals with the pipeline network operating parameters, and determine the spatiotemporal correlation data set. Historical matching errors refer to the actual spatial deviation of the theoretically associated pipeline network nodes in past data and the actual timestamp deviation of the theoretical synchronization time. Using historical matching errors as the basis for algorithm optimization, the calibration coefficient of the spatial coordinate association algorithm and the delay compensation value of the time alignment algorithm are automatically adjusted by analyzing the distribution of historical matching errors. The steps to implement dynamic alignment are as follows: Time dimension: Compare the trend consistency of the two types of signals through a sliding window and correct the timestamp deviation. The specific calculation formula is as follows: Δt offset =arg max τ p(t) Where ρ represents the trend correlation coefficient between stress-strain signal and pipeline network operation parameters, x t Represents the stress-strain signal value at time t, y t represents the network operation parameter value at time t, T represents the sliding window length, τ represents the time offset, Δt offset represents the optimal time offset, Represents x t 、y t The mean value in the time window T; Spatial dimension: Combine the pipeline topology to calibrate the coordinates to ensure that the stress and strain signal nodes coincide with the operating parameter nodes in physical space. The specific calculation formula for spatial dimension coordinate calibration is as follows: (x',y',z')=(x+Δx k ,y+Δy k ,z+Δz k ) Among them, x, y, z represent the theoretical installation coordinates of the sensor, x', y', z' represent the actual space coordinates after calibration, Δx k , Δy k , Δz k represents the coordinate correction of node k, where k represents the sensor node number.

5. The business process dynamic optimization decision system based on AI industrial big data processing according to claim 1 is characterized in that: The feature extraction module loads the historical feature importance evaluation results, performs intelligent transfer path analysis on the spatiotemporal correlation dataset, and automatically identifies the associated paths. The historical feature importance evaluation results are quantified values ​​of the contribution strength of the features obtained through historical data analysis to pipeline status assessment. Using the spatiotemporal correlation dataset as input, a graph neural network is used to construct a node-node stress transfer network model. By calculating the stress transfer efficiency and global impact of each path, the critical transfer path is automatically identified. The specific calculation formula of the stress transfer efficiency is as follows: Among them, η k→l represents the stress transfer efficiency from node k to node l, σ k (t), σ l (t) represents the stress value of nodes k and l at time t, represents the average stress value of nodes k and l in the time window T; The specific calculation formula of the global influence is as follows: Among them, δ p represents the global influence of path p, ∑ all represents the set of stress values ​​of all nodes, ∑ all\p represents the set of stress values ​​of all nodes after removing path p, and Var(·) represents the variance function, which measures the degree of discreteness of stress distribution.

6. The business process dynamic optimization decision system based on AI industrial big data processing according to claim 5 is characterized in that: The dynamic attenuation coefficient and diffusion coefficient of each key transmission path are calculated and dynamically corrected as the pipeline ages. The dynamic attenuation coefficient represents the energy attenuation degree of stress along the transmission path with distance. The specific calculation formula is as follows: a p =a p0 ·(1+λ·γ k ) Among them, α p represents the dynamic attenuation coefficient of path p, α p0 represents the initial attenuation coefficient of path p, λ represents the aging influence coefficient, γ k It represents the aging degree of node k on path p. The diffusion coefficient represents the diffusion range of stress from the main path to the surrounding nodes. The specific calculation formula is as follows: b p =b p0 ·(1+μ·g k ) Among them, β p represents the diffusion coefficient of path p, β p0 represents the initial diffusion coefficient of path p, μ represents the aging diffusion coefficient. Pipeline aging will change its mechanical properties. The stress transfer resistance of the aging node increases, resulting in an increase in the attenuation coefficient. At the same time, the stress distribution uniformity in the aging area decreases, resulting in an increase in the diffusion coefficient. The pipeline aging data is obtained in real time, and a mapping relationship between aging degree and coefficient correction amount is established.

7. The business process dynamic optimization decision system based on AI industrial big data processing according to claim 6 is characterized in that: The top-N feature combinations are screened from the dynamic attenuation coefficient and diffusion coefficient to determine the core transfer feature set. The influence of the quantitative features on the pipeline network status assessment task is ranked. The specific calculation formula is as follows: Among them, Imp(f) represents the contribution of feature f, f represents the feature to be evaluated, M represents the number of decision trees in random forest, OOB err,t Represents the out-of-bag error of the t-th tree, OOB err,t(f) It represents the out-of-bag error of the tth tree after randomly shuffling the feature f. The sorting criteria include the correlation between the feature and the maximum stress value and failure probability of the pipeline network, as well as the effective frequency of the feature in historical decision-making. After sorting from high to low by contribution, the top N features are selected for combination, where the value of N is determined by the data complexity, which is quantified by the number of nodes and the parameter fluctuation variance index. The core transfer feature set concentrates on the features that are most explanatory of the pipeline network status.

8. The business process dynamic optimization decision system based on AI industrial big data processing according to claim 1 is characterized in that: The dynamic adjustment decision module obtains the real-time transmission data of stress and strain of the current pipeline network, inputs the transmission-operation correlation model to obtain the initial adjustment parameter range, and calls the historical decision case library. Each case contains the stress and strain state D hist , adjust the parameter v hist 、P hist 、T hist , implementation effect E hist ,Through the case similarity matching algorithm, the effective parameter set verified by practice is screened out from the parameter range. The specific calculation formula is as follows: Among them, S represents the similarity, ranging from [0, 1], σ hist,k , ε hist,k The specific calculation formulas for stress and strain values ​​of historical cases and effective parameter screening are as follows: Valid={(v,P,T)|v∈[v low ,v hist ],P∈[P low ,P hist ],T∈[T low ,T hist ],S ≥S0,E hist ≥E0} Among them, Valid represents the set of valid parameter pairs, that is, the set of parameter combinations that meet all conditions after screening, | represents the following conditions are met, v low represents the lowest reasonable threshold value of parameter v, v hist represents the highest reasonable threshold of parameter v, P low represents the lowest reasonable threshold of parameter P, P hist represents the highest reasonable threshold of parameter P, T low represents the lowest reasonable threshold of parameter T, T hist Indicates the highest reasonable threshold of parameter T, S≥S0 indicates that the feature contribution of the parameter pair must meet the minimum requirement, E hist ≥E0 indicates that the historical validity of the parameter pair must meet the minimum requirement, and the valid parameter sets are prioritized to obtain the optimal adjustment plan.

Images