Simulation optimization method and system based on digital twin-based digital transformation platform

Abstract

The present application relates to the field of enterprise digital transformation and industrial internet, in particular to a digital transformation platform simulation optimization method and system based on digital twinning, which interacts with at least two target business systems, collects platform operation evaluation data sets, updates the digital twinning model to generate a unified time snapshot, forms a target control scheme according to the snapshot and decomposes cross-system writing actions, divides compensable / retriable, takes the confirmed version identifier before delivery, combines the data set to calculate the twinning consistency drift potential value and the writing action rollback risk potential value, determines the dynamic submission sequence, executes according to the drift potential value and the two water level lines, re-calculates after the reply and switches the execution mode, and the unified time snapshot drives the twinning deduction more accurately; based on the version identifier and the running index, the drift / rollback risk is quantized and dynamically sorted; two water level lines are controlled + reply re-evaluation, combined with compensation rollback and retry, to improve the safety and controllability of cross-system delivery and traceability.

Description

Technical Field

[0001] This invention relates to the field of enterprise digital transformation and industrial internet, and more specifically, to a simulation optimization method and system for digital transformation platforms based on digital twins. Background Technology

[0002] In existing enterprise digital transformation practices, business operations such as manufacturing, equipment maintenance, quality management, and supply chain collaboration are typically handled by different target business systems. Related platforms often aggregate multi-source operational data through API calls, message subscriptions, and data synchronization, and use this data to build or update digital twins to support decision-making activities such as process cycle optimization, production line load balancing, resource scheduling, and risk warning. While existing technologies can achieve data acquisition and simulation to a certain extent, uncertainties in the data link and version confirmation differences during the cross-business closed-loop execution phase can affect the reliability of the digital twin state. This is especially true when interface latency fluctuates, timeouts occur, and error rates rise, potentially causing discrepancies between the state snapshot obtained by the platform and the confirmed state of the target business system.

[0003] Existing technologies have also proposed various methods for cross-business deployment and consistency assurance. For example, in engineering, idempotent writes, status queries, and retry strategies are often used to reduce the risk of instruction loss due to occasional timeouts. Simultaneously, compensatory transactions or manual review release processes are also commonly used to roll back previous writes after some writes have taken effect, thereby reducing business disruptions caused by execution failures. Furthermore, some solutions set tiered alarm thresholds to monitor interface health or task failure rates, and limit the intensity of automated deployment when metrics are abnormal. These solutions are generally effective under stable network conditions and relatively simple dependencies.

[0004] However, there is still room for further optimization in collaborative write scenarios targeting at least two business systems: First, existing solutions mostly use interface availability or single failure rate indicators as gating criteria, and rarely incorporate confirmed version identifiers into consistency assessments, making it difficult to quantify the dispersion and evolution trend of version order of different target business systems within a time window; Second, cross-system write actions often have different reversibility and rollback costs. Although existing solutions support compensation and retries, they lack a unified quantitative characterization of the irreversibility of each write action, compensation success rate, and other historical execution characteristics, making it difficult to form a dynamic commit order that adapts to changes in risk; Third, gating strategies are often determined once before the task begins, and there is little reassessment of consistency risks and switching of execution methods after each write confirmation receipt is received. Therefore, they may be too aggressive when consistency drift rises rapidly, and too conservative after consistency is restored. Based on the aforementioned engineering facts, existing technologies require a simulation optimization closed-loop method that can construct a consistency risk representation by combining confirmed version identifiers and interface operation indicators, and construct a write rollback risk representation by combining historical rollback performance, thereby implementing hierarchical gating, step-by-step execution, and abnormal rollback retry in the dynamic submission order during the distribution process. Summary of the Invention

[0005] In view of the shortcomings of existing technologies, the purpose of this invention is to provide a simulation optimization method and system for digital transformation platforms based on digital twins.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The simulation optimization method for digital transformation platforms based on digital twins includes the following steps:

[0008] S1: The digital transformation platform interacts with at least two target business systems to collect platform operation evaluation datasets; based on the production operation data in the platform operation evaluation datasets, the digital twin model is updated to generate a snapshot of the current twin state with a unified time identifier;

[0009] S2: The digital transformation platform determines the target control scheme to be executed based on the current twin state snapshot; decomposes the target control scheme into a cross-system write action set, and divides the cross-system write action set into a compensable write action subset and a retryable write action subset;

[0010] S3: Before distributing the target control scheme across systems, the digital transformation platform obtains the confirmed version identifiers of each target business system and calculates the twin consistency drift potential value based on the confirmed version identifiers and the platform operation evaluation dataset; it calculates the rollback risk potential value of each write action based on the historical execution data in the platform operation evaluation dataset and determines the dynamic submission order based on the rollback risk potential value.

[0011] S4: The digital transformation platform, based on the comparison results of the twin consistency drift potential energy value with the first water level threshold and the second water level threshold, gating and issuing a set of cross-system write actions in a dynamic submission order; after executing any write action, it obtains a write confirmation receipt, recalculates the twin consistency drift potential energy value, and switches the dynamic gating and issuing execution mode based on the comparison results of the twin consistency drift potential energy value with the first water level threshold and the second water level threshold.

[0012] Furthermore, the cross-system write action set includes write actions targeting each target business system respectively; at least some of the write actions in the cross-system write action set are configured with corresponding compensation write actions, which are used to roll back the corresponding target business system to restore it to the state before the write action has been executed.

[0013] Furthermore, the twin-consistent drift potential value is obtained as follows: based on the interface latency, interface timeout rate, and interface error rate determined by the platform operation evaluation dataset, the interface trust weights of each target business system are generated; the confirmed version identifiers of each target business system are mapped to version sequences and smoothed within a preset sliding time window; based on the interface trust weights, the multi-statistic dispersion and time evolution term of the version sequence are weighted and fused to obtain the drift composite value, and the drift composite value is normalized to obtain the twin-consistent drift potential value.

[0014] Furthermore, the dispersion of the multistatistics must include at least two of the range, quantile difference, and variance, and the time evolution term must include at least the drift slope or drift acceleration within a preset sliding time window.

[0015] Furthermore, based on the historical execution data in the platform operation evaluation dataset, the irreversibility index and compensation success rate of each write action are calculated, and the rollback risk potential value of each write action is calculated based on the irreversibility index and compensation success rate; the compensation success rate is the ratio of the number of successful compensation write actions to the total number of compensation write actions triggered within the preset sliding time window.

[0016] Furthermore, the irreversibility index is calculated based on at least two of the following three factors: whether the corresponding compensation write action exists, the expected execution time of the corresponding compensation write action, and the scope of influence of the corresponding compensation write action.

[0017] Furthermore, the dynamic submission sequence gating includes: automatically executing a set of cross-system write actions according to the dynamic submission sequence when the twin consistency drift potential energy value is not greater than the first water level threshold; executing in at least two segments according to the dynamic submission sequence when the twin consistency drift potential energy value is greater than the first water level threshold but not greater than the second water level threshold; stopping automatic writing and outputting a target control scheme execution suggestion when the twin consistency drift potential energy value is greater than the second water level threshold, and continuously monitoring until the twin consistency drift potential energy value is not greater than the second water level threshold.

[0018] Furthermore, in the event of a write action failure or a write confirmation receipt timeout, the digital transformation platform performs the following exception handling: for the subset of compensable write actions that have already been executed, the corresponding compensable write actions are rolled back in reverse order of dynamic submission; for the subset of retryable write actions that have already been executed, the platform performs status queries and retrying until a write confirmation receipt is obtained; and updates the digital twin model and the confirmed version identifier based on the write confirmation receipt or the rollback and retry results.

[0019] Furthermore, the execution mode switching of dynamic gating includes continuing automatic execution, switching to segmented execution, or stopping automatic writing and outputting target control scheme execution suggestions.

[0020] Furthermore, the simulation and optimization system for a digital transformation platform based on digital twins includes:

[0021] The data acquisition module is used to interact with at least two target business systems to acquire platform operation evaluation datasets.

[0022] The twin update module is used to update the digital twin model based on the production operation data in the platform operation evaluation dataset and generate a snapshot of the current twin state with a unified time identifier.

[0023] The scheme decomposition module determines the target control scheme, decomposes the cross-system write action set into a compensable write action subset and a retryable write action subset, and determines the dynamic submission order.

[0024] The gating execution module calculates the twin-consistent drift potential energy value, dynamically gating it according to the first water level threshold and the second water level threshold, and then executes it. After writing the confirmation receipt, it recalculates and switches.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] This invention interacts with at least two target business systems to collect platform operation evaluation datasets and updates the digital twin model based on production operation data, generating a snapshot of the current twin state with a unified time identifier. This allows simulation optimization and control decisions to be based on a state benchmark with semantic alignment at the same time. This improves the consistency and traceability of multi-source state fusion, reduces the impact of state deviations caused by inconsistent sampling times and interface delay fluctuations on scheme deduction and execution judgment, and makes the target control scheme closer to actual production operation constraints.

[0027] The target control scheme is decomposed into a set of cross-system write actions, which are further divided into a compensable write action subset and a retryable write action subset. Before deployment, the confirmed version identifiers of each target business system are obtained, and the twin consistency drift potential value is calculated by combining the platform operation evaluation data with the centralized interface latency, interface timeout rate, interface error rate and other operation indicators. Then, based on historical execution data, the non-rollback characteristics and the performance of successful compensation are quantified to form the write action rollback risk potential value and determine the dynamic submission order. This achieves an adaptive orchestration that takes into account both consistency risk and rollback cost, so that the write actions proceed in an orderly manner from low risk to high risk under dependency constraints, reducing the impact scope and recovery cost in failure scenarios, and improving the security boundary and stability of cross-business collaborative writing.

[0028] During the execution phase, a hierarchical gating mechanism based on the twin consistency drift potential value and the first and second water level thresholds is introduced: when in a low-risk range, execution is automatic according to the dynamic submission order; when in a medium-risk range, execution is performed in at least two segments according to the dynamic submission order; when exceeding the high-risk boundary, automatic writing stops and a target control scheme execution suggestion is output, while continuous monitoring continues until the risk decreases; and after each write action, a write confirmation receipt is obtained, the twin consistency drift potential value is recalculated, and the execution mode is dynamically switched accordingly, forming an adaptive control with real-time closed-loop feedback. In the event of a write action failure or a write confirmation receipt timeout, this invention can trigger a compensation write action rollback in reverse according to the dynamic submission order for the already executed compensable write actions, query and retry the execution status of retryable write actions until a confirmation receipt is obtained, and synchronously update the digital twin model and the confirmed version identifier based on the receipt or rollback and retry results, thereby enhancing the controllability of anomaly handling, the verifiability of execution results, and the resilience of the overall process. Attached Figure Description

[0029] Figure 1 A flowchart illustrating the simulation optimization method for a digital transformation platform based on digital twins;

[0030] Figure 2 This is a structural block diagram of a simulation optimization system for a digital transformation platform based on digital twins.

[0031] Figure 3 A schematic diagram illustrating the calculation process for twin consistency drift potential energy;

[0032] Figure 4 A flowchart illustrating the dynamic submission order determination, two-level gating issuance, and anomaly handling for cross-system write actions. Detailed Implementation

[0033] Example 1: Refer to Figure 1 , Figure 3 , Figure 4 The simulation optimization method for digital transformation platforms based on digital twins includes the following steps:

[0034] S1: The digital transformation platform interacts with at least two target business systems to collect platform operation evaluation datasets; based on the production operation data in the platform operation evaluation datasets, the digital twin model is updated to generate a snapshot of the current twin state with a unified time identifier;

[0035] In one specific implementation, the digital transformation platform interacts with at least two target business systems to collect platform operation evaluation datasets. For example, it interacts with a manufacturing execution management business system and an equipment operation and maintenance management business system, respectively. Data is pulled or subscribed to and pushed according to a preset collection cycle. Data from different target business systems undergoes field mapping, noise reduction, verification, and time-series alignment. The platform operation evaluation dataset includes at least two categories: production operation data and operation evaluation data. Production operation data includes, for example, order information, work order status, process flow records, output and cycle time, work-in-process quantity, material batch and consumption records, and quality inspection data. The data includes test results and reasons for non-compliance, equipment operating status and alarm events, key process parameters, energy consumption metering data, and operational evaluation data such as interface call logs, interface latency, interface timeout rate, interface error rate, data missing and duplicate situations, event arrival order deviation, and execution results and write confirmation receipt records related to cross-system write actions in historical execution data. At the same time, the digital transformation platform assigns a unified time identifier to each piece of data entering the database and uses the unified time identifier to merge multi-source data into a consistent time slice, forming a traceable time-series platform operation evaluation dataset, providing a consistent data benchmark for subsequent updates to the digital twin model.

[0036] The digital transformation platform updates the digital twin model based on production operation data in the platform operation evaluation dataset. Specifically, the platform maps work order flow, process parameters, and quality results from the manufacturing execution management business system to process status variables in the digital twin model, and maps equipment start-up and shutdown, fault alarms, and maintenance records from the equipment operation and maintenance management business system to equipment status variables in the digital twin model. By combining a unified time identifier, the platform completes state merging and conflict resolution within the same time slice, ensuring the digital twin model remains continuous and consistent in the time dimension. After the update is completed, the digital transformation platform outputs a snapshot of the current twin state with a unified time identifier. The snapshot of the current twin state at least represents the production line load, equipment availability, work-in-process distribution, material constraints, key process parameters, and quality status corresponding to the unified time identifier.

[0037] S2: The digital transformation platform determines the target control scheme to be executed based on the current twin state snapshot; decomposes the target control scheme into a cross-system write action set, and divides the cross-system write action set into a compensable write action subset and a retryable write action subset;

[0038] In one specific implementation, the digital transformation platform determines the target control scheme to be executed based on the current twin state snapshot. Specifically, the digital transformation platform extracts key state variables such as capacity load, work-in-process distribution, equipment availability, material constraints, and quality status from the current twin state snapshot, constructs an objective function oriented towards delivery achievement, production line balance, and risk controllability, and performs simulation deduction and comparative evaluation of candidate control strategies under the premise that the constraints are met, thereby outputting the target control scheme. For example, when the current twin state snapshot indicates that the availability of a key piece of equipment has decreased and work-in-process has accumulated before that process, the target control scheme may include a combination of control strategies such as adjusting work order priority, temporarily rerouting to an alternative production line, and triggering maintenance resources in advance, so as to reduce queuing and avoid the spread of quality risks.

[0039] The digital transformation platform decomposes the target control scheme into a set of cross-system write actions. This set includes write actions targeting specific target business systems. Based on the reversibility and confirmation methods of the write actions' impact on business states, the platform further divides the set into a compensable write action subset and a retryable write action subset. For write actions that modify existing business data and require rollback capability, the platform configures corresponding compensatory write actions. These compensatory write actions are used to roll back the corresponding target business system to its pre-write state after the write action has been executed, such as prioritizing work orders. Write actions that adjust the level from the original value to the new value are included in the compensable write action subset, and compensatory write actions for writing back the original value are configured simultaneously to restore the state before the write. For write actions that are mainly based on confirmation receipts and allow repeated submissions through status queries without changing the final consistent result, the digital transformation platform includes them in the retryable write action subset. For example, when submitting a one-time execution instruction to the corresponding target business system and relying on write confirmation receipts to complete the closed loop, the digital transformation platform can perform status queries and retry without receiving write confirmation receipts, so as to ensure that the cross-system write action set can be implemented according to the target control scheme and remain controllable in abnormal situations.

[0040] S3: Before distributing the target control scheme across systems, the digital transformation platform obtains the confirmed version identifiers of each target business system and calculates the twin consistency drift potential value based on the confirmed version identifiers and the platform operation evaluation dataset; it calculates the rollback risk potential value of each write action based on the historical execution data in the platform operation evaluation dataset and determines the dynamic submission order based on the rollback risk potential value.

[0041] In one specific implementation, before distributing the target control scheme across systems, the digital transformation platform initiates version query and confirmation interactions with at least two target business systems to obtain the confirmed version identifiers of each target business system. Simultaneously, it extracts indicators related to interface status, such as interface latency, interface timeout rate, and interface error rate, from the platform's operation evaluation dataset. The confirmed version identifier is used to characterize the currently externally verifiable consistent version of each target business system, while the interface latency, interface timeout rate, and interface error rate are used to characterize the reliability of the version acquisition and status confirmation link. For example, for the production manufacturing execution management business system and the equipment operation and maintenance management business system, the digital transformation platform obtains the work order status version identifier and the equipment status version identifier, respectively, and extracts the latency distribution, timeout ratio, and error ratio of the corresponding interfaces in the most recent multiple sampling periods as input for subsequent consistency evaluation.

[0042] The digital transformation platform generates interface trust weights for each target business system based on interface latency, timeout rate, and error rate centrally determined in the platform operation evaluation data. The interface trust weights decrease as interface latency, timeout rate, or error rate increases to reduce the impact of low-trust links on consistency evaluation. Simultaneously, the digital transformation platform maps the confirmed version identifiers of each target business system to version sequences and smooths the version sequences within a preset sliding time window to weaken the amplification effect of instantaneous jitter or short-term rollback on the evaluation results. The smoothing process can use a weighted moving average or median filter within the sliding time window while maintaining uniform time identifier alignment, thereby obtaining a smooth version sequence for consistency calculation.

[0043] In one specific implementation, to avoid underestimating the risk of low-reliability links due to a decrease in interface trust weight, the digital transformation platform can introduce a conservative rule when calculating the twin consistency drift potential value: when the interface trust weight of any target business system falls below a preset lower limit or its interface timeout / error rate exceeds a preset alarm threshold, the digital transformation platform marks the target business system as "unreliable confirmation state" and increases the twin consistency drift potential value by a preset amount or directly switches the execution mode to segmented execution mode; when the unreliable confirmation state continues for more than a preset duration, the digital transformation platform stops issuing write actions across systems and outputs execution suggestions until the interface state recovers or manual confirmation is completed. Through the above conservative rule, gating misjudgments caused by unreliable observations can be suppressed when the version confirmation link deteriorates, improving the security and controllability of cross-system writes.

[0044] In one specific implementation, the digital transformation platform maps the confirmed version identifiers of each target business system to comparable version sequences. .

[0045] When the version identifier is confirmed to be a monotonically increasing sequence number, that sequence number can be directly used as the version sequence number;

[0046] Once the version identifier has been confirmed as a timestamp, the timestamp can be set to a preset time granularity. Quantify, for example, take Alternatively, the timestamp can be linearly scaled and used as the version sequence number;

[0047] Once the version identifier has been confirmed as a semantic version number, the major version number, minor version number, and revision number can be parsed and assigned to a preset base. Combining them yields integer ordinal positions, for example... Within a preset sliding time window, the digital transformation platform... Smoothing can be performed to reduce transient fluctuations. Smoothing methods such as moving average or exponential moving average can be used. , ;

[0048] in This represents the number of sample points within the sliding time window. For the smoothing coefficient, an example can be taken. And maintain alignment with the unified time marker.

[0049] In one specific implementation, the digital transformation platform uses the interface call logs within a preset sliding time window as the statistical scope, and calculates the number of calls respectively. Interface latency of the target business system Interface timeout rate Interface error rate The interface timeout rate is the ratio of the number of timeouts to the total number of calls, and the interface error rate is the ratio of the number of errors to the total number of calls; these metrics are then normalized to... Interval:

[0050]

[0051]

[0052]

[0053] in It can be determined by statistical values ​​or offline calibration values ​​within the same sliding time window.

[0054] Digital transformation platforms further enhance interface reliability. Based on this, an interface trust weight is generated. :

[0055] ;when In this case, the interface trust weight can be evenly distributed or the weight from the previous statistical period can be used as a fallback. Preferably, To predetermine non-negative coefficients, the example can be taken as follows: Timeout / error more directly reflects the risk of link unavailability and failure, and should be punished more than latency, so it is given higher weight.

[0056] The digital transformation platform uses interface trust weights to weight and fuse the dispersion of multiple statistics for version sequence positions with time evolution terms to obtain a comprehensive drift value. This comprehensive drift value is then normalized to obtain a twin-consistent drift potential value. The dispersion of multiple statistics includes at least two of the range, quantile difference, and variance. For example, the range and quantile difference of the smoothed version sequence positions are calculated separately within a preset sliding time window and combined with the variance to form a dispersion characterization. The time evolution term includes at least the drift slope or drift acceleration within the preset sliding time window. For example, the drift potential is obtained by fitting the time series of the smoothed version sequence. The drift slope is calculated and the change in slope is used to obtain the drift acceleration. The digital transformation platform weights the dispersion characterization and time evolution terms according to the interface credibility weight to form a drift comprehensive quantity. The drift comprehensive quantity is then mapped to a unified scale through a preset normalization interval, thereby obtaining a twin consistency drift potential energy value that can be used for subsequent gating decisions. For example, when the smooth version sequence of the manufacturing execution management business system and the equipment operation and maintenance management business system increases in dispersion and the drift slope continues to rise within a preset sliding time window, the twin consistency drift potential energy value increases accordingly to indicate an increase in cross-system consistency risk.

[0057] In one specific implementation, within a preset sliding time window, the first... The smooth version sequence of the target business systems is denoted as . And define the dispersion of multiple statistics. Time evolution item Overall drift volume Consistent drift potential with twin The following (among them) (Preset coefficients): By combining at least two of these terms, the coefficient of the unused term can be set to zero. For quantile difference, For variance; The sequence within this time window The maximum value (maximum version sequence); The sequence within this time window The minimum value (minimum version sequence); ;in For the sequence The drift slope is obtained by linear fitting within the sliding time window. The drift acceleration can be obtained from the slope of the first-order difference sequence or the second-order difference statistics; ;in As the interface's trust weight;

[0058]

[0059] in This is a truncation function that cuts the result to... Interval. As an example of a coefficient, it can be taken as... . Highlighting the range term and emphasizing the "worst deviation" across systems facilitates a rapid increase in risk when a few systems deviate quickly from their intended path. Taking the quantile difference is more resistant to outliers and characterizes the dispersion of the "majority sample"; The supplementary variance reflects the overall fluctuations but reduces sensitivity to noise. Prioritize slope to capture the direction and speed of continuous drift; Secondary acceleration is used to identify whether the drift is intensifying, but acceleration estimation is more susceptible to sampling jitter, so it has a lower weight. This allows dispersion to dominate, as consistency risks are primarily manifested in version order divergence within the window; at the same time, a 0.4 evolution term is retained to detect trend deterioration in advance, avoiding missed reports before dispersion has expanded.

[0060] In one specific implementation, and The distribution characteristics of the historical drift composite in the platform's evaluation dataset can be obtained offline through calibration.

[0061] Within the preset calibration period, collect a set of historical drift comprehensive quantity samples. and take low quantiles As High quantile As The low quantile and high quantile can be taken as examples. When historical samples are insufficient to stably estimate quantiles, a cold-start fallback approach can be used: a conservative interval given by the engineering boundary is used as the initial... It switches to quantile calibration results after the data accumulates to a preset sample size; and it can re-evaluate quantiles on a rolling basis according to a preset update cycle to adapt to long-term level changes in the interface.

[0062] The digital transformation platform calculates the irreversibility index and compensation success rate of each write action based on historical execution data in the platform operation evaluation dataset. It then calculates the rollback risk potential value for each write action based on these indices, and determines the dynamic submission order based on the rollback risk potential value. The irreversibility index is calculated based on at least two of the following: the existence of a corresponding compensation write action, the expected execution time of the corresponding compensation write action, and the scope of influence of the corresponding compensation write action. For example, a higher irreversibility index is assigned to write actions without a corresponding compensation write action, and a higher irreversibility index is assigned to write actions with longer expected execution times or a wider scope of influence covering more business objects. The compensation success rate is the ratio of the number of successful compensation write actions to the total number of times compensation write actions are triggered within a preset sliding time window. The digital transformation platform uses this to map the achievability of similar write actions in past rollbacks and combines the irreversibility index and compensation success rate to form the rollback risk potential value. For example, an increase in the irreversibility index and a decrease in the compensation success rate are both mapped to an increase in the rollback risk potential value.

[0063] When determining the dynamic submission order, the digital transformation platform can prioritize write actions with lower rollback risk potential and higher compensation success rate, and postpone write actions with higher rollback risk potential or lower compensation success rate and larger impact range. It can also combine the business dependencies between write actions to impose necessary order constraints, so as to form a dynamic submission order that is more in line with the risk control objective before the target control plan is issued. For example, write actions that adjust the priority of work orders are executed before write actions that lock the maintenance window, so as to reduce the probability of triggering high-cost rollbacks in abnormal situations.

[0064] In one specific implementation, for the first Each write operation, the digital transformation platform will have an irreversible index. The components are first normalized to Then perform a linear combination:

[0065] Let the indicator of whether the compensation write action exists be: (Existence) There is no way to take ),

[0066] The normalized value of the expected execution time of the compensation is

[0067]

[0068] The normalized value of the scope of compensation impact is

[0069]

[0070] but ;

[0071] in To ensure that the coefficients are preset to be non-negative and that the sum of the coefficients is one, the example can be taken as follows: . Ultimately: The existence of compensation determines reversibility; Secondly: The duration of compensation determines the recovery window and the cost of downtime; Lower: The scope of influence is often related to the duration and the estimation noise is large, so it is weakened.

[0072] Compensation success rate The number of successful compensations can be counted according to a preset sliding time window. Total number of compensation triggers get:

[0073] when hour ;

[0074] when At that time, it can be taken Alternatively, use Laplacian smoothing.

[0075]

[0076] This is to avoid zero denominator and reduce fluctuations in small samples.

[0077] Retreat risk potential value can be and The combined mapping yields: ;

[0078] in To pre-determine the trade-off coefficients, the example can be taken as follows: ; and accordingly A dynamic commit order is formed from smallest to largest (sorted under dependency constraints when there are dependencies).

[0079] S4: The digital transformation platform, based on the comparison results of the twin consistency drift potential energy value with the first water level threshold and the second water level threshold, gating and issuing a set of cross-system write actions in a dynamic submission order; after executing any write action, it obtains a write confirmation receipt, recalculates the twin consistency drift potential energy value, and switches the dynamic gating and issuing execution mode based on the comparison results of the twin consistency drift potential energy value with the first water level threshold and the second water level threshold.

[0080] In one specific implementation, after obtaining the twin consistency drift potential energy value, the digital transformation platform compares this value with the first water level threshold and the second water level threshold, respectively, and determines the gating strategy based on the dynamic submission order. The setting of the first and second water level thresholds can refer to existing hierarchical alarm and gating control approaches. Specifically, in statistical process control and risk gating practices, warning thresholds and blocking thresholds are typically set to form two water level levels. The first water level threshold defines the acceptable upper limit of consistency fluctuations, and the second water level threshold defines the security thresholds that require blocking automatic writes. Full boundary; In specific settings, the digital transformation platform can perform offline calibration based on the distribution characteristics of historical twin consistency drift potential energy values ​​in the platform operation evaluation dataset, the long-term levels of interface latency, interface timeout rate and interface error rate, and the tolerance of the target control scheme for business continuity. For example, the high percentile of the twin consistency drift potential energy value within the preset sliding time window, superimposed with a safety margin, can be used as the first water level threshold, and a higher percentile or the critical interval where the cost of previous abnormal rollbacks is significantly amplified can be used as the second water level threshold. This avoids execution jitter caused by frequent switching and can block in time when the consistency risk increases significantly.

[0081] The digital transformation platform dynamically gates and executes cross-system write action sets according to the submission order: when the twin consistency drift potential energy value is not greater than the first water level threshold, the cross-system write action set is automatically executed according to the dynamic submission order; when the twin consistency drift potential energy value is greater than the first water level threshold but not greater than the second water level threshold, it is executed in at least two segments according to the dynamic submission order, and a consistency review is performed after each segment; when the twin consistency drift potential energy value is greater than the second water level threshold, automatic writing stops and a target control scheme execution suggestion is output, and gated execution is resumed only after the twin consistency drift potential energy value is no greater than the second water level threshold; after executing any write action, the digital transformation platform obtains a write confirmation receipt and recalculates the twin consistency drift potential energy value, triggering the dynamically gated execution based on the recalculation result. The execution mode switching includes continuing automatic execution, switching to segmented execution, or stopping automatic writing and outputting execution suggestions for the target control scheme. For example, in a scenario where the production manufacturing execution management business system and the equipment operation and maintenance management business system are involved simultaneously, if the write confirmation receipt shows version convergence and the twin consistency drift potential value decreases and remains no greater than the first water level threshold after the first two write actions are executed, then the subsequent write actions will continue to be executed automatically in the dynamic submission order. If the interface latency increases after a certain write action is executed, causing the twin consistency drift potential value to rise and fall between the two water levels, then the execution mode will be switched to segmented execution immediately to reduce the risk of continuous write superposition. If the twin consistency drift potential value exceeds the second water level threshold, then the automatic writing will be stopped immediately and an execution suggestion will be given to first manually verify the version identifier consistency and then execute in segments at an opportune time.

[0082] In the event of a write operation failure or a write confirmation timeout, the digital transformation platform performs the following exception handling: For the subset of compensable write operations already executed, the corresponding compensable write operations are rolled back in reverse order of dynamic submission to restore the state before the write operation; for the subset of retryable write operations already executed, status queries and retrying are performed until a write confirmation receipt is obtained. Based on the write confirmation receipt, rollback, or retry results, the digital twin model and the confirmed version identifier are updated. Simultaneously, the number of rollback triggers, the number of successful compensable write operations, the number of retries, and the receipt timeout are recorded in the platform's operational evaluation dataset to support subsequent risk assessment. For example, if a work order priority adjustment write action for a production execution management business system has been executed, but the subsequent maintenance resource lock write action for an equipment operation and maintenance management business system times out, the digital transformation platform can first trigger the corresponding compensation write action to roll back the executed work order priority adjustment and restore the original priority. Then, it can query the execution status of the maintenance resource lock write action and retry according to the preset retry strategy until a write confirmation receipt is obtained. Finally, based on the rollback and receipt results, the digital twin model and the confirmed version identifier are updated synchronously to ensure that the set of cross-system write actions remains controllable and traceable in abnormal situations.

[0083] In one specific implementation, the first water level threshold With the second water level threshold Based on a sample set of historical twin consistency drift potential values Distribution characteristic quantization settings:

[0084] ;

[0085] in Describes the quantile function, and For safety margin;

[0086] Examples are acceptable For higher quantiles, For higher quantiles, and make This creates a more conservative barrier boundary. To avoid execution jitter caused by frequent switching, the digital transformation platform can be updated on a rolling basis according to a preset update cycle. And re-evaluate quantiles, or introduce minimum duration / minimum hysteresis width near the threshold to stabilize gating switching.

[0087] In one specific implementation, when the digital transformation platform executes in a segmented manner, it sequences the pending write actions in the dynamic submission order according to the rollback risk potential value from low to high. And risk-constrained segmentation is adopted: the minimum prefix length is selected. , making ;in, Indicates a write operation The potential value of the risk of pullback. To preset the risk limit for segmentation; prefix actions The first paragraph is the first action, and the remaining actions are the subsequent paragraphs.

[0088] After each execution phase ends, a stable period begins. During the stable period, at least one "version identifier acquisition - potential energy value recalculation" is performed, and the recalculated twin-consistent drift potential energy value is used. As a condition for continuation: when It can be switched to automatically execute subsequent actions; when Continue executing the next segment in a segmented manner and repeat the consistency check; when Automatic writing and execution suggestions will stop when the time is right, and monitoring will continue until... Execution can then be resumed; and the dynamic submission order can be updated based on the latest receipt information to reflect changes in risk before proceeding to the next segment.

[0089] In one specific implementation, the preset retry strategy includes at least a maximum number of retries. Receipt waiting timeout retreat interval sequence Total timeout limit For retryable write operations, when in If no confirmation receipt is received, the digital transformation platform first calls the status query interface to obtain the execution status of the write action: if the status is effective or submitted and confirmed, the action is recorded as successful and enters the gating judgment of the next action.

[0090] If the status is ineffective or unconfirmable, then press After retreating, the first Repeat the trial until it is achieved. Or the cumulative waiting time reaches The retreat interval can be calculated using exponential retreat. ;in For the initial retreat interval, As a backoff factor, if the retry limit or total timeout limit is reached and the operation still fails, the digital transformation platform classifies the action as a failure and triggers an exception handling process. To avoid state drift caused by repeated writes, the digital transformation platform assigns an idempotent key to each write action. When the idempotent key is sent to the target business system, it is carried along. The target business system performs deduplication and idempotency verification based on the idempotent key. The status query interface returns at least four categories: "effective, ineffective, processing, and unknown" to support the above retry and stop-loss determination.

[0091] Reference Figure 2Example 2: A simulation and optimization system for a digital transformation platform based on digital twins, comprising:

[0092] The data acquisition module establishes a data interaction channel between the digital transformation platform and at least two target business systems. It acquires data related to platform operation from each target business system according to a preset acquisition strategy and aggregates it to form a platform operation evaluation dataset. This dataset contains at least the operational information, interface status information, and historical execution data required for subsequent twin consistency drift potential value calculations. This ensures the platform has a unified and traceable data foundation for assessing the operational status of multiple target business systems, thus providing input for subsequent target control scheme determination, dynamic submission sequence determination, and gating execution.

[0093] The digital twin update module updates the digital twin model based on production operation data in the platform's operation evaluation dataset, ensuring that the digital twin model's state is synchronized with the production operation state of the target business system. After the update, it outputs a snapshot of the current digital twin state with a unified time stamp. This snapshot provides a "consistent" benchmark of the digital twin state, enabling the subsequent digital transformation platform to determine the target control scheme to be executed based on the current snapshot, thus constraining state consistency with a unified time stamp and reducing the impact of cross-system state differences on decision-making and execution.

[0094] The scheme decomposition module, after obtaining a snapshot of the current twin state, determines the target control scheme to be executed and decomposes it into a set of cross-system write actions. This allows the control intent to be translated into write actions that can be implemented across various target business systems. The scheme decomposition module further divides the set of cross-system write actions into a subset of compensable write actions and a subset of retryable write actions, enabling differentiated handling of different types of write actions in subsequent execution phases. Simultaneously, the scheme decomposition module calculates the rollback risk potential value of each write action based on historical execution data in the platform's operational evaluation dataset and determines the dynamic submission order based on this rollback risk potential value. This prioritizes reducing overall rollback risk and improving the stability and controllability of cross-system execution in the execution orchestration of the cross-system write action set.

[0095] The gating execution module obtains the confirmed version identifiers of each target business system before the target control scheme is distributed across systems. Based on these confirmed version identifiers and the platform operation evaluation dataset, it calculates the twin consistency drift potential energy value. Using the comparison between the twin consistency drift potential energy value and the first and second water level thresholds as the gating basis, the module gating and distributing the cross-system write action set according to the dynamic submission order and then executing it. After executing any write action, the gating execution module obtains a write confirmation receipt and recalculates the twin consistency drift potential energy value to reflect cross-system consistency changes in real time. Based on the comparison between the recalculated twin consistency drift potential energy value and the first and second water level thresholds, it dynamically switches the gating distribution execution mode, allowing selection between continuing automatic execution, switching to segmented execution, or stopping automatic writing and outputting target control scheme execution suggestions. This achieves dynamic risk control and adaptive execution strategies during the cross-system distribution process.

[0096] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A simulation optimization method for a digital transformation platform based on digital twins, characterized in that, Includes the following steps: S1: The digital transformation platform interacts with at least two target business systems to collect platform operation evaluation datasets; based on the production operation data in the platform operation evaluation datasets, the digital twin model is updated to generate a snapshot of the current twin state with a unified time identifier; S2: The digital transformation platform determines the target control scheme to be executed based on the current twin state snapshot; The target control scheme is decomposed into a set of cross-system write actions, and the set of cross-system write actions is divided into a subset of compensable write actions and a subset of retryable write actions. S3: Before distributing the target control scheme across systems, the digital transformation platform obtains the confirmed version identifiers of each target business system and calculates the twin consistency drift potential value based on the confirmed version identifiers and the platform operation evaluation dataset. The rollback risk potential value of each write action is calculated based on the historical execution data in the platform operation evaluation dataset, and the dynamic commit order is determined based on the rollback risk potential value. S4: The digital transformation platform, based on the comparison results of the twin consistency drift potential energy value with the first water level threshold and the second water level threshold, gating and issuing a set of cross-system write actions in a dynamic submission order; after executing any write action, it obtains a write confirmation receipt, recalculates the twin consistency drift potential energy value, and switches the dynamic gating and issuing execution mode based on the comparison results of the twin consistency drift potential energy value with the first water level threshold and the second water level threshold.

2. The simulation optimization method for a digital transformation platform based on digital twins according to claim 1, characterized in that, The cross-system write action set includes write actions targeting each target business system; at least some of the write actions in the cross-system write action set are configured with corresponding compensation write actions, which are used to roll back the corresponding target business system to restore it to the state before the write action was executed.

3. The simulation optimization method for a digital transformation platform based on digital twins according to claim 1, characterized in that, The twin-consistent drift potential value is obtained as follows: based on the interface latency, interface timeout rate, and interface error rate determined by the platform operation evaluation dataset, the interface trust weights of each target business system are generated; the confirmed version identifiers of each target business system are mapped to version sequences and smoothed within a preset sliding time window; based on the interface trust weights, the multi-statistic dispersion and time evolution term of the version sequence are weighted and fused to obtain the drift composite value, and the drift composite value is normalized to obtain the twin-consistent drift potential value.

4. The simulation optimization method for a digital transformation platform based on digital twins according to claim 3, characterized in that, The dispersion of the multistatistics should include at least two of the range, quantile difference and variance, and the time evolution term should include at least the drift slope or drift acceleration within a preset sliding time window.

5. The simulation optimization method for a digital transformation platform based on digital twins according to claim 1, characterized in that, Based on the historical execution data in the platform operation evaluation dataset, the irreversibility index and compensation success rate of each write action are calculated, and the rollback risk potential value of each write action is calculated based on the irreversibility index and compensation success rate. The compensation success rate is the ratio of the number of successful compensation write actions within the preset sliding time window to the total number of compensation write actions triggered.

6. The simulation optimization method for a digital transformation platform based on digital twins according to claim 5, characterized in that, The irreversible index is calculated based on at least two of the following three factors: whether the corresponding compensation write action exists, the expected execution time of the corresponding compensation write action, and the scope of the impact of the corresponding compensation write action.

7. The simulation optimization method for a digital transformation platform based on digital twins according to claim 1, characterized in that, The dynamic submission sequence gating includes: automatically executing a set of cross-system write actions according to the dynamic submission sequence when the twin consistency drift potential energy value is not greater than the first water level threshold; executing in at least two segments according to the dynamic submission sequence when the twin consistency drift potential energy value is greater than the first water level threshold but not greater than the second water level threshold; stopping automatic writing and outputting a target control scheme execution suggestion when the twin consistency drift potential energy value is greater than the second water level threshold, and continuously monitoring until the twin consistency drift potential energy value is not greater than the second water level threshold.

8. The simulation optimization method for a digital transformation platform based on digital twins according to claim 1, characterized in that, In the event of a write action failure or a write confirmation receipt timeout, the digital transformation platform performs the following exception handling: for the subset of compensable write actions that have already been executed, the corresponding compensable write actions are rolled back in reverse order of dynamic submission; for the subset of retryable write actions that have already been executed, the platform performs status queries and retry until a write confirmation receipt is obtained; and updates the digital twin model and the confirmed version identifier based on the write confirmation receipt or the rollback and retry results.

9. The simulation optimization method for a digital transformation platform based on digital twins according to claim 1, characterized in that, The execution mode switching of dynamic gating includes continuing automatic execution, switching to segmented execution, or stopping automatic writing and outputting target control scheme execution suggestions.

10. A simulation and optimization system for a digital transformation platform based on digital twins, applied to the simulation and optimization method for a digital transformation platform based on digital twins as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to interact with at least two target business systems to acquire platform operation evaluation datasets. The twin update module is used to update the digital twin model based on the production operation data in the platform operation evaluation dataset and generate a snapshot of the current twin state with a unified time identifier. The scheme decomposition module determines the target control scheme, decomposes the cross-system write action set into a compensable write action subset and a retryable write action subset, and determines the dynamic submission order. The gating execution module calculates the twin-consistent drift potential energy value, dynamically gating it according to the first water level threshold and the second water level threshold, and then executes it. After writing the confirmation receipt, it recalculates and switches.

Images