Erp optimal simulation method and system for robot tax response

Abstract

An ERP optimal simulation method and system for responding to a robot tax are disclosed. An ERP simulation system according to one embodiment may include: an AI-based data integration analysis engine that verifies the quality of data transmitted from an Enterprise Resource Planning (ERP) system and an external system; a deep learning-based prediction model engine that generates prediction result data through AI-based analysis on the quality-verified data; and a scenario-based optimization engine that generates a scenario for responding to a robot tax in an ERP system based on the generated prediction result data.

Description

Technology Field

[0001] The following description concerns ERP simulation technology for responding to the robot tax. Background Technology

[0003] Existing Enterprise Resource Planning (ERP) systems have limited capabilities for the integrated analysis of robot automation equipment operation data and tax data. There is a lack of functionality to analyze changes in robot tax policies in real-time in conjunction with a company's automation status. Consequently, it is difficult for companies to predict and respond to the burden of the robot tax, leading to increased financial risk.

[0004] The current system lacks the capability to simulate the tax implications of robot adoption in advance. Its ability to analyze and compare tax impacts under various automation scenarios is limited. Consequently, corporate investment decisions regarding automation are not adequately reflecting tax considerations.

[0005] There is a lack of analytical tools to determine the optimal balance between robotic automation and workforce management. It is difficult to calculate the optimal timing and scale of automation investment while considering factors such as tax credits and exemptions. As a result, companies sometimes incur unnecessary taxes or miss out on investment opportunities. The problem to be solved

[0007] This invention relates to an AI-based simulation method and system for responding to the robot tax in Enterprise Resource Planning (ERP) systems, aiming to optimize the robot tax burden resulting from the operation of automated facilities, support efficient decision-making, and automate regulatory compliance. means of solving the problem

[0009] The ERP simulation system may include: an AI-based data integration analysis engine that verifies the quality of data transmitted from an ERP (Enterprise Resource Planning) system and an external system; a deep learning-based prediction model engine that generates prediction result data through AI-based analysis on the data whose quality has been verified; and a scenario-based optimization engine that generates scenarios for responding to robot taxes in the ERP system based on the generated prediction result data.

[0010] The above AI-based data integration analysis engine can collect automation equipment data and personnel operation data transmitted from the above ERP system, and collect robot tax policy data and market trend data transmitted from the above external system.

[0011] The above AI-based data integration analysis engine includes performing preprocessing on data transmitted from the above ERP (Enterprise Resource Planning) system and an external system, and the preprocessing includes normalizing numerical values ​​of accounting data and financial data transmitted from the ERP system, verifying the consistency of the accounting data and financial data with normalized numerical values, standardizing code data of the accounting data and financial data with verified consistency, normalizing numerical data, correcting sensor data of a robot control system transmitted from the external system, converting the data format of the corrected sensor data of the robot control system, removing duplicate data from the sensor data of the robot control system with the converted data format, and processing missing values.

[0012] The above AI-based data integration analysis engine can distribute and store the data for which the above preprocessing has been performed in a distributed storage, perform real-time synchronization of the data stored in the distributed storage, and then manage the change history of each data.

[0013] The above AI-based data integration analysis engine can monitor the data quality level of the data for which real-time synchronization has been performed in real time, and automatically correct quality issues discovered by evaluating the data quality level through the monitoring.

[0014] The deep learning-based prediction model engine described above includes performing time series predictions for robot tax-related variables from the quality-verified data through a time series analysis artificial intelligence model, and the time series analysis artificial intelligence model may include an LSTM.

[0015] The deep learning-based prediction model engine described above can obtain the relationship between each variable in the prediction result data generated through the time series prediction through an attention mechanism, and can visualize the relationship between each variable obtained.

[0016] The above scenario-based optimization engine can generate scenarios for responding to robot taxes in an ERP system based on the prediction result data generated through Monte Carlo simulation.

[0017] The above scenario-based optimization engine can identify the range and confidence interval of expected results by executing multiple simulations based on the probability distribution of variables including robot activity time, throughput, error rate, and cost, identify bottlenecks, risk factors, and optimization opportunities in robot activities through the results of the simulations, and support robot tax planning and decision-making by reflecting the identified bottlenecks, risk factors, and optimization opportunities in ERP cost management.

[0018] The above scenario-based optimization engine can calculate the total robot process cost based on real-time data including robot process start / end times, execution times per process, waiting times, and error occurrence and recovery times, and cost allocation criteria.

[0019] The above scenario-based optimization engine can set a region of interest for robot activities through predictive analysis using the simulation, and realize risk management and resource allocation for the set region of interest for robot activities. The above scenario-based optimization engine can explore a solution space for determining robot tax through a genetic algorithm by performing verification on the generated scenario.

[0020] The above scenario-based optimization engine derives a solution to a multi-objective function based on the explored solution space and can evaluate risk for each scenario using the derived solution.

[0021] An ERP simulation method performed by an ERP simulation system may include: a step of verifying the quality of data transmitted from an ERP (Enterprise Resource Planning) system and an external system; a step of generating prediction result data through artificial intelligence-based analysis of the quality-verified data; and a step of generating a scenario for responding to a robot tax in the ERP system based on the generated prediction result data. Effects of the invention

[0023] Improvement of AI-based real-time prediction accuracy: Time series prediction accuracy of over 95% can be achieved through the combination of LSTM and attention mechanisms. Predictive power regarding complex robot tax policy changes can be improved through non-linear pattern learning. Continuous improvement of prediction performance is possible through real-time model updates.

[0024] Optimized Decision Support Features: Multi-objective optimization enables decision-making that simultaneously considers cost, efficiency, and risk. Scenario-based analysis allows for the proactive establishment of response strategies for various situations. Real-time monitoring and alerts provide immediate support for decision-making.

[0025] Scalable System Architecture: The modular system structure facilitates the addition of new features and expansion. The distributed processing architecture ensures high-capacity data processing performance. Standardized interfaces enable integration with various external systems.

[0026] Automated Compliance and Risk Management: Automated compliance is possible through real-time monitoring of policy changes. Potential risks can be identified in advance through AI-based risk assessment. The efficiency of regulatory response operations can be improved through the generation of automated reports.

[0027] Cost-effective operational optimization: Optimal automation strategies can be derived to minimize the burden of robot taxes. Operating costs can be reduced by optimizing the balance between human resources and robots. System operational efficiency can be improved through predictive maintenance. Brief explanation of the drawing

[0029] FIG. 1 is a drawing for explaining a simulation system in one embodiment. FIG. 2 is a diagram illustrating an AI-based data integration analysis engine in one embodiment. FIG. 3 is a diagram illustrating a deep learning-based prediction model engine in one embodiment. FIG. 4 is a diagram illustrating a scenario-based optimization engine in one embodiment. FIG. 5 is a flowchart illustrating the entire process of a core module in one embodiment. FIG. 6 is a flowchart illustrating the core module and peripheral module processes in one embodiment. FIG. 7 is a flowchart illustrating the detailed process of an optimization engine in one embodiment. FIGS. 8 to 10 are drawings for explaining the system architecture and cloud architecture in one embodiment. FIG. 11 is a diagram illustrating AI robot activity tracking and simulation operations in one embodiment. FIG. 12 is an example of applying Monte Carlo simulation to improve the effect of robot activity in one embodiment. Specific details for implementing the invention

[0030] Hereinafter, embodiments will be described in detail with reference to the attached drawings.

[0032] FIG. 1 is a drawing for explaining a simulation system in one embodiment.

[0033] The simulation system may include an AI-based data integration analysis engine, a deep learning-based prediction model engine, and a scenario-based optimization engine.

[0034] The AI-based data integration analysis engine may include a data preprocessing module, a data integration management module, and a data quality management module.

[0035] The data preprocessing module can be responsible for cleaning, normalizing, and standardizing all collected data. It can improve data quality by utilizing LSTM-based predictive models to handle missing values. It can ensure data reliability by applying autoencoder technology for outlier detection. Furthermore, the data preprocessing module can enhance the training efficiency of analysis models through data normalization.

[0036] The data integration management module can serve as a central hub for integrating and linking data from various sources. The data integration management module can operate a distributed database system for real-time data synchronization. The data integration management module can guarantee point-in-time data traceability for data version management.

[0037] The data quality management module can monitor data quality in real time through AI-based data verification algorithms. It can define data quality metrics and perform continuous quality improvement. The data quality management module can maintain data accuracy through automated data correction functions.

[0038] A deep learning-based prediction model engine may include a time series analysis module, an attention mechanism module, and a model training management module.

[0039] The time series analysis module can perform time series forecasting of variables related to robot tax by utilizing an LSTM network. The time series analysis module can predict future trends by learning patterns in multivariate time series data. The time series analysis module can quantify forecast uncertainty and utilize it for risk assessment.

[0040] The attention mechanism module can improve prediction accuracy by automatically learning the importance of variables. It can capture complex relationships through a multi-head attention structure. The attention mechanism module can provide explanatory power for prediction results through the visualization of attention weights.

[0041] The model training management module can perform real-time model updates through online training. It can monitor model performance metrics and automatically perform optimization. The model training management module can provide stable services through model version management.

[0042] A scenario-based optimization engine may include a genetic algorithm module, a multi-objective optimization module, and a scenario generation module.

[0043] The genetic algorithm module can search for optimal solutions through adaptive genetic algorithms. The genetic algorithm module can improve search efficiency through dynamic mutation rate adjustment. The genetic algorithm module can optimize the search of large-scale solution spaces through parallel processing.

[0044] The multi-objective optimization module can derive the optimal solution for a multi-objective function through Pareto optimization. The multi-objective optimization module can perform real-time optimization by dynamically incorporating constraints. The multi-objective optimization module can provide sensitivity analysis of optimization results.

[0045] The scenario generation module can generate various scenarios through Monte Carlo simulation. The scenario generation module can provide reliable results through scenario validation. The scenario generation module can evaluate uncertainty for each scenario through risk analysis.

[0046] In addition, peripheral technology area modules may include a data collection area and an interface area.

[0047] The data collection area may include an ERP data collector and an external data collector.

[0048] The ERP data collector can collect internal corporate data through real-time integration with the ERP system. The ERP data collector can perform automated collection through a data collection scheduler. The ERP data collector can verify the integrity of the collected data.

[0049] The external data collector can automatically collect external data, such as policy data and market trends. It can secure data from various sources through web crawling and API integration. The external data collector can verify the reliability of the collected data.

[0050] The interface area may include a user interface and a reporting module.

[0051] The user interface can visualize analysis results through an intuitive dashboard. The user interface can ensure security through user permission management. The user interface can notify important events through real-time notification features.

[0052] The reporting module can provide customized report templates. The reporting module can provide automated report generation capabilities. The reporting module can support output in various formats.

[0053] According to one embodiment, the robot tax burden can be optimized. Tax costs can be reduced by deriving the optimal balance point between a company's level of automation and the robot tax burden. Through an AI-based predictive model, measures to proactively respond to changes in robot tax policies can be presented. The financial burden on a company can be minimized by optimizing the timing and scale of automation investment.

[0054] According to one embodiment, decision support can be enhanced. The accuracy and timeliness of automation-related decisions can be improved through real-time data analysis. Support can be provided for establishing optimal automation strategies through various scenario simulations. Preemptive response measures for future situations can be presented through an AI-based prediction model.

[0055] According to one embodiment, operational efficiency can be maximized. Productivity can be improved by deriving the optimal balance point between human-robot operation. The operational efficiency of automated equipment can be improved through real-time monitoring. The equipment utilization rate can be maximized through predictive maintenance.

[0056] According to one embodiment, regulatory compliance can be automated. Changes in laws and policies related to the robot tax can be monitored in real time to analyze their impact. Potential regulatory violations can be identified in advance through AI-based risk assessment. Responses to regulatory agencies can be streamlined through the generation of automated reports.

[0057] According to one embodiment, system scalability can be ensured. Through a modular system structure, the addition and expansion of new functions can be facilitated. Interoperability with various external systems can be enabled through standardized interfaces. System performance and stability can be ensured through a distributed processing architecture.

[0058] FIG. 2 is a diagram illustrating an AI-based data integration analysis engine in one embodiment.

[0059] The AI-based data integration analysis engine may include a data preprocessing module, a data integration management module, and a data quality management module.

[0060] The data preprocessing module can improve the quality of source data and convert it into an analyzable form through data cleaning. The data preprocessing module can perform missing value correction using an LSTM-based prediction model (missing value processing). The data preprocessing module can perform autoencoder-based outlier detection and processing (outlier removal). The data preprocessing module can perform data uniqueness verification and duplicate removal (duplicate data removal). The data preprocessing module can generate a cleaned, high-quality dataset.

[0061] The data preprocessing module can unify the scales of data with different characteristics through data normalization (Norm). The data preprocessing module can apply Min-Max scaling or Z-score normalization (Numerical Normalization). The data preprocessing module can perform one-hot encoding or label encoding (Categorical Encoding). The data preprocessing module can perform time-unit standardization and sequence alignment (Time Series Normalization). The data preprocessing module can generate data in a format optimized for analysis models.

[0062] For example, the data preprocessing module can perform preprocessing on ERP system data and external system data. The data preprocessing module can perform numerical normalization of accounting and financial data. In this case, accounting and financial data can be a mandatory preprocessing item. The data preprocessing module can process missing values ​​in production and inventory data, remove outliers in personnel and payroll data, and sort time series of purchasing and procurement data. The data preprocessing module can verify data consistency by checking for missing essential fields, verifying referential integrity, and validating based on business rules. The data preprocessing module can standardize code data by organizing master data such as department codes and customer codes, managing code mapping tables between systems, and managing the history of code system changes. The data preprocessing module can normalize numeric data by unifying units (KRW, USD, etc.), standardizing decimal places, and applying rules for handling negative and positive numbers. In addition, the data preprocessing module can correct sensor data from robot control systems. The data preprocessing module can refine Manufacturing Execution System (MES) log data, standardize measurement data from quality control systems, and synchronize time-series data from IoT devices. In this context, sensor data from robot control systems can be an essential preprocessing item. The data preprocessing module can convert data formats through conversion to formats such as CSV, XML, and JSON, unification of character sets (encoding), and standardization of date and time formats. The module can eliminate duplicate data by removing duplicates based on key values, selecting the most recent data based on time, and checking for logical duplicates. The module can handle missing values ​​by applying default values ​​based on business rules, generating replacement values ​​using statistical methods, and excluding incomplete data.

[0063] The data preprocessing module can verify the quality and validity of preprocessed data through data validation. The data preprocessing module can validate the validity of data structure and format (schema validation). The data preprocessing module can verify compliance with business rules and constraints (domain validation). The data preprocessing module can validate the validity of relationships between data (referential integrity validation). The data preprocessing module can provide verified, highly reliable datasets.

[0064] The data integration management module requires a distributed storage for preprocessed data capable of handling large volumes of data, ensuring high availability, and supporting distributed processing. The module necessitates the efficient storage and management of preprocessed data, requires overcoming the physical limitations of a single repository, demands scalability to accommodate data growth, and must meet real-time data processing requirements. Furthermore, the module requires maintaining service continuity even in the event of system failure, minimizing the risk of data loss, guaranteeing 24 / 7 service availability, and making disaster recovery preparedness essential. Additionally, the module must support parallel training of artificial intelligence models, enable real-time data analysis, streamline the processing of complex queries, and handle large-scale batch jobs.

[0065] The Data Integration Management Module can perform efficient storage and management of large volumes of data through distributed storage. The Data Integration Management Module can implement sharding-based distributed storage (Distributed Data Storage). The Data Integration Management Module can perform storage applying high-efficiency compression algorithms (Data Compression). The Data Integration Management Module can operate an automated backup and recovery system (Backup and Recovery). The Data Integration Management Module can provide a stable and scalable data storage environment. The Data Integration Management Module can provide the latest data synchronized in real time.

[0066] Accordingly, data processing performance can be enhanced by improving processing speed through parallel processing, ensuring stability through load balancing, and increasing access speed through locality optimization. In this context, locality refers to the physical proximity between the location where data is stored and the computing resources that process it. Furthermore, system stability can be strengthened by guaranteeing availability through data replication, securing service continuity through fault isolation, and improving reliability through automatic recovery functions. Additionally, artificial intelligence analysis performance can be optimized by improving training data access speeds, supporting model training parallelization, and enhancing real-time analysis processing capabilities.

[0067] The data integration management module can perform real-time synchronization of multi-source data. The data integration management module can perform change tracking based on Change Data Capture (CDC) (change data detection). The data integration management module can perform real-time data replication and synchronization (data synchronization). The data integration management module can apply concurrency control and conflict resolution mechanisms (conflict resolution). The data integration management module can provide the latest synchronized data in real time.

[0068] The data integration management module can ensure data history management and traceability through version control. The data integration management module can automatically record data change history (version tracking). The data integration management module can support data restoration to a specific point in time (rollback support). The data integration management module can manage audit records for data changes (audit trail). The data integration management module can provide an environment where the entire data lifecycle can be tracked.

[0069] The data quality management module can monitor data quality levels in real time. The data quality management system can measure indicators such as accuracy, completeness, and consistency (quality indicator measurement). The data quality management system can set and manage thresholds for each quality indicator (threshold management). The data quality management system can generate automatic alerts when quality issues occur (alert generation). The data quality management system can provide real-time visibility into the status of data quality.

[0070] A data quality management system can quantitatively evaluate data quality levels through quality metrics. The data quality management system can perform quantitative measurements of various quality indicators (metric calculation). The data quality management system can analyze time-series changes in quality levels (trend analysis). The data quality management system can perform automatic reporting of quality analysis results (report generation). The data quality management system can provide objective evaluation results regarding data quality.

[0071] The data quality management system can automatically correct discovered quality issues. The data quality management system can perform automatic correction based on predefined rules (rule-based correction). The data quality management system can perform intelligent correction using machine learning models (AI-based correction). The data quality management system can track and manage the history of data correction (correction history management). The data quality management system can ensure the continuous maintenance of high-quality data.

[0072] FIG. 3 is a diagram illustrating a deep learning-based prediction model engine in one embodiment.

[0073] A deep learning-based prediction model engine may include a time series analysis module, an attention mechanism module, and a model training management module.

[0074] The TimeSeries analysis module can learn the long-term dependencies of time-series data and perform pattern recognition through time-series analysis AI models, such as LSTM networks. The TimeSeries analysis module can determine the learning importance of new information (Input Gate). The TimeSeries analysis module can determine whether to retain or remove existing information (Forget Gate). The TimeSeries analysis module can control the generation of final predicted values ​​(Output Gate). The TimeSeries analysis module can perform normalization of time-series data. The TimeSeries analysis module can perform sequence length optimization. The TimeSeries analysis module can perform multivariate time-series processing.

[0075] The time series analysis module can predict future values ​​based on output values, such as LSTM networks, through a prediction engine. The time series analysis module can perform predictions in units of 1 to 3 months through the short-term prediction module. The time series analysis module can perform predictions in units of 3 to 6 months through the medium-term prediction module. The time series analysis module can perform predictions of 6 months or longer through the long-term prediction module. The time series analysis module can set prediction intervals. The time series analysis module can calculate confidence intervals. The time series analysis module can perform correction of predicted values.

[0076] The time series analysis module can assess uncertainty by evaluating the reliability of forecast results and analyzing risks. The time series analysis module can calculate the reliability of forecast values ​​through the reliability evaluation module. The time series analysis module can identify potential risks through the risk analysis module. The time series analysis module can analyze the impact of each variable through the sensitivity analysis module.

[0077] The attention mechanism module can learn the relationships between various features in parallel through multi-head attention. The attention mechanism module can be responsible for extracting key features (Query Handler). The attention mechanism module can perform mapping related information (Key-Value Handler). The attention mechanism module can calculate the importance between features (Attention Score Calculator). The attention mechanism module can perform feature vector transformation. The attention mechanism module can perform parallel attention computation. The attention mechanism module can perform result integration.

[0078] The attention mechanism module can quantify the importance of each feature through weight calculation and reflect it in the prediction. The attention mechanism module can perform weight normalization through the normalization module. The attention mechanism module can perform weight adjustment through the scaling module. The attention mechanism module can perform weight updates through the update module. The attention mechanism module can perform initial weight setting. The attention mechanism module can perform dynamic weight adjustment. The attention mechanism module can derive optimal weights.

[0079] The attention mechanism module can visually represent the analysis results of the attention mechanism. The attention mechanism module can visualize weight distributions through a heatmap generator. The attention mechanism module can visualize relationships between features through a relationship map generator. The attention mechanism module can visualize changes over time through a time series plotter.

[0080] The Model Learning Management Module can improve model performance using real-time data through online learning. The Model Learning Management Module can handle real-time data processing through stream handlers. The Model Learning Management Module can perform incremental model updates through incremental learners. The Model Learning Management Module can measure learning effectiveness through a performance monitor. The Model Learning Management Module can perform real-time data collection. The Model Learning Management Module can perform incremental learning. The Model Learning Management Module can perform model updates.

[0081] The Model Training Management Module can optimize performance by continuously improving the model's predictive performance. The Model Training Management Module can perform model parameter optimization through the Hyperparameter Tuner. The Model Training Management Module can perform model performance verification through the Validation Module. The Model Training Management Module can reflect performance improvements through the Feedback Handler. The Model Training Management Module can measure performance metrics. The Model Training Management Module can apply optimization algorithms. The Model Training Management Module can provide result feedback.

[0082] The Model Training Management Module can systematically manage and track model versions. The Model Training Management Module can be responsible for model version management through a version controller. The Model Training Management Module can be responsible for recovering previous versions through a rollback manager. The Model Training Management Module can be responsible for recording performance by version through a performance logger. The Model Training Management Module can perform version history management. The Model Training Management Module can perform performance comparative analysis. The Model Training Management Module can select the optimal version.

[0083] FIG. 4 is a diagram illustrating a scenario-based optimization engine in one embodiment.

[0084] The scenario-based optimization engine may include a Genetic Algorithm Module, a multi-objective optimization module, and a scenario generation module.

[0085] The genetic algorithm module can perform solution space exploration for robot tax optimization through an Adaptive Genetic Algorithm. The genetic algorithm module can explore the optimal balance point between the level of automation and the tax burden. The genetic algorithm module can dynamically adjust the fitness function during the generational evolution process. The genetic algorithm module can implement diversity maintenance mechanisms to avoid local optima. The genetic algorithm module can generate various automation scenarios (generation of initial solution population). The genetic algorithm module can comprehensively evaluate cost, efficiency, and risk (fitness evaluation). The genetic algorithm module can preserve and combine the characteristics of superior solutions (selection and crossover). The genetic algorithm module can generate mutations to explore new solutions (mutation).

[0086] The genetic algorithm module can optimize the efficiency of solution space search through mutation adjustment. The genetic algorithm module can dynamically adjust the probability of mutation occurrence per generation. The genetic algorithm module can automatically adjust the search range based on the convergence of the solution. The genetic algorithm module can implement mutation strategies for escaping local optima. The genetic algorithm module can evaluate the diversity of the current solution population (convergence analysis). The genetic algorithm module can dynamically determine the probability of mutation occurrence (probability adjustment). The genetic algorithm module can induce mutations for selected solutions (mutation application).

[0087] The genetic algorithm module can perform efficient exploration of large-scale solution spaces through parallel processing. The genetic algorithm module can implement parallel evolution of multiple solution populations. The genetic algorithm module can optimize migration strategies between sub-solution populations. The genetic algorithm module can manage the efficient utilization of computing resources. The genetic algorithm module can divide the entire solution population into subgroups (solution population partitioning). The genetic algorithm module can perform independent evolution of each subgroup (parallel evolution). The genetic algorithm module can perform the exchange of superior solutions between subgroups (solution exchange).

[0088] The Multi-Objective Optimization Module can derive the optimal solution set of a multi-objective function through Pareto optimization. The Multi-Objective Optimization Module can analyze trade-offs between cost, efficiency, and risk. The Multi-Objective Optimization Module can generate and manage Pareto optimal solution sets. The Multi-Objective Optimization Module can perform optimal solution ranking for decision-makers. The Multi-Objective Optimization Module can perform multi-objective evaluation for each solution. The Multi-Objective Optimization Module can identify dominance relationships between solutions through Pareto dominance analysis. The Multi-Objective Optimization Module can derive a set of non-dominating solutions through the generation of Pareto fronts.

[0089] The multi-objective optimization module can perform optimization considering various constraints through constraint handling. The multi-objective optimization module can convert legal regulatory requirements into constraints. The multi-objective optimization module can reflect operational constraints in real time. The multi-objective optimization module can quantify and evaluate the degree of constraint violation. The multi-objective optimization module can verify whether a solution satisfies constraints through constraint validation. The multi-objective optimization module can calculate penalties for constraint violations. The multi-objective optimization module can modify the solution to satisfy constraints (solution correction).

[0090] The multi-objective optimization module can evaluate the stability and reliability of the optimal solution through sensitivity analysis. The multi-objective optimization module can analyze the impact of variations in input variables. The multi-objective optimization module can evaluate the robustness of the optimal solution. The multi-objective optimization module can calculate the importance of each risk factor. The multi-objective optimization module can generate variations in key input variables. The multi-objective optimization module can analyze changes in the optimal solution due to variations (impact analysis). The multi-objective optimization module can quantify the risk of variation through risk assessment.

[0091] The Scenario Generation Module can generate scenarios considering various uncertainties through Monte Carlo simulation. The Scenario Generation Module can model the probability distribution of key variables. The Scenario Generation Module can generate scenarios through random sampling. The Scenario Generation Module can calculate the probability of occurrence for each scenario. The Scenario Generation Module can estimate the probability distribution of key variables. The Scenario Generation Module can generate scenarios based on the probability distribution (sampling). The Scenario Generation Module can calculate the probability of occurrence for each scenario.

[0092] The scenario generation module can verify the realism and validity of generated scenarios through scenario validation. The scenario generation module can verify the logical consistency of scenarios. The scenario generation module can check whether realistic constraints are satisfied. The scenario generation module can filter out extreme scenarios. The scenario generation module can check for logical errors in scenarios through consistency checks. The scenario generation module can verify the satisfaction of realistic constraints through constraint validation. The scenario generation module can remove unrealistic scenarios through filtering.

[0093] The scenario generation module can quantitatively evaluate risks for each scenario through risk assessment. The scenario generation module can quantify financial risk. The scenario generation module can evaluate operational risk. The scenario generation module can analyze regulatory risk. The scenario generation module can identify key risk factors for each scenario (risk identification). The scenario generation module can quantify the impact of each risk factor (impact assessment). The scenario generation module can calculate the overall risk level (comprehensive assessment).

[0094] FIG. 5 is a flowchart illustrating the entire process of a core module in one embodiment.

[0095] In the embodiments, the operation of data transmission and reception between the data integration engine, deep learning model, optimization engine, and user interface will be described.

[0096] In the data collection and preprocessing stage, the data integration engine can collect data and perform cleaning on the collected data. After performing quality verification on the cleaned data, the data integration engine can perform data normalization.

[0097] In the prediction model execution phase, the data integration engine can transmit the preprocessed data to the deep learning model engine. The deep learning model engine can perform LSTM-based time series analysis using the preprocessed data and generate prediction results by applying an attention mechanism.

[0098] In the optimization analysis phase, the deep learning model engine can pass the generated prediction results to the optimization engine. The optimization engine can generate scenarios based on the prediction results. The optimization engine can perform optimization using genetic algorithms. The optimization engine can perform multi-objective optimization.

[0099] In the result generation phase, the optimization engine can transmit the optimized results to the user interface. The user interface can visualize the results and generate reports.

[0100] In the feedback and improvement phase, the user interface can provide feedback on model performance to the deep learning model engine. The deep learning model engine can retrain the model based on the feedback and provide feedback on data quality through the retrained model. The data integration engine can perform quality improvement.

[0101] FIG. 6 is a flowchart illustrating the core module and peripheral module processes in one embodiment.

[0102] In the embodiments, data transmission and reception operations between the ERP system, external system, core analysis system, data repository, user system, and report system will be described.

[0103] 1. Data Collection and Synchronization Process

[0104] The ERP system can transmit automation equipment data and workforce operation data to the core analysis system. External systems can transmit robot tax policy data and market trend data to the core analysis system. The core analysis system can integrate the received data and store it in a data repository.

[0105] 2. Real-time analysis process

[0106] The core analysis system can perform preprocessing on the received data, execute AI-based predictive analysis, and optimize scenarios. The core analysis system can store the analysis results in a data repository.

[0107] 3. Monitoring Notification Process

[0108] The core analysis system can transmit real-time analysis results to the user system. The user system can update the dashboard. The core analysis system can send risk alerts to the user system. The user system can request the generation of reports from the report system.

[0109] 4. Feedback and Adjustment Process

[0110] The user system can transmit user feedback to the core analytics system. The core analytics system can evaluate model performance based on user feedback and adjust parameters based on the evaluated model performance. The core analytics system can transmit optimized recommendations to the ERP system.

[0111] 5. Reporting Process

[0112] The reporting system can request analysis data from the data repository. The data repository can transmit data corresponding to the request to the reporting system. The reporting system can generate a report using the received data and transmit the generated report to the user system.

[0113] FIG. 7 is a flowchart illustrating the detailed process of an optimization engine in one embodiment.

[0114] In the scenario generation phase, the optimization engine can model the probability distribution through Monte Carlo simulation and verify the generated scenarios by performing random sampling.

[0115] In the scenario verification phase, the optimization engine can verify the logical consistency of the generated scenarios and check whether they satisfy the constraints.

[0116] In the genetic algorithm optimization phase, the optimization engine can perform genetic algorithm optimization on valid scenarios. After generating an initial solution population, the optimization engine can derive a set of candidate solutions by performing fitness evaluation, selection and crossover, and mutation application.

[0117] In the multi-objective optimization step, the optimization engine can derive an optimal solution set by performing Pareto optimization using the derived set of candidate solutions, processing constraints, and analyzing sensitivity.

[0118] In the risk assessment phase, the optimization engine can assess financial risk, assess operational risk, and transmit the final results of the regulatory risk assessment to the user interface.

[0119] In the result display stage, the user interface can visualize the results using the final results and generate a report.

[0120] FIGS. 8 to 10 are drawings for explaining the system architecture and cloud architecture in one embodiment.

[0121] Referring to Fig. 8, the system architecture for optimizing robot tax in manufacturing companies is shown and can be composed of a data source layer, a data processing layer, and a service layer.

[0122] Referring to Fig. 9, a hybrid cloud-based robot tax optimization system architecture is shown, and data can be transmitted and received through a hybrid connection between the cloud and an on-premise IDC.

[0123] Referring to Figure 10, a cloud-based robot tax optimization system architecture is shown, which can be composed of a client layer, a security layer, a microservices layer, and a data layer.

[0124] FIG. 11 is a diagram illustrating AI robot activity tracking and simulation operations in one embodiment.

[0125] The ERP simulation system can define Activity Categories. In this case, Activity Categories may include Machine Hours and Robot Process Hours. For Machine Hours, the Activity Category is MACH_HR, the unit is Hours, the tracking method is based on machine operating time, and cost elements may include electricity costs, depreciation costs, and maintenance costs. For Robot Process Hours, the Activity Category is ROB_HR, the tracking method is based on robot process execution time, and cost elements may include robot license costs, AI model operation costs, power consumption, and maintenance costs.

[0126] The ERP simulation system can define master data. The ERP simulation system can create activity types including AI_ROBOT_PROCESS, AI_ROBOT_IDLE, and AI_ROBOT_MAINTENANCE. The ERP simulation system can set cost element groups including AI Robot License Cost, AI Model Operation Cost, Robot Power Consumption, and Robot Maintenance Cost. The ERP simulation system can set statistical key figures (SKF) including Robot Processing Time, Robot Idle Time, Robot Error Rate, and Process Success Rate.

[0127] The ERP simulation system can generate an AI robot activity tracking mechanism. The ERP simulation system can collect real-time data including robot process start / end times, execution times per process, waiting times, and error occurrence and recovery times. The ERP simulation system can establish cost allocation criteria that include direct and indirect activity allocations. In this case, direct activity allocation includes actual robot operating time, while indirect activity allocation may include standard process time.

[0128] The ERP simulation system can calculate the total robot process cost through the activity and cost calculation module.

[0129] Total Robot Process Cost = (Robot License Cost / Total Available Time) * Actual Usage Time + (AI Model Operation Cost * Utilization Rate) + Power Consumption Cost + Maintenance Cost

[0130] The ERP simulation system can derive Key Performance Indicators (KPRs) through monitoring and analysis. The ERP simulation system can derive efficiency indicators, including robot utilization rate (%), process success rate, and cost efficiency. The ERP simulation system can derive quality indicators, including error rate, rework rate, and accuracy.

[0131] FIG. 12 is an example of applying Monte Carlo simulation to improve the effect of robot activity in one embodiment.

[0132] The ERP simulation system can generate various activity scenarios for AI robots through Monte Carlo simulation and derive optimal operational strategies by analyzing the results for each scenario. By running thousands of simulations based on the probability distributions of key variables such as robot activity time, throughput, error rate, and cost, the system can identify the range and confidence intervals of expected results. Through the simulation results, the system identifies bottlenecks, risk factors, and optimization opportunities in robot activities, and incorporates these into ERP cost management to support more accurate planning and decision-making. The system continuously compares and analyzes actual operational data with simulation results to improve model accuracy and adjust simulation parameters to align with the changing business environment. Through simulation-based predictive analysis, the system can optimize the ROI of robot activities and realize proactive risk management and resource allocation optimization. For specific processes where quality standards have declined, the system can improve based on information regarding actual quality issues by applying the most effective activity scenario from pre-simulated predictive scenarios to the AI ​​robot's next activity and measuring the results. Here, the simulation input variables include random variables and constraints. Random variables may include process execution time, probability of error, resource utilization, throughput variability, and latency. Constraints may include availability time, resource constraints, cost limits, quality criteria, and throughput requirements.

[0133] According to the embodiments, it can be applied to the automotive manufacturing industry. A tax optimization system for robotic automation equipment on an automotive assembly line can be established. The tax optimization system can monitor and analyze the level of automation of welding, painting, and assembly processes in real time. Through this, the annual tax burden associated with the operation of robotic automation equipment can be reduced, productivity can be improved through production line optimization, and regulatory compliance risks can be reduced through real-time monitoring.

[0134] According to the embodiments, it can be applied to the electronics manufacturing industry. It can optimize the operation of automated equipment in semiconductor and display production lines. It can analyze and improve the utilization of process robots requiring precision control. It can establish a robot-human operation strategy in a cleanroom environment. Through this, the automation efficiency of high-precision processes can be improved, annual costs related to robot taxes can be reduced, and production yield can be improved.

[0135] According to the embodiments, the operation of automated facilities in a logistics center can be optimized. Operational strategies for automated guided vehicles (AGVs) and robotic picking systems can be established. Measures to adjust the level of automation in response to seasonal changes in cargo volume can be presented. Through this, logistics processing efficiency can be improved, robot operating costs can be reduced, and the ability to respond to peak seasons can be enhanced.

[0136] The device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include a plurality of processing elements and / or a plurality of types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0137] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or instruct the processing unit independently or collectively. Software and / or data may be embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0138] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

[0139] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results can be achieved even if the described techniques are performed in a different order than described, and / or the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0140] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

Claim 1 An ERP simulation system comprising: an AI-based data integration analysis engine that verifies the quality of data transmitted from an Enterprise Resource Planning (ERP) system and an external system; a deep learning-based prediction model engine that generates prediction result data through artificial intelligence-based analysis of the quality-verified data; and a scenario-based optimization engine that generates scenarios for responding to robot taxes in the ERP system based on the generated prediction result data. Claim 2 An ERP simulation system according to claim 1, wherein the AI-based data integration analysis engine collects automation equipment data and personnel operation data transmitted from the ERP system, and collects robot tax policy data and market trend data transmitted from the external system. Claim 3 An ERP simulation system according to claim 1, wherein the AI-based data integration analysis engine comprises performing preprocessing on data transmitted from the Enterprise Resource Planning (ERP) system and an external system, and the preprocessing comprises normalizing numerical values ​​of accounting data and financial data transmitted from the ERP system, verifying the consistency of the accounting data and financial data in which the numerical values ​​are normalized, standardizing code data of the accounting data and financial data in which the consistency is verified, normalizing numerical data, correcting sensor data of a robot control system transmitted from the external system, converting the data format of the corrected sensor data of the robot control system, removing duplicate data from the sensor data of the robot control system in which the data format has been converted, and processing missing values. Claim 4 An ERP simulation system according to claim 1, wherein the AI-based data integration analysis engine distributes and stores the data on which the preprocessing has been performed in a distributed storage, performs real-time synchronization of the data stored in the distributed storage, and manages the change history of each data. Claim 5 An ERP simulation system according to claim 4, wherein the AI-based data integration analysis engine monitors the data quality level of the data for which real-time synchronization has been performed in real time, and automatically corrects quality issues discovered by evaluating the data quality level through the monitoring. Claim 6 An ERP simulation system according to claim 1, wherein the deep learning-based prediction model engine includes performing time series prediction for robot tax-related variables from the quality-verified data through a time series analysis artificial intelligence model, and wherein the time series analysis artificial intelligence model includes an LSTM. Claim 7 An ERP simulation system according to claim 6, wherein the deep learning-based prediction model engine acquires the relationship between each variable in the prediction result data generated through the time series prediction through an attention mechanism, and visualizes the acquired relationship between each variable. Claim 8 An ERP simulation system according to claim 1, wherein the scenario-based optimization engine generates a scenario for responding to a robot tax in an ERP system based on the prediction result data generated through Monte Carlo simulation. Claim 9 An ERP simulation system characterized in that, in claim 8, the scenario-based optimization engine identifies the range and confidence interval of expected results by executing multiple simulations based on the probability distribution of variables including robot activity time, throughput, error rate, and cost, identifies bottlenecks, risk factors, and optimization opportunities of robot activities through the results of executing the simulations, and supports robot tax planning and decision-making by reflecting the identified bottlenecks, risk factors, and optimization opportunities of robot activities in ERP cost management. Claim 10 In claim 9, the scenario-based optimization engine is characterized by calculating the total robot process cost based on real-time data including robot process start / end times, execution times per process, waiting times, error occurrence and recovery times, and cost allocation criteria. Claim 11 In claim 9, the scenario-based optimization engine is characterized by setting a region of interest for robot activities through predictive analysis using the simulation, and realizing risk management and resource allocation for the set region of interest for robot activities. Claim 12 An ERP simulation system according to claim 8, wherein the scenario-based optimization engine searches for a solution space to determine robot tax through a genetic algorithm while performing verification on the generated scenario. Claim 13 An ERP simulation system according to claim 12, wherein the scenario-based optimization engine derives a solution of a multi-objective function based on the explored solution space and evaluates risk for each scenario using the derived solution. Claim 14 An ERP simulation method performed by an ERP simulation system, comprising: a step of verifying the quality of data transmitted from an Enterprise Resource Planning (ERP) system and an external system; a step of generating prediction result data through artificial intelligence-based analysis of the quality-verified data; and a step of generating a scenario for responding to robot tax in the ERP system based on the generated prediction result data.

Images