Production simulation system, production simulation method and program

Abstract

To realize efficient production simulation. [Solution] The production simulation system includes a process information acquisition unit that acquires process information indicating the production process, a process division unit that divides the production process into multiple sub-processes based on information about inventory management locations included in the process information, a simulation model creation unit that creates a model and scenario for each sub-process, and a simulation execution unit that simulates each sub-process based on the model and scenario.

Description

[Technical Field]

[0001] The present invention relates to a production simulation system, a production simulation method, and a program. [Background technology]

[0002] A production line capacity evaluation technology has been proposed that virtually models a production line, predicts the behavior of workpieces flowing through the line, sets a target number of buffers in advance, runs a simulation, and evaluates the degree to which production volume has been achieved (see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2013 / 038754 Summary of the Invention [Problem to be solved by the invention]

[0004] Conventional manufacturing line simulations involve creating a manufacturing line buffer through trial and error, and then running a manufacturing line simulation each time to evaluate production volume, which results in a long design evaluation time.

[0005] The present invention has been made in view of the above background, and has an object to provide a technique that can realize an efficient production simulation. [Means for solving the problem]

[0006] The main invention of the present invention for solving the above-mentioned problems is a production simulation system comprising a process information acquisition unit that acquires process information indicating a production process, a process division unit that divides the production process into a plurality of sub-processes based on information on inventory management locations included in the process information, a simulation model creation unit that creates a model and a scenario for each of the sub-processes, and a simulation execution unit that simulates each sub-process based on the model and scenario.

[0007] Other problems and solutions disclosed in this application will be made clear in the section on preferred embodiments of the invention and the drawings. [Effects of the Invention]

[0008] According to the present invention, an efficient production simulation can be realized. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of the overall configuration of a production simulation system. [Figure 2] FIG. 2 is a diagram illustrating an example of the hardware configuration of a production simulation server 2. [Figure 3] FIG. 2 is a diagram illustrating an example of the software configuration of a production simulation server 2. [Figure 4] FIG. 10 is a diagram showing a processing flow of a probabilistic model creation process. [Figure 5] FIG. 10 is a diagram showing the processing flow of probabilistic case generation processing. [Figure 6] FIG. 2 is a diagram illustrating a processing flow in the production simulation system. DETAILED DESCRIPTION OF THE INVENTION

[0010] <System Overview> A production simulation system according to one embodiment of the present invention will be described below. The production simulation system of this embodiment divides a production process into multiple sub-processes based on inventory management locations, and executes simulations using both deterministic and stochastic models for each sub-process. Note that in addition to clear inventory management locations (e.g., stock shelves, racks, automated warehouses, etc.), any location where products are waiting to be input into the next process can also be considered an "inventory management location." For example, various names can be used for these locations, such as a "waiting location," "holding location," or "stagnation location." This system builds a probabilistic process sequence model using N-grams based on actual data collected from manufacturing systems such as manufacturing execution systems (MES), enterprise resource planning (ERP), supervisory control and data acquisition (SCADA), advanced planning and scheduling (APS), and Internet of Things (IoT), which handle production management, inventory management, material requirements calculation, enterprise resource planning, process management, work execution management, and equipment monitoring. It then executes multiple scenario trials using the Monte Carlo method in parallel, enabling efficient probabilistic evaluation of complex production systems, such as semiconductor manufacturing processes, job shops, flow shops, and mixed-flow production processes. Previously, there were no clearly established means or methods for probabilistically modeling uncertainties in production sites, such as process sequences, equipment operation, and worker decisions. While constructing probabilistic models has been possible in research, they have not been readily available for general users. Furthermore, because probabilistic models require numerous trials, it has been difficult to handle large-scale production models within realistic computational resources and timeframes. Therefore, the production simulation system of this embodiment provides a method for systematically modeling uncertainties such as process sequences, equipment operation, and worker decisions using a probabilistic model including N-grams, and for generating and analyzing process sequence information using this probabilistic model.Furthermore, in order to efficiently execute the large number of trials required when applying a probabilistic model, the system is equipped with an algorithm that parallelizes and optimizes model generation and trial execution, making it possible to perform probabilistic analysis of even large-scale production models within realistic computational resources and timeframes.

[0011] FIG. 1 is a diagram showing an example of the overall configuration of a production simulation system. The production simulation system of this embodiment is configured to include a production simulation server 2. The production simulation server 2 is communicatively connected to a user terminal 1, an MES server 3, and an ERP server 4 via a communication network. Note that the MES server 3 and the ERP server 4 are given as examples and are not limited to these. For example, when acquiring data such as SCADA, the MES server 3 can be communicatively connected to the SCADA server, and when MES data is not used, the MES server 3 can be omitted. The communication network is, for example, the Internet, and is constructed using a public telephone network, a mobile phone network, a wireless communication path, Ethernet (registered trademark), etc.

[0012] The user terminal 1 is a computer operated by a user, and may be, for example, a smartphone, a tablet computer, or a personal computer.

[0013] The production simulation server 2 may be a general-purpose computer such as a workstation or personal computer, or may be logically realized using cloud computing. In particular, in this embodiment, the production simulation server 2 is implemented on the cloud and receives parameters such as the number of simulation trials, the time constraint, or the upper cost limit, and dynamically controls the cloud computing resources.

[0014] The MES server 3 is a server that manages data for the manufacturing execution system and provides performance data for the production process. The ERP server 4 is a server that manages data for the enterprise resource planning system and provides data related to production planning and inventory management.

[0015] <Production simulation server 2> FIG. 2 is a diagram showing an example of the hardware configuration of the production simulation server 2. Note that the configuration shown in the figure is an example, and other configurations may also be used. The production simulation server 2 includes a CPU 201, a memory 202, a storage device 203, a communication interface 204, an input device 205, and an output device 206. The storage device 203 stores various data and programs, and is, for example, a hard disk drive, a solid state drive, or a flash memory. The communication interface 204 is an interface for connecting to a communication network, and is, for example, an adapter for connecting to Ethernet (registered trademark), a modem for connecting to a public telephone network, a wireless communication device for wireless communication, or a USB (Universal Serial Bus) connector or an RS232C connector for serial communication. The input device 205 is, for example, a keyboard, a mouse, a touch panel, a button, a microphone, or the like for inputting data. The output device 206 is, for example, a display, a printer, a speaker, or the like for outputting data. Each functional unit of the production simulation server 2, which will be described later, is realized by the CPU 201 reading a program stored in the storage device 203 into the memory 202 and executing it, and each storage unit of the production simulation server 2 is realized as part of the storage area provided by the memory 202 and the storage device 203.

[0016] 3 is a diagram showing an example of the software configuration of the production simulation server 2. The production simulation server 2 includes a process information acquisition unit 211, a process division unit 212, a simulation model creation unit 213, a simulation execution unit 214, a cloud control unit 215, and a storage unit 231.

[0017] <Production simulation server 2> The functional units of the production simulation server 2 will be described below.

[0018] The storage unit 231 is a storage device of the production simulation server 2. The storage unit 231 stores process information, partial process information, simulation models, scenario data, simulation results, N-gram models, probability maps, performance data, etc. The scenario data may include, for example, initial conditions (in-process data at the start of the simulation), production instructions (product input plans), factory operation information (operation plans for workers and each piece of equipment), etc. The storage unit 231 temporarily stores data acquired from the MES server 3 and ERP server 4 and manages it so that each functional unit can refer to it.

[0019] The process information acquisition unit 211 acquires process information indicating a production process. The process information acquisition unit 211 acquires manufacturing execution system data from the MES server 3 and acquires enterprise resource planning data from the ERP server 4. The process information acquisition unit 211 generates process information based on at least one of the acquired MES data and ERP data. Specifically, the process information acquisition unit 211 generates process information including information such as the product manufacturing process, equipment abnormality stop time for each process, processing time, transport time, inventory management location, and inter-process dependency relationships, and stores the generated process information in the storage unit 231.

[0020] The process division unit 212 divides a production process into multiple sub-processes based on information about inventory management locations included in the process information. The process division unit 212 identifies inventory management locations as division points and defines the process between each inventory management location as one sub-process. For example, in a semiconductor manufacturing process, if an inventory management location is set for each stage of the front-end process, middle-end process, and back-end process, the process division unit 212 divides the production process into three sub-processes using these inventory management locations as boundaries. The process division unit 212 stores information about each divided sub-process in the memory unit 231.

[0021] The simulation model creation unit 213 creates a model and a scenario for each partial process. The simulation model creation unit 213 creates both a deterministic model in which the process flow is uniquely defined and a stochastic model in which the process flow is dynamically determined.

[0022] The deterministic model is a model in which processing progresses according to a predefined, fixed process sequence. The simulation model creation unit 213 extracts a standard process flow from the process information and creates a deterministic model in which the standard equipment abnormality stop time, standard processing time, and standard transport time for each process are set.

[0023] The probabilistic model is created based on a model that extracts the occurrence probability of a series from past process sequence history and quantifies the branching probability between processes. The simulation model creation unit 213 creates the probabilistic model based on a model that learns N-grams of process sequences from past performance data.

[0024] In probabilistic modeling using N-grams, the simulation model creation unit 213 first executes the modeling part. Figure 5 shows the processing flow of the probabilistic model creation process. The simulation model creation unit 213 receives as input data CSV-formatted performance data in which the process order of a device or model is described by layer number (LAY_NO) (S501). The simulation model creation unit 213 fills missing cells with "-1" (S502) and initializes the row index (S503). The simulation model creation unit 213 acquires the sequence of the current row (S505) and initializes the element index (S506) until processing of all rows is completed (S504). The simulation model creation unit 213 extracts three consecutive elements (a, b, c) (S508) until processing is completed up to the maximum length (S507), and if none of them is "-1" (S509), counts trigram_counts((a, b))[c] (S510) and increments the element index (S511). The simulation model creation unit 213 increments the row index (S512), and when processing of all rows is completed, calculates the transition probability from trigram_counts (S513), and generates a probability map trigram_probs (S514).

[0025] The simulation model creation unit 213 selects the value of N as either 2 or 3 depending on the number of samples. When N=1, it is called a unigram, when N=2, it is called a bigram, and when N=3, it is called a trigram. For example, in a semiconductor manufacturing process, if the layer number transition pattern has a complex order including a reproduction process such as "1 ⇒ 2 ⇒ 3 ⇒ ... ⇒ 20 ⇒ 34 ⇒ 35 ⇒ 34 ⇒ 35 ⇒ ... ⇒ 999," the simulation model creation unit 213 uses trigrams to learn the occurrence probability of the next process from a combination of three consecutive processes.

[0026] In the probabilistic model, the simulation model creation unit 213 models the equipment abnormality stop time, processing time, or transportation time using a probability density function including a normal distribution, an exponential distribution, or a log-normal distribution. The simulation model creation unit 213 analyzes the distribution of the processing time for each process from past performance data, and selects the most suitable probability density function to model. For example, the processing time for a machining process can be modeled using a normal distribution, waiting time due to a failure can be modeled using an exponential distribution, and transportation time can be modeled using a log-normal distribution.

[0027] The simulation execution unit 214 simulates each partial process based on the model and scenario. During simulation using a probabilistic model, the simulation execution unit 214 probabilistically determines the next step of the process based on the model that has learned N-grams.

[0028] The simulation execution unit 214 executes the probabilistic case generation part. FIG. 5 shows the processing flow of the probabilistic case generation process. The simulation execution unit 214 first initializes the starting sequence [1, 2] (S601). The simulation execution unit 214 acquires the most recent two elements (a, b) (S602). If (a, b) exists in trigram_probs (S603), the simulation execution unit 214 samples the next candidate c from the probability distribution (S604) and adds c to the sequence (S605). The simulation execution unit 214 repeats the above process until the sequence length reaches the maximum length (S606), completing one sequence (S607). The simulation execution unit 214 repeats the above simulation a predetermined number of times (e.g., 100 times) (S608) and outputs all sequences to CSV (S609).

[0029] The simulation execution unit 214 executes multiple scenario trials using the Monte Carlo method. The simulation execution unit 214 performs random sampling from various probability distributions in the stochastic model, generates a large number of scenarios, and executes simulations. This allows the simulation execution unit 214 to obtain statistical evaluation results that take into account the uncertainty of the production system.

[0030] The simulation execution unit 214 executes the simulation of each partial process in a distributed parallel manner, sets a local tick for each distributed simulator, and synchronizes the overall time using the global tick. The simulation execution unit 214 divides the factory model at intermediate inventory locations and assigns each partial model to a processor unit. Each processor operates on a local tick (e.g., 1 second, 5 seconds, etc.), and the overall progress is synchronized by global tick control using the Clock module.

[0031] When updating the global tick, the simulation execution unit 214 integrates the internal state and event log of each distributed simulator and verifies the consistency of causality. The simulation execution unit 214 organizes the state information and event information collected from each distributed simulator in chronological order and checks whether the dependencies and constraints between processes are satisfied.

[0032] The simulation execution unit 214 dynamically adjusts the local tick interval or time granularity for each distributed simulator based on scenario variability or resource load. The simulation execution unit 214 monitors the processing load and event occurrence frequency of each sub-process, and optimizes overall calculation efficiency by setting finer time granularity for sub-processes with high loads and coarser time granularity for sub-processes with low loads.

[0033] The cloud control unit 215 receives the number of simulation attempts, the time constraint, or the upper cost limit as parameters and dynamically controls the cloud computing resources. The cloud control unit 215 calculates the required computing resources based on the simulation conditions specified by the user, and selects an appropriate number of instances and instance types in the cloud environment.

[0034] The cloud control unit 215 performs a short-term preliminary simulation before dividing the processes to estimate the load for each process, and automatically adjusts the division among the simulators so that the calculation load is even. The cloud control unit 215 evaluates the model division plan and the load through the pre-simulation, performs optimal processor allocation and instance placement, and moves on to actual execution. The cloud control unit 215 automatically controls the selection and optimal placement of cloud resources (e.g., AWS EC2 (Amazon Web Services Elastic Compute Cloud)) according to the execution environment.

[0035] FIG. 6 is a diagram illustrating a processing flow in the production simulation system.

[0036] The process information acquisition unit 211 acquires production process data from the MES server 3 or the ERP server 4 (S401) and generates process information based on the acquired data (S402). The process division unit 212 divides the production process into multiple sub-processes based on inventory management location information included in the process information (S403). The simulation model creation unit 213 learns an N-gram model from past performance data (S404) and creates a deterministic model and a probabilistic model for each sub-process (S405). The cloud control unit 215 executes a preliminary simulation to estimate the load for each process (S406) and adjusts the division into simulator groups so that the calculation load is even (S407). The simulation execution unit 214 sets a local tick for each distributed simulator (S408) and executes probabilistic case generation based on the N-gram model (S409). The simulation execution unit 214 executes distributed parallel simulations of multiple scenarios using the Monte Carlo method (S410), and integrates the results of each distributed simulator while synchronizing the overall time using a global tick (S411).

[0037] As described above, the production simulation system of this embodiment enables distributed parallel processing of simulations by dividing a production process into multiple subprocesses based on inventory management locations. Furthermore, while there was no means to generally express factory uncertainty or to perform probabilistic simulations, applying both N-gram-based probabilistic models and deterministic models to each subprocess allows for highly accurate prediction of the behavior of production systems, such as semiconductor manufacturing processes, where operating conditions and process sequences fluctuate stochastically or dynamically. Furthermore, tick-based distributed parallel simulation enables efficient simulation of even large-scale production systems. Furthermore, dynamic resource control in a cloud environment enables optimal allocation of computational resources under cost and time constraints. This significantly reduces design evaluation time and improves evaluation accuracy compared to conventional trial-and-error approaches.

[0038] Although the present embodiment has been described above, the above embodiment is intended to facilitate understanding of the present invention and is not intended to limit the present invention. The present invention may be modified or improved without departing from the spirit thereof, and equivalents thereof are also included in the present invention.

[0039] For example, the processing by each functional unit of the production simulation server 2 described above may be performed by any functional unit. Also, a different functional unit that executes part of the processing by each functional unit described above may be added. Also, the functional units of the production simulation server 2 may be provided in parallel and distributed across multiple computers.

[0040] Furthermore, the information stored in each storage unit of the production simulation server 2 may be stored by any storage unit. That is, the information stored in the above-mentioned multiple storage units may be stored in one storage unit, or part of the information stored in one storage unit may be stored in another storage unit.

[0041] <Variation 1> In the above-described embodiment, an example was shown in which a trigram (N=3) was used as the N-gram model, but the present invention is not limited to this. The simulation model creation unit 213 can dynamically adjust the value of N depending on the amount of performance data and the complexity of the process. Specifically, the simulation model creation unit 213 selects a bigram (N=2) when the number of performance data samples is small, and selects a trigram (N=3) or a value greater than that when there is a sufficient number of samples. The simulation model creation unit 213 evaluates the prediction accuracy for each N value and automatically selects the N value that shows the highest accuracy, thereby optimizing the prediction accuracy of the process sequence.

[0042] <Variation 2> In the above-described embodiment, an example of dividing processes based on inventory management locations was described, but the present invention is not limited to this. The process division unit 212 can divide processes based on the distribution of processing times, resource usage rates, or the strength of inter-process dependencies. Specifically, the process division unit 212 analyzes the average processing time and variance of each process and identifies boundary points where the variance of processing times changes significantly as division points. The process division unit 212 can also equalize resource loads by setting division points between processes where the utilization rates of resources (equipment or workers) change significantly. Furthermore, the process division unit 212 can minimize synchronization overhead during distributed simulation by quantifying the strength of inter-process dependencies and selecting division points between processes with weak dependencies.

[0043] <Variation 3> Although the above-described embodiment illustrates an example of a probabilistic simulation using the Monte Carlo method, the present invention is not limited thereto. The simulation execution unit 214 can improve sampling efficiency by using Latin Hypercube Sampling or a quasi-Monte Carlo method. Specifically, when sampling from a probability distribution, the simulation execution unit 214 applies Latin Hypercube Sampling to generate sample points that evenly cover the probability space even with a small number of samples. Furthermore, the simulation execution unit 214 can achieve statistical estimation that converges more quickly by using a quasi-Monte Carlo method that uses low-discrepancy sequences such as Sobol sequences or Halton sequences. This significantly reduces the number of simulation trials required to obtain results with the same accuracy.

[0044] <Variation 4> In the above-described embodiment, an example of synchronization control using a global tick was described, but the present invention is not limited to this. The simulation execution unit 214 can apply optimistic synchronization control or conservative synchronization control. In optimistic synchronization control, the simulation execution unit 214 allows each distributed simulator to proceed with parallel execution while temporarily ignoring causal constraints, and performs rollback processing if a violation of the causal relationship is detected. In conservative synchronization control, the simulation execution unit 214 pauses execution when there is a possibility that each distributed simulator may be affected by another simulator, and waits until a safe execution order is guaranteed. The simulation execution unit 214 can dynamically select the optimal synchronization control method based on the density of dependencies between processes and the frequency of event occurrence.

[0045] <Variation 5> Although the above-described embodiment illustrates an example of implementation in a cloud environment, the present invention is not limited to this. The production simulation system can be implemented in an edge computing environment or a hybrid cloud environment. In an edge computing environment, an edge server installed in a factory performs some of the functions of the process information acquisition unit 211 and the simulation execution unit 214, achieving near-real-time responsiveness. The cloud control unit 215 executes computationally intensive N-gram model learning and Monte Carlo simulation in the cloud environment and distributes the results to the edge server. In a hybrid cloud environment, highly confidential production data is processed on a private cloud and large-scale calculations are performed on a public cloud, thereby achieving both security and computational performance.

[0046] <Variation 6> In the above-described embodiment, examples of probability modeling using normal distribution, exponential distribution, and lognormal distribution were described, but the present invention is not limited thereto. The simulation model creation unit 213 can learn the distribution of equipment abnormality stoppage times, processing times, and transport times using machine learning. Specifically, the simulation model creation unit 213 models the processing time distribution as a weighted sum of multiple normal distributions using a Gaussian Mixture Model. Furthermore, the simulation model creation unit 213 estimates a probability density function directly from actual data without making parametric distribution assumptions using kernel density estimation. Furthermore, the simulation model creation unit 213 can learn complex multivariate distributions using generative models (such as a variational autoencoder or a generative adversarial network) using deep learning, thereby realizing probabilistic simulations that take into account correlations between processes.

[0047] <Variation 7> In the above-described embodiment, N-grams are used as a method for probabilistically modeling process sequences. However, this is not limiting, and it is also possible to apply sequence models using deep learning such as Markov models, Bayesian networks, probabilistic context-free grammars, conditional random fields, RNNs, LSTMs, and Transformers.

[0048] <Disclosures> The present disclosure also includes the following configurations. [Item 1] a process information acquisition unit that acquires process information indicating a production process; a process dividing unit that divides the production process into a plurality of sub-processes based on information on inventory management locations included in the process information; a simulation model creation unit that creates a model and a scenario for each of the partial processes; a simulation execution unit that simulates each partial process based on the model and scenario, Production simulation system. [Item 2] the process information acquisition unit acquires at least one of manufacturing execution system (MES) data and enterprise resource planning (ERP) data, and generates the process information based on the acquired data. Item 1. The production simulation system according to item 1. [Item 3] the simulation model creation unit creates, as the model, both a deterministic model in which a process flow is uniquely defined and a stochastic model in which a process flow is dynamically determined; Item 1. The production simulation system according to item 1. [Item 4] The probabilistic model is created based on a model that extracts a sequence occurrence probability from a past process sequence history and quantifies a branching probability between processes. Item 3. The production simulation system according to item 3. [Item 5] The probabilistic model is created based on a model that learns N-grams of process sequences from past performance data. Item 4. The production simulation system according to item 4. [Item 6] the simulation execution unit, during simulation using the probabilistic model, probabilistically determines the next step of the process based on a model that has learned the N-gram; Item 5. The production simulation system according to item 5. [Item 7] The stochastic model models the equipment abnormality stop time, processing time, or transportation time using a probability density function including a normal distribution, an exponential distribution, or a log-normal distribution. Item 1. The production simulation system according to item 1. [Item 8] the simulation execution unit executes a plurality of scenario trials using a Monte Carlo method; Item 1. The production simulation system according to item 1. [Item 9] the simulation execution unit executes a simulation of each of the partial processes in a distributed manner, sets a local tick in each distributed simulator, and synchronizes the overall time with a global tick; Item 1. The production simulation system according to item 1. [Item 10] the simulation execution unit integrates the internal states and event logs of each distributed simulator at the time of updating the global tick, and verifies consistency of causality; Item 9. The production simulation system according to item 9. [Item 11] the simulation execution unit dynamically adjusts a local tick interval or time granularity for each distributed simulator based on scenario variability or resource load; Item 9. The production simulation system according to item 9. [Item 12] The cloud control unit is implemented on the cloud and receives the number of simulation attempts, the time constraint, or the cost upper limit as parameters, and dynamically controls the cloud computing resources. Item 1. The production simulation system according to item 1. [Item 13] the cloud control unit executes a short-time preliminary simulation before dividing the processes to estimate the load for each process, and automatically adjusts the division among the simulator groups so that the calculation load is even. Item 13. The production simulation system according to item 12. [Item 14] Applied to probabilistic assessment of either semiconductor manufacturing processes, job shops, mixed-production processes, transportation planning or supply chains, Item 1. The production simulation system according to item 1. [Item 15] acquiring process information indicative of a production process; Dividing the production process into a plurality of sub-processes based on information on inventory management locations included in the process information; creating a model and a scenario for each of the partial processes; simulating each sub-process based on the model and scenario; A computer-implemented production simulation method. [Item 16] acquiring process information indicative of a production process; Dividing the production process into a plurality of sub-processes based on information on inventory management locations included in the process information; creating a model and a scenario for each of the partial processes; simulating each sub-process based on the model and scenario; A program that causes a computer to execute the following. [Explanation of symbols]

[0049] 1. User terminal 2. Production simulation server 3 MES Server 4. ERP Server

Claims

1. a process information acquisition unit that acquires process information indicating a production process; a process dividing unit that divides the production process into a plurality of sub-processes with the inventory management location as a boundary based on information on the inventory management location, including a buffer, a waiting location, or an automated warehouse, included in the process information; a simulation model creation unit that creates a model and a scenario for each of the partial processes; a simulation execution unit that executes a simulation of each partial process in a distributed and parallel manner based on the model and the scenario, the simulation execution unit sets a local tick in each distributed simulator, synchronizes the overall time using a global tick, and calculates at least one of production volume, processing load, cost, time, and resource allocation; Production simulation system.

2. The models include deterministic models in which the process flow is uniquely defined and stochastic models in which the process flow is dynamically determined. The production simulation system according to claim 1 .

3. The probabilistic model is created based on a model that extracts a sequence occurrence probability from a past process sequence history and quantifies a branching probability between processes. The production simulation system according to claim 2 .

4. The probability model is created based on a model that learns the probability of process sequences from past performance data using N-grams. The production simulation system according to claim 3 .

5. the simulation execution unit, during simulation using the probabilistic model, probabilistically determines the next step of the process based on the model learned using the N-gram; The production simulation system according to claim 4.

6. The stochastic model models the equipment abnormality stop time, processing time, or transportation time using a probability density function including a normal distribution, an exponential distribution, or a log-normal distribution. The production simulation system according to claim 2 .

7. the simulation execution unit executes a plurality of scenario trials in a distributed and parallel manner using a Monte Carlo method; The production simulation system according to claim 1 .

8. the simulation execution unit integrates the internal states and event logs of each distributed simulator at the time of updating the global tick, and verifies consistency of causality; The production simulation system according to claim 1 .

9. the simulation execution unit dynamically adjusts a local tick interval or time granularity for each distributed simulator based on scenario variability or resource load; The production simulation system according to claim 1 .

10. a cloud control unit implemented on the cloud, receiving the number of simulation attempts, the time constraint, or the cost upper limit as parameters, and dynamically controlling the cloud computing resources; The production simulation system according to claim 1 .

11. the cloud control unit executes a preliminary simulation before dividing the processes to estimate the calculation load for each process, and automatically adjusts the division among the simulator groups so that the calculation load is uniform; The production simulation system according to claim 10.

12. The production process includes a semiconductor manufacturing process. The production simulation system according to claim 1 .

13. acquiring process information indicative of a production process; Dividing the production process into a plurality of sub-processes with the inventory management locations as boundaries based on information on inventory management locations including buffers, waiting locations, or automated warehouses included in the process information; creating a model and a scenario for each of the partial processes; a step of executing a simulation of each partial process in a distributed and parallel manner based on the model and the scenario; The computer executes In the step of executing the simulation in a distributed and parallel manner, the computer sets a local tick in each distributed simulator, synchronizes the overall time using a global tick, and calculates at least one of production volume, processing load, cost, time, and resource allocation.

14. acquiring process information indicative of a production process; Dividing the production process into a plurality of sub-processes with the inventory management locations as boundaries based on information on inventory management locations including buffers, waiting locations, or automated warehouses included in the process information; creating a model and a scenario for each of the partial processes; a step of executing a simulation of each partial process in a distributed and parallel manner based on the model and the scenario; A program for causing a computer to execute the above, A program that, in the step of executing the simulation in a distributed and parallel manner, causes the computer to set a local tick in each distributed simulator, synchronize the overall time using a global tick, and calculate at least one of production volume, processing load, cost, time, or resource allocation.

Images