Design support system and design support method

Abstract

To design a generating process to obtain deliverables of appropriate quality at different scales.SOLUTION: A design support system 1 calculates a value for representing a non-uniformity of a state in a space area based on information on the state of each location when a process is performed on a raw material distributed in the space area; generates a state distribution model 108 that outputs data for representing the non-uniformity of the state after the process based on the value; inputs a process design parameter value to the state distribution model 108 so as to acquire a data value for representing a non-uniformity of the state after the process; acquires a quality parameter of the deliverable by inputting each data representing a possible state, which is identified from the data value, into a quality prediction model 105; and modifies the value of the process design parameter so that a quality of the deliverable falls within a predetermined range, based on each quality parameter.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a design support system and a design support method. [Background technology]

[0002] In the chemical and materials industries, technological considerations are carried out in stages from the start of product development to mass production. Specifically, we will conduct (1) laboratory experiments, (2) bench tests, (3) pilot plant tests, and (4) ) Plant design and operation. In the experiment, we carried out trial production at the beaker level while changing the combination of raw materials and various reaction conditions. Determine the basic properties and yield of the product. Bench testing produces the product obtained in the lab experiment. We will assemble small-scale equipment to perform various unit operations such as distillation or adsorption, and design processes. In the pilot plant test, a detailed physical property study and design will be carried out. A large-scale pilot plant will be constructed to determine the operating conditions, operation methods, and instrumentation systems. The plant design will be based on the technology established in the pilot plant. The bench test and pilot test will be integrated into the other phases. Sometimes they are combined.

[0003] The specifications of the equipment, facilities, and manufacturing processes vary greatly between these phases. For example, if the manufacturing process is a reaction process, the volume of the reactor increases as the phase advances. The product can be several tens to several hundreds of times.

[0004] Therefore, if the desired quality (physical properties and yield) is not achieved due to differences in specifications between phases, A specific example will be explained using a reaction process as an example. First, the volume of the reactor is large. When the temperature becomes too high, the surface area per unit volume decreases, and the efficiency of heat transfer from the outside decreases, or the agitation The surface area per unit volume that the blades come into contact with is reduced, which can reduce the mixing efficiency. As a result, heat transfer is delayed. In general, the reaction rate increases with temperature and concentration. The ratio and distribution of physical properties of by-products vary depending on the concentration of each substance and the reaction rate. Considering the above-mentioned heat transfer delay and the change in physical properties or yield due to unevenness in temperature or concentration, Specifications such as the size and shape of the agitator blades, and operating conditions such as the reaction temperature time profile The design work for each specification according to the progress of the phase is collectively called "S It's called Scale Up.

[0005] Scaling up is a time-consuming task. For example, the reaction conditions devised in the lab experiment If the conditions cannot be reproduced in bench tests, the properties or yields will need to be redesigned in laboratory experiments. In addition, pilot plant tests were conducted using a large capacity and long production time. In some cases, the work period may take six months to a year and a half. For manufacturers of chemicals and materials, scaling up is a bottleneck in product development.

[0006] For example, Non-Patent Document 1 describes a pilot scale-up technique for efficient scale-up. Immediately after completing the verification in the pilot plant test and starting production in the manufacturing facilities of the plant, Optimize plant operating conditions with as few experiments as possible to produce the desired products Specifically, a method is disclosed that combines transfer learning and Gaussian process regression. By using these data, we constructed a model (quality prediction model) that predicts the average and variance of quality from the operating conditions. By constructing a model and conducting experiments under conditions that may increase the objective function while taking into account variability, This discloses a method for obtaining good quality with a small number of experiments. [Prior art documents] [Non-patent literature]

[0007] [Non-Patent Document 1] Manabu Kano, Ryosuke Yoshizaki, Bayesian optimization of operating conditions during scale-up of manufacturing facilities, Chemical Engineering, Vol. 80, No. 12, pp. 769-772 (2016). Summary of the Invention [Problem to be solved by the invention]

[0008] In Non-Patent Document 1, the number of outputs of the quality prediction model is the same in the pilot plant and the actual plant. In other words, the KPIs to be optimized before and after scale-up are and the number of its components must be the same. However, the scale from the laboratory test The number of KPIs and their components often differs for each setup. For example, bench equipment, pyrotechnics, etc. In this plant, the heat transfer delay and the unevenness of temperature or concentration are taken into consideration. These are generally not considered in laboratory studies, but are necessary to optimize yield. In addition, quality items are considered as univariates in laboratory tests, but in bench, pilot, and full-scale tests, The runt scale uses multiple statistics (average, maximum, minimum, etc.) related to quality items to maximize the quality. In such cases, the method of Non-Patent Document 1 cannot be applied. stomach.

[0009] In addition, as an approach different from that of Non-Patent Document 1, a first-principles simulation on a plant scale has been proposed. Simulation (for reaction processes, coupled simulation of fluid and reaction, for extrusion processes, For example, a mixing simulation is performed to measure the concentration and temperature inside the process. It is also possible to precisely estimate the quality of the data and use the results to estimate the quality. The above simulation requires a huge calculation load, so we have to consider the equipment specifications, operating conditions, etc. Repeated trials with changes and simulations of the entire time period from start to finish of production It is difficult to carry out a

[0010] In this way, the data obtained from laboratory tests can be used for bench tests, pilot plants, Or, to efficiently grasp the quality and yield of the plant, and produce products of stable quality at the plant. It was difficult to create a mechanism to achieve this.

[0011] The present invention has been made in view of the above background, and its object is to provide a method for detecting a difference in a scale between different scales. In order to obtain deliverables of appropriate quality in the process of producing deliverables carried out in each of the above To provide a design support system and a design support method capable of carrying out the above process design. is located. [Means for solving the problem]

[0012] One aspect of the present invention for solving the above problem is to provide data representing a state, and a raw material in the state. Data representing a process to be performed on the raw material is input, and the raw material is obtained after the process. a storage device that stores a quality prediction model from which data representative of the quality of a deliverable to be obtained is output; and The state of each position in a spatial region when a process is performed on raw materials distributed in the spatial region. and representing the state heterogeneity in the spatial domain based on each of the acquired information. a value is calculated, and based on the calculated value representing the non-uniformity and data defining the process, Process design is data that defines the process to be performed on raw materials present in a spatial domain. The parameters are input and the state in the spatial domain during and after the process is executed is determined. A state distribution model construction process that generates a state distribution model that outputs data that represents the non-uniformity of and setting the value of the process design parameter, and setting the value of the set process design parameter By inputting the state distribution model, the values ​​of the set process design parameters are The process design parameters for which the values ​​are set during or after the process shown in FIG. A state distribution model execution process to obtain data values ​​representing the state heterogeneity in the spatial domain; From the acquired data values ​​representing the non-uniformity of the state in the spatial domain, A plurality of possible states are specified, and each data representing each specified state and the set profile are stored in the and inputting the process design parameter values ​​into the quality prediction model for each of the states. Thus, after the process indicated by the set process design parameter value, A manufacturing quality prediction process for obtaining quality parameters of the deliverables obtained from the manufacturing process and Based on each quality parameter, the quality of the product obtained from the raw material is adjusted to fall within a predetermined range. and performing a parameter search process to change the values ​​of the set process design parameters. The design support system is equipped with a control device that Effect of the Invention

[0013] According to the present invention, in the process of producing a product which is carried out at different scales, This allows for process design to obtain deliverables of appropriate quality. Configurations and effects other than those described above will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0014] [Figure 1] FIG. 1 is a diagram illustrating an example of a configuration of a design support system according to an embodiment of the present invention. [Diagram 2] FIG. 13 is a diagram showing an example of laboratory experiment data. [Diagram 3] FIG. 1 is a diagram illustrating an example of a configuration of a plant experiment. [Figure 4] FIG. 2 is a diagram showing an example of plant experiment data obtained by a plant experiment. [Diagram 5] FIG. 1 is a diagram illustrating an example of CAE simulation data. [Figure 6] FIG. 13 is a diagram illustrating an example of a state distribution model database. [Figure 7] FIG. 2 is a diagram illustrating an example of hardware included in an information processing device constituting the design support system. [Figure 8] FIG. 4 is a flow diagram illustrating an example of a first process design parameter determination process. [Figure 9] FIG. 13 is a diagram showing an example of a state distribution model construction screen. [Figure 10] FIG. 11 is a flow diagram illustrating details of a state distribution model construction process. [Figure 11] FIG. 11 is a flow diagram illustrating details of a state distribution generation process. [Figure 12] FIG. 11 is a flow diagram illustrating details of a manufacturing quality prediction process. [Figure 13] FIG. 11 is a flow diagram illustrating details of a process search process. [Figure 14] FIG. 13 is a diagram illustrating an example of a process search screen. [Figure 15] FIG. 11 is a flow diagram illustrating details of a state distribution model regeneration process. [Figure 16] FIG. 11 is a diagram illustrating an example of manufacturing performance data. [Figure 17] FIG. 11 is a flow diagram illustrating an example of a second process design parameter determination process. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The following are examples to explain the present invention, and some omissions and simplifications have been made to clarify the explanation. The present disclosure may be embodied in various other forms. Each component may be singular or plural.

[0016] The design support system according to the present embodiment performs a predetermined process (physical or chemical) on raw materials. Development of a product (an end product, e.g., resin or fiber) produced by carrying out a process (a technical process) For development and manufacturing, small-scale experiments (e.g., laboratory experiments, where small-scale processes are carried out) A small-scale trial was carried out at the beaker level, which has a spatial domain that is to be measured. Experiments are sometimes called "experiments.") and large-scale tests (bench tests, pilot plant tests, A spatial domain in which large-scale process operations are carried out, such as the plant design (operation) Large-scale testing or production of products in large-scale equipment. In the following, large-scale production is referred to as plant experiment. In some cases, this is used when implementing the following:

[0017] Specifically, the design support system was developed based on actual plant experiments (large-scale tests). The result data of the above or the coupled simulation using the coupled model described later (hereinafter, spatial prototyping) The data obtained by running the CAE (Computer Aided Engineering) Based on the simulation data, the process is carried out in a given area. Distribution of internal conditions in the spatial domain (spatial distribution of factors that affect the quality of the deliverables, such as temperature) The value that represents the non-uniformity of the distribution (in this embodiment, the average value and The state distribution model is used to calculate the distribution of the state of the variance ( Distribution of internal states in the spatial domain during the production of a product (under various equipment sizes and operating conditions) However, in this case, the value obtained by the plant experiment or the coupled simulation is output. Only a small amount of data can be collected.

[0018] Furthermore, the design support system takes into account the distribution of internal states as described above. Based on actual data from laboratory experiments (small-scale experiments) conducted on a small scale where no A quality prediction model is constructed to predict the quality (hardness, etc.) of the product obtained from the above.

[0019] The design support system is designed to estimate the external conditions of large-scale testing (e.g., the size of the plant equipment, etc.) are the shape, mixing conditions, or plant operating conditions. Hereinafter referred to as process design parameters.) By setting the contents of and inputting the set process design parameters into the state distribution model, A value that represents the unevenness or range of distribution of the internal state of the plant due to a large-scale test (in this embodiment) The output is the mean and variance of the internal state as a random variable. The support system 1 samples the state value (temperature value) based on this nonuniformity value. Each sampled value and the corresponding process design parameters are input to the quality prediction model. By inputting the information, the product will be manufactured under the specified conditions (temperature) and process design parameters. Predict the quality of each product.

[0020] Here, the original data of the input values ​​to the quality prediction model are random variables as described above. Similarly, the product quality values ​​output (predicted) from the quality prediction model also correspond to random variables. This allows the output to be a value that takes into account the expected variation in quality. It is now possible to understand deviations and variations in product quality when scaling up to mock tests. In other words, the design support system can predict the quality of the product manufactured in large-scale testing. The quality of the manufactured product must be within a specified range while taking into account the variation in the production process and the plant constraints, etc. It is possible to search for process design parameters that satisfy the above.

[0021] In the following, the process carried out on the raw material is referred to as a reaction process. The present invention also broadly applies to other types of processes, such as physical processes (e.g., mixing processes). For example, it may be in the manufacturing processes of continuous products such as chemicals, pharmaceuticals, or steel. In addition, for energy plants such as thermal power plants, it is important to understand the internal state of the equipment or facilities. It may also be a process involving.

[0022] <Design support system configuration> FIG. 1 is a diagram showing an example of the configuration of a design support system 1 according to this embodiment.

[0023] The design support system 1 includes a lab experiment data 21 that stores experimental data from a small-scale experiment, Plant experiment data 22 that stores experimental data from large-scale tests, and a reactor Manufacturing performance data 23 that stores data on the manufacturing performance of products manufactured on a full scale, such as 30 and is connected to.

[0024] The design support system 1 also includes a quality prediction model 105, a reaction characteristic model 106, and AE simulation data 107, state distribution model 108, and CAE simulation The system stores a state distribution model 109 and a state distribution model database 115.

[0025] (Lab experimental data) Here, FIG. 2 is a diagram showing an example of the laboratory experiment data 21. The laboratory experiment data 21 includes: This is data that stores the experimental conditions and results of small-scale experiments. The data 21 includes, for each small-scale experiment, an experiment ID 211 for identifying the experiment, The experimental conditions 212, the reaction characteristics 213 in the experiment, and the results obtained by the experiment Each data of the product (product) quality 214 is stored in association with each other.

[0026] The experimental conditions 212 and reaction characteristics 213 are data relating to the raw material or the process performed on the raw material. Specifically, the experimental conditions 212 are the composition of the raw materials used in the experiment (ingredients and their composition). The reaction characteristics 213 are determined by experiments under the experimental conditions. The measured reaction rate, heat of reaction, and viscosity of the solution are included. The hardness and elongation of the material obtained by the above method are included. Note that the data items explained here are only examples. The experimental conditions 212 and reaction characteristics 213 represent various physical or chemical processes. The quality 214 represents various physical or chemical properties of the product. There can be parameters.

[0027] The laboratory experimental data 21 is used as the raw material for the quality prediction model 105 and the reaction characteristic model 106. Used for creation.

[0028] (Plant experiment configuration and plant experiment data) Next, FIG. 3 is a diagram showing an example of a configuration of a plant experiment. FIG. 4 is a diagram showing an example of plant experiment data 22 obtained by a plant experiment. It is.

[0029] First, as shown in Figure 3, in this plant experiment, Thermometers 51 to 55 are installed at various points in the vessel (reactor 30). A control system information processing (which may be, for example, a design support system 1) controls the operation of the reactor 30. While the temperature is measured at predetermined intervals, the temperature of the thermometers 51 to 55 is measured. The temperature is recorded in the printer 22. The positions of the thermometers 51 to 55 may be registered in the runt experiment data 22. It is preferable to set the temperature measurement position so that the temperature values ​​measured at the positions vary.

[0030] Next, as shown in FIG. 4, the plant experiment data 22 is The ID of the plant experiment 221, the equipment specifications for the plant experiment 222, and the The composition of raw materials (raw material mix 223) and the results of plant experiments 224 are matched to each other. The results of the plant experiment 224 are the data obtained by measuring the temperature of each sensor (e.g., The parameters (temperature) measured by 51-55 in total are included. The time and temperature in the experiment are based on the CAE simulation data described below, although the data granularity is different. Simulation results for data 107 and data with the same structure as time and temperature in data 1074 The plant experimental data 22 is used to generate a state distribution model 108.

[0031] (CAE simulation data) Next, FIG. 5 is a diagram showing an example of the CAE simulation data 107. The simulation data 107 is a CAE simulation model 109 and a reaction characteristic model 106 This is a database that records the results of running a coupled simulation model that couples Specifically, the CAE simulation data 107 includes the following for each coupled simulation: The ID1071 identifies the coupled simulation and the data that defines the reaction process. The equipment specification 1072, which is the data, and the raw material specification set as the input value in the coupled simulation The composition (raw material mix 1073) and the result of the coupled simulation 1074 are associated I remember that.

[0032] The equipment specification 1072 is the process specification set as an input value in the coupled simulation. The vessel (reactor) in which the process is carried out or the equipment, devices, or facilities used in the process (hereinafter referred to as "the This is data on reactors, equipment, devices, and facilities (collectively referred to as equipment). Equipment Specifications 1072 specifies, for example, the shape of the reactor, the radius of the reactor, the height of the reactor, the installation The shape of the agitator blades, the area of ​​the agitator blades, and the heat removal method set in the reactor are included. 3 includes the ingredients of the raw materials and their blending ratios. Also, the results of the simulation 1074 are the flow velocity (3D flow velocity), temperature, and and viscosity.

[0033] (State distribution model database) FIG. 6 is a diagram showing an example of the state distribution model database 115. The database 115 stores data sources for generating each state distribution model 108. There are.

[0034] Specifically, the state distribution model database 115 stores the ID of the state distribution model 108, The type of data used to generate the state distribution model 108 (training data source 115 2), information identifying the contents of the data source (learning data ID 1153), and its status The model data 1154 stores the contents of the state distribution model 108. The data source 1152 may be, for example, CAE simulation data 107 or plant experiment data. The learning data ID 1153 is data indicating that the learning data is one of CA E Specific data in simulation data 107 or plant experiment data 22 Information to identify the source (specifically, data ID 2111 of plant experiment data 22 or C The data ID of the AE simulation data 107 is 1071).

[0035] Next, as shown in FIG. 1, the design support system 1 includes a quality prediction model construction unit 101 and a A response characteristic model construction unit 102, a CAE simulation unit 103, and a state distribution model construction unit 104, a state distribution generating unit 110, a manufacturing quality predicting unit 111, and a process searching unit 112. , and a user interface unit 113.

[0036] The quality prediction model construction unit 101 determines the quality prediction model under the specified experimental conditions based on the data of the small-scale experiment. A quality prediction model that predicts the quality of the product (outcome) manufactured based on the reaction characteristics of the raw materials and Generate 05.

[0037] Specifically, the quality prediction model 105 uses data representing a state (temperature) and a principle Reaction process performed on the material (experimental conditions 212 and reaction characteristics 213 of laboratory experimental data 21) ) is input, and the quality of the product produced from the raw materials in the reaction process is In this embodiment, the quality prediction model 105 outputs data representing the is generated and stored in advance.

[0038] In this embodiment, the quality prediction model 105 is generated based on the laboratory experimental data 21. In other words, the input value of the quality prediction model 105 is the actual value of the laboratory experimental data 21. The data corresponds to the test conditions 212 and the reaction characteristics 213. The value is data corresponding to the quality 214 of the laboratory experimental data 21.

[0039] The reaction characteristic model construction unit 102 generates a reaction characteristic model 106 . In this embodiment, the reaction characteristic model 106 is a model of the reaction characteristics from the start to the end of the large-scale test. The reaction characteristic model 106 calculates the change in the characteristics from the composition of the raw material to the completion of the reaction. The model uses reaction temperature or reaction time as input and reaction rate, reaction heat, or solution viscosity as output. Dell Inc.

[0040] In this embodiment, the reaction characteristic model 106 is based on the laboratory experimental data 21 and is The reaction characteristic model 106 is generated based on a learning algorithm. The input layer receives the force values, and one or more intermediate layers extract and output features from the input values ​​( A neural network (C NN: A model trained by a Convolutional Neural Network (NN, etc.). For example, reaction The characteristic model 106 is, for example, based on the experimental conditions 212 (corresponding to the input values) of the laboratory experimental data 21 and The output value is generated by machine learning using the response characteristic 213 (corresponding to the output value) as training data.

[0041] In this embodiment, it is assumed that the reaction characteristic model 106 is generated and stored in advance. The reaction characteristic model 106 may be, for example, a predetermined reaction stoichiometry, chemical descriptors, or molecular It is a mathematical formula that uses the amount distribution of a reaction as input variables and the reaction rate, reaction heat, or viscosity as output variables. It is also possible.

[0042] The CAE simulation unit 103 stores a CAE simulation model 109 . The CAE simulation unit 103 executes the CAE simulation model 109, CAE simulation data 107 is generated.

[0043] CAE Simulation Model 109 Predicts Fluid Behavior in Reactor During Large-Scale Testing Specifically, the CAE simulation model 109 is a model that calculates the reactor specifications ( The reactor is then calculated after the material movement process (fluid movement) using the input values ​​(shape, etc.) of the reactor. The governing equations (Navier-Stokes equations) are used to calculate the state (temperature, viscosity, etc.) at each time and position in the Equations).

[0044] The CAE simulation unit 103 is a CAE simulation model 109 and a reaction characteristic model. A coupled simulation model is executed by combining the The ration model is a process of raw materials distributed in the spatial region of a reactor (a flow reaction process). Data defining each of the flow reaction processes in the reactor at each time point after the start of the flow reaction process is input. This is a model that outputs the state of the raw material at a position.

[0045] In this embodiment, the reaction characteristic model construction unit 102 includes a state distribution model database 115 It is assumed that a coupled simulation model is generated in advance based on the above.

[0046] The coupled simulation model is a different type of simulation from the models described above. For example, if the process for manufacturing a product is an extrusion process, If the process includes a kneading simulation model, it may include a chemical process. In a process that does not involve a reaction (e.g., a steel manufacturing process), a reaction characteristic model construction unit 10 2 and the reaction characteristic model 106 are not required, and the coupled simulation model is The simulation model 109 may be configured as the only one.

[0047] In addition, coupled simulations on such a large-scale test scale have been performed using various commercial systems. Runs on simulator software or software such as OSS (Open Source Software) It is possible, but running such software requires a lot of time and effort. Fine-grained measurements in both time and space are required to accurately track reaction progress and viscosity changes. In addition, it is necessary to always calculate a solution that is consistent with the boundary conditions at the reactor walls and the agitator. Therefore, depending on the conditions, the calculation may not converge or a physically meaningless solution may be obtained. For this reason, such software may only be implemented to a very limited extent. These alone do not usually provide the information necessary to identify process design parameters. However, in this embodiment, even in such a situation (a limited number of Based on the results of these software (even if only simulation results are available), By combining the state distribution model 108 generated by the above and the quality prediction model 105, Process design parameters can be identified.

[0048] Next, the state distribution model construction unit 104 uses the CAE simulation data 107 or the plan The experimental data 22 is used to generate a state distribution model 108 .

[0049] The state distribution model 108 is data on the composition of the raw materials (ingredients of the raw materials and their blending ratios), and data defining the process to be performed on the raw materials present in the reactor (process design parameters data) is entered, and each time after the process is started (during or after the process is Data showing the non-uniformity of conditions within the reactor are output.

[0050] Here, the process design parameters in the input values ​​are data that define the manufacturing process. In this embodiment, for example, the specifications (shape, size, etc.) of the device or the operating conditions (Process methods such as stirring or cooling, or to facilitate reactions) The time profile of the state of the device (e.g., time profile of the device) is, but is not limited to, this.

[0051] The output values, which are data representing the heterogeneity of the state distribution model 108, are calculated as the process is executed. The distribution of the parameters may vary or vary depending on the quality prediction model. In this embodiment, the output value of the state distribution model 108 is the input value of the state distribution model 105. However, the temperature is not limited to this, and may be, for example, the concentration or density of the raw material or product in the reactor. In addition, the data representing heterogeneity is assumed to be a probability distribution (mean and variance). However, data showing other non-uniformities may be used. calculates the state of each position in the three-dimensional space, but in this embodiment, Represented by a probability distribution, which is a parameter that can represent the state (variability) as a whole This reduces the cost of computing or storing the state of each location.

[0052] The state distribution model 108 is, for example, a state (here, temperature ) probability density function f 0 (y(t)), and further, at time t, the It has the function of generating the state y(t) by performing sampling.

number

[0053] Here, state y(t) is the state value at time t, and σ(t) is the state y(t ) and μ(t) is the mean of state y(t) at time t.

[0054] Furthermore, the time evolution of σ(t) and μ(t) is expressed as the time evolution before time t, as shown in Equation 2. It is determined by the variance, mean, and process design parameters (external inputs) at the previous time. .

number

[0055] Here, θ(t) is the external input at time t and corresponds to the process design parameter. In the embodiment, θ(t) is the data of the device specifications (data of a fixed value regardless of time t) and is a vector consisting of the process variables (variables that depend on time t) and the process time t. d is the order. do.

[0056] Note that Equation 1 and Equation 2 are formulas for the case where state y is a univariate, but for the case where A similar description is possible.

[0057] Next, the state distribution generation unit 110 uses the state distribution model generated by the state distribution model construction unit 104 The raw material composition (raw material mix, etc.) and the values ​​of the process design parameters are input to the data processor 108. This represents the distribution (non-uniformity) of the state inside the reactor at each time after the start of the process. Output the data (i.e., mean and variance).

[0058] The production quality prediction unit 111 predicts the state distribution in the reactor outputted from the state distribution generation unit 110. From the data showing the distribution (non-uniformity) of the state, multiple values ​​of the specific states that can be taken inside the reactor are calculated. Calculate (sample) and compare the data of their conditions, the composition of the raw materials, and the process design parameters. By inputting the meter into the quality prediction model 105, the product generated after the process is completed can be Predict the quality of.

[0059] In this embodiment, the production quality prediction unit 111 randomly predicts the temperature data based on the mean and variance. The data of each sampled temperature is input to the quality prediction model105. By inputting these parameters and predicting the quality (hardness and elongation), the quality variation in the target process can be Specify the histogram or probability density function.

[0060] The process search unit 112 searches for a process so that the quality predicted by the manufacturing quality prediction unit 111 falls within a predetermined range. In other words, the process design parameters and the composition of raw materials are explored. The search unit 112 considers the variation in quality and also considers predetermined KPIs or constraints. Optimization is performed to find the probability (percentage) that the quality of the deliverables will fall within a specified range. This allows the contents of process design parameters and the composition of raw materials to be explored.

[0061] The user interface unit 113 is a unit for inputting parameters required for a series of processes performed by the design support system 1. The meter has a function to receive data from the user 40 (input function), and a function to receive calculation results and control results, etc. and a function to provide the information to the user 40 (output function).

[0062] Specifically, the input function may include, for example, a keyboard, a touch panel, a voice input device, and a gaze detection device. The input of information is received from a user 40 using an output device or the like. Information is output via a monitor display, printer, voice synthesizer, etc.

[0063] The user interface unit 113 is an information processing device different from the design support system 1. The information processing device may be, for example, a laptop type device operated by a user 40. Notebook, tablet, or desktop personal computers, smartphones, The devices are smartphones, or wearable devices such as goggles or wristwatches.

[0064] FIG. 7 shows an example of hardware included in an information processing device constituting the design support system 1. The information processing device 200 includes, for example, a processor 201 (control device) and A memory 202, an external storage device 203, a communication device 204, an output device 205, and an input device 206 and a read / write device 207, which communicate with each other via a communication line 208. The processor 201 is not limited to a CPU (Central Processing Unit), The external storage device 203 may be, for example, a hard disk drive. , flash memory devices, magneto-optical disk devices, optical disk devices, and other relatively large capacity devices. The communication device 204 is a device that stores data in a rewritable manner. rface Card) and other communication networks CN (Internet, LA N (Local Area Network), WAN, or dedicated line or other wireless or wired communication network) The output device 205 communicates with an external device via the user interface unit 113. The input device 20 is a device to be used, such as a monitor display, a printer, etc. 6 is a device used by the user interface unit 113, such as a keyboard, a pointing device, etc. The read / write device 207 is a non-temporary computer. The communication line 208 reads and writes information from and to the storage medium MM that stores the computer program. The bus may be a system bus that connects one computer, or a bus that connects multiple computers. In other words, the main functions of the design support system 1 can be shared among multiple computers. By installing the software on a computer and connecting the computers to each other via a communication network, It is also possible to realize a design support system 1.

[0065] Various data stored in the design support system 1 or used for processing are The server 201 reads the data from the memory 202 or the external storage device 203 and uses it. The functions of the functional units of the design support system 1 are implemented by a processor 201. A predetermined computer program stored in the external storage device 203 is loaded into the memory 202. This can be achieved by loading and executing the program.

[0066] The above-mentioned predetermined computer program is stored in the storage medium MM via the read / write device 207. or from the network via the communication device 204 to the external storage device 203. (downloaded), then loaded onto the memory 202 and The computer program may be executed by the read / write device 207. The data is then transferred from the storage medium MM or from the network via the communication device 204 to the memory 20 2 and executed by the processor 201.

[0067] In the following, the design support system 1 is assumed to be configured by one information processing device. As described below, all or part of these functions may be implemented on one or more computers, such as a cloud. The design support system is distributed among multiple computers and communicates with each other via a network. It may realize each function of system 1. Next, the processing performed by the design support system 1 will be described.

[0068] First, a state distribution model 108 is generated based on a CAE simulation model 109. In this case, the process design parameters are determined (hereinafter, the process design parameter first determination process). This section explains the process (called "fixed processing").

[0069] <First process for determining process design parameters> FIG. 8 is a flow diagram illustrating an example of the first process design parameter determination processing.

[0070] Design support system 1 is a system that calculates fixed values ​​of process design parameters assuming a plant experiment. Specifically, the user interface unit 113 receives a command from the user 40. The input may be a fixed value set in a large-scale test, or may be a preset value. A fixed value may be received from the database.

[0071] In this embodiment, the user interface unit 113 is As fixed values ​​of, the reactor shape is set to "cylinder" and the stirring blade shape is set to "propeller", Set the cooling method to "jacket."

[0072] The design support system 1 then calculates various values ​​based on the fixed values ​​of the process design parameters. Set multiple combinations and use the combinations as input values ​​to perform coupled stain analysis. Then, the simulation is performed (S802).

[0073] Specifically, first, the CAE simulation unit 103 receives a request from the user 40 for a large-scale test. The input of various values ​​of process design parameters to be set in the process is accepted.

[0074] For example, the CAE simulation unit 103 calculates the reactor radius r, the reactor height h, the agitation blade area The nominal values ​​r0, h0, and S0 of S are input from the user 40, and based on these input values, Generate conditions by varying r, h, and S. Each parameter is expected to vary significantly from its nominal value. For example, (r, h, S) = (0.25r0, 0.25h0, 0.25S 0), (0.5r0, 0.5h0, 0.5S0), (r, h, S), (2r0, 2h0, Generates (2S0), (4r0, 4h0, 4S0).

[0075] Then, the design support system 1 performs a coupled simulation on each of the values ​​set above. Input to the model (CAE simulation model 109 and reaction characteristic model 106) The design support system 1 performs a coupled simulation. The data output by the model is stored in CAE simulation data 107 .

[0076] Next, the design support system 1 stores the CAE simulation data 10 stored in S802. 7, a state distribution model construction process S803 is executed to generate a state distribution model 108 based on the The state distribution model construction process S803 will be described later in detail.

[0077] Then, the design support system 1 performs a process design to search for process design parameters. Candidate values ​​for the design parameters (external inputs θ0, θ1, ...) are set (S804). For example, The design support system 1 adds a predetermined increase or decrease value to the nominal value of the condition value of the process design parameter. value (with noise added) or accepts input of an initial value from the user 40. Good too.

[0078] Then, the design support system 1 calculates the candidate values ​​of the process design parameters set in S804. The state distribution model 108 is then executed by inputting the state distribution model 108 generated in S803. A state distribution generation process S80 is performed to calculate the state distribution (average and variance) in the reactor by 5. The state distribution generation process S805 will be described in detail later.

[0079] (State distribution model construction screen) FIG. 9 shows one example of a state distribution model construction screen 1200 displayed by the design support system 1. FIG.

[0080] The state distribution model construction screen 1200 is a screen for constructing a state distribution model based on the composition of raw materials and fixed values ​​of process design parameters. A precondition input field 1201 accepts input of the preconditions from the user, and a process design parameter input field 1202 accepts input of the preconditions from the user. a design parameter input field 1202 for receiving input of combinations of values ​​of each of the design parameters from the user; Based on the data entered in the prerequisites input field 1201 and the design parameters input field 1202, A state distribution model construction instruction specified by the user when generating the state distribution model 108 It has a fixed column 1203.

[0081] In addition, the state distribution model construction screen 1200 displays the state distribution model 108 generated A combination of values ​​of the process design parameters to be input is received from the user. A design parameter setting field 1208 and the input values ​​input in the design parameter setting field 1208 A state distribution model specified by a user when executing the state distribution model 108 based on The execution specification field 1204 and the graph showing the execution result of the state distribution model 108 are displayed. 12. It has a quality display column 1205 and a quality prediction display column 1206 which will be described later.

[0082] The graph displayed in the graph display field 1205 has time (elapsed time from the start of the reaction) on the horizontal axis. The vertical axis represents the state inside the reactor (here, temperature). This graph shows a first graph 1209 representing a range of state values ​​and the average value in that range. The graph also includes a second graph 1210 showing the change in the first graph 1209. A virtual sample is generated by connecting the sampled values ​​1212 from the range of Includes a graph of the state (temperature) over time (sample path graph 1211). Sampled values Details of 1212 and the sample path will be described later.

[0083] Also, in the graph display field 1205, a state limit value 1215 (the upper limit value of the temperature in this figure) is displayed. The limit value 1215 will be described in detail later.

[0084] Next, as shown in FIG. 8, the design support system 1 selects a state distribution from the state distribution output in S805. By inputting the information of each state sampled from the , and execute a manufacturing quality prediction process S806 for predicting the quality of the manufactured product. The quality prediction process S806 will be described in detail later.

[0085] Then, the design support system 1 calculates each quality of the product predicted in the manufacturing quality prediction process S806 ( From the quality distribution, the quality of the process design parameters and raw material composition set in S804 can be judged. The process search unit 112 executes the process search process S807 to be evaluated. is a distribution of quality for a given objective function that evaluates the quality of process design parameters, etc. Input the process design parameters and the composition of the raw materials, and obtain the values. The processing step S807 will be described in detail later.

[0086] When the change in the value of the objective function acquired in S807 becomes equal to or smaller than a predetermined value (the If the difference between the value acquired in S807 and the value acquired last time is equal to or less than a predetermined value, (S808: YE S), the process of S809 is executed, and the change in the value of the objective function acquired in S807 exceeds a predetermined value. If the process design parameter is different from the process design parameter (S808: NO), the process search unit 112 searches for a different process design parameter. The process of S804 is repeated to set the value.

[0087] In this case, the optimization process from S804 to S808 is performed using the so-called gradient method. However, other optimization methods, such as greedy search and random search, can also be used. We can solve the optimization problem by using

[0088] In S809, the design support system 1 uses the process changed in the process search process S807. A state distribution model regeneration process S for updating the state distribution model 108 based on the design parameters The details of the state distribution model regeneration process S809 will be described later. The first process for determining the design parameters is then completed.

[0089] Based on the above, we searched for process design parameters that are likely to achieve the target quality in actual manufacturing. It becomes possible to search. Next, a state distribution model construction process S803, a state distribution generation process S805, and a manufacturing quality prediction process The process search processing S806 and the process search processing S807 will be described in detail below.

[0090] <State distribution model construction process> FIG. 10 is a flow diagram illustrating the state distribution model construction process S803 in detail.

[0091] The CAE simulation unit 103 calculates S80 from the CAE simulation data 107. Extract one target process design parameter that meets the conditions set in 1 (S90 1).

[0092] In this embodiment, the CAE simulation unit 103 performs simulation using a reactor having a cylindrical shape and an agitating blade shape. The data ID for the propeller and jacket cooling system is simulated using CAE. The data is extracted from the application data 107.

[0093] Next, the CAE simulation unit 103 calculates the process design parameters extracted in S901. The coupled simulation was performed based on the raw material composition (raw material mix) that was set. From the CAE simulation data 107, the simulation results 1074 (each time The spatial distribution of temperature at the target point (S902) is obtained.

[0094] The CAE simulation unit 103 outputs the simulation result 107 obtained in S902. 4. Randomly sampling multiple locations from the sensor to obtain a time series of temperatures at the multiple locations. In this way, multiple virtual temperature time series (sample paths) are generated, and the generated sample paths are used as learning data. and stores the result (S903).

[0095] The CAE simulation unit 103 may perform sampling automatically or may perform sampling by a user. The user 40 instructs the state distribution model construction screen 1200 to designate data to be sampled. The sampled value 1212 may be received. The number of data to be sampled may be fixed or may be set according to the variance.

[0096] The CAE simulation unit 103 calculates the formula 3- 5, the state value y(t 0 )...y(t n-1 ) A rate process model is generated (S904). This identification is performed using a probabilistic identification method such as Gaussian process regression. This can be done using the technique.

number

number

number

[0097] Here, N N is an N-dimensional normal distribution, Equation 4 is the mean of each time point, and Equation 5 is the covariance matrix between time points. Also, μ represents the mean and V represents the variance.

[0098] The CAE simulation unit 103 calculates all process design parameters to be targeted in S90. Check whether the process design parameters have been extracted in step 1 (S905). If extracted in S901 (S905: YES), the CAE simulation unit 103 If there are any process design parameters that have not been extracted in S901 after performing the process in S906, (S905: NO), the CAE simulation unit 103 selects the process design parameters. The process of S902 is repeated to select the desired one.

[0099] In S906, the CAE simulation unit 103 performs the stochastic process model generated by the above processing. The state (mean μ(t), variance V(t, t)) and process Based on the design parameters (external input data), the state distribution model 108 shown in Equation 6 is Then, the state distribution model construction process S803 is completed.

[0100] The state distribution model 108 is identified using a general system identification method, such as the ARX model. This can be achieved by using the process identification and Hammerstein-Wiener model identification. It is possible to understand how the distribution of states changes as the design parameters change.

number

[0101] Here, in Equation 6, the component (first component) related to the average μ(t) is the energy - Determine the structure of differential equations from the theoretical equations of reaction heat based on the balance law and reaction rate equations, and The parameters are estimated using the mean μ(t), variance V(t, t), and external input data. In this case, the range of values ​​that the mean μ(t) can take is restricted, so Process design parameters in unknown (i.e., unproven by CAE simulation) areas It is expected that this will also make it easier to predict the state distribution for

[0102] <State distribution generation process> Next, FIG. 11 is a flow diagram illustrating details of the state distribution generation process S805.

[0103] The state distribution generating unit 110 uses the process design parameters acquired in S804 as external inputs. The state distribution generating unit 110 obtains the state (temperature) distribution (average The initial values ​​of μ(t) and variance V(t, t) are set (S1002). The initial value of the average is, for example, the initial condition (initial temperature, etc.) at the time of manufacturing the product. The initial variance (of temperature) is set to zero since it is uniform.

[0104] The state distribution generating unit 110 uses the external inputs and Based on the initial value of the state (temperature) distribution, the state distribution model shown in Equation 6 is integrated in the time direction. By doing so, the average and variance of the state (temperature) at each time are calculated (S1003). The state distribution generation process S805 ends.

[0105] <Production quality prediction processing> FIG. 12 is a flow diagram illustrating the details of the manufacturing quality prediction process S806.

[0106] First, the product quality prediction unit 111 sets the number of sampling times Ns (S1101). The sampling frequency is the number of times the state value is sampled at each time point. The quality prediction unit 111 may, for example, be recorded in a setting file created in advance by the user 40. The number of sampling times Ns is set by reading the value stored in the

[0107] In addition, the product quality prediction unit 111 determines the total number Nt of times (sampling times) at which quality prediction is performed and Those sample times (T 1 , T 2 , …T Nt Specifically, the manufacturing quality prediction The section 111 determines the range of possible sample times ( The product quality prediction unit 111 then identifies the time range from the start of the reaction to the end of the reaction. Nt is set by accepting input of each sample time point in this range from the user 40. do.

[0108] For example, the manufacturing quality prediction unit 111 displays the state distribution model construction screen 1200. Input of values ​​1212 at each time point on the graph 1211 of the sample path is accepted from the user 40 ( In the case of the figure, Nt=4.

[0109] Then, the production quality prediction unit 111 sets the index i for counting sampling to 1. This is set (S1103), and the process of S1104 is executed.

[0110] In S1104, the manufacturing quality prediction unit 111 performs the sample time t=T 1 , T 2 , …, T N t For each point in time, the probability distribution f(y(t)) obtained in the state distribution model construction process S803 The state value y(t)(T 1 , T 2 , …, T Nt )(child Here, the temperature value is extracted (sampled).

[0111] Next, the product quality prediction unit 111 sets the index j for the sample time to 1 (S 1105), and execute the processing of S1106.

[0112] In S1106, the manufacturing quality prediction unit 111 performs the process of predicting the manufacturing quality at the sample time T j The state value y(t j ) and inputting the preset composition of the raw materials into the quality prediction model 105 to obtain the quality prediction model By executing 105, the product quality Q is calculated (predicted). Note that Q is calculated by [q 1 , q 2 , , q k ] T and q k is the quality k at the end of the process (e.g., hardness, elongation, ) value.

[0113] The production quality prediction unit 111 stores the calculated quality Q (S1107).

[0114] The manufacturing quality prediction unit 111 performs the processes of S1106 and S1107 at each time (j=1, . . . , Nt) (S1108, S1111), the quality corresponding to each sample path is calculated. Calculate and store Q.

[0115] After that, the manufacturing quality prediction unit 111 performs the processes of S1104 to S1108 for each sample (i = 1, , Ns) (S1109, S1112) for each sample Calculate and store the quality Q for

[0116] After that, the product quality prediction unit 111 converts the distribution of the quality Q calculated by the above process into a histogram. The data is output in the form of a probability density function or the like (S1110).

[0117] For example, the manufacturing quality prediction unit 111 displays the quality prediction display field of the state distribution model construction screen 1200. In 1206, a graph 1213 relating to hardness and a graph 1214 relating to elongation are displayed. The graph 1213 for hardness has the hardness value on the horizontal axis and the probability of the hardness value appearing on the vertical axis. The graph 1214 relating to elongation has the horizontal axis representing the elongation value. The vertical axis of the graph represents the probability (probability density) of the occurrence of each growth value.

[0118] <Process search processing> Next, FIG. 13 is a flow diagram illustrating the process search processing S807 in detail.

[0119] The process search unit 112 generates a data indicating the range of acceptable products (the range of required product quality). The data is acquired (S1301).

[0120] For example, the process search unit 112 may receive from the user 40 the upper and lower limits of the range of non-defective products. In this embodiment, the process search unit 112 receives the hardness q 1 and Elongation q 2 The upper limit of (q 1,max , q 2,max ) and the lower limit (q 2,max , q 2,m in ) are accepted from the user 40.

[0121] The process search unit 112 performs the quality Q calculated in the manufacturing quality prediction process S806 and the Based on the range of non-defective products obtained in 1, each quality q 1 , q 2, ... is outside the range of good products The probability (proportion) of a defect (P(q 1 ), P(q 2 ), ...) are calculated (S1302 In this embodiment, the process search unit 112 calculates the defect rate by the following formula 7. It shall be so.

number

[0122] This allows for process design that takes into account quality distribution according to the variability of conditions within the process. This allows for parameter exploration.

[0123] Based on the defect rate calculated in S1302, the process search unit 112 selects a process having a lower defect rate. An objective function that minimizes the value and a constraint condition equation related to the objective function are generated. In the embodiment, the process search unit 112 calculates an objective function for each defect rate, which is shown in the following Equation 8: A constraint equation for the state of the device shown in Equation 9 is generated.

number

number

[0124] Here, a and b are constants. θ is the external input, Tmin is the minimum temperature, and Tmax is the temperature. T is the maximum value of the temperature. Note that T may be the average value of the temperature or the maximum value of the temperature. value (used for comparison with Tmax), minimum value of temperature (used for comparison with Tmin) In addition, the process search unit 112 may automatically set the objective function and the constraint condition formula. Alternatively, the objective function and the constraint condition equation may be input by the user 40. You may do so.

[0125] The process search unit 112 searches for a process that satisfies the constraint condition equation and minimizes the value of the objective function. External input θ 0 , θ 1 , ... is changed (S1303). The change of the external input θ in the process is, for example, stored in memory the calculated value of the objective function, and the external input θ 0 , θ 1 , ..., and apply the external input θ 0 , θ 1 , change ... Good.

[0126] The objective function and constraint condition formulas described here are merely examples, and the explanatory variables Anything that optimizes (quality or defect rate) is sufficient.

[0127] In addition, the shape of the reactor, the shape of the stirring blade, the cooling method, the combination of raw materials, and the mixing ratio We have explained the case where the process design parameters are searched for with the rate fixed, but these are also search targets. In this case, the reactor shape, agitator shape, cooling method, and raw material combination are generally Since the variables are distributed, the optimization problem shown in S1303 becomes a mixed integer programming problem. , it will be performed using a different solver than that shown in this embodiment.

[0128] (Process search screen) FIG. 14 shows the process displayed on the design support system 1 in the process search process S807. FIG. 14 is a diagram showing an example of a search screen 1400.

[0129] The process exploration screen 1400 displays the acceptable product range (upper and lower limits) for each type of quality. ) from the user 40, and an objective function is displayed. 14. The objective function definition unit 1402 includes an objective function definition input from a user 40, and a constraint condition formula The information required for generation (e.g., the range (upper and lower limits) of the state (temperature)) is displayed or controlled. A constraint definition unit 1403 that accepts input of information required for generating a constraint expression from a user 40. and a process search execution unit 1404 selected by the user when performing a search calculation. Prepare.

[0130] The process search screen 1400 also displays the distribution of each quality of the product calculated during the search calculation. A graph display section 1405 displays a graph of the number of defects, and a graph display section 1406 displays the defect rate calculated during the search calculation process. A defect rate display section 1406 is displayed, and the process design parameters calculated by the search calculation are displayed. and a process design parameter display section 1407 on which values ​​are displayed.

[0131] The graph display section 1405 shows the quality of each type of product on the horizontal axis, A graph 1408 is displayed with the probability (probability density) of that quality value on the vertical axis. Graph 1408 displays the range (upper and lower limits) of each quality.

[0132] <State distribution model regeneration process S809> FIG. 15 is a flow diagram illustrating details of the state distribution model regeneration process S809.

[0133] After the process search process S807 is executed, the user refers to the process search screen 1400. This allows us to confirm the expected quality of the manufactured product and to set the parameters that should be set in large-scale testing. Determine the process design parameters (e.g., specifications of the actual plant equipment) to determine the product Test production or actual production is carried out (S1501).

[0134] Then, the process exploration unit 112 searches for the manufactured product and the adopted process design parameters. The contents of the data are recorded in the manufacturing performance data 23 below (S1502). For example, The search unit 112 may receive data input from the user 40 or may search for the data in a predetermined experiment management system. The reactor 30 may be automatically fed with data from a DCS (distributed control system) that operates the reactor 30. You may obtain it.

[0135] (Manufacturing performance data) FIG. 16 is a diagram showing an example of the manufacturing performance data 23. The manufacturing performance data 23 includes data The ID 231, the type of raw material of the product 232, and the specifications of the equipment used to manufacture the product 233, which is the model data, and 234, which is the data on the operating conditions during the manufacture of the product. The data on the quality of the manufactured product 235 and the quality of the manufactured product 34 are stored in association with each other.

[0136] The product type 232 is data on the composition (mixture) of raw materials. The operation data 234 is For example, a target value given to a control device, a manipulated variable given to a process by a control device, or a process In this embodiment, the target value is the time series data of the controlled variable of the reactor. The target temperature, the operation amount is the jacket cold water temperature for temperature control, and the control amount is the temperature inside the reactor ( (i.e., state).

[0137] Quality 235 is not necessarily specified for each production of a product, but is a standard for the products inspected. Generally, there is variation in quality test values, so it is best to store them with a range. It is possible.

[0138] In the example of the process search screen 1400 in FIG. 14, the range of hardness among the product qualities is , it is out of the acceptable range set in S1301. In this case, regenerate the state distribution model. There is a possibility that the state distribution model 108 generated in process S809 is unreliable.

[0139] Next, as shown in FIG. 15, the process search unit 112 Based on this or automatically, the process design parameters are modified (S1503).

[0140] Specifically, first, the process search unit 112 searches the manufacturing performance data 23 and the The process design parameters were added to the plant experiment data22, and the plan with the added data was Based on the experimental data 22, the state distribution model 108 is retrained to create a new state distribution model 1 08. The process search unit 112 regenerates the new state distribution model 108. Each time, a process search process S807 is executed.

[0141] This allows the prediction of the state distribution model108 and product quality while actually repeating the manufacturing process. The accuracy of the measurement is improved, and the user 40 can perform appropriate measurements without repeating trial-and-error simulations. This enables appropriate design.

[0142] Second, based on plant experimental data22 rather than CAE simulation models109 A process for determining process design parameters when generating the state distribution model 108 (hereinafter (hereinafter referred to as the second process design parameter determination process) will be described.

[0143] <Process design parameter determination second step> FIG. 17 is a flow diagram illustrating an example of the second process design parameter determination processing.

[0144] The process of S1701 is similar to S801 of the first process design parameter determination process.

[0145] The design support system 1 then uses data from the results of the plant experiments (e.g., thermometer 51 The measurement data of 55 to 55) is acquired from the reactor 30, and the acquired data is referred to as plant experiment data 2 2 (S1702).

[0146] The state distribution model construction unit 104 performs the same procedure as in S803 based on the plant experiment data 22. Based on the above, a state distribution model 108 is generated (S1703). This process is similar to the processes in S804 to S808.

[0147] Based on both CAE simulation models109 and plant experimental data22 In this case, the time granularity of both data may be made to match. .

[0148] As described above, the design support system 1 of this embodiment is a reactor (specifically, an experimental The spatial domain of the reactor 30 used in the simulation or the spatial domain set in the coupled simulation) The temperature at each location in the reactor is calculated based on the temperature at each location in the reactor when the process is carried out on the raw material. The values ​​(average and variance) representing the heterogeneity of the state in the sample were calculated, and the calculated values ​​representing the heterogeneity were compared. Based on the data defining the process, process design parameters are input, and the process is The data (mean and variance) showing the heterogeneity of the conditions in the reactor after the process are output. Then, the design support system 1 generates a state distribution model 108 based on the set process design. By inputting the parameter values ​​into the state distribution model 108, the process Obtain data values ​​that represent the non-uniformity of the state in the reactor after the run, and From the data (mean and variance) representing the values, multiple samplings of possible states in the reactor are performed. The data from the sampled state and the process design parameter values ​​are used to create a quality prediction model. By inputting information about each state in the 105, the product quality parameters (hardness and elongation) can be Then, the design support system 1 calculates the product quality based on each quality parameter. The values ​​of the process design parameters are changed so that the quality of falls within a predetermined range.

[0149] That is, the design support system 1 samples process design parameters and random variables. The quality of the product is determined by inputting the value of each state data into the quality prediction model 105 and executing it. Based on the quality data, appropriate process design parameters are determined. Here, the value of can be determined by using sampling based on random variables. For example, the state distribution in a large process space assumed in a large-scale test can be easily and The quality of the product based on this state distribution can be expressed appropriately. For example, a quality prediction model 105 based on a small-scale experiment is used to predict the quality of the process. Design parameters can be determined.

[0150] In this way, according to the design support system 1 of this embodiment, The process design is to obtain a product of appropriate quality in the process of producing the product to be carried out. For example, experiments or simulations performed in each process can be While minimizing the number of iterations and re-experiments on small-scale processes, we aim to achieve quality KPIs. It is now possible to design processes that are within a specified range, and this reduces the risk of scale-up operations. This leads to reduced work time.

[0151] In addition, the design support system 1 of this embodiment is capable of calculating the composition of raw materials, the equipment used in the process, At least one of the specifications and the operating conditions of the equipment used in the process is input. A state distribution model 108 is generated.

[0152] This allows for the creation of a state distribution that can adequately predict changes in the distribution of states (such as temperature) in the process. A cloth model 108 can be generated.

[0153] In addition, the design support system 1 of this embodiment performs process setting in the state distribution model 108. Quality prediction model for processes carried out on a smaller scale than processes related to design parameters10 Remember 5.

[0154] This allows the manufacturing of products to be carried out in a variety of ways, from small-scale laboratory experiments to plant experiments. Even when scaling up to large-scale testing, the quality of the manufactured product must be verified at each experimental level. The prediction can be made appropriately with the bell.

[0155] In addition, the design support system 1 of this embodiment acquires the state distribution model 108 based on the execution thereof. Based on the quality parameters of each state, the variance (probability density) of the quality of the deliverables and the quality Calculates the probability (defective rate) within a specified range and outputs the calculated probability distribution and defective rate information. (Graph display area 1405 of process search screen 1400).

[0156] This allows accurate evaluation of the overall quality of products manufactured in large quantities.

[0157] In addition, the design support system 1 of this embodiment acquires the state distribution model 108 based on the execution thereof. Based on each quality parameter in each state, the quality of the product obtained from the raw materials falls within a specified range. The process design parameters are set to satisfy the specified constraints (operating conditions, constraint conditions). and the range of quality of the deliverables when the process design parameters are changed. Information is output (graph display section 1405 of process search screen 1400 or defect rate display section 1 406).

[0158] This allows the user 40 to understand the quality of the product to be manufactured while satisfying the necessary constraints. It is possible.

[0159] In addition, the design support system 1 of this embodiment can link the contents of the processes performed on the raw materials. The inputs to the simulation model allow the process to be analyzed during and after execution. The state of each position in the reactor is acquired, and based on the acquired state of each position inside the reactor, the non-uniformity The values ​​(mean and variance) representing the nonuniformity are calculated, and a state distribution model is created based on the calculated values ​​representing the nonuniformity. Generate rule 108.

[0160] In this way, state separation is performed based on the results of coupled simulations assuming large-scale tests, etc. By generating the cloth model 108, the state distribution model 108 can be easily generated. .

[0161] In addition, the design support system 1 of this embodiment performs the following operations during the execution of a predetermined process (plant experiment): Based on the state of each position in the reactor after the run, values ​​representing the non-uniformity (mean and variance) are calculated. A state distribution model 108 is generated based on the calculated value representing the non-uniformity.

[0162] In this way, the state distribution model 10 was developed based on the actual large-scale test (plant experiment). 8, a state distribution model 108 with appropriate content can be generated.

[0163] In addition, the design support system 1 of this embodiment calculates a state distribution model 10 by Gaussian process regression. Generates 8.

[0164] This allows us to create a state distribution model that calculates the probability distribution that is expected to actually occur. 08 can be generated.

[0165] In addition, the design support system 1 of this embodiment performs testing using the changed process design parameters. Obtain data on the quality of the product after experimental or actual manufacturing is carried out, and Then, new process parameters are generated, and a state distribution model is created based on the generated parameters. Regenerate 108.

[0166] In this way, the state distribution model 108 can be regenerated based on the manufacturing results of the actual product. Thus, the prediction accuracy of the state of the state distribution model 108 can be improved.

[0167] In addition, the design support system 1 of this embodiment performs the following on the basis of the regenerated state distribution model 108: , the values ​​of the process design parameters previously set are further changed.

[0168] In this way, for example, the state distribution model 11 is generated while the user 40 actually repeats the manufacturing of the product. 08 and the accuracy of product quality predictions has improved, and trial and error simulations have been repeated This makes it possible to design appropriate process design parameters without any need for additional calculations.

[0169] In addition, the design support system 1 of this embodiment includes a reaction process to be performed on the raw material. Handle the process.

[0170] This allows consideration of the various reaction processes that may produce the product.

[0171] In addition, the design support system 1 of this embodiment defines the reaction process of the raw materials distributed in the reactor. The data for determining the reaction process are input, and the original data for each position in the reactor during and after the reaction process are A coupled simulation model including a reaction characteristic model 106 that outputs the state of the material is generated. do.

[0172] This allows the creation of coupled simulation models that take into account various reaction processes that may be used to manufacture products. It is possible to generate a del.

[0173] In addition, the quality prediction model 105 of this embodiment uses data representing the composition of raw materials (mixture ratio, etc.) The design support system 1 further inputs data representing the composition of the raw materials. In addition, the quality parameters of the product are obtained by inputting them into a quality prediction model 105, and the process Change the process design parameters and the composition of the raw materials.

[0174] In this way, by using the composition of raw materials as a parameter for each model, it is possible to obtain information on the quality of the product. This allows optimization of the feedstock composition as well as process design parameters.

[0175] The present invention is not limited to the above-described embodiment, and may be modified without departing from the spirit and scope of the present invention. The present invention can be implemented using any desired components. The above-described embodiments and modifications are merely examples. The present invention is not limited to these contents as long as the characteristics of the invention are not impaired. Although various embodiments and modifications have been described above, the present invention is not limited to these. Other embodiments that are conceivable within the scope of the technical concept of the present invention are also included in the present invention. Included in the range.

[0176] For example, some of the hardware included in each device of this embodiment may be provided in another device. .

[0177] In addition, each program of each device may be provided in another device, or a certain program may be provided in multiple devices. It may consist of several programs, or several programs may be integrated into one program. Good too.

[0178] In addition, operating conditions other than those described in this embodiment can be set as process design parameters. For example, the target value in the manufacturing performance data 23 (such as the target temperature of the reactor) may be The operating conditions are generally defined as a time series, but the following equations can be used: Since the state distribution model 108 defined by Equation 8 can input external inputs at each time, ,Explore process design parameters by including various operating conditions in the process design parameters. It is possible. [Explanation of symbols]

[0179] 1 Design support system, 105 Quality prediction model, 106 Reaction characteristic model, 108 State Morphogenetic distribution model

Claims

1. Data representing the state and the process performed on the raw material at said state are input. and outputting data representative of the quality of the product obtained from the raw material after the process. A storage device that stores a quality prediction model; and The state of each position in a spatial region when a process is performed on raw materials distributed in the spatial region. and representing the state heterogeneity in the spatial domain based on each of the acquired information. a value is calculated, and based on the calculated value representing the non-uniformity and data defining the process, Process design is data that defines the process to be performed on raw materials present in a spatial domain. The parameters are input and the state in the spatial domain during and after the process is executed is determined. A state distribution model construction process that generates a state distribution model that outputs data that represents the non-uniformity of and, The values ​​of the process design parameters are set, and the set values ​​of the process design parameters are By inputting the state distribution model, the values ​​of the process design parameters set above are displayed. During or after the process is executed, the process design parameters are set to the values. A state distribution model execution process for acquiring data values ​​representing state heterogeneity in the intermediate region; From the acquired data values ​​representing the non-uniformity of the state in the spatial domain, A plurality of possible states are specified, and each data representing each specified state and the set profile are stored in the and inputting the process design parameter values ​​into the quality prediction model for each of the states. Thus, after the process indicated by the set process design parameter value, A manufacturing quality prediction process for obtaining quality parameters of the deliverables obtained from the manufacturing process and the ... Based on each of the acquired quality parameters, the quality of the product obtained from the raw material falls within a predetermined range. Parameter search to change the values ​​of the set process design parameters so that the set process design parameters fall within the range. A control device for executing the process A design support system equipped with:

2. The control device, in the state distribution model construction process, The composition of the raw materials, the specifications of the equipment used in the process, or the and the operating conditions of the equipment to be used. R, The design support system according to claim 1 .

3. The storage device stores a process design parameter in the state distribution model. A quality prediction model for a process that is performed on a smaller scale than the The design support system according to claim 1 .

4. The control device, in the parameter search process, Calculate the variability of the quality of the deliverables and the probability that the quality is within a predetermined range based on the data. and outputting the calculated variability and probability information. The design support system according to claim 1 .

5. The control device, in the parameter search process, Based on the meter, the quality of the product obtained from the raw material falls within a predetermined range and within a predetermined constraint. The values ​​of the set process design parameters are changed so as to satisfy outputting information regarding the range of quality of the deliverables when the parameter values ​​are changed; The design support system according to claim 1 .

6. The storage device stores data defining processes to be performed on ingredients distributed in a spatial domain. is input, and the state of each position in the spatial domain during or after the process is executed is output. storing a spatial process model to be The control device includes: The contents of the process to be performed on the raw materials distributed in the spatial domain are set, and the set process is By inputting the contents of the process into the spatial process model, the execution of the set process can be a spatial process model execution unit that acquires the state of each position in the spatial domain during or after execution; , A value representing the non-uniformity is calculated based on the state of each position in the spatial region obtained. The value representing the heterogeneity and the data representing the process entered into the spatial process model A state distribution model construction process for generating the state distribution model based on the data. The design support system according to claim 1 , further comprising:

7. A device that has a predetermined spatial area and is capable of carrying out a predetermined process on a raw material under predetermined conditions. The container further comprises: The control device, during the execution of the predetermined process in the state distribution model construction process, or calculating a value representing the non-uniformity based on the state of each position in the spatial region of the container after the execution. Based on the calculated value representing the non-uniformity and the data representing the contents of the predetermined process, , generating the state distribution model The design support system according to claim 1 .

8. The control device performs state distribution model construction processing by Gaussian process regression. The design support system according to claim 1 , wherein the system generates a cloth model.

9. The control device, in the state distribution model construction process, performing a process on said feedstock as indicated by said altered process design parameters; and obtaining quality data of the deliverables obtained by the above-mentioned method. generating new data defining the process to be performed, and comparing the generated new data with the calculation and regenerating the state distribution model based on the obtained non-uniformity value. The design support system according to claim 1 .

10. The control device controls the set process design based on the regenerated state distribution model. The design support system according to claim 9 , further comprising the step of changing values ​​of parameters.

11. The design of claim 1 , wherein the process comprises a reaction process performed on the raw material. Support system.

12. The control device receives data defining a reaction process of raw materials distributed in a spatial region. The state of the raw material at each position in the spatial region during or after the reaction process is output. performing a spatial process model construction process to generate a spatial process model to be The design support system according to claim 6.

13. The storage device stores a quality prediction model into which data representing the composition of the raw material is further input. Remember, The control device includes: In the manufacturing quality prediction process, data representing the composition of the raw materials is further The quality parameters of the deliverables are input for each of the states into the data Obtain information about the status, In the parameter search process, the set process design parameter values ​​and Changing the composition of raw materials; The design support system according to claim 1 .

14. A control device and data representing a state and a process performed on the raw material in the state. data representing the quality of the product obtained from the raw materials after the process is input; and a storage device for storing a quality prediction model from which data is output. A method of support comprising: The control device, The state of each position in a spatial region when a process is performed on raw materials distributed in the spatial region. and representing the state heterogeneity in the spatial domain based on each of the acquired information. a value is calculated, and based on the calculated value representing the non-uniformity and data defining the process, Process design is data that defines the process to be performed on raw materials present in a spatial domain. A parameter is input to represent the heterogeneity of the state in the spatial region after the process. A state distribution model construction process for generating a state distribution model for outputting data; The values ​​of the process design parameters are set, and the set values ​​of the process design parameters are By inputting the state distribution model, the values ​​of the process design parameters set above are displayed. After the process, the state in the spatial domain related to the process design parameter for which the value was set is A state distribution model execution process for acquiring data values ​​representing state heterogeneity; From the acquired data values ​​representing the non-uniformity of the state in the spatial domain, A plurality of possible states are specified, and each data representing each specified state and the set profile are stored in the and inputting the process design parameter values ​​into the quality prediction model for each of the states. Thus, after the process indicated by the set process design parameter value, A manufacturing quality prediction process for obtaining quality parameters of a product obtained from each of the states; Based on each of the acquired quality parameters, the quality of the product obtained from the raw material falls within a predetermined range. Parameter search to change the values ​​of the set process design parameters so that the set process design parameters fall within the range. Execute the process Design support method.

Images