Prediction device, prediction method, and program

Abstract

In this invention, prediction accuracy for a process including reactions in a chemical plant is improved. This prediction device comprises a process data processing unit for carrying out certain processing of process data acquired from a chemical plant and a prediction model generation unit for generating a prediction model having learned features of the process data acquired from the chemical plant on the basis of causal relationship information defining combinations of first process data as an explanatory variable, and second process data as a response variable or a value corresponding to the second process data from among the process data acquired from the chemical plant or the process data processed by the process data processing unit. Further, the process data processing unit determines a value corresponding to a reaction rate of a processing object in a predetermined period using the process data.

Description

[Technical field]

[0001] The present disclosure relates to a prediction device, a prediction method, and a program. [Background technology]

[0002] Conventionally, there have been proposed techniques for predicting the quality of products in a manufacturing process and controlling operations based on the prediction. For example, a prediction system has been proposed that predicts the characteristic values ​​of a product during manufacturing and calculates the control conditions of a subsequent manufacturing process based on the prediction results (Patent Document 1). This system includes a database that stores data measured in the manufacturing process and / or data indicating the state of the manufacturing process for each lot, a mathematical model creation unit that creates a mathematical model of the manufacturing process using the data stored in the database, a product characteristic prediction unit that predicts the characteristic values ​​of the product for the lot being manufactured by inputting actual values ​​for the processed process and representative values ​​obtained based on past lots for the unprocessed process into the mathematical model, and an optimal manufacturing condition calculation unit that calculates optimal manufacturing conditions for the unprocessed process to be controlled according to the prediction results by the product characteristic prediction unit, and the prediction by the product characteristic prediction unit and the calculation by the optimal manufacturing condition calculation unit are performed for each predetermined controlled process for the lot being manufactured. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6477423 [Patent Document 2] Patent No. 5751045 [Patent Document 3] JP 2018-120343 A [Patent Document 4] JP 2001-106703 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, there is a problem in that it is difficult to improve prediction accuracy, particularly in processes that include reactions in chemical plants. Therefore, an object of the present technology is to improve prediction accuracy for processes that include reactions in chemical plants. [Means for solving the problem]

[0005] The prediction device according to the present disclosure includes a process data processing unit that performs a predetermined processing process on process data obtained from a chemical plant, and a prediction model creation unit that creates a prediction model that learns the characteristics of the process data obtained from the chemical plant based on causal relationship information that defines a combination of a first process data as an explanatory variable and a second process data as a target variable or a value corresponding to the second process data, among the process data obtained from the chemical plant or the process data processed by the process data processing unit. In addition, the process data processing unit uses the process data to obtain a value corresponding to a reaction rate of a process target in a predetermined period.

[0006] In general, reactions are nonlinear and influenced by factors such as temperature, so it has been difficult to predict the characteristics of process data based on, for example, peak values, difference values, integral values, etc. of the process data. Values ​​according to the reaction rate as described above become explanatory variables that correlate with, for example, the amount of reaction per unit time. By creating a prediction model using values ​​according to the reaction rate, it is possible to improve the prediction accuracy for processes including reactions in chemical plants.

[0007] The value according to the reaction rate may be an integral value of the reaction rate. The value according to the reaction rate may be calculated using the Arrhenius equation, and the frequency factor and activation energy of the reaction rate constant may be determined according to the type of chemical reaction and the target to be treated in the chemical plant. Specifically, such values ​​may be adopted.

[0008] In addition, the process data processing unit may further obtain an integral value, a differential value, an average value, a maximum value, or a minimum value of the process data in a predetermined period, or an instantaneous value of the process data at a predetermined time point, depending on the type of the process data. By processing the process data in this manner, the prediction accuracy of the prediction model can be improved.

[0009] The prediction model may have a hierarchical structure including multiple prediction formulas and a second prediction formula including a predicted value calculated by the first prediction formula as an explanatory variable. For example, multiple causal relationships can be expressed by a function in this way.

[0010] The value according to the second process data may be a value obtained by sampling a plurality of second process data by a reduction method, and the range of acquisition timing of the first process data in the chemical plant and the calculation timing of the value according to the second process data may be associated based on the residence time of the processing object in the chemical plant, and the prediction model creation unit may create the prediction model. When the processing object is continuously processed in the chemical plant, the accuracy of prediction can be improved by appropriately associating the acquisition timing of the process data as the explanatory variable with the acquisition timing of the process data as the target variable.

[0011] The chemical plant may also perform a batch process in which the processing object is sequentially processed for each predetermined processing unit, followed by a continuous process in which the processing object is continuously processed, and the range of the completion timing of the batch process and the calculation timing of the value corresponding to the second process data may be associated based on the residence time of the processing object in the chemical plant, and the prediction model creation unit may create a prediction model. Even when the batch process and the continuous process are performed successively, the accuracy of prediction can be improved by appropriately associating the process data as the explanatory variable with the process data as the target variable.

[0012] The contents described in the means for solving the problem can be combined as much as possible without departing from the problem and technical idea of ​​the present disclosure. The contents of the means for solving the problem can be provided as a device such as a computer or a system including multiple devices, a method executed by a computer, or a program executed by a computer. A recording medium for storing the program may be provided. Effect of the Invention

[0013] According to the disclosed technology, it is possible to improve the prediction accuracy of product characteristic values. [Brief description of the drawings]

[0014] [Figure 1] FIG. 1 is a diagram illustrating an example of a system according to an embodiment. [Diagram 2] FIG. 2 is a schematic diagram showing an example of a process carried out by equipment included in a plant. [Diagram 3] FIG. 3 is a diagram for explaining an example of process data in a batch process. [Figure 4] FIG. 4 is a diagram showing an example of information registered in advance in the knowledge base. [Diagram 5] FIG. 5 is a diagram showing an example of a tag attribute table for a batch process created based on the knowledge base. [Figure 6] FIG. 6 is a diagram showing an example of a tag combination table created based on the knowledge base. [Figure 7] FIG. 7 is a diagram illustrating an example of a logic tree showing a configuration of a prediction model. [Figure 8] FIG. 8 is a diagram for explaining an example of process data in a continuous process. [Figure 9] FIG. 9 is a diagram for explaining the association between samples in the process inspection of a continuous process and serial numbers in a batch process. [Figure 10]FIG. 10 is a diagram for explaining the residence time from the sensor position in a continuous process to the sampling position in the process inspection. [Figure 11] FIG. 11 is a diagram showing an example of a tag attribute table for a continuous process, which is created based on a knowledge base. [Figure 12] FIG. 12 is a diagram illustrating an example of a logic tree showing a configuration of a prediction model. [Figure 13] FIG. 13 is a block diagram illustrating an example of a configuration of a prediction device. [Figure 14] FIG. 14 is a process flow diagram illustrating an example of a prediction process executed by the prediction device. [Figure 15] FIG. 15 is a diagram showing an example of a write array for a batch process. [Figure 16] FIG. 16 is a process flow diagram showing an example of the processing. [Figure 17] FIG. 17 is a diagram illustrating an example of a batch arrangement. [Figure 18] FIG. 18 is a diagram for explaining the process data or the predicted value thereof to be substituted into the prediction formula. [Figure 19] FIG. 19 is a diagram illustrating another example of process data to be substituted into the prediction formula or its predicted value. [Figure 20] FIG. 20 is a diagram showing an example of a write array that holds data whose type is "continuous." [Figure 21] FIG. 21 is a diagram showing an example of a write array for holding data whose type is "batch." [Figure 22] FIG. 22 is a diagram showing an example of a combination ID data array that holds process data after processing in a continuous process. [Diagram 23] FIG. 23 is a process flow diagram illustrating an example of a control process executed by the prediction device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, an embodiment of a prediction device will be described with reference to the drawings.

[0016] <Embodiment> FIG. 1 is a diagram illustrating an example of a system according to this embodiment. The system 100 includes a prediction device 1, a control station 2, and a plant 3. The system 100 is, for example, a distributed control system (DCS) and includes a plurality of control stations 2. That is, the control system of the plant 3 is divided into a plurality of sections, and each control section is controlled in a distributed manner by the control station 2. The control station 2 is an existing facility in the DCS, and receives status signals output from sensors and the like included in the plant 3, and outputs control signals to the plant 3. Then, actuators such as valves and other devices included in the plant 3 are controlled based on the control signals.

[0017] The prediction device 1 acquires a state signal (process data) of the plant 3 via the control station 2. The process data includes temperature, pressure, flow rate, etc. of the raw material or intermediate product to be processed, and set values ​​that determine the operating conditions of the equipment of the plant 3. The prediction device 1 also creates a prediction model based on a knowledge base that stores the correspondence between the assumed causes and the influences that appear as, for example, abnormalities. For example, a prediction formula for quality or cost is created using causal relationship information that defines a combination of process data (also called a factor system) that is an explanatory variable and process data (also called an influence system) that is an objective variable, which is created based on the knowledge base. The prediction device 1 can then predict characteristic values ​​that represent the quality of the product using the prediction formula and the process data, or predict characteristic values ​​of the product when the operating conditions of the plant 3 are changed. The prediction device 1 may also determine operating conditions that satisfy, for example, predetermined conditions for quality and cost. The prediction device 1 may also determine operating conditions that transition to a stable state in response to a change in state that appears as an influence, or determine operating conditions that cause the product to satisfy predetermined requirements, using the prediction formula and the process data. Moreover, the prediction device 1 may use an analysis value obtained in a predetermined process inspection as a response variable, instead of the process data of the influence system.

[0018] FIG. 2 is a schematic diagram showing an example of equipment provided in a plant or a process performed by the equipment. That is, the process includes a production process, which is a treatment, and a process equipment, which is an apparatus. In this embodiment, the process may include a batch process 31 and a continuous process 32. In the batch process 31, the processing target is sequentially processed for each predetermined processing unit, and processing such as receiving, holding, and discharging raw materials into each device is performed in sequence. In the continuous process 32, the processing target that is continuously introduced is continuously processed, and processing such as receiving, holding, and discharging raw materials is performed in parallel. In addition, the process may include multiple series 33 that perform the same processing in parallel.

[0019] The equipment for performing each process includes, for example, a reactor, a distillation apparatus, a heat exchanger, a compressor, a pump, a tank, and the like, which are connected via piping. Sensors, valves, and the like are provided at predetermined positions of the equipment and piping. The sensors may include a thermometer, a flow meter, a pressure gauge, a level gauge, a concentration meter, and the like. The sensors monitor the operating status of each equipment and output a status signal. The sensors provided in the plant 3 are each provided with a "tag," which is identification information for identifying each sensor. The prediction device 1 and the control station 2 manage the input / output signals to each equipment based on the tag.

[0020] <Batch process> FIG. 3 is a diagram for explaining an example of process data in a batch process. The left column of FIG. 3 shows a part of the process of the batch process 31 shown in FIG. 2. Specifically, the process includes a shredder 301, a cyclone 302, a pretreatment 303, a precooler 304, and a reactor 305. These processes are classified into a pretreatment process, a precooling process, and a reaction process. The right column of FIG. 3 shows an example of process data acquired in each process. In the pretreatment process, time-series data is acquired from sensors with tags 001 and 002. In the precooling process, time-series data is acquired from sensors with tags 003 and 004. In the reaction process, time-series data is acquired from sensors with tags 005, 006, and 007. In the batch process, a processing object associated with a serial number (also called a "production number") is intermittently processed. That is, the serial number is identification information for identifying processing objects to be processed collectively in the batch process. As shown in Fig. 3, time-series data on the processing object associated with the subsequent serial numbers is obtained as time passes. Note that time points t1 and t2 will be described later.

[0021] FIG. 4 is a diagram showing an example of information registered in advance in a knowledge base. The knowledge base is assumed to be stored in advance in a storage device of the prediction device 1. The table in FIG. 4 includes columns corresponding to each sensor and rows indicating the cause of the change in the output value of each sensor. That is, values ​​are registered in columns corresponding to sensors affected by causes such as "increase in the amount of auxiliary raw material A" and "decrease in the amount of auxiliary raw material A" shown in each row. The values ​​are registered with a positive or negative sign corresponding to the fluctuation in the output value of the sensor. Note that the combination of cause and effect is not necessarily one-to-one. That is, multiple causes may be associated with one effect, and the same cause may be associated with multiple effects.

[0022] The knowledge base is to be created in advance by the user based on, for example, HAZOP (Hazard and Operability Study). HAZOP is a method for comprehensively listing, in association with each other, the detection means at the monitoring points by the instrumentation devices constituting the plant, the control range (upper and lower limit thresholds, which are alarm setting points), deviations from the control range (abnormality, modulation), enumeration of assumed causes of deviations from the control range, logic (detection means) for determining which assumed cause caused the deviation, the impact of the deviation, measures to be taken when the deviation occurs, and actions for the measures. Note that the knowledge base may be created based on, in addition to HAZOP, FTA (Fault Tree Analysis), FMEA (Failure Mode and Effect Analysis), ETA (Event Tree Analysis), or methods based on these, methods similar to these, contents extracted from the results of interviews with operators, and contents extracted from work standards and technical standards.

[0023] 5 is a diagram showing an example of a tag attribute table for a batch process created based on a knowledge base. The tag attribute table defines a method for processing data obtained from a sensor corresponding to each tag. The tag attribute table may be a so-called database table or a file in a predetermined format such as CSV (Comma Separated Values). The tag attribute table is created in advance by a user and is read out by the prediction device 1.

[0024] The tag attribute table includes attributes such as tag, series, variety, primary processing, smoothing, and optimization of operating conditions. In the tag field, a tag that is identification information of a sensor is registered. In the series field, identification information for identifying a series of a process is registered. In the variety field, a type of a processing target is registered. In the prediction process, the prediction device 1 may set a parameter according to, for example, the variety of a processing target. In the primary processing field, information indicating a processing method of an output value of a sensor is registered. In addition, the attributes of the primary processing further include attributes of a batch process, a method, and a data interval. In the batch process field, identification information indicating a subdivided process in a batch process is registered. In the method field, information indicating a type of a data processing method is registered. The types include "instantaneous value", "average", "integral", "differential", "difference", "maximum", "minimum", "thermal history", and "none". "Instantaneous value" represents a value at the start or end of a data interval. "Average" represents an average value obtained by dividing the value of a period specified in a data interval by the number of data. "Integral" represents the total value of the values ​​in the period specified by the data interval. "Differential" represents the differential coefficient at the start or end of the data interval. "Difference" represents the difference between the values ​​at the start and end of the data interval. "Maximum" represents the maximum value in the period specified by the data interval. "Minimum" represents the minimum value in the period specified by the data interval. "Thermal history" is an example of the degree of progress of the reaction, and represents, for example, the integral value of the reaction rate in the period specified by the data interval. "None" is attached to the tag at the end of the batch and represents that no processing is performed. The attributes of the data interval further include attributes of start and end, and information indicating the timing of acquiring the output value of the sensor is registered in at least one of the start and end fields. The information indicating the timing may be defined, for example, based on a predetermined step that is determined in advance for each subdivided process. In the smoothing field, information indicating the necessity of performing a predetermined smoothing process on the data is registered. The attributes of the operating condition optimization further include the attributes of adjustment / monitoring, cost impact, management range, and setting range. In the adjustment / monitoring field, a type indicating whether the item is to be adjusted or monitored in the optimization process is registered.The cost impact field registers a predetermined cost per unit that is affected when an adjustment is made in the optimization process. The management range attribute further includes upper and lower limit attributes, and the upper and lower limit fields register information indicating an allowable range of the sensor output value. The setting range attribute further includes upper and lower limit attributes, and the upper and lower limit fields register information indicating a target range of the sensor output value. According to the information registered in the operating condition optimization field as described above, the prediction device 1 may perform multi-objective optimization or single-objective optimization.

[0025] 6 is a diagram showing an example of a tag combination table created based on a knowledge base. The tag combination table is information representing causal relationships obtained from the knowledge base, and defines combinations of process data serving as explanatory variables and process data serving as objective variables. The tag combination table may also be a so-called database table, or may be a file in a predetermined format such as CSV. The tag combination table is also created in advance by a user, and is read out by the prediction device 1.

[0026] The tag combination table includes attributes of combination ID, tag, factor / influence, causal relationship, learning period, and dependency relationship. In the field of combination ID, identification information indicating a group of causal relationships is registered. In the field of tag, a tag that is identification information of a sensor is registered. In the field of factor / influence, a type indicating whether it is a factor system or an influence system in the causal relationship (in other words, whether it is an explanatory variable or a target variable) is registered. In the field of causal relationship, a positive or negative type indicating a sign constraint for changing the output value of the factor system corresponding to the tag is registered in order to change the value of the influence system in a positive or negative direction. Therefore, the value of the causal relationship is registered in a record in which "factor" is registered in the field of factor / influence. In this embodiment, the prediction device 1 creates a prediction model by imposing a constraint (called a "sign constraint") that has a certain correspondence between the positive and negative directions of the change in the value of the factor system and the positive and negative directions of the change in the value of the influence system. For example, a prediction model is created so that the positive and negative change direction of the target variable is determined according to the positive and negative change direction of the explanatory variable. That is, the sign registered in the causal relationship field indicates whether the output value of the sensor corresponding to the tag of the record should be changed in a positive or negative direction in order to change the value of the influence system in a predetermined direction, either positive or negative. Also, the learning period field registers information for specifying the period of the process data used to create the prediction model. The information may be, for example, the number of the most recent serial numbers.

[0027] Based on the tag attribute table and the tag combination table as described above, the prediction device 1 creates a prediction model. FIG. 7 is a diagram showing an example of a logic tree showing the configuration of a prediction model. Each rectangle represents an output value or a predicted value of a sensor corresponding to a tag. The prediction model includes a prediction formula for predicting an output value of a sensor corresponding to an influential system connected by an arrow based on an output value of a sensor in an upstream process (left side in FIG. 7). The prediction model has a hierarchical structure including a plurality of prediction formulas, and includes other prediction formulas including a predicted value by a certain prediction formula as an explanatory variable. The prediction formula is created by combining records with the same combination ID in the tag combination table shown in FIG. 6. For example, a predetermined prediction formula is created using an output value or a predicted value of a sensor corresponding to a tag in which "factor" is registered in the factor / influence field as an explanatory variable, and an output value of a sensor corresponding to a tag in which "influence" is registered in the factor / influence field or a characteristic value of a product, for example, an analysis value obtained by a process inspection, as an objective variable.

[0028] Specifically, the prediction formula can be expressed, for example, by the following formula (1). Y(t) = a1(t)·x1(t)+a2(t)·x2(t)+···+a n (t)·x n (t)+a ar (t)·Y(t-1)+C (1) In addition, t is a value according to the serial number, Y(t) is a predicted value of the influence system, x(t) is the output value of the sensor of the factor system or its predicted value, a(t) is a coefficient of the factor system, and a ar (t) is the coefficient of the autoregressive term, and C is a constant term. The number of terms corresponding to the factor system is equal to the number of output values ​​of the original sensors connected by arrows in FIG. 7. The autoregressive term is a predicted value or an actual measurement value relating to a past serial number. The number of autoregressive terms is not limited to one, and the prediction formula may include autoregressive terms relating to a plurality of recent serial numbers.

[0029] Furthermore, the prediction device 1 performs a learning process for each serial number in a batch process to update the coefficients of the prediction formula. The coefficients are determined by performing a regression analysis using, as learning data, process data corresponding to a predetermined number of recent serial numbers set in the learning period field in Fig. 6. At this time, the coefficients are determined so as to satisfy the above-mentioned sign constraint.

[0030] <Continuous process> FIG. 8 is a diagram for explaining an example of process data in a continuous process. The left column of FIG. 8 shows a part of the process of the continuous process 32 shown in FIG. 2. Specifically, the process includes a tank 311 and a pump 312. The right column of FIG. 8 shows an example of process data acquired in each process. In the continuous process 32, time-series data associated with a tag and not associated with a serial number is continuously acquired from a sensor. In the example of FIG. 8, time-series data is acquired from each sensor with tags 102 and 103. In a continuous process, a device continuously receives a processing target and continuously processes it. When a continuous process is performed after a batch process, in order to link a processing target in the batch process with a processing target in the continuous process, in this embodiment, traceability information set in advance by a user is used. The traceability information includes a sampling interval and a residence time. The sampling interval represents an interval at which sampling is performed for a process inspection by, for example, a reduction method in the continuous process. The residence time represents a time during which a processing target resides from the completion of a batch process until it reaches a process included in the continuous process.

[0031] FIG. 9 is a diagram for explaining the correspondence between samples in a process inspection of a continuous process and serial numbers in a batch process. For example, the process inspection is performed at a predetermined interval, and the samples in the process inspection are reduced samples for a period corresponding to the interval. In addition, when a continuous process is performed after a batch process, the product of the batch process completed within a predetermined period is introduced as a processing target for the continuous process. Therefore, the reduced samples of the process data in the continuous process can be traced back to the residence time up to the sampling time and linked to the range of the completion time of the batch process to identify the corresponding serial numbers of the batch process. By such linking, when a batch process and a continuous process are performed consecutively, the accuracy of the prediction formula using the process data in the batch process can be improved.

[0032] FIG. 10 is a diagram for explaining the dwell time from the sensor position in the continuous process to the sampling position in the process inspection. The above-mentioned reduced sample is calculated, for example, as the average value of the process data obtained in the continuous process over a predetermined period. In order to associate the sampling time point at which this calculation is performed with the range of the acquisition time point of the process data to be averaged at that time point, in this embodiment, the dwell time of the interval between the position of the sensor that outputs the process data in the continuous process and the position where sampling for the process inspection is performed is set. This makes it possible to link the process data in the continuous process with the reduced sample of the process inspection. This linking makes it possible to improve the accuracy of the prediction formula that uses the process data in the continuous process.

[0033] 11 is a diagram showing an example of a tag attribute table of a continuous process created based on a knowledge base. The tag attribute table of a continuous process may be a so-called database table, or may be a file in a predetermined format such as CSV. The tag attribute table is created in advance by a user and is read by the prediction device 1.

[0034] The tag attribute table for a continuous process includes attributes for tag, type, residence time, batch-related tag, and operating condition optimization. Attributes with the same names as those in the tag attribute table for a batch process shown in Figure 5 will not be described. In the type field, the type of continuous, batch, or quality is registered. Among the types, "continuous" indicates that the tag indicated by each record is process data in a continuous process. "batch" indicates that the process data is in a batch process. "quality" indicates that the analysis value is a reduced sample in a process inspection.

[0035] Also, in the continuous process, the tag combination table as shown in FIG. 6 is used, and the prediction device 1 creates a prediction model based on the tag attribute table and the tag combination table. FIG. 12 is a diagram showing an example of a logic tree showing the configuration of a prediction model. In FIG. 12, each rectangle also represents the output value or predicted value of a sensor corresponding to a tag. The rectangle with the tag 101 indicates process data of a batch process. As described with reference to FIG. 9, the process data of the batch process is identified based on the residence time and the corresponding manufacturing number and series, and the average value of these is used as an explanatory variable of the prediction formula. The rectangles with the tags 102 and 103 indicate process data of a continuous process. As described with reference to FIG. 10, the process data of the continuous process is identified based on the residence time and the sampling interval, and the average value of the process data within the period is used as an explanatory variable of the prediction formula. The rectangle with the tag 104 is an analysis value of a process inspection (also called a quality process), and is a value corresponding to the process data obtained by, for example, a reduction method. In the continuous process, the prediction formula is also created by combining records with the same combination ID in the tag combination table shown in FIG. 6. The prediction formula is the same as that for the batch process, so the explanation will be omitted.

[0036] In this embodiment, since process data selected based on a knowledge base is used, a huge number of parameters for the entire plant are not required, and a prediction model can be created quickly from parameters with clear causal relationships. Furthermore, if a prediction model including an autoregressive term is created, prediction can be made that takes into account changes in the process data over time, which cannot be reflected by simply performing a simulation based on process data at a certain point in time.

[0037] Furthermore, without the sign constraint as described above, appropriate results may not be obtained when solving an inverse problem using a prediction model. In other words, when specifying a required quality and determining the corresponding operating conditions, there is a risk that a prediction model that outputs results that violate the principles of the process may be created. By imposing the sign constraint as described above, a prediction model that conforms to the principles of the process can be created. In other words, by using a prediction formula that satisfies the sign constraint, not only can it be possible to predict the values ​​of the influence system that are alternative indicators of the product quality, but it also becomes easy to know how to change the plant operating conditions to improve the quality.

[0038] <Control> The prediction device 1 may use the created prediction formula and process data to determine operating conditions for transitioning to a stable state in response to changes in state that have been affected, or to determine operating conditions under which the product satisfies a specified requirement, and control the plant 3 based on the operating conditions. For example, target values ​​may be set for some characteristic values, and allowable ranges may be set for other controllable process data, to determine preferred operating conditions. Furthermore, a unit cost may be set for at least some of the process data, and, for example, operating conditions that minimize costs or that satisfy an allowable range of costs may be determined.

[0039] As described above, the allowable range of each process data is set in the "management range" field of the table shown in Fig. 5. The target value of each process data is set in the "setting range" field. Furthermore, the cost per predetermined unit amount of each process data is set in the "cost impact (unit price)" field. Note that process data with "adjustment" registered in the "adjustment / monitoring" field represents a value that can be adjusted by controlling actuators or the like equipped in the plant 3. During control, the operating conditions of the adjustable process data are found based on conditions such as targets and allowable ranges.

[0040] In the example of Figure 5, the cost can be calculated by the sum of the products of the unit cost set in the "Cost Impact (Unit Cost)" field and the process data value corresponding to each tag. Then, the controllable process data value (operating condition) is calculated so that the cost, which is the objective function, is minimized.

[0041] Based on the set range, the predicted value calculated by the prediction formula is constrained to fall within the range registered in the "set range" in Figure 5. These process data are, for example, quality or quality substitute indexes, and the range can be said to be a target value determined by the required specifications.

[0042] Furthermore, the possible values ​​of the controllable process data are limited based on the management range. The process data corresponding to the tag with "Adjustment" registered in the "Adjustment / Monitoring" field is adjustable, but constraints are set due to setting limits determined according to, for example, the specifications of the plant 3.

[0043] In this embodiment, the prediction model created in the prediction process is also used as a constraint condition. That is, a constraint condition that represents a predetermined range is set for at least a part of the objective variable in the prediction formula, and optimal values ​​of the explanatory variables that bring the predicted value into the range of the constraint condition are searched for.

[0044] When a value that minimizes an objective function is found among variables that satisfy linear inequalities or linear equalities that serve as constraint conditions, optimization can be performed using so-called linear programming. Note that instead of cost, a predicted value of any process data may be used as the objective variable. In this case, an allowable range for cost may be set. Even if some of the constraint conditions or objective functions are nonlinear, they can be solved using existing nonlinear programming. Furthermore, multiple objective functions may be set and multi-objective optimization may be performed. As described above, the operating conditions can be found by solving the optimization problem.

[0045] For example, if the process data to be adjusted is the input amount of a secondary raw material, the calculated optimal solution becomes the set value as is. For example, if the process data to be adjusted is the integral value of the temperature of the processing object, actuators such as valves are adjusted so that the set value approaches the calculated optimal solution.

[0046] <Device configuration> FIG. 13 is a block diagram showing an example of the configuration of the prediction device 1. The prediction device 1 is a general computer, and includes a communication interface (I / F) 11, a storage device 12, an input / output device 13, and a processor 14. The communication I / F 11 may be, for example, a network card or a communication module, and communicates with other computers based on a predetermined protocol. The storage device 12 may be a main storage device such as a random access memory (RAM) or a read only memory (ROM), and an auxiliary storage device (secondary storage device) such as a hard disk drive (HDD), a solid state drive (SSD), or a flash memory. The main storage device temporarily stores a program read by the processor 14 and information transmitted / received between the processor 14 and other computers, and secures a working area for the processor 14. The auxiliary storage device stores a program executed by the processor 14, information transmitted / received between the processor 14 and other computers, and the like. The input / output device 13 is, for example, a user interface such as an input device such as a keyboard or a mouse, an output device such as a monitor, or an input / output device such as a touch panel. The processor 14 is an arithmetic processing device such as a central processing unit (CPU), and performs each process according to this embodiment by executing a program. 13, functional blocks are shown in the processor 14. That is, the processor 14 functions as a process data acquisition unit 141, a process data processing unit 142, a prediction model creation unit 143, a quality prediction unit 144, and a plant control unit 145 by executing a predetermined program.

[0047] The process data acquiring unit 141 acquires process data from a sensor provided in the plant 3 via, for example, the communication I / F 11 and the control station 2, and stores the process data in the storage device 12. As described above, the process data is associated with the sensor by a tag.

[0048] When creating a prediction model, the process data processing unit 142 processes the process data based on the tag attribute table of the batch process shown in Fig. 5 or the tag attribute table of the continuous process shown in Fig. 11. That is, the process data processing unit 142 extracts an instantaneous value at a specified timing, calculates an average value for a specified period, or calculates an integral value for a specified period based on information registered in the field of the primary process in the tag attribute table of the batch process. The process data processing unit 142 may also calculate an average value of process data corresponding to a specific tag, lineage, and serial number based on information registered in the field of the batch-related tag in the tag attribute table of the continuous process and, for example, the above-mentioned traceability information previously stored in the storage device 12, or calculate an average value of the process data of the continuous process for a period specified based on the traceability information.

[0049] The prediction model creation unit 143 creates a prediction model including a prediction formula such as that shown in the above-mentioned formula (1) based on the tag combination table shown in Fig. 6, for example, and stores the model in the storage device 12. For example, in a batch process, the prediction model creation unit 143 may update the coefficients of the prediction formula using data from a most recent predetermined period for each manufacturing number. Also, for example, in a continuous process, the prediction model creation unit 143 may update the coefficients of the prediction formula using the most recent data for each predetermined period.

[0050] The quality prediction unit 144 predicts the output value of a predetermined sensor and the analysis value of a process inspection by using the process data and the prediction model. The quality prediction unit 144 may calculate a predicted value after changing the operating conditions by using data based on any operating conditions and the prediction model.

[0051] The plant control unit 145 controls actuators such as valves and other devices provided in the plant 3, for example, via the communication I / F 11 and the control station 2. The plant control unit 145 may also determine operating conditions under which, for example, quality and cost satisfy predetermined conditions, and control the plant 3 based on these. The plant control unit 145 may also determine operating conditions for transitioning to a predetermined stable state, or operating conditions under which a product satisfies predetermined requirements, and control the plant 3 based on these.

[0052] The above-mentioned components are connected via a bus 15.

[0053] <Prediction processing (batch process)> FIG. 14 is a process flow diagram showing an example of a prediction process executed by the prediction device 1. The processor 14 of the prediction device 1 executes a predetermined program to execute the process shown in FIG. 14. The prediction process is executed for each serial number in the batch process, and at a predetermined sampling interval in the continuous process. It is assumed that the tag attribute table of the batch process shown in FIG. 5, the tag combination table shown in FIG. 6, the tag attribute table of the continuous process shown in FIG. 11, traceability information, and the like are created by a user and stored in the storage device 12 in advance. It is assumed that the process data acquisition unit 141 continuously acquires process data from a sensor equipped in the plant 3 via, for example, the communication I / F 11 and the control station 2, and temporarily or permanently stores the process data in the storage device 12. The process data is described according to a predetermined standard such as OPC.

[0054] The process data acquiring unit 141 of the prediction device 1 reads the setting information (FIG. 14: S1). In this step, the process data acquiring unit 141 reads the tag attribute table, the tag combination table, the traceability information, and the like from the storage device 12.

[0055] Moreover, the process data acquisition unit 141 reads the process data (FIG. 14: S2). In this step, the process data corresponding to the tag used in the prediction formula is extracted, for example, for each prediction formula, for each series, and for each subdivided process. FIG. 15 is a diagram showing an example of a write array for a batch process to write the data read out in this step. The write array for a batch process may be OPC data, a so-called database table, or a file in a predetermined format such as CSV. The table in FIG. 15 includes attributes of date and time, serial number, product type, step, and tag. In the date and time field, the date and time when the sensor outputs a measured value is registered. In the product type field, the manufacturing number is registered. In the product type field, the type of the processing target is registered. In the step field, information indicating a stage in the process, which is represented by a predefined step, is registered. In the tag field, the output value of the sensor corresponding to each tag is registered.

[0056] Furthermore, the process data processing unit 142 of the prediction device 1 performs a predetermined processing process on the process data (S3 in FIG. 14). Details of this step will be described with reference to FIG. 16. FIG. 16 is a processing flow diagram showing an example of the processing process. When the process data for each prediction formula, each sequence, and each subdivided process is extracted into the write array shown in FIG. 15, the process data processing unit 142 executes the processing shown in FIG. 16 for each record of the write array.

[0057] The process data processing unit 142 reads a record from the write array (FIG. 16: S11). In this step, one record is read in sequence from a table such as that shown in FIG. 15. The process data processing unit 142 also processes data according to the primary processing method (FIG. 16: S12). In this step, the tag attribute table shown in FIG. 5 is referenced, and based on the type registered in the "method" field of the "primary processing" of the corresponding tag, for example, an instantaneous value, an average value, an integral value, a differential coefficient, a difference, a maximum value, a minimum value, a thermal history, or the process data itself is obtained. FIG. 17 is a diagram showing an example of a batch array into which the processing result of this step is written. The batch array may also be OPC data, a so-called database table, or a file in a predetermined format such as CSV. The table in FIG. 17 includes each attribute of the serial number, the end date and time, and the tag. The serial number is registered in the serial number field. The end date and time field is registered with the date and time when the batch process of the serial number is completed. The tag field is registered with the process data after processing.

[0058] Here, the calculation of the thermal history will be described. The thermal history is information that indicates the degree of progress of a reaction, for example, for a general chemical reaction such as depolymerization, acetylation, or deacetylation. In this embodiment, the thermal history is calculated as an integral value of the reaction rate for a predetermined period of time. For example, as shown in the following formula (2), it is calculated from the integral value of the reaction rate equation using the Arrhenius equation.

number

[0059] The process data processing unit 142 also performs a predetermined data cleansing process (FIG. 16: S13). The data cleansing process is a process for removing outliers, and various methods can be adopted. For example, a moving average is calculated using the most recent data. Also, the difference between the moving average and the actual measurement value is taken to obtain a standard deviation σ (also called error variance) that represents the variation in the difference. Then, values ​​that do not fall within a predetermined confidence interval, such as the interval from the mean value of the probability distribution -3σ to the mean value of the probability distribution +3σ (also called the 3σ interval), are removed. Similarly, values ​​that do not fall within the 3σ interval for the difference between the actual measurement values ​​before and after may be removed. The data cleansing process is performed on, for example, instantaneous values ​​or process data at the end of a batch.

[0060] The process data processing unit 142 also performs a predetermined smoothing process (FIG. 16: S14). The smoothing process is performed on tags for which "required" is registered in the smoothing field in the tag attribute table shown in FIG. 5. The smoothing process may be, for example, a process of calculating a moving average of a predetermined number of recent values ​​after data cleansing, or may be another method capable of smoothing data. This ends the processing process in FIG. 16, and the process returns to the process in FIG. 14.

[0061] After that, the prediction model creation unit 143 of the prediction device 1 performs a prediction model construction process (FIG. 14: S4). In this step, a prediction formula constituting a prediction model is created based on the tag combination table shown in FIG. 6. Specifically, the processed process data of the tags with the same combination ID is read, and the processed process data is applied to the prediction formula (for example, the above-mentioned formula (1)) based on the type registered in the cause / influence field, and the coefficient and constant term of the prediction formula are determined by regression analysis. At this time, the processed process data is subject to learning of the most recent data according to the value registered in the learning period field. The prediction model creation unit 143 may also search for a preferable value for the learning period. For example, a correlation coefficient is calculated using the created prediction model and process data, and the learning period is set so that the correlation coefficient is improved. In addition, the coefficient of the prediction formula and the like are determined under the constraint that there is a certain correspondence between the direction of change in the value of the factor system and the direction of change in the value of the influence system based on the code registered in the causal relationship field. Then, the prediction model creation unit 143 stores the created prediction formula in the storage device 12.

[0062] The quality prediction unit 144 of the prediction device 1 performs prediction processing using the prediction model created in the prediction model construction processing and the process data or its predicted value (FIG. 14: S5). For convenience, the prediction processing is shown in the processing flow of FIG. 14, but the quality prediction unit 144 can perform prediction processing using the prediction model and process data at any time. In this step, the quality prediction unit 144 reads out the most recent prediction model and process data, and substitutes the process data or its predicted value into the prediction formula included in the prediction model to obtain a predicted value corresponding to an arbitrary influence system.

[0063] Fig. 18 is a diagram for explaining the process data or the predicted value thereof to be substituted into the prediction formula. When predicting the value of tag 007 shown in Fig. 4, actual measured values ​​or predicted values ​​of sensors corresponding to each tag are used at time t1 in Fig. 3 as shown in Fig. 18. That is, when predicting the value of tag 007 of serial number 003, the known actual measured values ​​of tags 001 to 004 of serial number 003 and the unknown predicted values ​​of tags 005 and 006 of serial number 003 are substituted into the prediction formula for tag 007.

[0064] Fig. 19 is a diagram showing another example for explaining the process data or the predicted value thereof to be substituted into the prediction formula. When predicting the value of tag 007 shown in Fig. 4, the actual measurement value or the predicted value of the sensor corresponding to each tag is used at time t2 in Fig. 3 as shown in Fig. 19. That is, when predicting the value of tag 007 of serial number 004, the known actual measurement values ​​of tags 001 and 002 of serial number 004, the actual measurement value of tag 004 of serial number 003, and the unknown predicted values ​​of tags 003, 005, and 006 of serial number 004 are substituted into the prediction formula for tag 007. Here, as shown in Fig. 4, since there is no prediction formula for the value of tag 004, the actual measurement value of the most recent serial number is used.

[0065] Furthermore, the values ​​input to the prediction model are not limited to process data, and may be data based on any operating conditions, for example. In this way, it is possible to predict the results when the operating conditions of the plant 3 are changed. As described above, the quality prediction unit 144 uses the prediction model and the sensor output value or its predicted value to calculate, for example, a predicted value of predetermined process data.

[0066] Furthermore, the quality prediction unit 144 may obtain a predetermined confidence interval for the calculated predicted value or the actual measured value of the process data by the above-mentioned data cleansing, and display the predetermined confidence interval and the predicted value or the actual measured value on a graph on the input / output device 13 such as a monitor. In this way, the user can visually grasp the tendency, and can use it as a basis for deciding whether or not to change the operating conditions of the plant 3.

[0067] In addition, the plant control unit 145 may automatically change the operating conditions of the plant 3 based on the calculated predicted value, or may output information suggesting a change in the operating conditions to the user via the input / output device 13.

[0068] <Prediction processing (continuous process)> The prediction process shown in Fig. 14 is also performed in the continuous process. The following mainly describes the differences from the batch process. Note that the date and time when the batch process ended and the transfer (transfer) of the processing target to the continuous process was completed for each series and each production number is stored in the storage device 12.

[0069] In reading the process data (S2 in FIG. 14), in the case of a continuous process, rolling is continuously performed in which new data is written and old data is deleted, not in units of serial numbers. Also, the data structure of the write array shown in FIG. 15 may be changed based on the information registered in the type field of the tag attribute table shown in FIG. 11. FIG. 20 is a diagram showing an example of a write array that holds data of type "continuous". The write array that holds data of type "continuous" may be configured by deleting the serial number and the step from the table in FIG. 15, for example. Also, FIG. 21 is a diagram showing an example of a write array that holds data of type "batch". The write array that holds data of type "batch" is, for example, a table that holds a serial number and process data of the serial number, and a table like that in FIG. 21 is created for each system. The data registered in the table in FIG. 21 may be processed data registered in the batch array in FIG. 17.

[0070] Also, the processing of the process data (S3 in FIG. 14) is performed, for example, based on the timing of the process inspection. The timing of the process inspection is defined as a sampling interval in the traceability information. FIG. 22 is a diagram showing an example of a combined ID data array that holds the process data after processing in the continuous process. The combined ID data array may also be OPC data, a so-called database table, or a file in a predetermined format such as CSV. In the table in FIG. 22, in the process inspection ID field including the attributes of the process inspection ID, sampling, and tag, identification information for identifying each process inspection is registered. The sampling attribute further includes attributes of the start date and time and the end date and time, and the start date and time and the end date and time of sampling of the process inspection by the reduction method are registered in the start date and time and the end date and time fields, respectively. In the tag field, an analysis value corresponding to each tag is registered. Here, for a tag in which "continuous" is registered in the type field of the tag attribute table shown in FIG. 11, the average value of the process data from the processing time to the time going back to the sampling interval in the traceability information is registered. In addition, for tags in which “batch” is registered in the type field of the tag attribute table shown in FIG. 11, the average value of the process data corresponding to the manufacturing numbers for which transfer was completed from the time of processing back to the time of the sampling interval in the traceability information is registered.

[0071] In addition, in the prediction model construction process (S4 in FIG. 14), for example, when the prediction model creation unit 143 acquires an analysis value of a process inspection, the prediction model of the analysis value is updated. In this step, a prediction formula constituting a prediction model is also created based on the tag combination table shown in FIG. 6. In addition, in a continuous process, as shown in FIG. 9, the prediction model creation unit 143 may associate the process data of the cause system with the process data of the influence system based on the completion timing of the batch process and a predetermined residence time, and may learn the characteristics of the process data obtained from the plant 3. In addition, as shown in FIG. 10, the prediction model creation unit 143 may associate the process data of the cause system with the process data of the influence system based on the difference between the acquisition timing of the process data of the cause system and the acquisition timing of the process data of the influence system, and may learn the characteristics of the process data obtained from the plant 3.

[0072] The quality prediction unit 144 of the prediction device 1 performs prediction processing using the prediction model created in the prediction model construction process and the process data or its predicted value (FIG. 14: S5). In this step, the quality prediction unit 144 reads the most recent prediction model and process data, and substitutes the process data or its predicted value into the prediction formula included in the prediction model to obtain a predicted value of a value corresponding to an arbitrary influence system.

[0073] <Control processing> FIG. 23 is a process flow diagram showing an example of the control process executed by the prediction device 1. The processor 14 of the prediction device 1 executes a predetermined program to execute the process shown in FIG. 23. The control process is executed at any timing, for example, after the prediction model is updated. In the control process, the tag attribute table of the batch process shown in FIG. 5, the tag combination table shown in FIG. 6, the tag attribute table of the continuous process shown in FIG. 11, traceability information, and the like are stored in the storage device 12 in advance. In addition, the process data acquisition unit 141 continuously acquires process data from a sensor provided in the plant 3 via, for example, the communication I / F 11 and the control station 2, and temporarily or permanently stores the process data in the storage device 12.

[0074] The plant control unit 145 of the prediction device 1 reads the setting information (FIG. 23: S21). In this step, the plant control unit 145 reads the tag attribute table, tag combination table, traceability information, and the like from the storage device 12. In the control process, the information read includes the control target expressed, for example, as an objective function of an optimization problem, and the allowable range of control expressed, for example, as a constraint condition of the optimization problem. Here, the objective function is to minimize the cost calculated using the unit price registered in the cost impact field in FIG. 5. In addition, the values ​​registered in the management range and setting range fields are the constraint conditions.

[0075] Furthermore, the process data acquiring unit 141 reads the process data (FIG. 23: S22). The processing of this step is the same as S2 in FIG. 14. In this step, the process data corresponding to the tag used in the prediction formula is extracted, for example, for each prediction formula, for each series, and for each subdivided process. Furthermore, the output value of the sensor is registered in the write array shown in FIG. 15 or FIG. 20.

[0076] Furthermore, the process data processing unit 142 performs a predetermined processing process on the process data (FIG. 23: S23). The processing of this step is similar to S3 in FIG.

[0077] Then, the plant control unit 145 performs a calculation process for the optimization problem (FIG. 23: S24). In this step, the operating conditions that minimize or maximize the objective function under the read constraint conditions are obtained. For example, when the cost is minimized based on the settings shown in FIG. 5, the cost is obtained by the following formula (3). Cost = (Process data of tag 001 × Unit price) + (Process data of tag 002 × Unit price) + (Process data of tag 004 × Unit price) (3)

[0078] In addition, as constraint conditions, the information registered in the "Setting range" in Fig. 5 is used. Specifically, the following conditions are set: Lower limit of tag 005 <= Predicted value of tag 005 <= Upper limit of tag 005 Lower bound of tag 007 ≦ Predicted value of tag 007 ≦ Upper bound of tag 007

[0079] In addition, as other constraints, the information registered in the "management range" in Figure 5 is used. Specifically, the following conditions are set: Lower limit of tag 001 <= Value of tag 001 <= Upper limit of tag 001 Lower limit of tag 002 <= Value of tag 002 <= Upper limit of tag 002 Lower limit of tag 003 <= Value of tag 003 <= Upper limit of tag 003 Lower limit of tag 005 <= Value of tag 005 <= Upper limit of tag 005

[0080] In this embodiment, the prediction model created in the prediction process is also used as a constraint. For example, when a prediction formula is defined between process data in the logic tree shown in FIG. 7 and FIG. 12, downstream process data is calculated based on upstream process data based on the prediction formula constructed in the prediction process. In addition, the process data corresponding to the tag that is located at the upstream end and has "monitoring" registered in the "adjustment / monitoring" field in FIG. 5 is acquired in S22, and the processed value is used in S23. Then, when the above-mentioned setting range and management range are set for the objective variable in each prediction formula, a constraint condition that represents these ranges is set, and an optimal value of the explanatory variable that makes the predicted value fall within the range of the constraint condition is searched for.

[0081] The optimization problem described above can be solved by using existing solutions. The process data corresponding to the tags with "Adjustment" registered in the "Adjustment / Monitoring" field in Figure 5 are the adjustment targets. In other words, the process data with tags 001, 003, 004, and 006 are the adjustment targets, and the target tags do not necessarily match the constraint conditions.

[0082] When the optimization problem is solved and the operating conditions including the set values ​​of the process data to be adjusted are obtained, the plant control unit 145 controls the plant 3 according to the operating conditions (FIG. 23: S25). In this step, the plant control unit 145 outputs data indicating the operating conditions to the control station 2 via the communication I / F 11. Then, the operation of the plant 3 is controlled according to the control signal from the control station 2. Note that, for example, when the multi-objective optimization problem is solved in S24, multiple candidates for the operating conditions may be presented to the user via the input / output device 13, and the plant 3 may be controlled based on the operating conditions selected by the user.

[0083] <Modification> Each configuration and their combinations in each embodiment are merely examples, and addition, omission, substitution, and other modifications of the configurations are possible as appropriate within the scope of the present invention. The present disclosure is not limited by the embodiments, but is limited only by the scope of the claims. In addition, each aspect disclosed in this specification can be combined with any other feature disclosed in this specification.

[0084] Although the above embodiment has been described with reference to a chemical plant, the present invention can be applied to a manufacturing process in a general production facility. For example, a process similar to the batch process in the embodiment can be applied by using a lot number as a processing unit instead of the serial number of the batch process in the embodiment.

[0085] At least a part of the functions of the prediction device 1 may be distributed to multiple devices, or the same functions may be provided in parallel by multiple devices. For example, a model creation device that creates a prediction model, a prediction device that performs prediction using the created prediction model, and a control device that controls production equipment using the created prediction model may be different. Also, at least a part of the functions of the prediction device 3 may be provided on a so-called cloud.

[0086] Furthermore, although the above-mentioned formula (1) is a linear model including an autoregressive term, the present invention is not limited to such an example. For example, a model that does not include an autoregressive term can also be adopted. The model may be linear or nonlinear. It may be a single formula, or a state space model incorporating periodic fluctuations such as seasonal fluctuations may be adopted. However, it is preferable that the model satisfies a sign constraint. In other words, the coefficients of the prediction formula and the like are determined under the constraint that there is a certain correspondence between the direction of fluctuation of the value of the factor system and the direction of fluctuation of the value of the influence system.

[0087] The present disclosure also includes a method and a computer program for executing the above-mentioned processes, and a computer-readable recording medium having the program recorded thereon. The recording medium having the program recorded thereon enables the above-mentioned processes by causing a computer to execute the program.

[0088] Here, a computer-readable recording medium refers to a recording medium that stores information such as data and programs through electrical, magnetic, optical, mechanical, or chemical action and can be read by a computer. Among such recording media, those that can be removed from a computer include flexible disks, magneto-optical disks, optical disks, magnetic tapes, memory cards, etc. Furthermore, recording media that are fixed to a computer include HDDs, SSDs (Solid State Drives), ROMs, etc. [Explanation of symbols]

[0089] 1: Prediction device 11: Communication I / F 12: Storage device 13: Input / output devices 14: Processor 141: Process data acquisition unit 142: Process Data Processing Department 143: Prediction Model Creation Department 144: Quality Prediction Department 145: Plant Control Unit 2: Control Station 3: Plant

Claims

1. a process data processing unit that performs a predetermined processing process on process data obtained from a chemical plant; a prediction model creation unit that creates a prediction model of quality represented by the objective variable by learning features of the process data obtained from the chemical plant based on causal relationship information that is created based on a knowledge base and defines a combination of a first process data set that is an explanatory variable and a second process data set that is an objective variable; Equipped with the first process data includes process data that has been subjected to the predetermined processing, and is a value associated with the objective variable as a parameter that influences the objective variable in the causal relationship information, the second process data is process data obtained from the chemical plant and is a value that serves as an alternative index of product quality; the predetermined processing step is a step of obtaining a value corresponding to a reaction rate of a processing target in a predetermined period of time by using the process data obtained from the chemical plant; The value according to the reaction rate is the integral value of the reaction rate. Prediction device.

2. The value according to the reaction rate is calculated using the Arrhenius equation, and a frequency factor of the reaction rate constant and an activation energy are determined according to the type of chemical reaction and the target to be treated in the chemical plant. The prediction device according to claim 1 .

3. The predetermined processing step further includes a step of calculating an integral value, a differential value, an average value, a maximum value, or a minimum value of the process data in a predetermined period, or an instantaneous value of the process data at a predetermined time point, depending on the type of the process data. The prediction device according to claim 1 .

4. The prediction model has a hierarchical structure including a plurality of prediction formulas, and has a second prediction formula including a predicted value calculated by a first prediction formula as an explanatory variable. The prediction device according to claim 1 .

5. the second process data is a value obtained by sampling a plurality of second process data by a reduction method; The range of acquisition timing of the first process data in the chemical plant is associated with the execution timing of the sampling based on a residence time of the processing object in the chemical plant, and the prediction model creation unit creates the prediction model. The prediction device according to claim 1 .

6. The chemical plant performs a batch process in which a processing object is sequentially processed for each predetermined processing unit, and then a continuous process in which the processing object is continuously processed; The range of the completion timing of the batch process and the execution timing of sampling in the continuous process are associated with each other based on the residence time of the object to be processed in the chemical plant, and the prediction model creation unit creates the prediction model. The prediction device according to claim 1 .

7. a process data processing step for performing a predetermined processing process on process data obtained from the chemical plant; a prediction model creation step of creating a prediction model of quality represented by the objective variable by learning features of the process data obtained from the chemical plant based on causal relationship information created based on a knowledge base, the causal relationship information defining a combination of a first process data set as an explanatory variable and a second process data set as an objective variable; The prediction device executes the above. the first process data includes process data that has been subjected to the predetermined processing, and is a value associated with the objective variable as a parameter that influences the objective variable in the causal relationship information, the second process data is process data obtained from the chemical plant and is a value that serves as an alternative index of product quality; the predetermined processing step is a step of obtaining a value corresponding to a reaction rate of a processing target in a predetermined period of time by using the process data obtained from the chemical plant; The value according to the reaction rate is the integral value of the reaction rate. Forecasting methods.

8. a process data processing step for performing a predetermined processing process on process data obtained from the chemical plant; a prediction model creation step of creating a prediction model of quality represented by the objective variable by learning features of the process data obtained from the chemical plant based on causal relationship information created based on a knowledge base from the process data obtained from the chemical plant or the process data processed in the process data processing step, the causal relationship information defining a combination of a first process data as an explanatory variable and a second process data as an objective variable or a value corresponding to the second process data; on the computer, the first process data includes process data that has been subjected to the predetermined processing, and is a value associated with the objective variable as a parameter that influences the objective variable in the causal relationship information, the second process data is process data obtained from the chemical plant and is a value that serves as an alternative index of product quality; the predetermined processing is a process of obtaining a value corresponding to a reaction rate of a processing target in a predetermined period of time by using the process data obtained from the chemical plant, The value according to the reaction rate is the integral value of the reaction rate. program.

Images