Prediction device, prediction method, and program

Abstract

The present invention aims at improving prediction accuracy for a process including reactions in a chemical system. A prediction device includes a process data processing section that performs prescribed processing on process data obtained from the chemical system, and a prediction model generation section that generates a prediction model that has learned features of the process data obtained from the chemical system based on a causal relationship defined for a combination of first process data as an explanatory variable and second process data as a target variable or a value corresponding to the second process data among the process data obtained from the chemical system or the process data processed by the process data processing section. Further, the process data processing section uses the process data to find a value corresponding to a reaction speed of a processing target in a prescribed period.

Description

Technical Field

[0001] This disclosure relates to prediction devices, prediction methods, and procedures. Background Technology

[0002] Previously, techniques for predicting product quality or controlling actions based on predictions during manufacturing processes have been proposed. For example, a prediction system (Patent Document 1) was proposed that predicts product characteristic values ​​during the manufacturing process and calculates control conditions for subsequent manufacturing processes based on the prediction results. This system includes: a database storing data measured in each batch of the manufacturing process and / or data representing the state of the manufacturing process; a numerical model generation unit that generates a numerical model of the manufacturing process using the data stored in the database; a product characteristic prediction unit that, for batches under manufacturing, inputs actual values ​​for processed processes into the numerical model and representative values ​​obtained based on past batches into the numerical model for unprocessed processes to predict product characteristic values; and an optimal manufacturing condition calculation unit that calculates the optimal manufacturing conditions for the process of the object to be controlled in the unprocessed process based on the prediction results obtained by the product characteristic prediction unit. This system performs the prediction by the product characteristic prediction unit and the calculation by the optimal manufacturing condition calculation unit for each batch under manufacturing for each specified controlled process.

[0003] Existing technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent No. 6477423

[0006] Patent Document 2: Japanese Patent No. 5751045

[0007] Patent Document 3: Japanese Patent Application Publication No. 2018-120343

[0008] Patent Document 4: Japanese Patent Application Publication No. 2001-106703 Summary of the Invention

[0009] The problem the invention aims to solve

[0010] However, particularly in processes involving reactions within chemical units, there is a challenge in improving prediction accuracy. Therefore, the purpose of this technology is to improve prediction accuracy for processes involving reactions within chemical units.

[0011] Technical solution

[0012] The prediction apparatus disclosed herein includes: a process data processing unit that performs prescribed processing on process data obtained from a chemical unit; and a prediction model generation unit that generates a prediction model based on causal relationship information learned from the characteristics of the process data obtained from the chemical unit. This causal relationship information defines a combination of first process data (as an explanatory variable) and second process data (as a target variable) or values ​​corresponding to the second process data from the process data obtained from the chemical unit or the process data processed by the process data processing unit. Furthermore, the process data processing unit uses the process data to determine a value corresponding to the reaction rate of the object being processed within a specified period.

[0013] Reactions are influenced by factors such as temperature and are typically non-linear; therefore, it is difficult to predict the characteristics of process data based on peak values, difference values, integral values, etc., of the process data. Values ​​corresponding to the reaction rate, as described above, are explanatory variables, for example, related to the amount of reaction per hour. Predictive models can be generated using values ​​corresponding to the reaction rate to improve the accuracy of predictions for processes including reactions in chemical units.

[0014] Furthermore, the value corresponding to the reaction rate can also be the integral value of the reaction rate. Alternatively, it can be set such that the value corresponding to the reaction rate is calculated using the Arrhenius formula, with the frequency factor and activation energy of the reaction rate constant determined according to the type of chemical reaction and the object being processed in the chemical unit. Specifically, such a value can be used.

[0015] Alternatively, the process data processing department can also calculate the integral, differential, average, maximum, or minimum values ​​of the process data within a specified period, or the instantaneous values ​​of the process data at a specified time point, based on the type of the process data. Such processing of the process data can improve the prediction accuracy of the prediction model.

[0016] Alternatively, the predictive model can be a hierarchical structure comprising multiple predictors, with a second predictor that includes the predicted value calculated by the first predictor in the explanatory variables. For example, multiple causal relationships can be represented by functions in this way.

[0017] Alternatively, the value corresponding to the second process data can be obtained by sampling multiple second process data using a reduction method. The prediction model generation unit, based on the residence time of the processed object within the chemical unit, establishes a correspondence between the acquisition timing of the first process data in the chemical unit and the calculation timing of the value corresponding to the second process data, thereby generating a prediction model. When the processed object is continuously processed in the chemical unit, the accuracy of the prediction can be improved by appropriately establishing a correspondence between the acquisition timing of the process data as an explanatory variable and the acquisition timing of the process data as the objective variable.

[0018] Alternatively, the chemical unit can be configured to perform batch processing of the processed object sequentially according to each prescribed processing unit, followed by a continuous processing step where the processed object is processed continuously. The prediction model generation unit, based on the residence time of the processed object within the chemical unit, establishes a correspondence between the completion timing range of the batch processing step and the calculation timing of the corresponding value of the second process data, and generates a prediction model. Even when performing both batch and continuous processing steps, the accuracy of the prediction can be improved by appropriately establishing a correspondence between the process data as explanatory variables and the process data as target variables.

[0019] It should be noted that the contents of the technical solution can be combined as much as possible without departing from the problem and technical concept of this disclosure. Furthermore, the contents of the technical solution can be provided as a computer or other device, or a system comprising multiple devices, a method executed by a computer, or a program that causes a computer to execute. It should also be noted that it can be provided as a recording medium for storing the program.

[0020] Invention Effects

[0021] Based on the publicly available technology, the accuracy of predicting the characteristic values ​​of the product can be improved. Attached Figure Description

[0022] Figure 1 This is a diagram illustrating an example of a system implementation.

[0023] Figure 2 This is a schematic diagram illustrating an example of a process performed by the machinery of the unit.

[0024] Figure 3 This is a diagram illustrating an example of process data in a batch process.

[0025] Figure 4 This is a diagram representing an example of information pre-registered in the knowledge base.

[0026] Figure 5 This is a diagram representing an example of a tag attribute table for batch processes generated based on a knowledge base.

[0027] Figure 6 This is a diagram representing an example of a tag combination table generated based on a knowledge base.

[0028] Figure 7 This is a diagram representing an example of a logic tree that constitutes a predictive model.

[0029] Figure 8 This is a diagram used to illustrate an example of process data in a continuous process.

[0030] Figure 9It is a diagram used to illustrate the correspondence between samples in continuous process inspections and production numbers in batch processes.

[0031] Figure 10 It is a graph used to illustrate the dwell time from the sensor position in a continuous process to the sampling position in the process inspection.

[0032] Figure 11 This is a diagram representing an example of a tag attribute table for a series of processes generated based on a knowledge base.

[0033] Figure 12 This is a diagram representing an example of a logic tree that constitutes a predictive model.

[0034] Figure 13 This is a block diagram illustrating an example of the configuration of a prediction device.

[0035] Figure 14 This is a flowchart illustrating an example of the predictive processing performed by the predictive device.

[0036] Figure 15 This is a diagram illustrating an example of the arrangement of write operations used in batch processes.

[0037] Figure 16 This is a flowchart illustrating an example of processing.

[0038] Figure 17 This is a diagram illustrating an example of batch arrangement.

[0039] Figure 18 It is a graph used to illustrate the process data or its predicted values ​​substituted into the predictive formula.

[0040] Figure 19 This is a diagram illustrating other examples of process data or their predicted values ​​used in a predictive formula.

[0041] Figure 20 This is a diagram illustrating an example of how writing data with the category "continuous" is done using an arrangement.

[0042] Figure 21 This is a diagram illustrating an example of how writing data categorized as "batch" is done using an arrangement.

[0043] Figure 22 This is a diagram illustrating an example of how process data is combined with ID data to maintain a continuous workflow.

[0044] Figure 23 This is a flowchart illustrating an example of the control processing performed by the predictive device. Detailed Implementation

[0045] Hereinafter, embodiments of the prediction device will be described with reference to the accompanying drawings.

[0046] <Implementation Method>

[0047] Figure 1 This diagram illustrates an example of the system described in this embodiment. System 100 includes a prediction device 1, a control station 2, and a unit 3. System 100 is, for example, a distributed control system (DCS), including multiple control stations 2. That is, the control system of the unit 3 is divided into multiple zones, and each control zone is distributedly controlled by the control stations 2. The control stations 2 are existing devices in the DCS, receiving status signals output from sensors or other components of the unit 3, or outputting control signals to the unit 3. Then, based on the control signals, the actuators such as valves and other machines of the unit 3 are controlled.

[0048] The prediction device 1 acquires the status signals (process data) of the unit 3 via the control station 2. The process data includes temperature, pressure, flow rate, etc., of the raw materials and intermediate products being processed, as well as setpoints that determine the operating conditions of the machines in the unit 3. Furthermore, the prediction device 1 generates a prediction model based on a knowledge base storing the correspondence between assumed causes and, for example, the effects of anomalies. For example, it generates predictive formulas for quality and cost using causal relationship information defined by combining process data (also called the primary cause system) generated from the knowledge base as explanatory variables and process data (also called the influence system) as objective variables. The prediction device 1 can then use the predictive formulas and process data to predict characteristic values ​​representing product quality, etc., or to predict the characteristic values ​​of the product under changed operating conditions of the unit 3. Alternatively, the prediction device 1 can determine, for example, the operating conditions under which quality and cost meet specified conditions. Furthermore, the prediction device 1 can use the predictive formulas and process data to determine the operating conditions for transitioning to a stable state or for determining the operating conditions under which the product meets specified requirements, in response to changes in the state as an influence. In addition, the prediction device 1 can also use the analytical values ​​obtained from the specified process inspection as the target variable to replace the process data affecting the system.

[0049] Figure 2This is a schematic diagram illustrating an example of a machine possessed by a unit or a process performed by that machine. That is, the process is assumed to include a production process as a treatment and a process machine as an apparatus. In this embodiment, the process may include batch operations 31 and continuous operations 32. In batch operations 31, the objects to be processed are processed sequentially according to each defined processing unit, for example, sequentially receiving, holding, and discharging raw materials from each machine. In continuous operations 32, continuously introduced objects to be processed are processed continuously, for example, receiving, holding, and discharging raw materials in parallel. Furthermore, the process may also include multiple series 33 performing the same treatment in parallel.

[0050] The machines performing the various processes include, for example, reactors, distillation units, heat exchangers, compressors, pumps, tanks, etc., which are connected via piping. Furthermore, sensors, valves, etc., are installed at designated locations on the machines and piping. Sensors may include thermometers, flow meters, pressure gauges, levels, concentration meters, etc. In addition, the sensors monitor the operating status of each machine and output status signals. Furthermore, the sensors provided with Unit 3 are equipped with "tags" to identify each sensor. Then, the predictive device 1 and control station 2 manage the input and output signals to each machine based on these tags.

[0051] <Batch Process>

[0052] Figure 3 This is a diagram illustrating an example of process data in a batch process. Figure 3 The columns on the left represent Figure 2 This is part of the batching process 31 shown. Specifically, the process includes: a pulverizer 301, a hydrocyclone 302, a pretreatment unit 303, a precooler 304, and a reactor 305. Furthermore, these processes are divided into a pretreatment step, a precooling step, and a reaction step. Figure 3 The column on the right represents an example of process data acquired in each process. In the pretreatment process, timing data is acquired from sensors labeled 001 and 002. In the precooling process, timing data is acquired from sensors labeled 003 and 004. In the reaction process, timing data is acquired from sensors labeled 005, 006, and 007. Furthermore, in the batching process, processing objects corresponding to production numbers (also called "product numbers") are processed intermittently. That is, the production number is identification information used to identify processing objects processed centrally in the batching process. For example... Figure 3 As shown, over time, data on the timing of processing objects corresponding to subsequent production numbers is obtained. It should be noted that time points t1 and t2 will be described later.

[0053] Figure 4This is a diagram representing an example of information pre-registered in the knowledge base. Let the knowledge base be pre-stored in the storage device of prediction device 1. Figure 4 The table includes columns corresponding to each sensor and rows indicating the reasons for changes in the sensor's output value. Specifically, values ​​are recorded in the columns corresponding to sensors affected by reasons such as "increase in the amount of by-product A" or "decrease in the amount of by-product A" shown in each row. Values ​​are recorded with a positive or negative sign corresponding to the change in the sensor's output value. It should be noted that the combination of cause and effect is not limited to a one-to-one relationship. That is, for one effect, multiple causes may be established, and the same cause may be associated with multiple effects.

[0054] The knowledge base can be pre-generated by the user based on HAZOP (Hazard and Operability Study). HAZOP is, for example, a method used to: establish and comprehensively list the relationships between sensing units based on monitoring points of the equipment constituting the unit, management ranges (thresholds of upper and lower limits, alarm setpoints), deviations from management ranges (abnormalities, modulations), a list of hypothetical causes of deviations from management ranges, logic (sensing units) for determining which hypothetical cause caused the deviation, the effects of the deviation, the actions taken in the event of a deviation, and the actions taken in response to those actions. It should be noted that, not limited to HAZOP, the knowledge base can also be generated based on FTA (Fault Tree Analysis), FMEA (Failure Mode and Effect Analysis), ETA (Event Tree Analysis), or similar methods, content extracted from operator feedback, or content extracted from operational standards and technical standards.

[0055] Figure 5 This diagram illustrates an example of a label attribute table for batch processes generated based on a knowledge base. The label attribute table defines the processing method for data obtained from sensors corresponding to each label. It should be noted that the label attribute table can be a table in a database or a file in a predefined format such as CSV (Comma Separated Values). Furthermore, the label attribute table is pre-generated by the user and read out by the prediction device 1.

[0056] The label attribute table includes attributes for label, series, variety, primary processing, smoothing, and operating condition optimization. The label field records the label used as sensor identification information. The series field records identification information for the series used to determine the process. The variety field records the category of the object being processed. Alternatively, the predictive device 1 can set parameters corresponding to the variety of the object being processed in predictive processing. The primary processing field records information on the processing method representing the sensor's output value. Furthermore, the primary processing attributes include attributes for batch operations, method, and data interval. The batch operation field records identification information for the subdivided operations within the batch operation. The method field records information on the category of the data processing method. Categories include: "Instantaneous Value," "Average," "Integral," "Derivative," "Differential," "Maximum," "Minimum," "Thermal History," and "None." "Instantaneous Value" represents the value at the start or end of the specified data interval. "Average" represents the average value obtained by dividing the value of the specified period by the number of data points. "Integral" represents the sum of the values ​​of the specified period. "Derivative" represents the derivative coefficient at the start or end of the specified data interval. "Difference" indicates the difference between the values ​​at the start and end of the specified data interval. "Maximum" indicates the maximum value within the specified period. "Minimum" indicates the minimum value within the specified period. "Thermal History" is an example of the degree of reaction progress, such as the integral value of the reaction rate within the specified period. "None" is a label appended to the end of a batch, indicating no processing is performed. The data interval attributes also include start and end attributes, with at least one of the start and end fields indicating the timing information for acquiring the sensor's output value. This timing information can also be defined, for example, based on predetermined steps for each subdivided process. The smoothing field indicates whether specified smoothing processing for the data is required. The operating condition optimization attributes also include attributes for adjustment / monitoring, cost impact, management scope, and set range. The adjustment / monitoring field indicates whether the data is subject to adjustment or monitoring during optimization. The cost impact field indicates the cost per specified unit affecting adjustments made during optimization. The management range attributes also include upper and lower limit attributes, with the upper and lower limit fields recording information indicating the permissible range of the sensor's output value. The setting range attributes also include upper and lower limit attributes, with the upper and lower limit fields recording information indicating the target range for the sensor's output value. Alternatively, the prediction device 1 can be configured to perform multi-purpose optimization or single-purpose optimization based on the information recorded in the fields for optimizing operating conditions as described above.

[0057] Figure 6This diagram illustrates an example of a tag combination table generated based on a knowledge base. The tag combination table represents causal relationship information obtained from the knowledge base, defining the combination of process data as descriptive variables and process data as objective variables. The tag combination table can also be a table in a database, or a file in a predefined format like CSV. Furthermore, the tag combination table can also be pre-generated by the user and read out by the prediction device 1.

[0058] The tag combination table includes attributes such as combination ID, tag, primary cause / effect, causal relationship, learning period, and dependency relationship. The identification information representing the set of causal relationships is entered in the combination ID field. The tag field enters the tag as identification information for the sensor. The primary cause / effect field enters the category indicating whether it is the primary cause system or the influencing system in the causal relationship (in other words, whether it is the explanatory variable or the target variable). The causal relationship field enters the positive or negative category of the constraint on the sign of the change in the output value of the primary cause system corresponding to the tag, which should be made to change the value of the influencing system in a positive or negative direction. Therefore, the causal relationship value is entered in the record with "primary cause" in the primary cause / effect field. In this embodiment, the prediction device 1 generates a prediction model by applying a constraint (called a "sign constraint") with a certain correspondence between the positive and negative directions of the change in the value of the primary cause system and the positive and negative directions of the change in the value of the influencing system. For example, the prediction model is generated in a way that determines the positive or negative direction of the change in the target variable based on the positive or negative direction of the change in the explanatory variable. That is, the sign in the causal relationship field indicates which direction the output value of the sensor corresponding to the tag of that record needs to be changed—positive or negative—in order for the value affecting the system to change in a specified direction (positive or negative). Furthermore, the learning period field records information used to determine the period for which process data is used to generate the predictive model. This information could, for example, be the number of most recent production numbers.

[0059] The prediction device 1 generates a prediction model based on the label attribute table and label combination table as described above. Figure 7 This is a diagram showing an example of a logic tree representing the structure of a prediction model. Each rectangle represents the output or predicted value of a sensor corresponding to a label. The prediction model includes components for upstream (in...) Figure 7 The output values ​​of sensors in the process (left side) are used to predict the output values ​​of sensors affecting the system at the destination connected by arrows. Furthermore, the prediction model is a hierarchical structure comprising multiple predictions, including other predictions that include the predicted values ​​obtained from one prediction in the explanatory variables. Figure 6The label combination table shown uses records with the same combination ID to generate a predictive formula. For example, the output value of the sensor corresponding to the label with "Main Cause" in the Main Cause / Influence field, or its predicted value, is used as the explanatory variable, and the output value of the sensor corresponding to the label with "Influence" in the Main Cause / Influence field, or the characteristic value of the product obtained as an analysis value through process inspection, is used as the objective variable to generate the prescribed predictive formula.

[0060] Specifically, the prediction can be represented, for example, by the following formula (1).

[0061] Y(t)=a1(t)·x1(t)+a2(t)·x2(t)+……+a n (t)·x n (t)+a ar (t)·Y(t-1)+C (1)

[0062] It should be noted that t is the value corresponding to the production number, Y(t) is the predicted value affecting the system, x(t) is the sensor output value or its predicted value of the main cause system, and a(t) is the coefficient of the main cause system. ar (t) represents the coefficient of the autoregressive term, and C is the constant term. The term corresponding to the principal cause system is only included in... Figure 7 The number of output values ​​from the source sensors connected by arrows. Furthermore, the autoregressive term is a predicted or measured value from past production numbers. There is no limit to one autoregressive term; the prediction may also include autoregressive terms from multiple recent production numbers.

[0063] Furthermore, the prediction device 1 performs learning processing for each production number in the batch process, updating the coefficients of the prediction formula. For example, the coefficients will be set to... Figure 6 The process data corresponding to the most recent specified production number in the field during the learning period is used as learning data, and regression analysis is performed to determine the coefficients. Here, we assume that the coefficients are determined in a manner that satisfies the aforementioned sign restrictions.

[0064] <Continuous Process>

[0065] Figure 8 This is a diagram used to illustrate an example of process data in a continuous process. Figure 8 The columns on the left represent Figure 2 This is part of the process of the continuous step 32 shown. Specifically, the process includes tank 311 and pump 312. Figure 8 The column on the right represents an example of process data acquired in each process. In continuous process 32, data is continuously acquired from sensors that corresponds to tags but not to production numbers. Figure 8In the example, timing data is acquired from sensors labeled 102 and 103. In a continuous process, the machine continuously receives and processes objects. When a continuous process follows a batch process, traceability information pre-defined by the user in this embodiment is used to associate the objects processed in the batch process with those processed in the continuous process. The traceability information includes sampling intervals and dwell times. The sampling interval represents the interval at which sampling for process inspection is performed in the continuous process, for example, using a reduction method. The dwell time represents the time the object remains from the completion of the batch process until it reaches the process included in the continuous process.

[0066] Figure 9 This diagram illustrates the correspondence between samples from process inspections in continuous processes and production numbers in batch processes. For example, process inspections are performed at predetermined intervals, and the samples from these inspections serve as reduced samples corresponding to those intervals. Furthermore, when a continuous process follows a batch process, the products obtained from the batch process completed within the predetermined period are imported as the processing objects for the continuous process. Therefore, the reduced samples of process data in the continuous process can be correlated with the range of completion times of the batch processes by tracing the dwell time back to the sampling point, thus determining the corresponding production number group for the batch process. By establishing this correlation, the accuracy of predictive formulas utilizing process data from the batch processes can be improved when both batch and continuous processes are performed consecutively.

[0067] Figure 10 This is a graph illustrating the dwell time from the sensor position in a continuous process to the sampling position during process inspection. The aforementioned reduced sample is, for example, calculated as the average value of process data obtained during a specified period in the continuous process. In order to establish a correspondence between the sampling time point where this calculation is performed and the range of acquisition time points for which the process data is averaged at that time point, in this embodiment, the dwell time of this interval is set for the interval between the sensor position that outputs process data in the continuous process and the sampling position for process inspection. This allows for the establishment of a correlation between the process data in the continuous process and the reduced sample of the process inspection. By establishing such a correlation, the accuracy of predictive formulas using process data from the continuous process can be improved.

[0068] Figure 11 This diagram illustrates an example of a tag attribute table for a series of processes generated based on a knowledge base. The tag attribute table for a series of processes can also be a table in a database, or a file in a predefined format like CSV. Furthermore, the tag attribute table is pre-generated by the user and read out by the prediction device 1.

[0069] The label attribute table for continuous processes includes labels, categories, dwell time, batch-associated labels, and attributes for optimizing operating conditions. It should be noted that for... Figure 5 For attributes with the same name in the label attribute table for batch operations, their descriptions are omitted. Enter the category (Continuous, Batch, or Quality) in the Category field. "Continuous" in the Category field indicates that the labels shown for each record represent process data from a continuous operation. "Batch" indicates process data from a batch operation. "Quality" indicates the analysis values ​​of a reduced sample during process inspection.

[0070] Furthermore, in continuous processes, the prediction device 1 also utilizes, for example... Figure 6 The label combination table shown is used to generate a prediction model based on the label attribute table and the label combination table. Figure 12 This is a diagram illustrating an example of a logic tree representing the structure of a predictive model. Figure 12 In the diagram, each rectangle represents the output or predicted value of the sensor corresponding to the tag. The rectangle labeled 101 represents the process data for a batch operation. (For example, using...) Figure 9 As explained, the process data for batch operations is based on dwell time to determine the corresponding production number and series, and these averages are used as explanatory variables for the forecast. Rectangles labeled 102 and 103 represent process data for consecutive operations. (If using...) Figure 10 As explained, process data for consecutive processes are determined based on dwell time and sampling intervals to define the corresponding periods, with the average process data within each period used as a predictive explanatory variable. The rectangle labeled 104 represents the analytical values ​​for process checks (also known as quality checks), for example, values ​​corresponding to the process data obtained through reduction. In consecutive processes, this will also be... Figure 6 The label combination table shown contains records with the same combination ID to generate the prediction formula. The prediction formula is the same as the batch process, so its description is omitted.

[0071] In this embodiment, process data selected based on a knowledge base is used. Therefore, predictive models can be generated quickly based on parameters with clear causal relationships, without requiring a large number of parameters for the entire unit. Furthermore, if a predictive model including autoregressive terms is generated, predictions can be made that take into account the temporal changes in process data that cannot be reflected by simulations based solely on process data at a single point in time.

[0072] Furthermore, without the sign restrictions described above, using predictive models to solve inverse problems sometimes fails to yield satisfactory results. That is, when determining the corresponding operating conditions to specify the required quality, predictive models may produce results that violate the principles of the process. By imposing the sign restrictions described above, predictive models that adhere to the principles of the process can be generated. In other words, by using predictive formulas that satisfy the sign restrictions, it is not only possible to predict the values ​​of the system's influence as substitute indicators for the product, but also to easily understand how to change the unit's operating conditions to improve quality.

[0073] <Control>

[0074] Alternatively, the predictive device 1 can use the generated predictive formulas and process data to determine the operating conditions for transitioning to a stable state or for ensuring the product meets specified requirements, based on changes in the state that are influencing the process, and control the unit 3 based on these operating conditions. For example, it can determine target values ​​for a portion of the characteristic values ​​and allowable ranges for other controllable process data, thus determining the preferred operating conditions. Furthermore, it can be configured to pre-set unit prices for at least a portion of the process data, for example, determining the operating conditions that minimize costs or satisfy allowable cost ranges.

[0075] As mentioned above, in Figure 5 The table shown uses the "Management Scope" field to set the allowable range for each process data point. The "Set Range" field sets the target value for each process data point. Furthermore, the "Cost Impact (Unit Price)" field sets the cost per specified unit quantity for each process data point. It should be noted that process data marked "Adjust" in the "Adjust / Monitor" field indicates values ​​that can be adjusted using actuators or similar devices provided with control unit 3. During control, operating conditions for adjustable process data are determined, for example, based on targets and allowable ranges.

[0076] exist Figure 5 In the example, the cost can be calculated by summing the unit price set in the "Cost Impact (Unit Price)" field and the product of the process data values ​​corresponding to each label. Then, the values ​​of controllable process data (operating conditions) are calculated to minimize the cost as the objective function.

[0077] Furthermore, restrictions are imposed based on a set range to ensure that the predicted values ​​calculated through the predictive algorithm fall within the logged-in range. Figure 5 The "set range" refers to the range of process data, such as quality or quality substitute indicators, where the range can be described as a target value set according to the required specifications.

[0078] Furthermore, the possible values ​​of controllable process data are limited based on the scope of management. Process data corresponding to those labeled "Adjust" in the "Adjust / Monitor" field can be adjusted, but restrictions can be set, for example, by setting limits based on parameters such as those of Unit 3.

[0079] Furthermore, in this embodiment, the prediction model generated during the prediction process is also used as a constraint. That is, constraints representing a specified range are set for at least a portion of the objective variable in the prediction formula, and the optimal value of the descriptive variable within the range of the constraints is searched for.

[0080] Optimization can be performed using linear programming, where the objective function is minimized among variables satisfying linear inequalities and equations that serve as constraints. It should be noted that the objective variable can also be any predicted value of process data, not cost. In this case, the acceptable range can be determined for cost. Furthermore, even if the constraints or part of the objective function are nonlinear, it can be solved using existing nonlinear programming methods. Additionally, multiple objective functions can be defined for multi-objective optimization. As described above, operating conditions can be determined by solving the optimization problem.

[0081] For example, if the process data for the object being adjusted is the input amount of auxiliary raw materials, the calculated optimal solution is directly used as the setpoint. Alternatively, if the process data for the object being adjusted is the integral value of the temperature of the object being processed, actuators such as valves are adjusted to approximate the calculated optimal solution.

[0082] <Device Composition>

[0083] Figure 13This is a block diagram illustrating an example of the configuration of the prediction device 1. The prediction device 1 is a general-purpose computer, comprising: 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 interface card (NIC) or a communication module, which communicates with other computers based on a defined protocol. The storage device 12 may be a main storage device such as RAM (Random Access Memory) or ROM (Read Only Memory), and an auxiliary storage device (secondary storage device) such as HDD (Hard-Disk Drive), SSD (Solid State Drive), or flash memory. The main storage device temporarily stores programs read by the processor 14, information transmitted and received between other computers, or ensures the working area of ​​the processor 14. The auxiliary storage device stores programs executed by the processor 14, information transmitted and received between other computers, etc. The input / output device 13 may be a user interface such as an input device like a keyboard or mouse, an output device like a monitor, or an input / output device like a touch panel. The processor 14 is a processing unit such as a CPU (Central Processing Unit), which performs various processes in this embodiment by executing programs. Figure 13 In the example, functional blocks are shown within 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 generation unit 143, a quality prediction unit 144, and a unit control unit 145 by executing prescribed programs.

[0084] The process data acquisition unit 141 acquires process data from sensors provided with the unit 3, for example, via communication I / F 11 and control station 2, and stores it in storage device 12. As described above, the process data is associated with the sensors via tags.

[0085] When generating the prediction model, the process data processing unit 142 is based on... Figure 5 The label attribute table of the batch process shown or Figure 11 The process data is processed using the label attribute table of the continuous processes shown. Specifically, the process data processing unit 142 extracts instantaneous values ​​at specified times, calculates average values ​​over a specified period, or calculates integral values ​​over a specified period based on information from the fields of the primary process in the label attribute table logged to the batch processes. Furthermore, the process data processing unit 142 can also calculate average values ​​of process data corresponding to specified labels, systems, and production numbers, or calculate average values ​​for a period determined based on traceability information, for the process data of the continuous processes, based on information from the fields of batch-associated labels logged to the label attribute table of the continuous processes and, for example, the traceability information pre-stored in the storage device 12.

[0086] Prediction model generation unit 143, for example, based on Figure 6 The label combination table shown is used to generate a prediction model including the prediction formula shown in formula (1) above, and the model is stored in storage device 12. The prediction model generation unit 143 may, for example, update the coefficients of the prediction formula using data from the most recent specified period for each production number in a batch process. In addition, the prediction model generation unit 143 may, for example, update the coefficients of the prediction formula using the most recent data for each specified period in a continuous process.

[0087] The quality prediction unit 144 uses process data and prediction models to predict specified sensor output values ​​and process inspection analysis values. It should be noted that the quality prediction unit 144 can also use data and prediction models based on arbitrary operating conditions to calculate predicted values ​​after changes in operating conditions.

[0088] The unit control unit 145 controls, for example, the valves and other actuators, and other machinery of the unit 3 via communication I / F 11 and control station 2. Furthermore, the unit control unit 145 can determine, for example, the operating conditions that satisfy specified quality and cost requirements, and control the unit 3 accordingly. Additionally, the unit control unit 145 can determine the operating conditions for transitioning to a specified stable state, or the operating conditions for ensuring that products meet specified requirements, and control the unit 3 accordingly.

[0089] The components described above are connected via bus 15.

[0090] Predictive Processing (Batch Processing)

[0091] Figure 14 This is a flowchart illustrating an example of the prediction processing performed by prediction device 1. The processor 14 of prediction device 1 executes a prescribed program to perform... Figure 14 The processing is shown below. Predictive processing is performed per production number in batch processes and at specified sampling intervals in continuous processes. It should be noted that this is assumed to be user-generated. Figure 5 The label attribute table for batch processes shown below. Figure 6 The label combination table shown Figure 11 The label attribute table and traceability information of the continuous processes shown are pre-stored in the storage device 12. Furthermore, the process data acquisition unit 141 continuously acquires process data from sensors on the unit 3 via, for example, communication I / F 11 and control station 2, and stores it temporarily or permanently in the storage device 12. The process data is described, for example, according to a specified specification such as OPC.

[0092] The process data acquisition unit 141 of the prediction device 1 reads the setting information. Figure 14(S1). In this step, the process data acquisition unit 141 reads the tag attribute table, tag combination table, traceability information, etc. from the storage device 12.

[0093] In addition, the process data acquisition unit 141 reads in the process data ( Figure 14 (S2). In this step, the process data corresponding to the labels used for the predictions are extracted, for example, by each prediction, by each series, and by each subdivided process. Figure 15 This diagram illustrates an example of a write arrangement used in batch operations to write data read in this step. The write arrangement for batch operations can be OPC data, a so-called database table, or a file in a specified format like CSV. Figure 15 The table includes attributes for date and time, product number, variety, step, and label. The date and time of the sensor output measurement should be entered in the date and time field. The production number should be entered in the product number field. The category of the processed object should be entered in the variety field. The step field should enter information representing the stage in the process as defined by predefined steps. The label field should enter the sensor output value corresponding to each label.

[0094] Furthermore, the process data processing unit 142 of the prediction device 1 performs prescribed processing on the process data. Figure 14 :S3). Use Figure 16 Let me explain the details of this step. Figure 16 This is a process flow diagram illustrating an example of processing. Process data processing unit 142 is in... Figure 15 When extracting process data by each prediction, each series, and each subdivided operation from the write arrangement shown, the following operations are performed on each record of the write arrangement: Figure 16 The processing shown.

[0095] Process data processing unit 142 reads records from the writing array ( Figure 16 (S11). In this step, from... Figure 15 The table shown reads one record at a time. Furthermore, the process data processing unit 142 processes the data according to a single processing method. Figure 16 (S12). In this step, refer to... Figure 5 The label attribute table shown is based on the category of the "method" field of the "one-time processing" of the corresponding label, such as calculating instantaneous value, average value, integral value, differential coefficient, difference, maximum value, minimum value, thermal history, or process data itself. Figure 17 This is a diagram illustrating an example of batching the processing results written to this step. Batching can also be OPC data, a so-called database table, or a file in a specified format like CSV. Figure 17 The table includes product number, end date and time, and label attributes. Enter the production number in the product number field. Enter the end date and time of the batch process for this product number in the end date and time field. Enter the processed process data in the label field.

[0096] Here, the calculation of thermal history will be explained. Thermal history is, for example, information indicating the extent to which a reaction proceeds in a general chemical reaction such as depolymerization, acetylation, and deacetylation. In this embodiment, the thermal history is calculated as an integral value of the reaction rate over a specified period. For example, as shown in the following formula (2), the calculation is performed based on the integral value of the reaction rate formula using the Arrhenius equation.

[0097] [Formula 1]

[0098]

[0099] Here, A is the frequency factor. E is the activation energy. R is the gas constant. Furthermore, A(t) and B(t) are concentration terms, and m and n are the number of reactions. These values ​​are defined according to the reaction and the object being processed. Additionally, t represents a specified interval within the process. T(t) is the temperature at that step, obtained as process data. Through such processing, the heat received by the object during the specified period can be used for quality prediction.

[0100] In addition, the process data processing department 142 performs the prescribed data cleaning process ( Figure 16 (S13) Data cleaning is the process of eliminating outliers, and various methods can be used. For example, a moving average can be calculated using the most recent data. Furthermore, the difference between the moving average and the measured value is taken to determine the standard deviation σ (also called the error variance), representing the unevenness of the difference. Then, for example, values ​​that do not fall within a defined reliable interval (also called the 3σ interval) from the mean of the probability distribution to the mean of the probability distribution +3σ. Similarly, values ​​that do not fall within the 3σ interval can be excluded based on the difference between consecutive measured values. Data cleaning is performed, for example, on instantaneous values ​​and process data at the end of a batch.

[0101] In addition, the process data processing department 142 performs the prescribed smoothing process ( Figure 16 S14). Smoothing processing is aimed at... Figure 5 The label attribute table shown lists the "required" labels in the smoothing field. Furthermore, smoothing can also be, for example, calculating a moving average of the most recent specified values ​​after data cleaning, or other methods that can smooth the data. That concludes the explanation. Figure 16 The processing and return Figure 14 The processing.

[0102] Then, the prediction model generation unit 143 of the prediction device 1 performs prediction model construction processing. Figure 14 (S4). In this step, based on Figure 6 The label combination table shown is used to generate a predictive formula that constitutes the predictive model. Specifically, processed process data with labels attached to the same combination ID is read out, and the processed process data is substituted into the predictive formula (e.g., the formula (1) mentioned above) based on the category of the field registered to the main cause / influence. The coefficients and constant terms of the predictive formula are determined by regression analysis. At this time, the processed process data uses the most recent data as the learning object based on the value of the field registered to the learning period. It should be noted that the predictive model generation unit 143 can also be configured to search for preferred values ​​based on the size of the learning period. For example, the correlation coefficient is calculated using the generated predictive model and process data, and the learning period is set in a way to improve the correlation coefficient. In addition, the coefficients of the predictive formula are determined based on the sign of the field registered to the causal relationship, under the constraint that there is a certain correspondence between the direction of change of the value of the main cause system and the direction of change of the value of the influence system. Then, the predictive model generation unit 143 stores the generated predictive formula in the storage device 12.

[0103] The quality prediction unit 144 of the prediction device 1 performs prediction processing using the prediction model and process data or their predicted values ​​generated in the prediction model building process. Figure 14 (S5). For convenience, in Figure 14 The processing flow shows predictive processing; however, the quality prediction unit 144 can perform predictive processing at any point in time using the predictive model and process data. In this step, the quality prediction unit 144 reads the most recent predictive model and process data, substitutes the process data or its predicted value into the predictive formula included in the predictive model, and calculates the predicted value corresponding to any influencing system.

[0104] Figure 18 It is a graph used to illustrate the process data or its predicted values ​​substituted into the predictive formula. In the prediction... Figure 4 In the case of the value of label 007 shown, such as Figure 18 As shown, in Figure 3 At time point t1, the measured or predicted values ​​of the sensors corresponding to each tag are used. That is, when predicting the value of tag 007 with production number 003, the measured values ​​of tags 001 to 004 with production numbers 003 and the predicted values ​​of tags 005 and 006 with unknown production numbers 003 are substituted into the prediction formula for tag 007.

[0105] Figure 19 These are diagrams illustrating other examples of process data substituted into a predictive formula or its predicted values. In the prediction... Figure 4In the case of the value of label 007 shown, such as Figure 19 As shown, in Figure 3 At time point t2, the measured or predicted values ​​of the sensors corresponding to each tag are used. That is, when predicting the value of tag 007 with production number 004, the known measured values ​​of tags 001 and 002 with production number 004 and the unknown predicted values ​​of tags 003, 005, and 006 with production number 004 are substituted into the prediction formula for tag 007. Here, as... Figure 4 As shown, there is no predictive value for tag 004, so the measured value of the most recent production number is used.

[0106] Furthermore, the values ​​input to the prediction model are not limited to process data; for example, they can be based on data from arbitrary operating conditions. In this case, it is possible to predict the results under changed operating conditions of Unit 3. As described above, the quality prediction unit 144 uses the prediction model and the output values ​​or predicted values ​​of the sensors to, for example, calculate the predicted values ​​of the specified process data.

[0107] Alternatively, the quality prediction unit 144 can calculate the predicted value or the measured value of the process data, and then determine the specified reliability range obtained in the data cleaning process described above. The input / output device 13, such as a monitor, can then display the specified reliability range and the predicted or measured value on a graph. In this case, the user can visually grasp the trend and use it as material to determine whether the operating conditions of unit 3 should be changed.

[0108] In addition, the unit control unit 145 can automatically change the operating conditions of the unit 3 based on the calculated predicted values, or output information to the user requesting changes to the operating conditions via the input / output device 13.

[0109] <Predictive Processing (Continuous Processes)>

[0110] In continuous processes, it is also carried out Figure 14 The predictive processing is shown below. The following explanation focuses on the differences from batch processing. It should be noted that, assuming the batch processing ends by each series and each production number, the date and time of completion of the transfer (pipetting) of the processed object to the subsequent process is stored in storage device 12.

[0111] In the process data reading ( Figure 14 In S2), in the case of continuous processes, data is continuously rolled over by deleting old data as new data is written, without using production numbers as the unit. Alternatively, it can be based on login... Figure 11 The information in the category field of the label attribute table shown is used to change the information. Figure 15 The data structure shown is arranged for writing. Figure 20This diagram illustrates an example of an arrangement used to write data that maintains the category of "continuous". For example, an arrangement can be used to write data that maintains the category of "continuous". Figure 15 The product number and process details were deleted from the table. Additionally, Figure 21 This diagram illustrates an example of an arrangement for maintaining the writing of data categorized as "batch". An example of such an arrangement is a table that maintains the production number and the process data for that production number, generated for each system as follows: Figure 21 The table shown. Log in Figure 21 The data in the table can also be logged in. Figure 17 The processed data arranged in batches.

[0112] In addition, the processing of process data ( Figure 14 S3) For example, it is performed based on the timing of process inspections. The timing of process inspections is defined as the sampling interval in the traceability information. Figure 22 This diagram illustrates an example of a combined ID data arrangement that maintains the processed process data in a continuous workflow. The combined ID data arrangement can also be OPC data, a database table, or a file in a specified format like CSV. Figure 22 The table records identification information for each process inspection in the Process Inspection ID field, which includes attributes such as Process Inspection ID, Sampling, and Label. The Sampling attributes also include Start Date and Time, and End Date and Time. The Start Date and Time and End Date and Time fields record the start date and time of sampling for process inspections performed using the reduction method, respectively. The Label field records the analysis values ​​corresponding to each label. Here, for... Figure 11 The label attribute table shown has a "continuous" label in the category field, which represents the average value of process data from the processing time point to the time point of the sampling interval traced in the traceability information. Furthermore, for example... Figure 11 The label attribute table shown has the label "Batch" in the category field. The label is the average value of the process data corresponding to the production number where the pipetting was completed from the processing time point to the time point when the sampling interval is traced in the traceability information.

[0113] In addition, the prediction model building process ( Figure 14 S4) For example, when the prediction model generation unit 143 obtains the analysis value of the process inspection, the prediction model for that analysis value is updated. In this step, it is also based on... Figure 6 The labels shown are combined to generate the predictive formulas that constitute the predictive model. Furthermore, in continuous processes, it is also possible to... Figure 9As shown, the prediction model generation unit 143, based on the completion timing of batch processes and a predetermined dwell time, establishes a correspondence between the process data of the main cause system and the process data of the influencing system, and learns the characteristics of the process data obtained from unit 3. Alternatively, it can also be as follows... Figure 10 As shown, the prediction model generation unit 143 establishes a correspondence between the process data of the main cause system and the process data of the influencing system based on the difference between the acquisition timing of the process data of the main cause system and the acquisition timing of the process data of the influencing system, and learns the characteristics of the process data obtained from the unit 3.

[0114] The quality prediction unit 144 of the prediction device 1 performs prediction processing using the prediction model and process data or their predicted values ​​generated in the prediction model building process. Figure 14 (S5). In this step, the quality prediction unit 144 reads the most recent prediction model and process data, substitutes the process data or its predicted value into the prediction formula included in the prediction model, and calculates the predicted value corresponding to any influencing system.

[0115] <Control Processing>

[0116] Figure 23 This is a flowchart illustrating an example of the control processing performed by the prediction device 1. The processor 14 of the prediction device 1 executes, for example, a predetermined program. Figure 23 The process is shown. Control processing can be executed at arbitrary time intervals, such as after updating the prediction model. In the control processing, it is also set to... Figure 5 The label attribute table for batch processes shown below. Figure 6 The label combination table shown Figure 11 The label attribute table and traceability information of the continuous process shown are pre-stored in the storage device 12. In addition, the process data acquisition unit 141 continuously acquires process data from the sensors of the unit 3 via, for example, communication I / F 11 and control station 2, and stores it temporarily or permanently in the storage device 12.

[0117] The unit control unit 145 of the prediction device 1 reads the setting information. Figure 23 (S21). In this step, the unit control unit 145 reads the tag attribute table, tag combination table, traceability information, etc. from the storage device 12. During control processing, the read information includes, for example, the control objective represented as a goal function of an optimization problem, and, for example, the control tolerance range represented as a constraint condition of an optimization problem. Here, it is assumed that logging in will be used... Figure 5 The cost impact field is used to calculate the minimum cost based on the unit price, which is the objective function. Furthermore, the values ​​of fields set to the management scope and defined scope become constraints.

[0118] In addition, the process data acquisition unit 141 reads in the process data ( Figure 23 (S22). The processing in this step is related to... Figure 14 The same as S2. In this step, process data corresponding to the labels used for prediction are extracted, for example, by each prediction, by each series, and by each subdivided process. Furthermore, in Figure 15 or Figure 20 The output values ​​of the sensor are logged in the arrangement shown in the diagram.

[0119] In addition, the process data processing unit 142 performs prescribed processing on the process data. Figure 23 (S23). The processing in this step is related to... Figure 14 It is the same as S3.

[0120] Then, the unit control unit 145 performs optimization problem calculations. Figure 23 (S24). In this step, the operating conditions that minimize or maximize the objective function are determined under the read constraints. For example, based on... Figure 5 The cost is calculated using the following formula (3) under the condition that the setting shown minimizes the cost.

[0121] Cost = (Process data of label 001 × Unit price) + (Process data of label 002 × Unit price) + (Process data of label 004 × Unit price) (3)

[0122] In addition, by logging in Figure 5 The information of the "setting range" is used as a restriction condition. Specifically, the conditions are set as follows.

[0123] The lower bound of label 005 ≤ the predicted value of label 005 ≤ the upper bound of label 005

[0124] The lower bound of label 007 ≤ the predicted value of label 007 ≤ the upper bound of label 007

[0125] In addition, by logging in Figure 5 The information regarding the "scope of management" is used as another constraint. Specifically, the conditions are set as follows.

[0126] The lower limit of label 001 ≤ the value of label 001 ≤ the upper limit of label 001

[0127] The lower limit of label 002 ≤ the value of label 002 ≤ the upper limit of label 002

[0128] The lower limit of label 003 ≤ the value of label 003 ≤ the upper limit of label 003

[0129] The lower limit of label 005 ≤ the value of label 005 ≤ the upper limit of label 005

[0130] Furthermore, in this embodiment, the prediction model generated during the prediction process is also used to constrain the conditions. For example, in situations such as... Figure 7 and Figure 12 In the logic tree shown, where the prediction formula is defined between process data, the downstream process data is calculated based on the prediction formula constructed in the prediction processing, using the upstream process data. Furthermore, in S22, the process data located at the upstream end is obtained and related to... Figure 5 The process data corresponding to the "monitor" tag is logged in the "Adjustment / Monitoring" field, and the processed value is used in S23. Then, with the above-mentioned setting range and management range set for the target variable in each prediction, the constraints representing these ranges are set, and the optimal value of the description variable that falls within the range of the constraints is searched.

[0131] The optimization problem described above can be solved using existing solutions. It should be noted that this will be related to... Figure 5 In the "Adjustment / Monitoring" section, the process data corresponding to the "Adjustment" tag are used as the adjustment objects. That is, the process data with tags 001, 003, 004, and 006 are the adjustment objects, and the tags of the objects do not necessarily match the constraints.

[0132] When solving optimization problems and determining the operating conditions, including the setpoints of the process data of the adjustment objects, the unit control unit 145 controls the unit 3 according to the operating conditions. Figure 23 (S25). In this step, the unit control unit 145 outputs data representing the operating conditions to the control station 2 via the communication I / F11. Then, it controls the operation of the unit 3 according to the control signals from the control station 2. It should be noted that, for example, if the multi-purpose optimization problem is solved in S24, multiple candidates for operating conditions can be presented to the user via the input / output device 13, and the unit 3 can be controlled based on the operating conditions selected by the user.

[0133] <Variation Example>

[0134] It should be noted that the various components and combinations thereof in each embodiment are merely examples. Without departing from the spirit of the invention, appropriate additions, omissions, substitutions, and other modifications to the components can be made. This disclosure is not limited to the embodiments, but only to the claims. Furthermore, the various solutions disclosed in this specification can be combined with any other features disclosed in this specification.

[0135] Furthermore, while the above embodiments have been described using a chemical unit as an example, they can also be applied to manufacturing processes in general production equipment. For instance, batch numbers can be used instead of production numbers for batch operations in the embodiments as processing units, and processing according to batch operations in the embodiments can be applied.

[0136] Alternatively, at least a portion of the function of the prediction device 1 can be distributed among multiple devices, or multiple devices can provide the same function in parallel. For example, the model generation device that generates the prediction model, the prediction device that uses the generated prediction model to make predictions, and the control device that uses the generated prediction model to control the production equipment can also be different. Furthermore, at least a portion of the function of the prediction device 3 can be located in a so-called cloud.

[0137] Furthermore, the above formula (1) is a linear model that includes an autoregressive term, but it is not limited to such an example. For example, a model that does not include an autoregressive term can also be used. In addition, the model can be linear or nonlinear. Furthermore, it can be a single formula, for example, a state-space model that incorporates periodic variations such as seasonal variations can also be used. However, it is preferred to use a model that satisfies the sign restriction. That is, the coefficients of the predictive formula are determined under the restriction that there is a certain correspondence between the direction of the change in the value of the main cause system and the direction of the change in the value of the influencing system.

[0138] Furthermore, this disclosure includes a method for performing the above-described processing, a computer program, and a computer-readable recording medium on which the program is recorded. The recording medium on which the program is recorded can perform the above-described processing by causing a computer to execute the program.

[0139] Here, computer-readable recording media refers to recording media that can accumulate information such as data and programs through electrical, magnetic, optical, mechanical, or chemical processes and can be read from a computer. Examples of such recording media that can be removed from a computer include floppy disks, magneto-optical disks, optical disks, magnetic tapes, and memory cards. Furthermore, recording media that are fixed to a computer include HDDs, SSDs (Solid State Drives), and ROMs.

[0140] Explanation of reference numerals in the attached figures

[0141] 1: Prediction device

[0142] 11: Communication I / F

[0143] 12: Storage device

[0144] 13: Input / output devices

[0145] 14: Processor

[0146] 141: Process Data Acquisition Department

[0147] 142: Process Data Processing Department

[0148] 143: Predictive Model Generation Department

[0149] 144: Quality Forecasting Department

[0150] 145: Unit Control Department

[0151] 2: Control Station

[0152] 3: Generating Unit

Claims

1. A prediction device comprising: The process data processing department performs prescribed processing on the process data obtained from the chemical unit; and The prediction model generation unit generates a prediction model based on causal relationship information learned from the features of the process data obtained from the chemical unit, wherein... The causal relationship information is generated based on a knowledge base, which defines the combination of the first process data as an explanatory variable and the second process data as the objective variable. The prediction model is a prediction model of the quality represented by the objective variable. The first process data includes process data from processes that have undergone the specified processing, and is a value established corresponding to the target variable as a parameter that affects the target variable in the causal relationship information. The second process data is obtained from the chemical unit and serves as a substitute indicator for predicting the quality of the product. The specified processing procedure is a process that uses the process data obtained from the chemical unit to determine a value corresponding to the reaction rate of the object being processed within a specified period. The value corresponding to the reaction rate is an integral value obtained by further integrating the reaction rate over the specified period.

2. The prediction device according to claim 1, wherein, The value corresponding to the reaction rate is the integral value of the reaction rate.

3. The prediction device according to claim 2, wherein, The values ​​corresponding to the reaction rate are calculated using the Arrhenius formula, which determines the frequency factor and activation energy of the reaction rate constant based on the type of chemical reaction and the object being processed in the chemical unit.

4. The prediction device according to any one of claims 1 to 3, wherein, The process data processing unit also calculates the integral value, differential value, average value, maximum value or minimum value of the process data within a specified period, or calculates the instantaneous value of the process data at a specified time point, based on the category of the process data.

5. The prediction device according to any one of claims 1 to 3, wherein, The prediction model is a hierarchical structure that includes multiple prediction formulas, and has a second prediction formula that includes the predicted value calculated by the first prediction formula in the explanatory variables.

6. The prediction device according to any one of claims 1 to 3, wherein, The value corresponding to the second process data is obtained by sampling multiple second process data using a reduction method. The prediction model generation unit generates the prediction model by establishing a correspondence between the acquisition timing range of the first process data in the chemical unit and the calculation timing of the corresponding value of the second process data, based on the residence time of the processed object in the chemical unit.

7. The prediction device according to any one of claims 1 to 3, wherein, The chemical unit performs batch processing of the object to be processed in sequence according to each prescribed processing unit, and then performs continuous processing of the object to be processed continuously. The prediction model generation unit generates the prediction model by establishing a correspondence between the range of completion timing of the batch process and the calculation timing of the corresponding value of the second process data based on the residence time of the processed object in the chemical unit.

8. A prediction method, wherein, Prediction device execution: Process data processing steps involve performing prescribed processing procedures on process data obtained from the chemical unit; and The prediction model generation step involves generating a prediction model based on causal relationship information learned from the features of the process data obtained from the chemical unit. This causal relationship information is generated based on a knowledge base and defines the combination of first process data as a descriptive variable and second process data as a target variable. The prediction model is a prediction model for the quality represented by the target variable. The first process data includes process data from processes that have undergone the specified processing, and is a value established corresponding to the target variable as a parameter that affects the target variable in the causal relationship information. The second process data is obtained from the chemical unit and serves as a substitute indicator for predicting the quality of the product. The specified processing procedure is a process that uses the process data obtained from the chemical unit to determine a value corresponding to the reaction rate of the object being processed within a specified period. The value corresponding to the reaction rate is an integral value obtained by further integrating the reaction rate over the specified period.

9. A program product, wherein, To make the computer perform: Process data processing steps involve performing prescribed processing procedures on process data obtained from the chemical unit; and The predictive model generation step involves generating a predictive model based on causal relationship information learned from the features of the process data obtained from the chemical unit. This causal relationship information is generated based on a knowledge base, which defines the combination of first process data as a descriptive variable and second process data as a target variable. The predictive model is a predictive model of the quality represented by the target variable. The first process data includes process data from processes that have undergone the specified processing, and is a value established corresponding to the target variable as a parameter that affects the target variable in the causal relationship information. The second process data is obtained from the chemical unit and serves as a substitute indicator for predicting the quality of the product. The specified processing procedure is a process that uses the process data obtained from the chemical unit to determine a value corresponding to the reaction rate of the object being processed within a specified period. The value corresponding to the reaction rate is an integral value obtained by further integrating the reaction rate over the specified period.

Images