Optimization system

Abstract

To provide a technology for reducing the management cost of observation data while properly operating an actual device.SOLUTION: The optimization system 1 includes a real apparatus 2 and a virtual apparatus 4. The real apparatus 2 includes a thermostat 21, a control unit 23, a temperature sensor 25, and a data transmission unit 27. The control unit 23 controls the incubator 21 according to the gain matrix g. The data transmission unit 27 transmits the temperature data acquired by the temperature sensor 25 to the virtual device 4 at a set transmission rate. The virtual device 4 includes a data storage unit 41, a system optimization unit 43, and a system update unit 45. The system optimization unit 43 optimizes the gain matrix g of the control unit 23 and the transmission rate of the data transmission unit 27 in the real apparatus 2 based on the temperature data stored in the data storage unit 41. The system updating unit 45 updates the control unit 23 and the data transmitting unit 27 based on the gain matrix g and the transmission rate optimized by the system optimizing unit 43.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The subject matter disclosed herein relates to an optimization system. [Background technology]

[0002] Using a digital twin, which is a virtual model that digitally reproduces an actual physical device (real device), the operation of the real device is monitored in real time and the control parameters that control the real device are optimized (for example, Patent Document 1).

[0003] In a digital twin, observation data is collected from sensors attached to the actual equipment and transmitted to the virtual equipment. In the virtual equipment, a predictive model operates based on the collected observation data, and future operation and status are calculated as predicted values. These predicted values ​​are used to set parameters to optimize the operation of the actual equipment.

[0004] Digital twins also undergo model validation, a process in which observed data is compared with the output of a predictive model to verify its accuracy. If the accuracy of a predictive model declines, updating the model enables more reliable control. Digital twins are a technology that achieves real-time control and optimization through continuous data transmission and feedback between real and virtual equipment, supporting ideal operation of real equipment. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2022-185394 Summary of the Invention [Problem to be solved by the invention]

[0006] In order to properly operate a digital twin, it is necessary to continuously acquire observation data and manage it so that it can be analyzed. However, if all data acquired by sensors is stored, various issues can arise, such as high-load data transfer processing, bandwidth pressure due to increased data communication volume, and increased data storage costs. Furthermore, the need for preprocessing for analysis increases computational costs. On the other hand, if the amount of data transferred is reduced too much, the accuracy of the digital twin will decrease, making it difficult to properly operate the actual equipment.

[0007] An object of the present invention is to provide a technique for reducing the management costs of observation data while properly operating actual equipment. [Means for solving the problem]

[0008] In order to solve the above problem, a first aspect is an optimization system comprising a real device and a virtual device, wherein the real device comprises a control object, a control unit that controls the control object in accordance with control parameters, an observation unit that observes the state of the control object, and a data transmission unit that transmits the observation data acquired by the observation unit to the virtual device at a set transmission rate, and the virtual device comprises a data storage unit that stores the observation data sent from the data transmission unit, a system optimization unit that optimizes the control parameters of the control unit in the real device and the transmission rate of the data transmission unit based on the observation data stored in the data storage unit, and a system update unit that updates the control unit and the data transmission unit based on the control parameters and the transmission rate optimized by the system optimization unit.

[0009] A second aspect is an optimization system of the first aspect, wherein the system optimization unit optimizes the control parameters using a prediction model that predicts the observation data, and the system optimization unit increases the transmission rate when the prediction accuracy of the prediction model decreases.

[0010] A third aspect is an optimization system of the second aspect, in which, when the prediction accuracy of the prediction model decreases, the system optimization unit updates the prediction model using the observation data collected after increasing the transmission rate.

[0011] A fourth aspect is an optimization system according to any one of the first to third aspects, wherein the system update unit updates the transmission rate by updating a transmission interval, which is a period for transmitting the observation data.

[0012] A fifth aspect is the optimization system of the fourth aspect, wherein the data transmission unit selects the latest observation data from the plurality of observation data acquired during the transmission interval and transmits it to the virtual device.

[0013] A sixth aspect is the optimization system of the fourth aspect, wherein the data transmission unit calculates a representative value of the plurality of pieces of observation data acquired during the transmission interval and transmits the representative value to the virtual device. [Effects of the Invention]

[0014] According to the optimization systems of the first to fourth aspects, the transmission load can be appropriately controlled by optimizing the transmission rate, thereby reducing data management costs.

[0015] According to the optimization system of the second aspect, when the prediction accuracy of the prediction model decreases, the transmission rate can be increased to collect more observation data.

[0016] According to the optimization system of the third aspect, by increasing the transmission rate, it is possible to collect observation data in more detail, and therefore it is possible to appropriately update the prediction model using the collected observation data.

[0017] According to the optimization system of the fourth aspect, the transmission rate can be increased by shortening the transmission interval, and the transmission rate can be decreased by lengthening the transmission interval.

[0018] According to the optimization system of the fifth aspect, the latest values ​​of a plurality of observation data are transmitted, making it possible to grasp the most recent situation.

[0019] According to the optimization system of the sixth aspect, it is possible to grasp the overall trend of the observation data acquired during the transmission interval. [Brief explanation of the drawings]

[0020] [Figure 1] FIG. 1 is a block diagram illustrating an optimization system according to an embodiment. [Figure 2] FIG. 4 is a diagram showing changes over time in temperature data output from a temperature sensor. [Figure 3] 2 is a flowchart showing a process executed in the optimization system shown in FIG. 1. DETAILED DESCRIPTION OF THE INVENTION

[0021] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings. Note that the components described in the embodiment are merely examples and are not intended to limit the scope of the present invention. In the drawings, the dimensions and numbers of each part may be exaggerated or simplified as necessary to facilitate understanding.

[0022] <1. Embodiment> 1 is a block diagram showing an optimization system 1 according to an embodiment. The optimization system 1 includes a real device 2 and a virtual device 4. The optimization system 1 optimizes the real device 2 by simulating the operation of the real device 2 in the virtual device 4.

[0023] <Actual device 2> The actual device 2 is a device having a control object and a control unit that controls the control object based on control parameters, and is a device for achieving a preset target state. Here, the actual device 2 is a device intended to adjust temperature, and is equipped with an incubator 21 as a control object. However, the actual device may be configured for purposes other than maintaining temperature, such as position control or pressure control.

[0024] The practical apparatus 2 is used, for example, in a substrate processing apparatus for processing substrates. The substrates are, for example, semiconductor wafers, glass substrates for liquid crystal displays, glass substrates for plasma displays, glass for magnetic / optical disks, ceramic substrates, glass substrates for organic EL displays, glass substrates for solar cells, or various substrates to be processed for electronic devices such as silicon substrates, flexible substrates, and printed circuit boards. However, the practical apparatus 2 may also be used in apparatuses other than substrate processing apparatuses.

[0025] The real device 2 has an incubator 21, a control unit 23, a temperature sensor 25, and a data transmission unit 27. The incubator 21 is a device for adjusting the temperature of an object (e.g., a chemical solution) to a constant value, and has a heater such as an electric heating wire. The control unit 23 controls the incubator 21 in accordance with control parameters. The temperature sensor 25 measures the temperature of the incubator 21 or the temperature of the object whose temperature is adjusted by the incubator 21. The temperature sensor 25 is an example of an "observation unit" that observes the state of the incubator 21 (the controlled object). The temperature data acquired by the temperature sensor 25 is an example of "observation data." The data transmission unit 27 transmits the temperature data (observation data) acquired by the temperature sensor 25 to the data accumulation unit 41 of the virtual device 4 at a preset transmission rate.

[0026] The control unit 23 is configured as a feedback controller such as a PID controller, and has a comparator that compares a predetermined target temperature with the observed temperature input from the temperature sensor 25. The control unit 23 generates a required control signal y using a gain matrix g based on the result of the comparison by the comparator, and outputs the control signal y to the incubator 21. In this way, the control unit 23 adjusts the temperature of the incubator 21 so that it approaches the target temperature. The gain matrix g is an example of a "control parameter," and is updated as appropriate by a system optimization unit 43 (more specifically, a parameter search unit 431) and a system update unit 45 of the virtual device 4, which will be described later.

[0027] Before transmitting the temperature data to the data storage unit 41, the data transmission unit 27 performs a predetermined pre-processing to adjust the amount of data transmitted per unit time (hereinafter referred to as the "transmission rate").

[0028] FIG. 2 shows the temperature data s output from the temperature sensor 25. t It should be noted that in this specification, an alphabetic symbol with an overline attached thereto is referred to as "(alphabetic symbol) overline."

[0029] Here, the output of the temperature sensor 25 at time t is s t The temperature sensor 25 detects the temperature at intervals T s When data is output N times per sampling period, the set of sensor data S output from the temperature sensor 25 is 0→N is expressed by the following equation:

[0030]

number

[0031] The data transmission unit 27 transmits data at a transmission time b t (b t <T s ), transmission interval T t (T s ≦T tThe data transmission unit 27 receives the S 0→N The data to be transmitted to the data storage unit 41 is selected from the data to be transmitted at a transmission time b t After waiting for the transmission interval T t The latest sensor data received at that time is transmitted every mth time. m Overline (m is a natural number), then s m The overline is t≦(b t +(m-1)T t ) the sensor data s corresponding to the maximum t that satisfies t For example, in the example shown in Fig. 2, the first s1 overline selects s2, which is the latest data, from the sensor data s1 and s2.

[0032] The transmission data may be sensor data other than the latest one (for example, the first sensor data) among several pieces of sensor data. t It may be a representative value (average value, median value, mode value, maximum value, minimum value) of several pieces of sensor data received during the period. For example, if the transmission data is the average value, the first transmission data s1 overline shown in Figure 2 is calculated using the following formula:

[0033]

number

[0034] The transmission rate of the data transmission unit 27 is updated by the system update unit 45 of the virtual device 4. The transmission rate is updated at a transmission interval T t The transmission interval T t By dynamically changing the data rate, it becomes possible to transmit an optimal amount of data to the data storage unit 41.

[0035] <Virtual Device 4> Returning to FIG. 1 , the virtual device 4 is a device that simulates the real device 2 and optimizes the operation of the real device 2 in the optimization system 1. The virtual device 4 is a computer that includes a processor such as a CPU and a storage unit configured with a memory such as a RAM and a storage such as an HDD (Hard Disk Drive). The virtual device 4 may be configured with a single computer, or may be configured with multiple computers connected to each other via a network line such as a LAN (Local Area Network) or the Internet. A computer program executable by a computer is stored in the storage unit. The computer program is provided to the virtual device 4, for example, in a state where it is recorded on a non-transitory recording medium.

[0036] The virtual device 4 includes a data storage unit 41, a system optimization unit 43, and a system update unit 45. The data storage unit 41, the system optimization unit 43, and the system update unit 45 are functions realized by a processor that executes a computer program and a storage unit.

[0037] The data accumulation unit 41 accumulates the temperature data transmitted from the data transmission unit 27. The data accumulation unit 41 accumulates the temperature data, the observation time, and the gain matrix g (control parameter) of the control unit 23 at the time the temperature data was observed in a memory unit, in association with each other.

[0038] The system optimization unit 43 optimizes the control parameters of the control unit 23 in the actual device 2 and the transmission rate of the data transmission unit 27 in the actual device 2 based on the temperature data accumulated in the data accumulation unit 41. The system optimization unit 43 includes a prediction model M1, a parameter search unit 431, a model verification unit 433, and a model update unit 435.

[0039] The prediction model M1 is a mathematical or logical model that represents the operation of the incubator 21, which is a physical device, and is implemented by a computer program. For example, a regression model or a neural network can be used as the prediction model M1. The inputs to the prediction model M1 are the observed temperature T (observed value) at the previous time, the control signal y that is the output of the control unit 23, and the gain matrix g (control parameter). The output of the prediction model M1 is the predicted temperature T' (predicted value), which is the temperature at the next time.

[0040] The parameter search unit 431 optimizes the gain matrix g using the prediction model M1 so that the output from the temperature sensor 25 is maintained at the target temperature. That is, the parameter search unit 431 searches for a gain matrix g that reduces the difference between the predicted temperature T′ and the target temperature, using a technique such as gradient descent. When a more appropriate gain matrix g is found, the parameter search unit 431 instructs the system update unit 45 to set the gain matrix g in the control unit 23. In this case, the system update unit 45 updates the control unit 23 based on the gain matrix g (control parameter) optimized by the system optimization unit 43.

[0041] The model verification unit 433 verifies the accuracy of the prediction model M1. Specifically, the model verification unit 433 calculates a control signal y based on the gain matrix g of a data set (gain matrix g and observed temperature T) at a certain time received from the data accumulation unit 41, and inputs this to the current prediction model M1 to obtain a predicted temperature T' at the next time. The model verification unit 433 then compares the predicted temperature T' with the actual observed temperature T. Note that if data on the observed temperature T at the desired time step is not available, the temperature may be calculated by linear interpolation or the like.

[0042] If the difference between the observed temperature T and the predicted temperature T′ exceeds a predetermined accuracy reference value, the model verification unit 433 determines that the prediction accuracy of the prediction model M1 is low. In this case, the model verification unit 433 increases the transmission rate of the data transmission unit 27 (i.e., the transmission interval T tThe system update unit 45 instructs the system update unit 45 to shorten the transmission time (shorten the transmission time). This causes the transmission rate of the data transmission unit 27 to be set to a high rate. That is, the system update unit 45 updates the data transmission unit 27 based on the transmission rate optimized by the system optimization unit 43.

[0043] Furthermore, if it is determined that the prediction accuracy of the prediction model M1 is low, the model update unit 435 updates the parameters of the prediction model M1. The model update unit 435 corrects the parameters of the prediction model M1 by performing re-learning based on the values ​​of the observed temperature T and the predicted temperature T′.

[0044] On the other hand, if the difference between the observed temperature T and the predicted temperature T′ is equal to or less than the accuracy reference value, the model verification unit 433 determines that the prediction accuracy of the prediction model M1 is high. In this case, the model verification unit 433 instructs the data transmitter 27 to lower the transmission rate (i.e., the transmission interval T t In this case, the system update unit 435 instructs the system update unit 45 to set the transmission rate of the data transmission unit 27 to a low rate. In addition, in this case, the model update by the model update unit 435 is omitted.

[0045] <Optimization System 1 Processing> Next, the processing executed by the optimization system 1 according to the embodiment will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the processing executed in the optimization system 1 shown in Fig. 1.

[0046] In the optimization system 1, the transmission rate of the data transmission unit 27 is set to an initial value (step S1). Then, the data transmission unit 27 starts transmitting the temperature data measured by the temperature sensor 25 at the set transmission rate. As a result, the data accumulation unit 41 starts accumulating the temperature data (step S2).

[0047] When the accumulation of temperature data is started in step S2, the model verification unit 433 verifies the prediction model M1 (step S3). That is, the model verification unit 433 calculates a control signal y based on the gain matrix g from the data set (gain matrix g and observed temperature T) at a certain time received from the data accumulation unit 41, and inputs this to the current prediction model M1 to obtain a predicted temperature T'. Then, the model verification unit 433 compares the observed temperature T with the predicted temperature T'.

[0048] Next, the model verification unit 433 determines whether the difference between the observed temperature T and the predicted temperature T' obtained by the comparison exceeds a predetermined accuracy standard value (step S4). If the difference between the observed temperature T and the predicted temperature T' exceeds the predetermined accuracy standard value (Yes in step S4), the model verification unit 433 instructs the system update unit 45 to increase the transmission rate of the data transmission unit 27. As a result, the transmission rate of the data transmission unit 27 is set to a high rate (step S5).

[0049] After step S5, the model update unit 435 updates the parameters of the prediction model M1 (step S6). That is, the model update unit 435 corrects the parameters of the prediction model M1 by performing re-learning based on the observed temperature T, the predicted temperature T′, the control signal y, and the gain matrix g.

[0050] On the other hand, if the difference between the observed temperature T and the predicted temperature T′ is equal to or smaller than the accuracy reference value (No in step S4), the model verification unit 433 instructs the system update unit 45 to lower the transmission rate of the data transmission unit 27. As a result, the transmission rate of the data transmission unit 27 is set to a low rate (step S7).

[0051] In addition, when the transmission rate is high, the transmission interval T t is the "first transmission interval," and the transmission interval T t is defined as the "second transmission interval," the second transmission interval of the low rate is longer than the first transmission interval of the high rate.

[0052] Next, the parameter search unit 431 of the system optimization unit 43 searches for an optimal gain matrix g using the prediction model M1 based on the temperature data accumulated by the data accumulation unit 41. Then, when the optimal gain matrix g is found (Yes in step S8), the system optimization unit 43 instructs the system update unit 45 to update the gain matrix g of the control unit 23. As a result, the gain matrix g of the control unit 23 is updated (step S9).

[0053] After step S9, if the process in the substrate processing apparatus is to be continued, the process returns to step S3 and continues. On the other hand, if the process in the substrate processing apparatus is completed, the process in the optimization system 1 also ends.

[0054] <Effects> As described above, the optimization system 1 according to the embodiment includes a real device 2 and a virtual device 4. The real device 2 includes an incubator 21 (control target), a control unit 23, a temperature sensor 25 (observation unit), and a data transmission unit 27. The control unit 23 controls the incubator 21 according to a gain matrix g (control parameter). The temperature sensor 25 observes the state of the incubator 21. The data transmission unit 27 transmits temperature data (observation data) acquired by the temperature sensor 25 to the virtual device 4 at a set transmission rate. The virtual device 4 includes a data storage unit 41, a system optimization unit 43, and a system update unit 45. The data storage unit 41 stores the temperature data sent from the data transmission unit 27. The system optimization unit 43 optimizes the gain matrix g of the control unit 23 in the real device 2 and the transmission rate of the data transmission unit 27 based on the temperature data stored in the data storage unit 41. The system update unit 45 updates the control unit 23 and the data transmission unit 27 based on the gain matrix g and transmission rate optimized by the system optimization unit 43. With this configuration, by optimizing the transmission rate, it is possible to reduce the amount of data communication except when detailed observation data is required. This allows the transmission load to be appropriately controlled. Furthermore, the reduction in data communication amount allows data storage costs to be reduced. Therefore, it is possible to reduce the management costs of observation data while operating the actual device 2 appropriately.

[0055] Furthermore, the system optimization unit 43 optimizes the gain matrix g (control parameters) using the prediction model M1 that predicts temperature data (observation data). When the prediction accuracy of the prediction model M1 decreases, the system optimization unit 43 increases the transmission rate. With this configuration, when the prediction accuracy of the prediction model M1 decreases, it is possible to collect more observation data.

[0056] Furthermore, when the prediction accuracy of the prediction model M1 decreases, the system optimization unit 43 updates the prediction model M1 using temperature data (observation data) collected by increasing the transmission rate. With this configuration, by increasing the transmission rate, it is possible to collect observation data in more detail, and therefore the prediction model can be appropriately updated using the collected observation data.

[0057] The system update unit 45 also updates the transmission interval T t By updating the transmission interval T t By shortening the transmission interval T t By lengthening the time, the transmission rate can be reduced.

[0058] The data transmission unit 27 transmits data at a transmission interval T t Among the temperature data (observation data) acquired during the period, the latest temperature data is selected and transmitted to the virtual device 4. With this configuration, the latest observation data can be transmitted, allowing the current situation to be properly grasped.

[0059] The data transmission unit 27 transmits data at a transmission interval T t A representative value may be calculated based on several pieces of temperature data (observation data) acquired during the transmission interval T t By transmitting a representative value of a plurality of temperature data acquired during the transmission interval, it is possible to grasp the overall trend of the observation data acquired during the transmission interval.

[0060] <2. Variations> Although the embodiments have been described above, the present invention is not limited to the above and various modifications are possible.

[0061] In the above embodiment, the transmission interval T t However, the transmission interval T t may be finely adjusted within a predetermined range. For example, the transmission interval T t may be adjusted.

[0062] Although the present invention has been described in detail, the above description is merely illustrative in all respects and does not limit the present invention. It is understood that countless variations not illustrated can be envisioned without departing from the scope of the present invention. The configurations described in the above embodiments and variations can be combined or omitted as appropriate as long as they are not mutually inconsistent. [Explanation of symbols]

[0063] 1: Optimization system 2: Actual device 4: Virtual device 21: Incubator (control object) 23: Control section 25: Temperature sensor (observation part) 27: Data transmission section 41: Data storage unit 43: System Optimization Department 45: System Update Department M1: Prediction model

Claims

1. An optimization system comprising a real device and a virtual device, The real device is A control object; a control unit that controls the controlled object in accordance with a control parameter; an observation unit that observes a state of the controlled object; a data transmission unit that transmits the observation data acquired by the observation unit to the virtual device at a set transmission rate; and The virtual device is a data storage unit that stores the observation data sent from the data transmission unit; a system optimization unit that optimizes the control parameters of the control unit and the transmission rate of the data transmission unit in the actual device based on the observation data stored in the data storage unit; a system update unit that updates the control unit and the data transmission unit based on the control parameters optimized by the system optimization unit and the transmission rate; An optimization system having:

2. 2. The optimization system of claim 1, the system optimization unit optimizes the control parameters using a prediction model that predicts the observation data; The system optimization unit increases the transmission rate when the prediction accuracy of the prediction model decreases.

3. 3. The optimization system of claim 2, an optimization system, wherein when the prediction accuracy of the prediction model decreases, the system optimization unit updates the prediction model using the observation data collected after increasing the transmission rate.

4. 4. The optimization system according to claim 1, further comprising: The system update unit updates the transmission rate by updating a transmission interval, which is a period for transmitting the observation data.

5. 5. The optimization system of claim 4, The data transmission unit selects the latest observation data from the plurality of observation data acquired during the transmission interval and transmits the latest observation data to the virtual device.

6. 5. The optimization system of claim 4, The data transmission unit calculates a representative value of the plurality of pieces of observation data acquired during the transmission interval and transmits the representative value to the virtual device.

Images