Method for optimizing a work system, and a control system

Abstract

According to the present invention, a management device optimizes a work system. More specifically, this management device: sets a control parameter set that is made up of one or more parameters for controlling actions of the work system (step S10); controls the actions of the work system on the basis of the control parameter set (step S12); acquires an observation value relating to the work system in relation to execution of the actions based on the control parameter set (step S14); calculates a posterior distribution of a model function in which one or more parameters are variables by using the control parameter set and the observation value (step S20); and updates an optimization function for identifying an optimal parameter set as the control parameter set, using calculation results of the posterior distribution (step S24).

Description

[Technical Field]

[0001] This invention relates to the optimization of a work system. [Background technology]

[0002] Conventionally, various technologies have been proposed for optimizing work systems for producing and / or processing objects. For example, Japanese Patent Publication No. 6931946 (Patent Document 1) discloses a production process optimization system that obtains the execution results of an execution procedure by the entity executing the production process, obtains evaluation results (e.g., human fragrance evaluation) of the execution results, and uses the evaluation results to optimize variable parameters in the execution procedure. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Patent No. 6931946 [Overview of the project] [Problems that the invention aims to solve]

[0004] Generally, optimizing a work system requires users to perform a wide range of tasks, including adjusting parameters used to control the system. Therefore, technologies that reduce the user's workload are always in demand.

[0005] This invention was conceived in view of the aforementioned circumstances, and its purpose is to provide a technology for reducing the user's burden in optimizing the work system. [Means for solving the problem]

[0006] A method for optimizing a work system according to a certain aspect of this disclosure comprises the steps of: setting a control parameter set consisting of one or more parameters for controlling the operation of the work system; controlling the operation of the work system based on the control parameter set; obtaining observed values ​​from the work system related to the execution of the operation based on the control parameter set; calculating the posterior distribution of a model function with one or more parameters as variables using the control parameter set and the observed values; and updating an optimization function for identifying the optimal parameter set as the control parameter set using the calculation result of the posterior distribution.

[0007] A control system according to a certain aspect of this disclosure comprises a work system and a management device that controls the operation of the work system. The management device sets a control parameter set consisting of one or more parameters for controlling the operation of the work system, controls the operation of the work system based on the control parameter set, the work system performs an operation based on the control parameter set, the management device obtains observed values ​​from the work system related to the execution of an operation based on the control parameter set, calculates a posterior distribution of a model function with one or more parameters as variables using the control parameter set and the observed values, and updates an optimization function to identify the optimal parameter set as the control parameter set using the result of the calculation of the posterior distribution. [Effects of the Invention]

[0008] In accordance with certain aspects of this disclosure, the user's burden for optimizing the work system is reduced. [Brief explanation of the drawing]

[0009] [Figure 1] This is a schematic diagram of a control system 1 according to one embodiment of the present disclosure. [Figure 2] This figure shows an example of the hardware configuration of the management device 100. [Figure 3] This figure shows an example of the search results based on four sets of observations. [Figure 4] It is a diagram showing an example of search results based on five sets of observed values. [Figure 5] It is a diagram showing an example of search results based on eleven sets of observed values. [Figure 6] It is a diagram showing an equation representing the assumption that the model function follows a Gaussian process. [Figure 7] It is a diagram showing an example of a kernel function. [Figure 8] It is a diagram showing a mathematical formula for calculating the average value. [Figure 9] It is a diagram showing a mathematical formula for calculating the variance value. [Figure 10] It is a diagram showing an example of an optimization function. [Figure 11] It is a flowchart of a process related to the optimization of a control parameter set, which is executed by the management device 100.

Embodiments for Carrying Out the Invention

[0010] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals and their descriptions are not repeated.

[0011] [Configuration of Control System] FIG. 1 is a schematic configuration diagram of a control system 1 according to an embodiment of the present disclosure. The control system 1 includes a management device 100 and a work system 200. The management device 100 performs processing for optimizing the operation of the work system 200. The work system 200 produces and / or processes a target substance. In the example of FIG. 1, the work system 200 cultures bacteria. That is, the work system 200 generates new bacteria using bacterial strains.

[0012] The work system 200 includes an incubator 210, a centrifuge 220, a sample transfer robot 230, a liquid dispenser 240, a plate reader 250, a mass spectrometer 260, and a liquid chromatograph 270.

[0013] The management device 100 controls the operations of each element of the operating system 200. More specifically, when a well plate in which strains and a culture medium have been introduced into each well is set in the incubator 210, the management device 100 causes the incubator 210 to perform bacterial culture.

[0014] After the start of the culture in the incubator 210, when a given time (assuming that a given degree of cell growth and metabolism has progressed) has elapsed, the management device 100 causes the sample transfer robot 230 to take out the well plate from the incubator 210. Then, the management device 100 causes the sample transfer robot 230 to move the well plate for pre-treatment for measuring the metabolic components of the sample. More specifically, the management device 100 causes the sample transfer robot 230 to move the well plate to the centrifuge 220, and causes the centrifuge 220 to sink the bacterial cells in each well. Also, the management device 100 moves the sample transfer robot 230 to the liquid dispenser 240, and causes the liquid dispenser 240 to add a reagent to each well.

[0015] Thereafter, the management device 100 causes the sample transfer robot 230 to move the well plate to the plate reader 250, and causes the plate reader 250 to perform cell density measurement for each well. Also, the management device 100 causes the sample transfer robot 230 to move the well plate to the mass spectrometer 260, and causes the mass spectrometer 260 to perform mass spectrometry for each well. Also, the management device 100 causes the sample transfer robot 230 to move the well plate to the liquid chromatograph 270, and causes the liquid chromatograph 270 to generate a liquid chromatogram for each well.

[0016] In control system 1, cell density and metabolic component measurements are performed for the culture in the incubator 210, and these measurement results can be used to evaluate the productivity of the target substance. In control system 1, the reproducibility of pre-treatment work, which is prone to variability, can be improved by automating a series of experimental sequences. In addition, the throughput of the experimental process realized by the work system 200 can be improved by removing the well plates from the incubator 210 at a scheduled time and performing metabolic component measurements.

[0017] [Hardware configuration of the management device] Figure 2 shows an example of the hardware configuration of the management device 100. The management device 100 is comprised of, for example, a general-purpose computer. In the example in Figure 1, the management device 100 includes a processor 101, memory 102, input / output interface (I / F) 103, and communication I / F 104.

[0018] The processor 101 executes various programs for the control device 100 to perform various processes. The memory 102 non-temporarily stores the programs executed by the processor 101 and various data necessary for program execution. The input / output interface 103 is an interface for the processor 101 to communicate with the incubator 210, centrifuge 220, sample transport robot 230, liquid dispenser 240, plate reader 250, mass spectrometer 260, and liquid chromatograph 270. The communication interface 104 is an interface for the processor 101 to communicate with devices outside the control system 1 via a network.

[0019] The management device 100 further includes a display 110 and an input device 120. The display 110 displays the results of the calculations performed by the processor 101, and the input device 120 (mouse, keyboard, touch sensor, etc.) accepts data input operations to the processor 101.

[0020] [Control parameter set] In the control system 1, the management device 100 utilizes one or more parameters to control each element within the work system 200. In this specification, the set of parameters comprised of "one or more parameters" is also referred to as the "control parameter set".

[0021] Specific examples of parameters that make up the control parameter set are as follows: One example of a parameter related to the incubator 210 is the culture temperature.

[0022] One example of a parameter related to the centrifuge 220 is the rotational speed of the centrifuge. Another example is the rotational time.

[0023] An example of a parameter for the sample transport robot 230 is the travel speed when moving the well plate.

[0024] An example of a parameter for the liquid dispenser 240 is the amount of reagent injected into each well. An example of a parameter for the plate reader 250 is the specification of the type of measurement (luminescence measurement, fluorescence intensity measurement, absorbance measurement (ultraviolet), absorbance measurement (visible), time-resolved fluorescence measurement, and / or fluorescence polarization measurement).

[0025] One example of a parameter for the mass spectrometer 260 is the voltage of the input electrode of the quadrupole electrostatic electrode, while another example is the voltage of the output electrode.

[0026] An example of a parameter related to the liquid chromatograph 270 is the pumping speed of the mobile phase pump.

[0027] [Evidence] In the control system 1, the management device 100 operates the work system 200 according to a control parameter set, and acquires observational values ​​to estimate the behavior of the object (contents of each well) in relation to the execution of the operation according to the control parameter set. The observational values ​​may be acquired before the start of the operation in the work system 200, during the operation, and / or after the completion of the operation.

[0028] The observed values ​​may be a collection of one or more indices. An indice is, for example, a measurement result representing the behavior of the work in the work system 200. One example of an indice is the absorbance of each well measured by the plate reader 250. Another example is the absorbance at a given wavelength in a chromatography produced by the liquid chromatograph 270. Yet another example is the peak intensity (relative intensity) of a given mass number in a mass spectrum produced by the mass spectrometer 260.

[0029] The control device 100 uses the observed values ​​to identify the optimized control parameter set. The control device 100 then operates the work system 200 according to the optimized control parameter set and acquires new observed values. The control device 100 repeats the process of operating the work system 200 according to the control parameter set, acquiring new observed values, and optimizing the control parameter set using the control parameter set and the new observed values.

[0030] The control device 100 may generate a new index from multiple indices constituting the observed values ​​in order to reduce the dimensionality of the observed values ​​used for optimizing the control parameter set, and then use this new index instead of the multiple indices to optimize the control parameter set. This enables the control device 100 to perform optimization that takes multiple indices into consideration in the calculation for optimizing the control parameter set (calculation of the posterior distribution of the model function).

[0031] One example of a new indicator is the amount of target substance produced per cell. The control device 100 generates the amount of target substance produced per cell by, for example, dividing the amount of target substance produced by the liquid chromatograph 270, which is identified from the liquid chromatograph, by the cell yield, which is identified based on the cell density measurement results by the plate reader 250.

[0032] Another example of a new indicator is the sum of the amounts of two components. The control device 100 may, for example, identify the amount of each of the two components from the liquid chromatograph produced by the liquid chromatograph 270, and then use the identified amounts of each of the two components to calculate the sum of the amounts of the two components.

[0033] [Bayesian optimization] The control device 100 utilizes Bayesian optimization as an example of an optimization method for the control parameter set of the work system 200. That is, the control device 100 searches for the optimal control parameter set based on Bayesian optimization.

[0034] Figure 3 shows an example of the search results based on four sets of observations. In Figure 3, the black dots represent each set of observations actually obtained in connection with the execution of the operation of the working system 200. The solid curve represents the mean of the posterior distribution of the model function estimated by Bayesian optimization. The shaded area represents the uncertainty (or variance) of the posterior distribution of the model function. The thick vertical solid line labeled "Next" represents the next identified set of control parameters.

[0035] In the example shown in Figure 3, the control device 100 also estimates the model function of the working system 200 from four sets of observed values ​​corresponding to each of the four sets of control parameters, and calculates the mean and variance of the posterior distribution of the model function. More specifically, to calculate the posterior distribution of the model function shown in Figure 3, the working system 200 performs four operations (bacterial culture experiments) according to each of the four sets of control parameters. The control device 100 acquires one set of observed values ​​in relation to each operation. Therefore, the control device 100 acquires four sets of observed values ​​acquired in relation to each of the four operations. The control device 100 then uses these four sets of observed values ​​to calculate the posterior distribution of the model function, and also calculates the mean and variance of the posterior distribution.

[0036] The control device 100 then identifies the following set of control parameters, which are expected to improve the model to more closely resemble the true value, based on the calculated mean and variance values.

[0037] Furthermore, the control device 100 performs calculations under the assumption that the model of the work system 200 follows a Gaussian process.

[0038] The control device 100 further acquires new observations by operating the work system 200 under conditions that conform to the "next set of control parameters" identified above. The control device 100 then re-estimates the posterior distribution of the model function, adding the newly acquired observations to the previously acquired observations. By repeating this process, the control device 100 can obtain a set of control parameters that yields observations corresponding to more favorable results.

[0039] A control parameter set that yields a more favorable result in the operating system 200 is the control parameter set that maximizes the optimization function described later with reference to Figure 10. A more favorable result is, for example, a higher bacterial yield in bacterial culture.

[0040] Figure 4 shows an example of the search results based on 5 sets of observations. Figure 4 shows the state where 1 set of observations has been added to the initial 4 sets of observations shown in Figure 3. Figure 5 shows an example of the search results based on 11 sets of observations. Figure 5 shows the state where 7 sets of observations have been added to the initial 4 sets of observations shown in Figure 3, that is, 6 sets of observations have been added to the 5 sets of observations shown in Figure 4.

[0041] The control device 100 identifies the control parameter set within the range where the observed value is highest in the model function, as the next control parameter set, as shown in Figure 5 as the "priority search range". In the graphs shown in Figures 3 to 5, a higher observed value corresponds to a higher bacterial yield in the culture.

[0042] [Mean and variance of the posterior distribution of the model function] Next, we will explain how to calculate the mean and variance of the posterior distribution of the model function described above.

[0043] Figure 6 shows the equation representing the assumption that the model function follows a Gaussian process. Equation (1) shown in Figure 6 contains three terms. The leftmost term and the middle term are connected by "~", and the middle term and the rightmost term are also connected by "~".

[0044] The leftmost term, f, represents the model function. The subscripts attached to f correspond to each set of control parameters. For example, "1," "n," and "i" in Figure 6 represent the first, nth, and ith control parameter sets, respectively. Then, f1 represents the observed value corresponding to the first control parameter set. n represents the observed value corresponding to the nth control parameter set, and f i This represents the observed value corresponding to the i-th control parameter set.

[0045] The μ shown in the middle term represents the mean parameter under the Gaussian assumption. The subscript attached to μ has the same meaning as the subscript attached to f.

[0046] N represents a normal distribution. K represents the kernel function. K is denoted by two subscripts, i and j. Kij represents the kernel function (covariance function) of the control parameters fi and fj.

[0047] Figure 7 shows an example of a kernel function. In Figure 7, the kernel function is given by equation (2). In equation (2), x represents each set of control parameters. For example, x i represents the i-th control parameter set. σ represents the variance calculated from the mean parameter μ (the regression function μ(x) under the Gaussian assumption). l represents the covariance amplitude.

[0048] Again, refer to FIG. 6. In the rightmost term, the subscript D represents the acquired data group. Also, the subscript T represents transposition.

[0049] The management device 100 calculates the mean value and variance value of the posterior distribution of f(x) for any xi according to the assumptions shown in FIG. 6.

[0050] FIG. 8 is a diagram showing a mathematical formula for calculating the mean value. The management device 100 calculates the mean value E(p(f i |f D , K, μ)) according to Equation (3) in FIG. 8. In Equation (3), p represents a probability function.

[0051] FIG. 9 is a diagram showing a mathematical formula for calculating the variance value. The management device 100 calculates the variance value V(p(f i}|f D , K, μ)) according to Equation (4) in FIG. 9. [[ID=!25]]

[0052] [Optimization Function] The management device 100 uses an optimization function in the search for a control parameter set based on Bayesian optimization. An example of the optimization function is an acquisition function. The management device 100 uses the UCB (Upper Confidence Bound) function as an example of the acquisition function.

[0053] FIG. 10 is a diagram showing an example of the optimization function. In Equation (5) in FIG. 10, μ(x) represents a regression function of the Gaussian assumption, σ(x) represents a function of variance, κ represents a predetermined constant, and x represents a control parameter set. Note that the value of σ(x) in Equation (5) can also be expressed as the variance value V(p(f i |f D , K, μ)). ?

[0054] In Equation (5), the function UCB(x) includes the mean value of the posterior distribution by including μ(x), and includes the variance value of the posterior distribution by including σ(x).

[0055] The control device 100 identifies the value of x that maximizes the value of UCB(x) in equation (5), and identifies the control parameter set corresponding to the identified value of x as a new control parameter set.

[0056] [Termination Conditions] As described above, the control device 100 repeatedly operates the work system 200 according to the control parameter set, acquires new observed values, and optimizes the control parameter set using the new observed values. The control device 100 may also have termination conditions set for the above repetition.

[0057] An example of a termination condition includes a criterion related to observed values. For example, if the termination condition is set to be that the yield of the target substance identified by the observed values ​​has reached a predetermined amount, the control device 100 may determine that the termination condition has been met if the yield of the target substance identified from the observed values ​​has reached the predetermined amount, and may terminate the process related to the optimization of the control parameter set without operating the work system 200 any further.

[0058] Other examples of termination conditions include criteria related to the number of times the control parameter set has been optimized (the number of times the control parameter set has been updated). For example, if the termination condition is set to reach 100 optimizations, the management device 100 may determine that the termination condition is met if the control parameter set has been updated 100 times, and may terminate the process related to the optimization of the control parameter set without further operation of the work system 200.

[0059] Further examples of termination conditions include criteria relating to the relationship between observed values ​​before and after the optimization of the control parameter set. For example, suppose the termination condition is set such that the yield of the target substance identified by the observed values ​​after the control parameter set has been updated has increased by 10% or more compared to the yield of the target substance identified by the observed values ​​before the control parameter set was updated. In this case, if the updated yield has increased by 10% compared to the yield before the update, the control device 100 may determine that the termination condition has been met and terminate the process related to the optimization of the control parameter set without operating the work system 200 any further. The termination condition may also be that the change in observed values ​​has saturated in repeated updates. For example, the termination condition may be that, in the most recent five updates, the change in the yield of the target substance is less than 10% compared to the oldest yield among the five most recent updates.

[0060] As described above, in the control system 1, the management device 100 terminates the process related to the optimization of the control parameter set when the termination condition is met. This can prevent the work system 200 from operating more than necessary. For example, if the usage fee for the control system 1 is set according to the number of times the work system 200 operates, the user can avoid being charged unexpected usage fees due to the work system 200 operating more than necessary by setting the termination condition in advance.

[0061] [Process Flow] Figure 11 is a flowchart of the process related to the optimization of the control parameter set, which is performed by the management device 100. The process shown in Figure 11 is realized, for example, by the processor 101 executing a given program, and is started when the user instructs the management device 100 to start the process.

[0062] In step S10, the management device 100 acquires the control parameter set. In step S10 of one implementation example, the management device 100 acquires the control parameter set by receiving input of the control parameter set from the user.

[0063] In step S12, the control device 100 instructs the work system 200 to perform an operation according to the control parameter set acquired in step S10 (or updated in step S28, described later).

[0064] In step S14, the control device 100 acquires observational values ​​from the work system 200 related to the operation that was instructed to be performed in step S12.

[0065] In step S16, the control device 100 determines whether the termination condition for the iteration of the optimization of the control parameter set is met. If the control device 100 determines that the termination condition is met (YES in step S16), it proceeds to step S30. If the control device 100 determines that the termination condition is not met (NO in step S16), it proceeds to step S18.

[0066] In step S18, the control device 100 acquires a model parameter set (reads it from memory 102). The model parameter set refers to a set of data used for estimating the model function. The model parameter set includes the control parameter set used in the instructions in step S12 and the observed values ​​acquired in step S14. The model parameter set may also include the control parameter set used in past operations of the control device 100 and the observed values ​​related to those operations. The control parameter set used in past operations of the control device 100 and the observed values ​​related to those operations may be stored in memory 102.

[0067] More specifically, as explained with reference to Figures 3 to 5, multiple sets of "control parameter sets and observed values" may be used to estimate the model function. In step S18, in addition to the most recent "control parameter set and observed values," i.e., the control parameter set from step S12 and the observed values ​​from step S14, past "control parameter sets and observed values" may also be used.

[0068] In step S20, the control device 100 uses the model parameter set obtained in step S18 to calculate the posterior distribution of the model function.

[0069] In step S22, the control device 100 calculates the mean and variance of the posterior distribution calculated in step S20.

[0070] In step S24, the control device 100 updates the optimization function (for example, equation (5) in Figure 10) using the mean and variance values ​​calculated in step S22.

[0071] In step S26, the control device 100 identifies a new set of control parameters that maximizes the value of the optimization function updated in step S24.

[0072] In step S28, the control device 100 replaces the parameter set used to instruct the operation of the work system 200 with the control parameter set identified in step S26. This updates the control parameter set used for the instructions. Subsequently, the control device 100 returns control to step S12. As a result, the work system 200 performs the operation again based on the updated control parameter set in response to the instructions in step S12. In step S14, the control device 100 acquires observational values ​​related to the operation based on the updated control parameter set.

[0073] Meanwhile, in step S30, the control device 100 notifies the result of the optimization of the control parameter set and terminates the process shown in Figure 11.

[0074] In the process shown in Figure 11, if the control device 100 determines in step S16 that the termination condition has been met, it proceeds to step S30 in order to avoid updating the control parameter set in step S28. In this sense, the control in step S16 is an example of a step to avoid updating the control parameter set.

[0075] One example of the information reported in step S30 is the latest control parameter set and the observed values ​​obtained in relation to the operation according to the latest control parameter set. Another example is the control parameter set and the observed values ​​obtained in relation to the operation according to the control parameter set for each experiment that has been repeated in the process shown in Figure 11 up to that point.

[0076] In the process shown in Figure 11, the control device 100 performs the above notification in step S30 when it avoids updating the control parameter set. In step S0, it may also be notified that "the update of the control parameter set has been avoided and the optimization of the control parameter set has been completed." In this sense, the control in step S30 is an example of a step that notifies that the update of the control parameter set has been avoided.

[0077] One example of a notification method is the display of data on the display 110. Another example is the sending of a notification from the management device 100 to an address stored regarding the optimization process of the control parameter set (for example, the address of the user who instructed the optimization).

[0078] In the embodiment described above, the management device 100 instructs the work system to perform an operation based on a set of control parameters, acquires observed values ​​related to the work system in connection with the execution of the operation, calculates the posterior distribution of the model function using the acquired observed values, and updates the optimization function using the results of the calculated posterior distribution (e.g., mean and variance). In other words, in this embodiment, user intervention is not required in the process from instructing the work system to perform an operation to updating the optimization function. This can reduce the user's effort in optimizing the work system.

[0079] [Differentiation] In control system 1, the operation of work system 200 was for the purpose of generating bacteria. In this respect, bacteria are merely one example of the target substance that can be generated. The target substance generated in work system 200 is not limited to bacteria. Furthermore, the operation of work system 200 is not limited to the generation of the target substance, but may also involve processing the target substance.

[0080] More specifically, the work system 200 may be a system for producing a solid material having desired properties. For example, one example is a device that sets control parameters for the component ratios constituting the solid material and uses tensile strength as an indicator. Another example is optimizing for an indicator that takes both properties into account, such as the sum of tensile strength and toughness.

[0081] [Pattern] Those skilled in the art will understand that the above-described exemplary embodiments are specific examples of the following embodiments.

[0082] (Section 1) An optimization method for a work system according to one embodiment may include the steps of: setting a control parameter set consisting of one or more parameters for controlling the operation of the work system; controlling the operation of the work system based on the control parameter set; obtaining observed values ​​from the work system related to the execution of the operation based on the control parameter set; calculating a posterior distribution of a model function with the one or more parameters as variables using the control parameter set and the observed values; and updating an optimization function for identifying the optimal parameter set as the control parameter set using the calculation result of the posterior distribution.

[0083] According to the work system optimization method described in paragraph 1, the user's burden for optimizing the work system is reduced.

[0084] (Clause 2) In the optimization method for the work system described in paragraph 1, the optimization function may include the mean and variance of the posterior distribution.

[0085] According to the optimization method for the work system described in Section 2, a function suitable for optimizing the work system is used as the optimization function.

[0086] (3) The method for optimizing the work system described in paragraph 1 or 2 may further comprise the step of updating the control parameter set with a set of parameters that maximizes the value of the updated optimization function.

[0087] According to the optimization method for the work system described in Section 3, an optimal set of control parameters is provided.

[0088] (Clause 4) The method for optimizing the work system described in paragraph 3 may further include a step of avoiding updating the control parameter set in accordance with predetermined conditions.

[0089] According to the optimization method for the work system described in Section 4, the work system will not operate more than necessary.

[0090] (Clause 5) In the optimization method of the work system described in paragraph 4, the predetermined conditions may include at least one of the following: a criterion relating to the observed values, a criterion relating to the number of updates, and a criterion relating to the relationship between the observed values ​​before and after the update.

[0091] According to the optimization method for the work system described in Section 5, the work system will not operate more than necessary in accordance with specific criteria.

[0092] (Clause 6) The method for optimizing the work system described in paragraph 4 may further include a step of notifying that the update of the control parameter set has been avoided.

[0093] According to the optimization method for the work system described in Section 6, the user can easily recognize that the update of the control parameter set has been avoided.

[0094] (Clause 7) In the method for optimizing a work system described in any one of Clauses 1 to 6, the step of calculating the posterior distribution of the model function may include generating one index from a plurality of indexes included in the observed values, and calculating the posterior distribution using at least one index.

[0095] According to the optimization method for the work system described in Section 7, the dimensionality of the observed values ​​is reduced in the optimization of the control parameter set, thereby reducing the computational load for optimizing the control parameter set.

[0096] (Clause 8) In the method for optimizing the work system described in any one of paragraphs 1 to 7, the operation may include generating a target substance from an object or processing a target substance.

[0097] According to the optimization method for the work system described in Section 8, the user's burden is reduced in optimizing the system for generating or processing the target substance.

[0098] (Clause 9) A control system according to one embodiment comprises a work system and a management device for controlling the operation of the work system, wherein the management device sets a control parameter set consisting of one or more parameters for controlling the operation of the work system, controls the operation of the work system based on the control parameter set, the work system executes an operation based on the control parameter set, the management device obtains observed values ​​from the work system related to the execution of the operation based on the control parameter set, calculates a posterior distribution of a model function with the one or more parameters as variables using the control parameter set and the observed values, and updates an optimization function for identifying the optimal parameter set as the control parameter set using the calculation result of the posterior distribution.

[0099] According to the control system described in Section 9, the user's burden for optimizing the work system is reduced.

[0100] (Clause 10) In the control system described in Clause 9, the optimization function may include the mean and variance of the posterior distribution.

[0101] According to the control system described in Section 10, a function suitable for optimizing the work system is used as the optimization function.

[0102] (Clause 11) In the control system described in paragraph 9 or 10, the control device may update the control parameter set with a set of parameters that maximizes the value of the updated optimization function.

[0103] According to the control system described in Section 11, an optimal set of control parameters is provided.

[0104] (Clause 12) In the control system described in paragraph 11, the management device may avoid updating the control parameter set in accordance with predetermined conditions.

[0105] According to the control system described in paragraph 12, the operation of the work system is prevented from operating more than necessary.

[0106] (Clause 13) In the control system described in paragraph 12, the predetermined conditions may include at least one of the following: a criterion relating to the observed values, a criterion relating to the number of updates, and a criterion relating to the relationship between the observed values ​​before and after the update.

[0107] According to the control system described in paragraph 13, the operation of the work system is prevented from operating unnecessarily, in accordance with specific criteria.

[0108] (Clause 14) In the control system described in paragraph 12, the management device may notify that the update of the control parameter set has been avoided.

[0109] According to the control system described in Section 14, the user can easily recognize that the update of the control parameter set has been avoided.

[0110] (Clause 15) In the control system described in any one of Clauses 9 to 14, calculating the posterior distribution of the model function may include generating one index from a plurality of indexes included in the observed values, and calculating the posterior distribution using at least one index.

[0111] According to the control system described in Section 15, the dimensionality of the observed values ​​is reduced in the optimization of the control parameter set, thereby potentially reducing the computational load for optimizing the control parameter set.

[0112] (Clause 16) In the control system described in any one of paragraphs 9 to 14, the operation may include generating a target substance from an object or processing a target substance.

[0113] According to the control system described in paragraph 16, the user's burden is reduced in optimizing the system for generating or processing the target substance.

[0114] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of this disclosure is indicated by the claims rather than by the description of the embodiments above, and all modifications within the meaning and scope of the claims are intended to be included. Furthermore, each technique in the embodiments is intended to be practiced individually or, as far as possible, in combination with other techniques in the embodiments. [Explanation of symbols]

[0115] 1 control system, 100 management devices, 101 processor, 102 memory, 110 display, 120 input devices, 200 work systems, 210 incubators, 220 centrifuges, 230 sample transport robots, 240 liquid dispensers, 250 plate readers, 260 mass spectrometers, 270 liquid chromatographs.

Claims

1. The steps include setting a control parameter set consisting of one or more parameters for controlling the operation of the work system, A step of controlling the operation of the work system based on the control parameter set, The steps include obtaining observed values ​​from the aforementioned work system related to the execution of the operation based on the control parameter set, A step of calculating the posterior distribution of a model function with one or more parameters as variables using the control parameter set and the observed values, The process includes the step of updating an optimization function for identifying the optimal parameter set as the control parameter set using the calculation results of the posterior distribution, The step of calculating the posterior distribution of the aforementioned model function is: To generate one indicator from multiple indicators included in the aforementioned observed values, A method for optimizing a work system, comprising: calculating the posterior distribution using at least one of the aforementioned indicators.

2. The method for optimizing a work system according to claim 1, wherein the optimization function includes the mean and variance of the posterior distribution.

3. The method for optimizing a work system according to claim 1, further comprising the step of updating the control parameter set with a parameter set that maximizes the value of the updated optimization function.

4. The method for optimizing a work system according to claim 3, further comprising the step of avoiding updating the control parameter set in accordance with predetermined conditions.

5. The method for optimizing a work system according to claim 4, wherein the predetermined conditions include at least one of the following: a criterion for the observed values, a criterion for the number of updates, and a criterion for the relationship between the observed values ​​before and after the update.

6. The method for optimizing a work system according to claim 4, further comprising the step of notifying that the update of the control parameter set has been avoided.

7. The method for optimizing the work system according to claim 1, wherein the operation includes generating a target substance from an object or processing a target substance.

8. Work system and, The system comprises a control device for controlling the operation of the aforementioned work system, The aforementioned control device is A control parameter set consisting of one or more parameters for controlling the operation of the work system is established. Based on the aforementioned control parameter set, the operation of the work system is controlled. The aforementioned work system is The operation is performed based on the aforementioned control parameter set. The aforementioned control device is Observed values ​​related to the execution of the operation based on the control parameter set are obtained from the aforementioned work system. Using the control parameter set and the observed values, the posterior distribution of the model function with the one or more parameters as variables is calculated. Using the calculation results of the aforementioned posterior distribution, the optimization function for identifying the optimal parameter set as the control parameter set is updated. Calculating the posterior distribution of the aforementioned model function is: To generate one indicator from multiple indicators included in the aforementioned observed values, A control system comprising: calculating the posterior distribution using at least one of the aforementioned indicators.

9. The control system according to claim 8, wherein the optimization function includes the mean and variance of the posterior distribution.

10. The aforementioned control device is The control system according to claim 8, wherein the control parameter set is updated with a parameter set that maximizes the value of the updated optimization function.

11. The aforementioned control device is The control system according to claim 10, which avoids updating the control parameter set in accordance with predetermined conditions.

12. The control system according to claim 11, wherein the predetermined conditions include at least one of the following: a criterion for the observed values, a criterion for the number of updates, and a criterion for the relationship between the observed values ​​before and after the update.

13. The aforementioned control device is The control system according to claim 11, which provides notification that the update of the control parameter set has been avoided.

14. The control system according to claim 8, wherein the operation includes generating a target substance from an object or processing a target substance.

Images