# US 11,520,310 B2 · Generating control settings for a chemical reactor > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,520,310 B2 - Document ID: US11520310B2 - Status: Granted - Kind: B2 - Application number: 16/444,565 - Filed: 2019-06-18 - Published: 2020-12-24 - Granted: 2022-12-06 - Earliest claimed: 2019-06-18 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Dmitry Zubarev; Victoria A. Piunova; Nathaniel H. Park; James L. Hedrick; Sarath Swaminathan - Web page: https://patentgrep.com/patent/US11520310B2 ## CPC - G05B 19/4155: characterised by program execution, i.e. part program or machine function execution, e.g. selection of a program - G05B 13/0265: {the criterion being a learning criterion} - G06N 3/0455: Auto-encoder networks; Encoder-decoder networks - G06N 3/0475: Generative networks - G06N 3/094: Adversarial learning - G06N 20/00: Machine learning - G05B 2219/32375: Petrinet synthesis tool - G06N 3/045: Combinations of networks - G06N 3/047: Probabilistic or stochastic networks - G06N 3/08: Learning methods - G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Techniques regarding autonomously controlling one or more chemical reactors using generative machine learning models are provided. For example, one or more embodiments described herein can comprise a system, which can comprise a memory that can store computer executable components. The system can also comprise a processor, operably coupled to the memory, and that can execute the computer executable components stored in the memory. The computer executable components can comprise a model component that can build a generative machine learning model based on training data regarding a past chemical reactor operation. The generative machine learning model can generate a recommended chemical reactor control setting for experimental discovery of a polymer. ## Claims ### Claim 1 A system, comprising: a memory; a processor, operably coupled to the memory, wherein the processor: builds a generative machine learning model based on training data regarding a past chemical reactor operation, wherein the generative machine learning model generates a recommended chemical reactor control setting for experimental discovery of a polymer with a target characteristic; measures a characteristic of the polymer synthesized by the chemical reactor; determines whether the characteristic is within a defined range of the target characteristic; and updates the training data, with the recommended chemical reactor control setting, based on the determination that the characteristic is within the defined range. ### Claim 2 (depends on 1) The system of claim 1, wherein the generative machine learning model is a variational autoencoder model trained via a gradient decent algorithm that optimizes a loss function with respect to parameters of an encoder and a decoder. ### Claim 3 (depends on 2) The system of claim 2, wherein the generative machine learning model generates the recommended chemical reactor control setting by sampling a latent variable from a latent space and decoding the latent variable via the decoder. ### Claim 4 (depends on 1) The system of claim 1, wherein the generative machine learning model is via a generative adversarial network model comprising a generator network and a discriminator network. ### Claim 5 (depends on 4) The system of claim 4, wherein the generator network upsamples a vector of noise to generate new data, and wherein the discriminator network is a binomial classifier that analyzes the training data and the new data. ### Claim 6 (depends on 5) The system of claim 5, wherein the generator network achieves a trained state based on the discriminator network being unable to discern between the training data and the new data, and wherein the recommended chemical reactor control setting is comprised within the new data generated by the generator network while the generator network is in the trained state. ### Claim 7 (depends on 1) The system of claim 1, wherein the training data comprises a control setting implemented by a chemical reactor during the past chemical reactor operation. ### Claim 8 (depends on 1) The system of claim 1, wherein the processor: operates a chemical reactor in accordance with the recommended chemical reactor control setting to synthesize the polymer. ### Claim 9 (depends on 1) The system of claim 1, wherein the target characteristic comprises molecular weight. ### Claim 10 (depends on 1) The system of claim 1, wherein the target characteristic comprises retention time. ### Claim 11 A computer-implemented method, comprising: generating, by a system operatively coupled to a processor, a generative machine learning model based on training data regarding a past chemical reactor operation, wherein the generative machine learning model generates a recommended chemical reactor control setting for experimental discovery of a polymer with a target characteristic; measuring, by the system, a characteristic of the polymer synthesized by the chemical reactor; determining, by the system, whether the characteristic is within a defined range of the target characteristic; and updating, by the system, the training data, with the recommended chemical reactor control setting, based on the determination that the characteristic is within the defined range. ### Claim 12 (depends on 11) The computer-implemented method of claim 11, wherein the generative machine learning model is a variational autoencoder, and wherein the computer-implemented method further comprises: training, by the system, the generative machine learning model using a gradient decent algorithm that optimizes a loss function with respect to parameters of an encoder and a decoder. ### Claim 13 (depends on 12) The computer-implemented method of claim 12, wherein the generative machine learning model generates the recommended chemical reactor control setting by sampling a latent variable from a latent space and decoding the latent variable via the decoder. ### Claim 14 (depends on 11) The computer-implemented method of claim 11, wherein the generative machine learning model is via a generative adversarial network comprising a generator network and a discriminator network. ### Claim 15 (depends on 14) The computer-implemented method of claim 14, further comprising: training, by the system, the generative adversarial network by upsampling a vector of noise to generate new data and analyzing the training data along with the new data to discern a categorical difference between the training data and the new data. ### Claim 16 A computer program product for controlling a chemical reactor, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to: generate, by a system operatively coupled to the processor, a generative machine learning model based on training data regarding a past chemical reactor operation, wherein the generative machine learning model generates a recommended chemical reactor control setting for experimental discovery of a polymer with a target characteristic, wherein the generative machine learning model is via a generative adversarial network model comprising a generator network and a discriminator network, and wherein the generator network upsamples a vector of noise to generate new data, and wherein the discriminator network is a binomial classifier that analyzes the training data and the new data; measure, by the system, a characteristic of the polymer synthesized by the chemical reactor; determine, by the system, whether the characteristic is within a defined range of the target characteristic; and update, by the system, the training data with the recommended chemical reactor control setting, based on the determination that the characteristic is within the defined range. ### Claim 17 (depends on 16) The computer program product of claim 16, wherein the program instructions cause the processor to generate the generative machine learning model in a cloud computing environment. ### Claim 18 (depends on 16) The computer program product of claim 16, wherein the training data comprises a control setting implemented by the chemical reactor during the past chemical reactor operation. ### Claim 19 (depends on 16) The computer program product of claim 16, wherein the generative machine learning model is a variational autoencoder, and wherein the program instructions further cause the processor to train, by the system, the generative machine learning model using a gradient decent algorithm that optimizes a loss function with respect to parameters of an encoder and a decoder.