# US 11,906,951 B2 · Method and system for managing model updates for process models > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,906,951 B2 - Document ID: US11906951B2 - Status: Granted - Kind: B2 - Application number: 17/477,076 - Filed: 2021-09-16 - Published: 2023-03-16 - Granted: 2024-02-20 - Earliest claimed: 2021-09-16 - Assignees: Saudi Arabian Oil Company - Applicants: Saudi Arabian Oil Company - Inventors: Mohammed K Alanazi; Talal A. Altook; Gabriel Winter; Abdulaziz M Alnutaifi - Web page: https://patentgrep.com/patent/US11906951B2 ## CPC - G05B 19/4183: {characterised by data acquisition, e.g. workpiece identification} - G05B 19/41885: {characterised by modeling, simulation of the manufacturing system} - G05B 2219/32128: Gui graphical user interface - G05B 2219/32335: Use of ann, neural network - G05B 2219/35499: Model of process, machine and parameters - 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) ## Abstract A method may include obtaining acquired process data regarding a plant process that is performed by a plant system. The method may further include obtaining from a process model, simulated process data regarding the plant process. The method may further include determining drift data for the process model based on a difference between the acquired process data and the simulated process data. The drift data may correspond to an amount of model drift associated with the process model. The method may further include determining whether the drift data satisfies a predetermined criterion. The method further includes determining, in response to determining that the drift data fails to satisfy the predetermined criterion, a model update for the process model. ## Claims ### Claim 1 A method, comprising: obtaining, by a computer processor, first acquired process data regarding a first plant process that is performed by a plant system; obtaining, by the computer processor and from a first process model, first simulated process data regarding the first plant process, wherein the first process model is a surrogate model that numerically approximates the first plant process using black-box modelling; determining, by the computer processor, first drift data for the first process model based on a difference between the first acquired process data and the first simulated process data, wherein the first drift data corresponds to an amount of model drift associated with the first process model; determining, by the computer processor, whether the first drift data satisfies a first predetermined criterion; determining, by the computer processor and in response to determining that the first drift data fails to satisfy the first predetermined criterion, a model update for the first process model; obtaining, by the computer processor, aggregated drift data from a plurality of plant systems for a plurality of process models, wherein the aggregated drift data comprises the first drift data and the plurality of process models comprises the first process model; determining, by the computer processor, an update schedule for the plurality of process models using the aggregated drift data; transmitting, by the computer processor and automatically based on the update schedule and in response to determining the model update, a plurality of update commands to the plurality of plant systems, wherein the plurality of update commands comprises a first request to a first plant system among the plurality of plant systems for updating the first process model using the model update; and updating, by the first plant system, the first process model using the model update in response to receiving the first request. ### Claim 2 (depends on 1) The method of claim 1, further comprising: obtaining, by the computer processor, second drift data regarding a second process model and third drift data regarding a third process model; determining, by the computer processor, a plurality of model statuses using the second drift data and the third drift data; and presenting, by the computer processor, the plurality of model statuses within a graphical user interface. ### Claim 3 (depends on 1) The method of claim 1, wherein the first acquired process data corresponds to a plurality of inputs to the first plant process and at least one output of the first plant process. ### Claim 4 (depends on 1) The method of claim 1, wherein the first process model generates the first simulated process data based on one or more predetermined simulation modes for the first plant process, and wherein the one or more predetermined simulation modes correspond to a plurality of equipment parameters and a plurality of input parameters for performing the first plant process. ### Claim 5 (depends on 1) The method of claim 1, further comprising: obtaining, by a drift analyzer coupled to a second process model, second acquired process data and second simulated process data regarding a second plant process, wherein the second acquired process data corresponds to a plurality of outputs by the second plant process; and determining, by the drift analyzer and using the second acquired process data and the second simulated process data, second drift data based on a key performance indicator of the second plant process, wherein the second drift data is used to determine whether the second process model satisfies a second predetermined criterion. ### Claim 6 (depends on 1) The method of claim 1, wherein the model update corresponds to a retraining operation using a machine-learning algorithm for the first process model. ### Claim 7 (depends on 1) The method of claim 1, wherein the first process model is an artificial neural network. ### Claim 8 (depends on 1) The method of claim 1, wherein the first predetermined criterion is a drift threshold. ### Claim 9 (depends on 1) The method of claim 1, wherein the first plant process is a petro-chemical process for a refining operation. ### Claim 10 A system, comprising: a first user device; a plurality of plant systems comprising a first plant system, the first plant system comprising a first process model and first plant equipment, wherein the first process model is a surrogate model that numerically approximates a first plant process using black-box modelling; and a model drift manager comprising a computer processor and coupled to the plurality of plant systems and the first user device, wherein the model drift manager comprises functionality for: obtaining first acquired process data regarding a first plant process that is performed by the first plant system; obtaining, from the first process model, first simulated process data regarding the first plant process; determining first drift data for the first process model based on a difference between the first acquired process data and the first simulated process data; determining whether the first drift data satisfies a first predetermined criterion; determining, in response to determining that the first drift data fails to satisfy the first predetermined criterion, a model update for the first process model; obtaining aggregated drift data from a plurality of plant systems for a plurality of process models, wherein the aggregated drift data comprises the first drift data and the plurality of process models comprises the first process model; determining an update schedule for the plurality of process models using the aggregated drift data; and transmitting, automatically based on the update schedule and in response to determining the model update, a plurality of update commands to the plurality of plant systems, wherein the plurality of update commands comprises a first request to a first plant system among the plurality of plant systems for updating the first process model using the model update, and wherein the first plant system updates the first process model using the model update in response to receiving the first request. ### Claim 11 (depends on 10) The system of claim 10, further comprising: a second plant system comprising a second process model and second plant equipment; and a drift analyzer coupled to the second process model, wherein the drift analyzer determines second drift data based on a key performance indicator of a second plant process using second acquired process data and second simulated process data, and wherein the second drift data is used to determine whether the second process model satisfies a second predetermined criterion. ### Claim 12 (depends on 10) The system of claim 10, wherein the model drift manager further comprises functionality for: obtaining second drift data regarding a second process model and third drift data regarding a third process model; determining a plurality of model statuses using the second drift data and the third drift data; and presenting, by the first user device, the plurality of model statuses within a graphical user interface. ### Claim 13 (depends on 10) The system of claim 10, wherein the model drift manager further comprises functionality for: transmitting, in response to determining that the first process model has not been updated within a predetermined period of time, a second request to a second user device to update the first process model, wherein the first request is based on a first escalation level, and wherein the second request is based on a second escalation level. ### Claim 14 (depends on 10) The system of claim 10, wherein the first process model is an artificial neural network, wherein the first predetermined criterion is a drift threshold, and wherein the first plant process is a petro-chemical process for a refining operation.