System and method for mitigating data drift in industrial plant

Abstract

Embodiments of the present disclosure relate to a system and method for mitigating data drift in an industrial plant. The method includes monitoring, by a processor (202), one or more process parameters associated with an industrial plant (102). The method includes detecting, by the processor (202), a drift in the one or more process parameters based on a deviation from the one or more predefined process parameters. The method includes determining, by the processor (202), one or more drift contexts and process contexts based on the drift and the one or more process parameters. The method includes determining, by the processor (202), a sampling policy from a plurality of sampling policies based on the one or more drifts and the process context, the sampling policy for sampling the one or more process parameters using a first artificial intelligence (AI) model (216). The method includes training, by the processor (202), a second AI model (302) for mitigating data drift based on the sampling policy.

Description

Technical Field

[0001] The present disclosure relates to industrial plants. More particularly, the present disclosure relates to mitigating data drift in industrial plants. Background Art

[0002] Operating conditions in industrial plants change over time due to changes in raw materials, changes in operating processes, equipment wear, etc. Such changes in operating conditions can cause drift in one or more process parameters associated with the industrial plant. When operating conditions change, the performance of any artificial intelligence (AI) and / or machine learning (ML) model may degrade because the model was trained using one or more process parameters before the drift. If model performance degrades, any automation planned in the processing plant based on model predictions will stagnate. This may lead to undesirable results such as poor product quality, low yield, equipment failure, plant shutdowns, etc. Conventional methods for mitigating drift and reducing AI model degradation due to drift of one or more process parameters use data sampling to collect and learn new characteristics of one or more process parameters. However, if the data sampling is inaccurate, it may result in insufficient training and / or even negative training. Therefore, data sampling needs to be performed accurately to train AI models for mitigating drift.

[0003] The information disclosed in the background section of this disclosure is only for enhancement of understanding of the general background of the invention and should not be taken as an acknowledgment or any form of suggestion that this information constitutes the prior art already known to those skilled in the art. Summary of the Invention

[0004] The foregoing summary is illustrative only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features will become apparent by reference to the drawings and the following detailed description.

[0005] In one embodiment, a method for mitigating data drift in an industrial plant is disclosed. The method includes monitoring, by a processor, one or more process parameters associated with the industrial plant. The method includes detecting, by the processor, drift in the one or more process parameters based on deviations from one or more predefined process parameters. The method includes determining, by the processor, one or more drift contexts and one or more process contexts based on the drift and the one or more process parameters. The method includes determining, by the processor, a sampling strategy from a plurality of sampling strategies based on the one or more drift contexts and the one or more process contexts, the sampling strategy for sampling the one or more process parameters using a first artificial intelligence (AI) model. Furthermore, the method includes training, by the processor, a second AI model for mitigating data drift based on the sampling strategy, wherein the second AI model is configured to control one or more processes in the industrial plant.

[0006] In one embodiment, a system for mitigating data drift in an industrial plant is disclosed. The system includes a memory storing processor-executable instructions. The system includes a processor configured to execute the processor-executable instructions stored in the memory and, thereby, to monitor one or more process parameters associated with the industrial plant. The processor is configured to detect drift in the one or more process parameters based on a deviation from one or more predefined process parameters. The processor is configured to determine one or more drift contexts and one or more process contexts based on the drift and the one or more process parameters. The processor is configured to determine a sampling strategy from a plurality of sampling strategies based on the one or more drift contexts and the one or more process contexts, the sampling strategy for sampling the one or more process parameters using a first AI model. The processor is configured to train a second AI model for mitigating data drift based on the sampling strategy, wherein the second AI model is configured to control one or more processes in the industrial plant.

[0007] The foregoing summary is illustrative only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features will become apparent by reference to the drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate exemplary embodiments and, together with the description, serve to explain the principles disclosed. Like numerals are used throughout the drawings to refer to similar features and components. Some embodiments of at least one of the apparatus and method according to embodiments of the present subject matter will now be described, by way of example only, and with reference to the accompanying drawings, in which:

[0009] Figure 1 illustrates an environment in which some embodiments of the present disclosure may be practiced;

[0010] Figure 2 A system for mitigating data drift in an industrial plant according to an embodiment of the present disclosure is illustrated;

[0011] Figure 3 A block diagram for training a second AI model deployed in an industrial factory according to an embodiment of the present disclosure is illustrated;

[0012] Figure 4 A flowchart illustrating training a first AI model and a second AI model according to an embodiment of the present disclosure is illustrated; and

[0013] Figure 5 A flow chart of a method for mitigating data drift in an industrial plant according to an embodiment of the present disclosure is illustrated.

[0014] Figure 6 Illustrated is a block diagram of an exemplary computer system 600 for implementing embodiments consistent with the present disclosure.

[0015] The accompanying drawings depict embodiments of the present disclosure for illustrative purposes only. Those skilled in the art will readily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles disclosed herein. DETAILED DESCRIPTION

[0016] In this document, the word “exemplary” is used herein to mean “serving as an example, instance, or illustration.” Any embodiment or implementation of the subject matter described herein as “exemplary” is not necessarily to be construed as advantageous over other embodiments.

[0017] While the present disclosure is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and will be described in detail below. However, it should be understood that it is not intended to limit the present disclosure to the particular forms disclosed, but rather, the present disclosure is intended to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure.

[0018] The terms "comprises," "comprising," or any other variations thereof are intended to cover a non-exclusive inclusion, such that an arrangement, apparatus, or method that comprises a list of components or steps includes not only those components or steps but may also include other components or steps not expressly listed or inherent to such arrangement, apparatus, or method. In other words, a listing of one or more elements in an apparatus, system, or device beginning with "comprises...a" does not, without more constraints, preclude the presence of other or additional elements in the apparatus, system, or device.

[0019] In the following detailed description of the embodiments of the present disclosure, reference is made to the accompanying drawings which form a part of the present disclosure, in which specific embodiments in which the present disclosure may be practiced are shown by way of illustration. These embodiments are described in sufficient detail to enable those skilled in the art to practice the present disclosure, and it should be understood that other embodiments may be utilized and may be changed without departing from the scope of the present disclosure. Therefore, the following description should not be considered restrictive.

[0020] It should be noted that the embodiments of the present disclosure are explained with reference to a reinforcement learning (RL) process, wherein the first artificial intelligence (AI) model is an RL model. However, it should be noted that any other AI model (such as a rule-based system, a genetic algorithm, etc.) can be used as the first AI model.

[0021] As used herein, the term "AI model(s)" refers to models that are trained on a set of data to recognize certain patterns and / or make certain decisions without further human intervention. AI models apply different algorithms to relevant data inputs to achieve the tasks and / or outputs they have been trained for. In some embodiments, the "AI model(s)" may be continuous learning models.

[0022] The following references Figure 1-Figure 5 Various embodiments of the present disclosure are explained.

[0023] Figure 1 An environment 100 is illustrated in which some embodiments of the present disclosure may be practiced. The environment 100 illustratively depicts an industrial plant 102 and a system 104. Some examples of industrial plant 102 may include, but are not limited to, nuclear power plants, oil refineries, chemical plants, fertilizer plants, manufacturing plants, power plants, and the like.

[0024] Data drift in the industrial plant 102 occurs due to deviations of one or more process parameters from one or more predefined process parameters. For example, in a steel manufacturing plant, the iron content of the iron ore may drift, i.e., one or more predefined process parameters for the iron content may change from 60% to 70%. Thus, there is a 10% drift in the iron content. One or more AI models in the steel manufacturing plant (such as Figure 3 The second AI model 302 in FIG. 3 can be trained with one or more process parameters to perform one or more actions, such as automated actions, including but not limited to mixing chemicals, generating notifications, operating actuators, controlling generators, and the like. While the one or more actions are performed based on the one or more process parameters, any drift in the one or more process parameters needs to be accounted for by the one or more AI models. The present disclosure provides a method and system for mitigating the effects of drift in one or more process parameters.

[0025] In one embodiment, data drift may include concept drift. Concept drift in this article refers to the deviation of the underlying target, i.e., the input and output target parameters of the model, a set of targets and / or AI models in the industrial plant 102 (such as Figure 3 The deviation of the underlying relationship between the targets of the second AI model 302 in the AI ​​model 302 is measured. It should be noted that both data drift and concept drift can lead to performance degradation of the AI ​​model. In one embodiment, the performance deviation of the AI ​​model relative to a set of targets, goals, and / or objectives is measured using key performance indicators (KPIs).

[0026] In one example, industrial plant 102 may be a chemical plant. In a chemical plant, the sulfur content of the raw materials used may drift. Specifically, one or more predefined process parameters for sulfur content may be 50%, but the sulfur content detected from one or more process parameters may be 70%. Consequently, the sulfur content of the raw materials may drift by 20%, resulting in a drift in the chemical reactions of the chemical plant. This drift in sulfur content may also cause a drift in the expected output product, resulting in reduced efficiency due to the drift.

[0027] In one embodiment, the system 104 is deployed in an industrial plant 102 to mitigate drift. Figure 2 The operations performed by the system 104 are explained in detail.

[0028] Figure 2 1 is a diagram illustrating a system for mitigating data drift in an industrial plant 102 according to an embodiment of the present disclosure. A system 104 is deployed in the industrial plant 102 for mitigating drift in the industrial plant 102. The system 104 is depicted as including a processor 202, a memory 204, an input / output (I / O) module 206, and a communication interface 208. In some embodiments, the system 104 can be deployed on premises at the industrial plant 102, or can be remotely deployed and connected to the industrial plant 102. For example, the system 104 can be deployed in the cloud and configured to receive one or more process parameters from the industrial plant 102 via a communication network.

[0029] It should be noted that in some embodiments, system 104 may include more or fewer components than depicted herein. The various components of system 104 may be implemented using hardware, software, firmware, or any combination thereof. Furthermore, the various components of system 104 may be operatively coupled to one another. More specifically, the various components of system 104 may be capable of communicating with one another using a communication channel medium (such as a bus, an interconnect, etc.).

[0030] In one embodiment, the processor 202 may be embodied as a multi-core processor, a single-core processor, or a combination of one or more multi-core processors and one or more single-core processors. For example, the processor 202 may be embodied as one or more of various processing devices, such as a coprocessor, a microprocessor, a controller, a digital signal processor (DSP), a processing circuit system with or without an accompanying DSP, or various other processing devices, including a microcontroller unit (MCU), a hardware accelerator, a dedicated computer chip, etc. The processor 202 may include, but is not limited to, a drift detection module 212, a performance analyzer 218, etc.

[0031] The drift detection module 212 can detect drift between one or more process parameters and one or more predefined process parameters. The performance analyzer 218 can determine a performance metric of the AI ​​model. In addition, the performance analyzer 218 determines a reward value for the AI ​​model by comparing the performance metric with a predefined performance metric (i.e., a reference performance metric). The deviation of the performance of the AI ​​model is determined based on the reward value. If the reward value is maximized, the performance of the AI ​​model is considered to be optimal.

[0032] In one embodiment, memory 204 can store machine-executable instructions 205 (referred to herein as instructions 205) and a first AI model 216. Some examples of first AI model 216 may include, but are not limited to, RL models, rule-based systems, genetic algorithms, and the like. In one embodiment, processor 202 is embodied as an executor of software instructions. Thus, processor 202 can execute instructions 205 stored in memory 204 to perform one or more operations described herein.

[0033] The memory 204 may be any type of memory accessible to the processor 202 to execute corresponding functions. For example, the memory 204 may include one or more volatile or non-volatile memories or a combination thereof. For example, the memory 204 may be embodied as a semiconductor memory such as a flash memory, a mask ROM, a PROM (Programmable ROM), an EPROM (Erasable PROM), a RAM (Random Access Memory), etc.

[0034] In one embodiment, the processor 202 is configured to execute the instructions 205 and use the first AI model 216 to: (1) monitor one or more process parameters associated with the industrial plant 102, (2) detect a drift in the one or more process parameters based on a deviation from one or more predefined process parameters, (3) determine one or more drift contexts 222 and one or more process contexts 224 based on the drift and the one or more process parameters, (4) determine a sampling strategy from a plurality of sampling strategies 220 based on the one or more drift contexts 222 and the one or more process contexts 224 for sampling the one or more process parameters using the first AI model 216, and (5) train a second AI model 302 (e.g., based on the sampling strategy) Figure 3 ) is used to mitigate data drift, where the second AI model 302 is configured to control one or more processes in the industrial plant 102.

[0035] In one embodiment, the I / O module 206 may include a mechanism configured to receive input from and provide output to an operator of the system 104 (not shown). As used herein, the term "operator of the system 104" may refer to one or more individuals directly or indirectly associated with the management system 104. To enable receiving input and providing output to the system 104, the I / O module 206 may include at least one input interface and at least one output interface. In one example, the operator of the system 104 may configure the system 104 via at least one input interface. Examples of input interfaces may include, but are not limited to, a keyboard, a mouse, a joystick, a keypad, a touch screen, soft keys, a microphone, and the like. Examples of output interfaces may include, but are not limited to, a display (such as a light emitting diode display, a thin film transistor (TFT) display, a liquid crystal display, an active matrix organic light emitting diode (AMOLED) display), a microphone, a speaker, a ringer, and the like.

[0036] In one embodiment, the communication interface 208 may include mechanisms configured to communicate with other entities in the environment 100 (e.g., the industrial plant 102, the database 210, etc.). In one embodiment, the communication interface 208 of the system 104 receives one or more process parameters to detect drift and determines one or more process contexts 224 and one or more drift contexts 222.

[0037] System 104 is depicted as being in operational communication with a database 210. In one embodiment, database 210 is configured to store a plurality of sampling strategies 220, a plurality of drift contexts 222, and a plurality of process contexts 224. In one embodiment, database 210 is configured to store various AI prediction models based on different specifications. The various AI models may include, but are not limited to, reinforcement learning (RL) models, rule-based systems, genetic algorithms, and the like.

[0038] One or more drift contexts 222 may include information related to various types of drift. The drift contexts 222 in the database 210 may include multiple drift contexts 222 that may be simulated using a simulator, obtained during the training phase of the first AI model 216, and the like.

[0039] The process context 224 in the database 210 includes a plurality of process contexts, wherein the plurality of process contexts 224 may include information based on a type of industrial plant and process conditions associated with the industrial plant.

[0040] Database 210 may include multiple storage units, such as hard disks and / or solid-state disks in a Redundant Array of Inexpensive Disks (RAID) configuration. In some embodiments, database 210 may include a storage area network (SAN) and / or a network-attached storage (NAS) system. In one embodiment, database 210 may correspond to a distributed storage system, wherein individual databases are configured to store information, such as multiple sampling strategies 220, multiple drift contexts 222, multiple process contexts 224, and the like.

[0041] refer to Figure 3 Drift mitigation in industrial plant 102 by system 104 is explained.

[0042] Figure 3 A block diagram for training a second AI model 302 deployed in an industrial plant 102 according to an embodiment of the present disclosure is illustrated. The block diagram 300 includes some components of the system 104 and the second AI model 302. The second AI model 302 is deployed in the industrial plant 102 to mitigate drift. The system 104 is deployed in the industrial plant 102 to train the second AI model 302 in the industrial plant 102. In some embodiments, the system 104 is pre-trained before being deployed in the industrial plant 102. Figure 4 The steps involved in training the system 104 are further explained.

[0043] The drift detection module 212 monitors one or more process parameters associated with the industrial plant 102. The one or more process parameters vary based on the industrial plant 102. For example, in a steel manufacturing plant, the one or more process parameters may include, but are not limited to, the composition of the raw materials, the composition of the steel, temperature, pressure, etc. In another example, in a chemical plant, the one or more process parameters may include, but are not limited to, the chemical composition of the raw materials, released intermediate gases, temperature, pressure, etc.

[0044] The drift detection module 212 detects a drift in one or more process parameters based on a deviation from one or more predefined process parameters. The drift detection module 212 monitors one or more process parameters associated with the industrial plant 102 and detects when a deviation occurs from the one or more predefined process parameters, which may be stored in the memory 204. For example, in a steel manufacturing plant, there may be a drift in the iron composition of the iron ore, i.e., one or more predefined process parameters for the iron composition may be 60%, but the composition detected from the one or more process parameters may be 70%. Thus, there is a 10% drift in the iron composition. The manufacturing units of the steel manufacturing plant are trained on iron ore with an iron composition of 60%, and thus, the steel manufacturing is changed to be dominated by the new iron ore composition.

[0045] The drift detection module 212 determines one or more drift contexts 222 and one or more process contexts 224 based on the drift and one or more process parameters. The one or more drift contexts 222 include information related to the type of drift. For example, the one or more drift contexts 222 may include information related to the cause of the drift, such as changes in raw material quality, changes in raw material composition, etc. The one or more process contexts 224 are based on the type of industrial plant 102 and the process conditions associated with the industrial plant 102. For example, the one or more process contexts 224 may include information such as wear and tear on equipment, malfunctioning sensors, factors affecting temperature and / or pressure changes, etc. The one or more drift contexts 222 and one or more process contexts 224 determined by the drift detection module 212 are given as input to the first AI model 216.

[0046] First AI model 216 determines at least one sampling strategy from plurality of sampling strategies 220 for sampling one or more process parameters using first AI model 216 based on one or more drift contexts 222 and one or more process contexts 224 .

[0047] The term "sampling" as used herein refers to the sampling of data used to train an AI model. Sampling is the selection of a subset of data from a statistical population to estimate the characteristics of the entire population. Sampling enables the AI ​​model to process smaller, more manageable subsets of data rather than trying to analyze the entire data population. With representative samples, the AI ​​model can build and run analytical models more quickly while still producing accurate results. In the present disclosure, sampling of one or more process parameters is performed to train the second AI model 302 for mitigating drift due to deviations in one or more process parameters. The first AI model 216 is configured to train the second AI model 302 based on at least one sampling strategy 220. The training of the AI ​​models mentioned herein may include, but is not limited to, retraining AI models, fine-tuning AI models, adapting AI models, etc., which process data to ensure that the AI ​​model continuously learns to perform specific tasks (such as selecting the best sampling strategy for mitigating drift).

[0048] Some examples of the plurality of sampling strategies 220 may include, but are not limited to, a random sampling strategy, a stratified sampling strategy, a systematic sampling strategy, a biased representative sampling strategy, an event-driven sampling strategy, and the like.

[0049] Embodiments of the present disclosure will be explained below with reference to an RL model as the first AI model 216. The first AI model 216 is pre-trained for one or more drift contexts 222 and one or more process contexts 224. However, it should be noted that any other AI model (such as a rule-based system, a genetic algorithm, etc.) can be used as the first AI model 216.

[0050] Genetic algorithms are defined as computational optimization techniques inspired by the principles of natural selection and genetics. They are used to solve complex problems by simulating the process of evolution to iteratively improve a population of potential solutions. In the present disclosure, a first AI model 216 can be trained using a genetic algorithm, such that the first AI model 216 determines a plurality of sampling strategies 220 from a database. In some embodiments, the sampling strategies can be considered a population, and a fitness score is provided for each sampling strategy. In some embodiments, a crossover of the sampling strategies 220 can be performed, and a fitness score can be derived for the new population resulting from the crossover. The sampling strategy 220 with the highest fitness score can be selected. For each of the plurality of sampling strategies 220, the first AI model 216 trains a second AI model 302. The performance metric of the second AI model 302 is evaluated using a predefined performance metric. Based on the evaluation of the performance metric, a fitness value is assigned to each of the plurality of sampling strategies 220. The sampling strategy with the highest fitness value is considered the optimal sampling strategy.

[0051] Rule-based learning is a method of identifying, learning, or evolving "rules" for storage, manipulation, or application. The defining characteristic of a rule-based learning model is the identification and utilization of a set of relational rules that collectively represent the knowledge captured by the system. Rule-based learning models use a set of pre-written rules to make decisions and solve problems. Developers create rules based on human expert knowledge that enable the system to process input data and produce results. In the present disclosure, the first AI model 216 can be a rule-based learning model in which a set of rules are pre-written for selecting the best sampling strategy from multiple sampling strategies 220.

[0052] The first AI model 216 trains the second AI model 302 based on the sampling strategy to mitigate data drift, wherein the second AI model 302 is configured to control one or more processes in the industrial plant 102. The second AI model 302 can be any AI model. In some embodiments, a single AI model can be used to implement the above functions.

[0053] Performance analyzer 218 in processor 202 compares the performance metric of second AI model 302 with a predefined performance metric. When the performance metric of second AI model 302 is less than the predefined performance metric, processor 202 generates at least one new sampling strategy. When the performance metric of second AI model 302 is greater than the predefined performance metric, the sampling strategy used to train second AI model 302 is deemed to be the optimal sampling strategy.

[0054] For example, if the second AI model 302 is deployed in a steel manufacturing plant, the performance metric may be based on comparing the final composition of the steel to a predefined desired composition. For example, one or more process parameters may be found to be 80% iron and 5% carbon. However, one or more predefined process parameters are 99% iron and less than 0.25% carbon. Therefore, the performance metric of the second AI model 302 is less than the predefined performance metric.

[0055] In one embodiment, performance analyzer 218 generates a reward value for first AI model 216. When the performance metric of second AI model 302 is greater than a predefined performance metric, the reward value is positive, and when the performance metric of second AI model 302 is less than the predefined performance metric, the reward value is negative. For example, assuming the initial reward value for first AI model 216 is 50, if the performance metric of second AI model 302 is less than the predefined performance metric, performance analyzer 218 may generate a reward value of 25. In another example, if the performance metric of second AI model 302 is greater than the predefined performance metric, performance analyzer 218 may generate a reward value of 75. The goal of system 104 is to obtain the maximum reward value. The larger the reward value, the more optimal the sampling strategy is considered.

[0056] For example, if the industrial plant 102 includes a stirred tank reactor, the one or more predefined process parameters may be: a normally distributed reactant concentration of approximately 1.0 molar with a small standard deviation, a reaction temperature of 800°C, an operating pressure of 2 bar, etc. The drift detection module 212 may detect a drift in the normally distributed concentration, where the concentration of the normally distributed concentration is approximately 0.8 M due to a change in the supplier. The drift in the normally distributed concentration also causes the pressure to increase from 2 bar to 5 bar. Here, the drift detection module 212 determines one or more drift contexts 222 as "input data drift" and one or more process contexts 224 as information about "pressure increase (from 2 bar to 5 bar)". The first AI model 216 may determine a sampling strategy from the one or more sampling strategies 220. The second AI model 302 deployed in the stirred tank reactor is trained using a sampling strategy with an initial reward value of 900.

[0057] The performance analyzer 218 compares the performance metric of the second AI model 302 with the predefined performance metric. For example, after training the second AI model 302 using the sampling strategy in a stirred tank reactor, if the normal distribution is approximately 1.0 mol, the deviation is small, and the operating pressure is 2 bar (the same as one or more predefined process parameters), then the performance metric is greater than the predefined performance metric, and the new reward value is 1000. Therefore, the reward value is called a positive reward value.

[0058] In one embodiment, the reward value is based on a performance metric.

[0059] In one embodiment, when the first AI model 216 fails to select the best sampling strategy from the plurality of predefined sampling strategies 220, the processor 202 generates a new sampling strategy. The plurality of predefined sampling strategies 220 may include, but are not limited to, a random sampling strategy, a stratified sampling strategy, a systematic sampling strategy, a biased representative sampling strategy, an event-driven sampling strategy, etc. The new sampling strategy is generated such that when the second AI model 302 is trained using the sampling strategy, the performance metric of the second AI model 302 is greater than the predefined performance metric and the reward value is maximized.

[0060] refer to Figure 4 The method 400 for training the first AI model 216 and the second AI model 302 is further explained.

[0061] Figure 4 A flowchart for training the first AI model 216 and the second AI model 302 according to an embodiment of the present disclosure is illustrated.

[0062] At 402, the processor 202 receives training data including a plurality of drift contexts 222 and a plurality of process contexts 224 from the database 210. The processor 202 may receive the plurality of drift contexts 222 and the plurality of process contexts 224 from another entity (not limited to the database 210). In one embodiment, the plurality of drift contexts 222 and the plurality of process contexts 224 may be simulated using a simulator. The plurality of drift contexts 222 include information related to the type of drift. For example, the plurality of drift contexts 222 may include information related to the quality of raw materials, changes in the composition of raw materials, etc. The plurality of process contexts 224 are based on the type of industrial plant 102 and the process conditions associated with the industrial plant 102. For example, the plurality of process contexts 224 may include information such as wear and tear of equipment, malfunctioning sensors, factors affecting temperature and pressure changes, etc.

[0063] At 404, the processor 202 selects a sampling strategy from the plurality of predefined sampling strategies 220 for each combination of the drift context 222 and the process context 224 in the training data. The plurality of predefined sampling strategies 220 may include, but are not limited to, a random sampling strategy, a stratified sampling strategy, a systematic sampling strategy, a biased representative sampling strategy, an event-driven sampling strategy, and the like.

[0064] At 406 , the processor 202 trains the second AI model 302 for drift mitigation based on the sampling strategy.

[0065] At 408 , performance analyzer 218 in processor 202 evaluates the performance metric of second AI model 302 by comparing the performance metric of second AI model 302 to a predefined performance metric.

[0066] At 410 , performance analyzer 218 determines whether the performance metric of second AI model 302 is greater than or less than a predefined performance metric. When the performance metric of second AI model 302 is less than the predefined performance metric, new sampling strategy 220 is selected from database 210 .

[0067] At 412 , when the performance metric of second AI model 302 is greater than the predefined performance metric, sampling strategy 220 used to train second AI model 302 is stored in database 210 along with corresponding process context 224 and drift context 222 .

[0068] refer to Figure 4 The disclosed method 400 may be implemented using software comprising computer-executable instructions stored on one or more computer-readable media (e.g., non-transitory computer-readable media such as one or more optical disks, volatile memory components (e.g., dynamic random access memory (DRAM) or SRAM), or non-volatile memory or storage components (e.g., a hard drive or a solid-state non-volatile memory component such as a flash memory component) and executed on a computer (e.g., any suitable computer such as a laptop, netbook, netbook, tablet computing device, smartphone, or other mobile computing device). For example, such software may be executed on a single local computer.

[0069] refer to Figure 5 The method 500 for mitigating data drift in the industrial plant 102 is further explained.

[0070] Figure 5 A flow chart of a method for mitigating data drift in an industrial plant 102 according to an embodiment of the present disclosure is illustrated.

[0071] At 502 , the processor 202 monitors one or more process parameters associated with the industrial plant 102 .

[0072] At 504 , the processor 202 detects a drift in one or more process parameters based on a deviation from one or more predefined process parameters.

[0073] At 506 , the processor 202 determines one or more drift contexts 222 and one or more process contexts 224 based on the drift and the one or more process parameters.

[0074] At 508 , the processor 202 determines a sampling strategy from the plurality of sampling strategies 220 based on the one or more drift contexts 222 and the one or more process contexts 224 , the sampling strategy for sampling the one or more process parameters using the first AI model 216 .

[0075] At 510 , the processor 202 trains the second AI model 302 based on the sampling strategy for mitigating data drift, wherein the second AI model 302 is configured to control one or more processes in the industrial plant 102 .

[0076] refer to Figure 5 The disclosed method 500 may be implemented using software comprising computer-executable instructions stored on one or more computer-readable media (e.g., non-transitory computer-readable media such as one or more optical disks, volatile memory components (e.g., DRAM or SRAM), or non-volatile memory or storage components (e.g., a hard drive or solid-state non-volatile memory components such as flash memory components) and executed on a computer (e.g., any suitable computer such as a laptop, netbook, netbook, tablet computing device, smartphone, or other mobile computing device). For example, such software may be executed on a single local computer.

[0077] Figure 6 A block diagram of an exemplary computer system 600 for implementing embodiments consistent with the present disclosure is illustrated. Computer system 600 may be, but is not limited to, system 104 or second AI model 302. Computer system 600 may include a central processing unit ("CPU" or "processor") 601. Processor 601 may include at least one data processor for performing processing. Processor 601 may include specialized processing units such as an integrated system (bus) controller, a memory management control unit, a floating point unit, a graphics processing unit, a digital signal processing unit, and the like.

[0078] The processor 601 may be configured to communicate with one or more input / output (I / O) devices 608 and 609 via an I / O interface 607. The I / O interface 607 may employ communication protocols / methods such as, but not limited to, audio, analog, digital, mono, RCA, stereo, IEEE-1394, serial bus, Universal Serial Bus (USB), infrared, PS / 2, BNC, coaxial, component, composite, Digital Video Interface (DVI), High-Definition Multimedia Interface (HDMI), RF antenna, S-Video, VGA, IEEE 802.n / b / g / n / x, Bluetooth, cellular (e.g., Code Division Multiple Access (CDMA), High Speed ​​Packet Access (HSPA+), Global System for Mobile Communications (GSM), Long Term Evolution (LTE), WiMax, etc.), and the like.

[0079] Using the I / O interface 607, the computer system 600 can communicate with one or more I / O devices 608 and 609. For example, the input device 608 can be an antenna, a keyboard, a mouse, a joystick, an (infrared) remote control, a camera, a card reader, a fax machine, a dongle, a biometric reader, a microphone, a touch screen, a touchpad, a trackball, a stylus, a scanner, a storage device, a transceiver, a video device / source, etc. The output device 609 can be a printer, a fax machine, a video display (e.g., a cathode ray tube (CRT), a liquid crystal display (LCD), a light emitting diode (LED), a plasma, a plasma display panel (PDP), an organic light emitting diode display (OLED)), etc.), an audio speaker, etc.

[0080] In some embodiments, the processor 601 can be configured to communicate with external components, such as external computer systems, servers, and network components. The network interface 610 can utilize connection protocols including, but not limited to, direct connection, Ethernet (e.g., twisted pair 10 / 100 / 1000BaseT), Transmission Control Protocol / Internet Protocol (TCP / IP), Token Ring, IEEE 802.11a / b / g / n / x, and the like.

[0081] In some embodiments, the processor 601 may be configured to communicate with a memory 603 (e.g., RAM, ROM, etc.) via a storage interface 602. The storage interface 602 may be connected to a memory 603 including, but not limited to, a memory drive, a removable disk drive, etc., using a connection protocol such as Serial Advanced Technology Attachment (SATA), Integrated Drive Electronics (IDE), IEEE-1394, Universal Serial Bus (USB), Fibre Channel, Small Computer System Interface (SCSI), etc. The memory drive may also include a drum, a magnetic disk drive, a magneto-optical drive, an optical disk drive, a redundant array of independent disks (RAID), a solid-state memory device, a solid-state drive, etc.

[0082] The memory 603 may store a collection of program or database components, including but not limited to a user interface 604, an operating system 605, a web browser 606, etc. In some embodiments, the computer system 600 may store user / application data, such as data, variables, records, etc. described in this disclosure. Such a database may be implemented as a fault-tolerant, relational, scalable, secure database, such as or

[0083] Operating system 605 may facilitate resource management and operation of computer system 600. Examples of operating systems include, but are not limited to, APPLE OS X, UNIX-like system distributions (for example, BERKELEYSOFTWARE DISTRIBUTION TM (BSD), FREEBSD TM , NETBSD TM 、OPENBSD TM etc.), LINUXDISTRIBUTIONS TM (For example, RED HAT TM UBUNTU TM 、KUBUNTU TM etc.), IBM TM OS / 2, MICROSOFT TM WINDOWS TM (XP TM VISTA TM / 7 / 8, 10, etc.), iOS TM 、 Android TM 、 OS, etc.

[0084] In some embodiments, computer system 600 may implement a program component stored in a web browser 606. Web browser 606 may be a hypertext viewing application such as INTERNET GOOGLE TM CHROME TM 、 etc. Secure web browsing can be provided using Secure Hypertext Transfer Protocol (HTTPS), Secure Sockets Layer (SSL), Transport Layer Security (TLS), etc. The web browser 606 can utilize web browsers such as AJAX, DHTML, In some embodiments, the computer system 600 may implement a program component stored in a mail server. The mail server may be an Internet mail server such as Microsoft Exchange. The mail server may utilize a web server such as Active Server Pages (ASP), C++ / C#, .NET、CGI SCRIPTS、 PHP, Mail servers can utilize facilities such as Internet Message Access Protocol (IMAP), Messaging Application Programming Interface (MAPI), Exchange, Post Office Protocol (POP), Simple Mail Transfer Protocol (SMTP), and other communication protocols. In some embodiments, the computer system 600 may implement a mail client stored program component. The mail client may be a mail viewing application such as MAIL, wait.

[0085] Some advantages of the present disclosure include providing a method and system for mitigating drift in industrial plants. By mitigating drift in industrial plants, the efficiency of the industrial plants is improved. Furthermore, the present disclosure provides a method for fine-tuning AI models in industrial plants, thereby reducing the time required to retrain the entire model for drift data. The present disclosure also eliminates the need for sufficient data to efficiently retrain the AI ​​model for drift in one or more process parameters. Consequently, the system is also robust, easy to implement, and requires no additional data.

[0086] The described operations can be implemented as methods, systems or products that use standard programming and / or engineering techniques to produce software, firmware, hardware or any combination thereof. The described operations can be implemented as code maintained in a "non-transitory computer-readable medium", where a processor can read and execute the code from the computer-readable medium. The processor is at least one of a microprocessor and a processor capable of processing and executing queries. Non-transitory computer-readable media may include media such as magnetic storage media (e.g., hard drives, floppy disks, tapes, etc.), optical storage (CD-ROMs, DVDs, optical disks, etc.), volatile and non-volatile memory devices (e.g., EEPROMs, ROMs, PROMs, RAMs, DRAMs, SRAMs, flash memories, firmware, programmable logic, etc.). In addition, non-transitory computer-readable media may include all computer-readable media except transient ones. The code for implementing the above operations can also be implemented in hardware logic (e.g., integrated circuit chips, programmable gate arrays (PGAs), application-specific integrated circuits (ASICs), etc.).

[0087] The steps shown are intended to explain the exemplary embodiments shown, and it should be expected that ongoing technological developments will change the way specific functions are performed. These examples are given herein for illustration and not limitation. In addition, for ease of description, the boundaries of the functional building blocks are arbitrarily defined herein. As long as the specified functions and their relationships are properly performed, alternative boundaries can be defined. Based on the teachings contained herein, those skilled in the relevant art will understand that the alternatives described herein (including equivalents, extensions, variations, deviations, etc.) are within the scope and spirit of the disclosed embodiments. In addition, words such as "comprising," "having," "containing," and "including" and other similar forms are equivalent in meaning and are open, because one or more items following any of these words do not represent an exhaustive list of these items, nor are they limited to the listed items. It must also be noted that, unless the context clearly dictates otherwise, the singular forms "a," "an," and "the" used herein include plural references.

[0088] In addition, one or more computer-readable storage media may be used to implement embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory that can store information or data that is readable by a processor. Thus, a computer-readable storage medium can store instructions for execution by one or more processors, including instructions for causing (multiple) processors to perform steps or stages consistent with the embodiments described herein. The term "computer-readable medium" should be understood to include tangible items, but not carrier waves and transient signals, i.e., non-transient. Examples include random access memory (RAM), read-only memory (ROM), volatile memory, non-volatile memory, hard drives, CD-ROMs, DVDs, flash drives, disks, and any other known physical storage media.

[0089] Finally, the language used in the specification is primarily selected for readability and teaching purposes and may not be intended to define or limit the subject matter of the present invention. Therefore, the disclosure of the embodiments of the present disclosure is intended to illustrate but not limit the scope of the present disclosure.

[0090] Regarding the use of substantially any plural and / or singular terms herein, those skilled in the art can convert the plural to the singular and / or the singular to the plural as the context and / or application requires. For clarity, various singular / plural arrangements may be explicitly described herein.

Claims

1. A method for mitigating data drift in an industrial plant (102), comprising: monitoring, by a processor (202), one or more process parameters associated with the industrial plant (102); detecting, by the processor (202), a drift in the one or more process parameters based on a deviation from one or more predefined process parameters; determining, by the processor (202), one or more drift contexts (222) and one or more process contexts (224) based on the drift and the one or more process parameters; determining, by the processor (202), a sampling strategy from a plurality of sampling strategies (220) based on the one or more drift contexts (222) and the one or more process contexts (224), for sampling the one or more process parameters using a first artificial intelligence (AI) model (216); as well as A second AI model (302) is trained by the processor (202) based on the sampling strategy for mitigating data drift, wherein the second AI model (302) is configured to control one or more processes in the industrial plant (102).

2. The method of claim 1, wherein the first AI model (216) is trained using one of a reinforcement learning (RL) process, a rule-based system, and a genetic algorithm.

3. The method of claim 2, wherein training the first AI model (216) and the second AI model (302) comprises: The processor (202) provides the first AI model (216) with training data including a plurality of drift contexts (222) and a plurality of process contexts (224); Configuring, by the processor (202), the first AI model (216) to train the second AI model (302) using at least one sampling strategy from a plurality of predefined sampling strategies (220) on the training data; as well as An optimal sampling strategy is selected by the processor (202) for each combination of drift context and process context (224) in the training data based on a performance metric of the second AI model (302).

4. The method according to claim 3, further comprising: When the first AI model (216) fails to identify an optimal sampling strategy from the plurality of predefined sampling strategies (220), at least one new sampling strategy is generated by the processor (202).

5. The method according to claim 1, further comprising: comparing, by the processor (202), a performance metric of the second AI model (302) with a predefined performance metric; as well as When the performance metric of the second AI model (302) is less than the predefined performance metric, at least one new sampling strategy is generated by the processor (202).

6. The method according to claim 3, further comprising: generating, by the processor (202), a reward value for the first AI model (216) based on the performance metric, wherein When the performance metric of the second AI model (302) is greater than a predefined performance metric, the reward value is a positive value; and When the performance metric of the second AI model (302) is less than the predefined performance metric, the reward value is a negative value.

7. The method according to claim 1, wherein the plurality of sampling strategies (220) are: a random sampling strategy, a stratified sampling strategy, a systematic sampling strategy, a biased representative sampling strategy, and an event-driven sampling strategy.

8. The method of claim 1, wherein the one or more drift contexts (222) include information related to a drift type.

9. The method of claim 1, wherein the one or more process contexts (224) are based on a type of industrial plant (102) and process conditions associated with the industrial plant (102).

10. A system for mitigating data drift in an industrial plant (102), comprising: Processor (202); as well as a memory, wherein the memory stores processor-executable instructions that, when executed, cause the processor (202): monitoring one or more process parameters associated with an industrial plant (102); detecting a drift in the one or more process parameters based on a deviation from one or more predefined process parameters; determining one or more drift contexts (222) and one or more process contexts (224) based on the drift and the one or more process parameters; determining a sampling strategy from a plurality of sampling strategies (220) based on the one or more drift contexts (222) and the one or more process contexts (224) for sampling the one or more process parameters using a first AI model (216); as well as A second AI model (302) is trained based on the sampling strategy for mitigating data drift, wherein the second AI model (302) is configured to control one or more processes in the industrial plant (102).

11. The system of claim 10, wherein the processor (202) is configured to train the first AI model (216) using one of a reinforcement learning (RL) process, a rule-based system, and a genetic algorithm.

12. The system of claim 11, wherein the processor (202) is configured to train the first AI model (216) and the second AI model (302), wherein the processor is configured to: Providing training data including a plurality of drift contexts (222) and a plurality of process contexts (224) to the first AI model (216); configuring the first AI model (216) to train the second AI model (302) using at least one sampling strategy from a plurality of predefined sampling strategies (220) on the training data; and An optimal sampling strategy is selected for each combination of drift context and process context (224) in the training data based on a performance metric of the second AI model (302).

13. The system of claim 12, wherein the processor (202) is further configured to: When the first AI model (216) fails to identify an optimal sampling strategy from the plurality of predefined sampling strategies (220), at least one new sampling strategy is generated.

14. The system of claim 10, wherein the processor (202) is further configured to: comparing a performance metric of the second AI model (302) to a predefined performance metric; and When the performance metric of the second AI model (302) is less than the predefined performance metric, at least one new sampling strategy is generated.

15. The system of claim 12, wherein the processor (202) is further configured to: generating a reward value for the first AI model (216) based on the performance metric, wherein When the performance metric of the second AI model (302) is greater than a predefined performance metric, the reward value is a positive value; and When the performance metric of the second AI model (302) is less than the predefined performance metric, the reward value is a negative value.

16. The system of claim 10, wherein the processor (202) is configured to determine the sampling strategy from the plurality of sampling strategies (220) selected from a group of strategies, the group of strategies comprising a random sampling strategy, a stratified sampling strategy, a systematic sampling strategy, a biased representative sampling strategy, and an event-driven sampling strategy.

17. The system of claim 10, wherein the processor (202) is configured to detect the one or more drift contexts (222), wherein the one or more drift contexts (222) include information related to a drift type.

18. The system of claim 10, wherein the processor (202) is configured to detect the one or more process contexts (224), wherein the one or more process contexts (224) are based on a type of industrial plant (102) and process conditions associated with the industrial plant (102).

Images