Method and system for mitigating data drift in industrial systems

Abstract

The present disclosure relates to a method and a system for mitigating data drift in process plants. The method comprises receiving metadata of a first process in a first process plant, when a data drift is detected in the first process due to a change in an input scenario in the first process plant. The data drift affects first model parameters associated with a first Artificial Intelligence (AI) model deployed in the first process. Further, the method comprises identifying a second process plant configured with a second AI model trained on the input scenario, based on the metadata. Furthermore, the method comprises receiving second model parameters of the second AI model from the second process plant, upon the identifying. Thereafter, the method comprises updating the first model parameters of the first AI model based on the second model parameters, for mitigating the data drift in the first process plant.

Description

TITLE: “METHOD AND SYSTEM FOR MITIGATING DATA DRIFT IN INDUSTRIAL SYSTEMS”TECHNICAL FIELD

[0001] The present disclosure generally relates to data analytics in industrial systems. More particularly, the present disclosure relates to a method and a system for mitigating data drift in industrial systems.BACKGROUND

[0002] Soft sensors are widely used in industrial systems for estimating key variables, product quality, and the like. Soft sensors are virtual sensors that alleviate the need for more expensive hardware sensors by accurate predictions and are relied upon heavily in the industrial systems. Soft sensors, with predictive models, provide scenarios in which estimations can drive decisionmaking and improve the reliability of monitoring systems. Soft sensors are used to deal with the problems such as small number of samples, nonlinearity in process data, and so on.

[0003] Currently in process industries, predictive models / Artificial Intelligence (Al) models based soft sensors are gaining attraction due to the non-physical nature and low-operational cost. The soft sensors have the ability to perform real-time analyzing, monitoring, and control of industrial processes where no hardware sensor is available. A soft sensor is developed by training a predictive model with historical data. The trained predictive model is deployed at a process plant in the industrial system to analyze the process data and monitor the process plant.

[0004] Operating conditions in a process plant change over time due to variations in raw material, changes in operating procedures, pattern changes in data, equipment wear-and-tear, and the like. Due to changes in the operating conditions, there is a drift in data provided to the predictive model based soft sensor. The performance of the predictive model based soft sensor may degrade as the models are trained using the one or more process parameters before the drift. This can lead to undesired outcomes such as poor product quality, poor yields, equipment failures, plant shutdowns, and the like.

[0005] The information disclosed in this background of the disclosure section is only for enhancement of understanding of the general background of the invention and should not betaken as an acknowledgement or any form of suggestion that this information forms the prior art already known to a person skilled in the art.SUMMARY

[0006] In an embodiment, the present disclosure discloses a method for mitigating data drift in process plants. The method comprises receiving metadata of a first process in a first process plant, when a data drift is detected in the first process. The data drift is detected due to a change in an input scenario in the first process plant. The data drift affects first model parameters associated with a first Artificial Intelligence (Al) model deployed in the first process. Further, the method comprises identifying a second process plant configured with a second Al model trained on the input scenario, based on the metadata. Furthermore, the method comprises receiving second model parameters of the second Al model from the second process plant, upon the identifying. Thereafter, the method comprises updating the first model parameters of the first Al model based on the second model parameters, for mitigating the data drift in the first process plant.

[0007] In an embodiment, the present disclosure discloses a system for mitigating data drift in process plants. The system comprises a processor and a memory. The processor is configured to receive metadata of a first process in a first process plant, when a data drift is detected in the first process. The data drift is detected due to a change in an input scenario in the first process plant. The data drift affects first model parameters associated with a first Artificial Intelligence (Al) model deployed in the first process. Further, the processor is configured to identify a second process plant configured with a second Al model trained on the input scenario, based on the metadata. Furthermore, the processor is configured to receive second model parameters of the second Al model from the second process plant, upon the identifying. Thereafter, the processor is configured to update the first model parameters of the first Al model based on the second model parameters, for mitigating the data drift in the first process plant.

[0008] The foregoing summary is illustrative only and is not intended to be in any way limiting. 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 ACCOMPANYING DRAWINGS

[0009] The novel features and characteristics of the disclosure are set forth in the appended claims. The disclosure itself, however, as well as a preferred mode of use, further objectives, and advantages thereof, will best be understood by reference to the following detailed description of an illustrative embodiment when read in conjunction with the accompanying figures. One or more embodiments are now described, by way of example only, with reference to the accompanying figures wherein like reference numerals represent like elements and in which:

[0010] Figure 1 illustrates an exemplary environment for mitigating data drift in process plants, in accordance with some embodiments of the present disclosure;

[0011] Figure 2 illustrates a detailed diagram of a system for mitigating data drift in process plants, in accordance with some embodiments of the present disclosure;

[0012] Figures 3, 4A and 4B show exemplary illustrations for mitigating data drift in process plants, in accordance with some embodiments of the present disclosure;

[0013] Figure 5 shows an exemplary flow chart illustrating method steps for mitigating data drift in process plants, in accordance with some embodiments of the present disclosure; and

[0014] Figure 6 shows a block diagram of a general-purpose computing system for mitigating data drift in process plants, in accordance with embodiments of the present disclosure.

[0015] It should be appreciated by those skilled in the art that any block diagram herein represents conceptual views of illustrative systems embodying the principles of the present subject matter. Similarly, it will be appreciated that any flow charts, flow diagrams, state transition diagrams, pseudo code, and the like represent various processes which may be substantially represented in computer readable medium and executed by a computer or processor, whether or not such computer or processor is explicitly shown.DETAILED DESCRIPTION

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

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

[0018] The terms “comprises”, “comprising”, or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a setup, device or method that comprises a list of components or steps does not include only those components or steps but may include other components or steps not expressly listed or inherent to such setup or device or method. In other words, one or more elements in a system or apparatus proceeded by “comprises. . . a” does not, without more constraints, preclude the existence of other elements or additional elements in the system or apparatus.

[0019] Predictive model based soft sensors are virtual sensors that alleviate the need for more expensive hardware sensors by accurate predictions and are relied upon heavily in the industrial systems. Operating conditions in a process plant change over time due to variations in raw material, changes in operating procedures, pattern changes in data, equipment wear-and-tear, and the like. Due to changes in the operating conditions, the deployed predictive model based soft sensor starts to produce wrong predictions leading to break in automations planned based on the predictions at the process plant, which in turn might lead to plant shutdown. Conventional ways to mitigate the drift and reduce the predictive model degradation is the manual inspection of the issue and redeployment after retraining the predictive model with new data. Another conventional way is to train the predictive model with all real time scenarios and sufficient data. However, the said approach demands to collect the data from various plants which leads to data privacy issues. This whole cycle of wrong predictions, retraining and redeploying the model incurs huge engineering effort and operational cost.

[0020] The present disclosure provides a method and a system for mitigating data drifts in process plants. In the present disclosure, metadata of a process in a process plant is collected when a data drift is detected in the process due to a change in scenario in the process plant. An Al model deployed in the first process and its parameters are affected due to the data drift. The metadata is used to identify another plant with an Al model trained on similar scenario. As the Al model of another process plant is trained on the similar scenario, parameters of the Al modelof another process plant are used to update the parameters of the Al model of the process plant. The Al model is trained using the updated parameters. In this way, the Al model affected by the data drift is trained using the parameters of another Al model, and thereby the data drift is mitigated autonomously.

[0021] The present disclosure enables collaborative learning between Al models of different plants by mutual knowledge transfer between the plants via the system, to mitigate the data drift. In the present disclosure, the system collects only the metadata of the process affected by the data drift and identifies another plant with similar scenario to mitigate the data drift. In this way, sensitive data of the plants are not collected, thus preserving data privacy.

[0022] Figure 1 illustrates an exemplary environment 100 for mitigating data drift in process plants, in accordance with embodiments of the present disclosure. The environment 100 comprises a system 102, a process plant 1, and a process plant 2. Figure 1 illustrates only two process plants i.e., plant 1 and plant 2. However, the environment 100 comprises a plurality of process plants, and thus the illustration should be considered as limiting. The present description is explained with respect to the process plant 1 and process plant 2 for simplicity purposes. The process plant 1 and the process plant 2 are also referred as a first process plant and a second process plant, respectively. Some examples of a process plant may include, but are not limited to, a nuclear plant, an oil refinery, a chemical plant, a fertilizer plant, a manufacturing plant, a power plant, and the like.

[0023] The first process plant and the second process plant comprise a first soft sensor 104i, and a second soft sensor 1042, respectively. The first Al model 106i and the second Al model IO62 are deployed in the first soft sensor 104i and the second soft sensor 1042, respectively. Soft sensors are predictive / inferential models that use easily measured variables to estimate process variables that are hard to measure due to technological limitations, large measurement delays, operational costs, and the like. Soft sensors convert various inputs from simple hardware sensors and combines them to mimic the output of a more complex hardware sensor. Soft sensors provide real-time estimates of process variables to improve process control and performance. Predictive learning based soft sensors are developed by training predictive / Artificial Intelligence (Al) models using data historians. The term “Al model(s)” used herein refers to a model that has been trained on a set of data to recognize certain patterns and / or make certain decisions without further human intervention. Al models apply different algorithms to relevantdata inputs to achieve the tasks and / or output they have been trained for. The soft sensors may be developed using Al techniques such as Principal Component Analysis (PC A), PLS, deep learning, Support Vector Machine / Regression (SVM / SVR), Fuzzy Inference System (FIS), Neuro Fuzzy System (NFS), and the like.

[0024] A data drift in a process plant occurs due to deviation in operating conditions over time due to variations in raw material, changes in operating procedures, pattern changes in data, equipment wear-and-tear, and the like. Due to changes in the operating conditions, there is a drift in data provided to the Al model based soft sensor. The present description is explained considering such data drift occurs in the first process plant. In an embodiment, the data drift may include a concept drift, a covariate shift, a prior probability shift, and the like. The concept drift herein refers to a deviation in an underlying goal i.e., deviation in the underlying relationship between input and output target parameters of the model, a set of targets, and / or objectives for the Al model. The covariate shift occurs when there is a change in the input data long with the shift in the relationship between the parameters. The prior probability shift occurs when the proportion of the different classes in the data changes over time. A person skilled in the art will appreciate that the data drift can be any kind of drift in the data provided to the Al model, and the above-stated types of data drifts should not be considered as limiting.

[0025] In the present disclosure, the system 102 is configured to mitigate the data drift in the process plants. Herein, the system 102 is configured to receive metadata of a first process in the first process plant, when a data drift is detected in the first process. The data drift is detected when an input scenario changes in the first process plant. As the first Al model 106i is not trained for the changed input scenario, the data drift affects first model parameters associated with the first Al model 106i . Then, the system 102 identifies a second process plant in which a second Al model IO62 is deployed and trained on the similar input scenario, based on the metadata. Herein, the system 102 identifies the second process plant based on operating conditions of the first process plant in the metadata. The system 102 receives second model parameters of the second Al model 1062from the second process plant. The system 102 updates the first model parameters of the first Al model 1061 based on the second model parameters, for mitigating the data drift in the first process plant. Hence, the present disclosure enables mutual knowledge transfer between the process plants to mitigate the data drift. In some embodiments, the system 102 may be deployed in the premises of the process plant or may be deployed remotely and connected to the process plant. For example, the system 102 may be deployed ina cloud server and configured to communicate with the process plants over a communication network.

[0026] Figure 2 illustrates a detailed diagram 200 of the system 102 for mitigating data drifts in process plants, in accordance with some embodiments of the present disclosure. The system 102 may include Central Processing Units 206 (also referred as “CPUs” or “one or more processors 206”), Input / Output (I / O) interface 202, and a memory 204. In some embodiments, the memory 204 may be communicatively coupled to the one or more processors 206. The memory 204 stores instructions executable by the one or more processors 206. The one or more processors 206 may comprise at least one data processor for executing program components for executing user or system-generated requests. The memory 204 may be communicatively coupled to the one or more processors 206. The memory 204 stores instructions, executable by the one or more processors 206, which, on execution, may cause the one or more processors 206 to mitigate data drifts in process plants.

[0027] In an embodiment, the memory 204 may include one or more modules 210 and computation data 208. The one or more modules 210 may be configured to perform the steps of the present disclosure using the computation data 208, to mitigate data drifts in process plants. In an embodiment, each of the one or more modules 210 may be a hardware unit which may be outside the memory 204 and coupled with the system 106. As used herein, the term modules 210 refers to an Application Specific Integrated Circuit (ASIC), an electronic circuit, a Field- Programmable Gate Arrays (FPGA), Programmable System-on-Chip (PSoC), a combinational logic circuit, and / or other suitable components that provide described functionality. The one or more modules 210 when configured with the described functionality defined in the present disclosure will result in a novel hardware. Further, the I / O interface 202 is coupled with the one or more processors 206 through which an input signal or / and an output signal is communicated. For example, the system 102 may receive the metadata of the first process from the first process plant using the I / O interface 202. Also, the system 102 may receive the second model parameters from the second process plant using the I / O interface 202.

[0028] In one implementation, the modules 210 may include, for example, an input module 220, a plant identification module 222, a parameter update module 224, and auxiliary modules 226. It will be appreciated that such aforementioned modules 210 may be represented as a singlemodule or a combination of different modules. In one implementation, the computation data 208 may include, for example, input data, identification data, parameter data, and auxiliary data 218.

[0029] In an embodiment, the input module 220 may be configured to receive metadata of a first process in the first process plant. The input module 220 may receive the metadata when a data drift is detected in the first process. The data drift may occur due to a change in an input scenario in the first process plant. The change in the input scenario may be due to a deviation in operating conditions over time due to variations in raw material, changes in operating procedures, pattern changes in data, equipment wear-and-tear, and the like. In such a case, there may be a new scenario in the first process plant. In an example, consider the first soft sensor 104i deployed in the first process plant is a Hotmeal Sulphur soft process. The first process plant may be a cement manufacturing plant. In the cement manufacturing plant, there may be a drift in sulfur content in raw materials used in the first process. For example, sulfur content used in the first process during training of the Al model 106i in the first process may be 70%. Whereas the current sulfur content may be 50%. Thus, there is a 20% drift in sulfur content in the raw materials used thereby leading to a drift in chemical reactions in the first process. Due to this drift in the sulfur content, there may be a data drift in the data provided to the first Al model 106i, thus causing degradation in performance of the first Al model 106i . In another example, the data drift may occur due to a change in setpoints defined in a process.

[0030] In an embodiment, the data drift is detected in the first process using one or more data drift detection methods such as sequential analysis methods, machine learning model-based approach, time distribution methods such as Population Stability Index, Kolmogorov-Smirnov (K-S) test, and the like. The data drift is detected by receiving first model parameters associated with the first Al model 106i from the first soft sensor 104i. In an example, the first model parameters may be received periodically to detect the data drift. The data drift may be detected by a computing device 302i in the first process plant, as shown in Figure 3. The first soft sensor 104i in the first process plant comprises components such as data collector, data processor, the first Al model 106i, and the like. In an embodiment, the computing device 302i may be a part of the first soft sensor 104i and may be an internal component of the first soft sensor 104i along with the above-mentioned components. In another embodiment, the computing device 302i may be within the first process plant, but external to the first soft sensor 104i .

[0031] In an embodiment, the computing device 302i may comprise a drift detection module 304i and a communication module 3061. The drift detection module 304i may receive the firstmodel parameters from the first soft sensor 104i . Further, the drift detection module 304i may detect the data drift using the one or more data drift detection methods to verify performance of the first Al model 106i . The drift detection module 304i may notify the communication module 3061 when the data drift is detected. In an example, the drift detection module 304i may raise a flag when the data drift is detected. The communication module 3061 periodically checks flags from drift detection module 304i. The input module 220 receives model degradation information from the communication module 3061, when the data drift is detected. Similarly, a computing device 3022 comprsing a drift detection module 3042 and a communiction module 3062 may be deployed in the second process plant. These components are not explained again for the sake of brevity.

[0032] The first model parameters of the first Al model 106i are internal variables that are learned or estimated from data during training and are continuously optimized. These parameters are initialized to certain values. As training progresses, the values are updated continuously. In an example, consider the first Al model 1061 is a linear regression model, where y = a x + b. The variables ‘a’ and ‘b’ are the first model parameters. In another example, consider the first Al model 106i is a neural network model. The first model parameters include weights and biases. In yet another example, consider the first Al model 106i is a clustering model. The first model parameters include centroids of clusters.

[0033] Referring back to Figure 2, the input module 220 receives the metadata associated with the first process, when the data drift is detected in the first process. The metadata corresponds to data of the first process that is affected by the data drift. Herein, the input module 220 receives information related to the process that is affected by the data drift only and does not receive other process data of the first process plant. Referring to the above-stated example, the metadata associated with the first process may include information about composition of raw materials used in the first process. The metadata of the first process may be stored as the input data 212 in the memory 204.

[0034] In an embodiment, the plant identification module 222 may be configured to receive the input data 212 from the input module 220. Further, the plant identification module 222 may be configured to identify a second process plant configured with a second Al model IO62 trained on the input scenario, based on the metadata. Herein, the plant identification module 222 identifies operating conditions associated with the input scenario in the first process plant, based on the metadata. Referring to the above- stated example, the plant identification module 222identifies change in composition of the raw materials used in the first process. In another example, consider the data drift is due to a change in the raw material used due to unavailability. In such a case, the plant identification module 222 identifies a list of raw materials used in the first process along with composition. In yet another example, consider the data drift is due to a change in setpoint values of variables in a first process. In such case, the plant identification module 222 identifies operating parameters of the variables of the first process.

[0035] Then, the plant identification module 222 may be configured to compare the operating conditions associated with the input scenario in the first process plant with pre-stored data. In an embodiment, the prestored data may be a shared database associated with the system 102. The shared database may comprise operating conditions associated with each of a plurality of scenarios in a plurality of process plants. For instance, the shared database contains scenarios from different operating states of processing units from various similar plants. In an embodiment, the plant identification module 222 entails a loosely bounded search over operating conditions (for instance, operating ranges), as the operating conditions / states of processes may differ even in similar process plants due to errors in various process parameters. For example, the plant identification module 222 entails a loosely bounded search over a variable range of between 79.5-delta and 116+delta, instead of 79.5 and 116. The loosely bounded search over operating conditions / ranges provides similar scenarios from various plants operating in the matching operating ranges. For example, the plant identification module 222 may compare the operating conditions associated with the input scenario in the first process plant with that of other process plants from the plurality of plants. The plant identification module 222 may identify plant 2, plant 4, and plant 5 associated with Al models trained on similar scenarios.

[0036] In an embodiment, the plant identification module 222 performs probability distribution of process data of the plurality of plants. For instance, plant identification module 222 samples multivariate process variables data from the input scenario (i.e., the scenario data from the new scenario that is running in the first process plant) and matches to probability distributions of the scenarios of the other plants (plant 2, plant 4, and plant 5). In an embodiment, the plant identification module 222 may use a multivariate distribution matching technique such as Maximum Mean Discrepancy (MMD), a univariate statistical test such as Kolmogorov-Smirnov test (KS test) to sample and match the scenarios. A person skilled in the art will appreciate that other techniques may be employed to sample and match the scenarios. The plant identification module 222 identifies the second process plant configured with the second Al model IO62 trainedon the input scenario based on the comparison. The plant identification module 222 may identify multiple process plants configured with Al models trained on similar scenarios. The identification of one process plant is explained for simplicity purposes. Referring to the abovestated example, the plant identification module 222 may identify that the second process plant is associated with the second Al model IO62 and is trained on similar scenario i.e., similar composition of raw materials. Data related to the identification of the second process plant with the second Al model IO62 is stored as the identification data 214 in the memory 204.

[0037] In an embodiment, the input module 220 may be configured to receive the identification data 214 from the plant identification module 222. Further, the input module 220 may be configured to receive second model parameters of the second Al model IO62 from the second process plant, upon the identifying. The second model parameters may comprise weights, biases, and the like. In an example, consider that the first Al model IO61 is a neural network model. The second Al model IO62 may also be a neural network model. The second model parameters may include weights and biases. In another example, the first Al model 1061 and the second Al model IO62 are clustering models. The second model parameters may include centroids of clusters. In an embodiment, the input module 220 may be configured to receive the first model parameters of the first Al model 1061 from the first process plant. The first model parameters and the second model parameters may be stored as the parameter data 216 in the memory 204.

[0038] In an embodiment, the parameter update module 224 may be configured to receive the parameter data 216 from the input module 220. Further, the parameter update module 224 may be configured to update the first model parameters of the first Al model 1061 based on the second model parameters. Herein, the parameter update module 224 may be configured to update the first model parameters based on the second model parameters using an aggregation strategy. Model aggregation or fusion using averaging the first model parameters and the second model parameters may lead to model performance divergence. Instead of simply fusing the first model parameters and the second model parameters, the present disclosure proposes mutual knowledge transfer between the process plants. Herein, the parameter update module 224 computes soft targets or outputs of the first Al model IO61 using the first model parameters and the second model parameters. In an embodiment, the parameter update module 224 computes a gradient update by minimizing the loss function that contains distance between two soft targets. A person skilled in the art will appreciate that any other methods may be used to update the first model parameters using the second model parameters.

[0039] In an embodiment, the parameter update module 224 identifies a plurality of aggregation strategies to update the first model parameters with the second model parameters. Then, the parameter update module 224 updates the first model parameters using the aggregation strategy of the plurality of aggregation strategies selected based on characteristics of the first model parameters and the second model parameters. The selected aggregation strategy impacts convergence rate and performance of the first Al model 106i . Hence, using a fixed aggregation strategy for all the cases may also lead to degradation in model performance. The present disclosure adapts the aggregation strategy from the plurality of aggregation strategies based on characteristics such as momentum of the first model parameters and the second model parameters for faster convergence. For instance, the parameter update module 224 tracks momentum value periodically and switches to the aggregation strategy which has highest change in momentum value for previous period. In an embodiment, the plurality of aggregation strategies may include federated learning strategies such as FedAvg, FedYogi, FedAdam, and the like. A person skilled in the art will appreciate that any other aggregation strategies may be used, and the above-mentioned strategies should not be considered as limiting.

[0040] The parameter update module 224 may update the first model parameters of the first Al model 106i using the second model parameters of the second Al model IO62. In an example, weights of the first Al model IO61 may be updated based on weights of the second Al model IO62. The first Al model IO61 may be retrained using the updated first model parameters to mitigate the data drift. Referring to the above- stated example, the data drift in the first Al model IO61 occurred due to change in composition of the raw material may be mitigated upon retraining. As stated above, the plant identification module 222 may identify multiple process plants configured with Al models trained on similar scenarios. In such case, the parameter update module 224 may receive model parameters from identified process plants. The parameter update module 224 may use a suitable aggregation strategy to obtain a consolidated model from all similar scenarios. The parameter update module 224 may update the first model parameters using the said aggregation strategy and transmit the updated model parameters to the first process plant. In an example, consider that the plant identification module 222 may identify a second process plant and a third process plant configured with Al models trained on similar scenarios. In such a case, the parameter update module 224 may receive model parameters from the second process plant and the third process plant. The parameter update module 224 may use a suitable aggregation strategy to obtain a consolidated model from all similar scenarios of the secondprocess plant and the third process plant. The parameter update module 224 may update the first model parameters using the said aggregation strategy and transmit the updated model parameters to the first process plant. The examples of identification of process plants with similar scenarios herein are provided by considering one process plant and two process plants. However, the present disclosure is applicable similarly to identification of more than two process plants with similar scenarios.

[0041] The present disclosure can be used to mitigate data drift occurred due to various reasons. For instance, the present disclosure can be used for model adaptation based on requirements such as performance improvement, version update, adding new features based on feedback from a user, and the like. The present disclosure can be used for training / retraining the Al model in such cases.

[0042] The auxiliary data 218 may store data, including temporary data and temporary files, generated by the one or more modules 210 for performing the various functions of the system 102. The one or more modules 210 may also include the auxiliary modules 226 to perform various miscellaneous functionalities of the system 102. The auxiliary data 218 may be stored in the memory 204. It will be appreciated that the one or more modules 210 may be represented as a single module or a combination of different modules.

[0043] In an embodiment, the system 102 may be implemented in a server 402 (for instance, cloud server), as shown in Figures 4A and 4B. The computing devices 302i and 3022 in the process plants may be implemented as client devices 4041 and 4042. In an embodiment, the server 402 may receive generalized model parameters of all the plurality of plants and store in the shared database along with the scenarios (as illustrated in 400a in Figure 4A). The server 402 may receive specific model parameters from a plant (the second process plant) and communicate the updated model parameters to the first process plant when the data drift occurs (as illustrated in 400b in Figure 4B).

[0044] Figure 5 shows an exemplary flow chart illustrating method steps for mitigating data drifts in process plants, in accordance with some embodiments of the present disclosure. As illustrated in Figure 5, the method 500 may comprise one or more steps. The method 500 may be described in the general context of computer executable instructions. Generally, computer executable instructions can include routines, programs, objects, components, data structures,procedures, modules, and functions, which perform particular functions or implement particular abstract data types.

[0045] The order in which the method 500 is described is not intended to be construed as a limitation, and any number of the described method blocks can be combined in any order to implement the method. Additionally, individual blocks may be deleted from the methods without departing from the scope of the subject matter described herein. Furthermore, the method can be implemented in any suitable hardware, software, firmware, or combination thereof.

[0046] At step 502, the metadata of the first process in the first process plant is received. The metadata is received when a data drift is detected in the first process. The metadata corresponds to data of the first process that is affected by the data drift. The data drift may occur due to a change in an input scenario in the first process plant. In an embodiment, the data drift is detected by receiving first model parameters associated with the first Al model 106i from the first soft sensor 104i . The data drift may affect the first model parameters of the first Al model 106i.The first model parameters of the first Al model 106i are internal variables that are learned or estimated from data during training and are continuously optimized.

[0047] At step 504, a second process plant configured with a second Al model IO62 trained on the input scenario is identified, based on the metadata. Herein, the system 102 identifies operating conditions associated with the input scenario in the first process plant, based on the metadata. Then, the system 102 may be configured to compare the operating conditions associated with the input scenario in the first process plant with pre-stored data. In an embodiment, the prestored data may be a shared database associated with the system 102. The shared database may comprise operating conditions associated with each of a plurality of scenarios in a plurality of process plants. The system 102 identifies the second process plant configured with the second Al model IO62 trained on the input scenario based on the comparison.

[0048] At step 506, the second model parameters of the second Al model IO62 may be received from the second process plant, upon the identifying. The second model parameters may comprise weights, biases, and the like. In an embodiment, the system 102 may be configured to receive the first model parameters of the first Al model IO61 from the first process plant.

[0049] At step 508, the first model parameters of the first Al model IO61 may be updated based on the second model parameters. Herein, the system 102 may be configured to update the firstmodel parameters based on the second model parameters using an aggregation strategy. In an embodiment, the system 102 identifies a plurality of aggregation strategies to update the first model parameters with the second model parameters. Then, the system 102 updates the first model parameters using the aggregation strategy of the plurality of aggregation strategies selected based on characteristics of the first model parameters and the second model parameters.COMPUTER SYSTEM

[0050] Figure 6 is a diagram of an exemplary computer system 600 for implementing embodiments consistent with the present disclosure. In an embodiment, the computer system 600 may be used to realize the system 102 of Figure 2. Hence, the computer system 600 may be used to mitigate data drifts in process plants, in accordance with embodiments of the present disclosure. The computer system 600 may comprise a Central Processing Unit 604 (also referred as “CPU” or “processor”). The processor 604 may comprise at least one data processor. The processor 604 may include specialized processing units such as integrated system (bus) controllers, memory management control units, floating point units, graphics processing units, digital signal processing units, etc. The processor 604 may be used to realize the processor 206 described in Figure 2.

[0051] The processor 604 may be disposed in communication with one or more input / output (I / O) devices (not shown) via I / O interface 602. The I / O interface 602 may employ communication protocols / methods such as, without limitation, audio, analog, digital, monoaural, RCA, stereo, IEEE (Institute of Electrical and Electronics Engineers) -1394, serial bus, universal serial bus (USB), infrared, PS / 2, BNC, coaxial, component, composite, digital visual interface (DVI), high-definition multimedia interface (HDMI), Radio Frequency (RF) antennas, 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), longterm evolution (LTE), WiMax, or the like), etc.

[0052] Using the VO interface 602, the computer system 600 may communicate with one or more VO devices. For example, the input device 620 may be an antenna, keyboard, mouse, joystick, (infrared) remote control, camera, card reader, fax machine, dongle, biometric reader, microphone, touch screen, touchpad, trackball, stylus, scanner, storage device, transceiver, video device / source, etc. The output device 622 may be a printer, fax machine, video display (e.g., cathode ray tube (CRT), liquid crystal display (LCD), light-emitting diode (LED), plasma,Plasma display panel (PDP), Organic light-emitting diode display (OLED) or the like), audio speaker, etc.

[0053] The processor 604 may be disposed in communication with the communication network 618 via a network interface 606. The network interface 606 may communicate with the communication network 618. The network interface 606 may employ connection protocols including, without limitation, direct connect, Ethernet (e.g., twisted pair 10 / 100 / 1000 Base T), transmission control protocol / internet protocol (TCP / IP), token ring, IEEE 802.11a / b / g / n / x, etc. The communication network 618 may include, without limitation, a direct interconnection, local area network (LAN), wide area network (WAN), wireless network (e.g., using Wireless Application Protocol), the Internet, etc. The network interface 606 may employ connection protocols include, but not limited to, direct connect, Ethernet (e.g., twisted pair 10 / 100 / 1000 Base T), transmission control protocol / internet protocol (TCP / IP), token ring, IEEE 802.11a / b / g / n / x, etc.

[0054] The communication network 618 includes, but is not limited to, a direct interconnection, an e-commerce network, a peer to peer (P2P) network, local area network (LAN), wide area network (WAN), wireless network (e.g., using Wireless Application Protocol), the Internet, WiFi, and such. The first network and the second network may either be a dedicated network or a shared network, which represents an association of the different types of networks that use a variety of protocols, for example, Hypertext Transfer Protocol (HTTP), Transmission Control Protocol / internet Protocol (TCP / IP), Wireless Application Protocol (WAP), etc., to communicate with each other. Further, the first network and the second network may include a variety of network devices, including routers, bridges, servers, computing devices, storage devices, etc. The computer system 600 may communicate with the plurality of plants in the industrial system via the communication network 618.

[0055] In some embodiments, the processor 604 may be disposed in communication with a memory 610 (e.g., RAM, ROM, etc. not shown in Figure 5) via a storage interface 608. The storage interface 608 may connect to memory 610 including, without limitation, memory drives, removable disc drives, etc., employing connection protocols such as serial advanced technology attachment (SATA), Integrated Drive Electronics (IDE), IEEE- 1394, Universal Serial Bus (USB), fiber channel, Small Computer Systems Interface (SCSI), etc. The memory drives may further include a drum, magnetic disc drive, magneto-optical drive, optical drive, Redundant Array of Independent Discs (RAID), solid-state memory devices, solid-state drives, etc.

[0056] The memory 610 may store a collection of program or database components, including, without limitation, user interface 612, an operating system 614, web browser 616 etc. In some embodiments, computer system 600 may store user / application data, such as, the data, variables, records, etc., as described in this disclosure. Such databases may be implemented as fault- tolerant, relational, scalable, secure databases such as Oracle ® or Sybase®. The memory 610 may be used to realize the memory 204 described in Figure 2. The memory 610 is communicatively coupled to the processor 604. The memory 610 stores instructions, executable by the one or more processors 604, which, on execution, may cause the processor 604 to mitigate data drifts in process plants, in accordance with embodiments of the present disclosure.

[0057] The operating system 614 may facilitate resource management and operation of the computer system 600. Examples of operating systems include, without limitation, APPLE MACINTOSH1* OS X, UNIXR, UNIX-like system distributions (E.G., BERKELEY SOFTWARE DISTRIBUTION™ (BSD), FREEBSD™, NETBSD™, OPENBSD™, etc.), LINUX DISTRIBUTIONS™ (E.G., RED HAT™, UBUNTU™, KUBUNTU™, etc.), IBM™ OS / 2, MICROSOFT™ WINDOWS™ (XP™, VISTA™ / 7 / 8, 10 etc.), APPLERIOS™, GOOGLERANDROID™, BLACKBERRY1* OS, or the like.

[0058] In some embodiments, the computer system 600 may implement the web browser 616 stored program component. The web browser 616 may be a hypertext viewing application, for example MICROSOFT1* INTERNET EXPLORER™, GOOGLE1* CHROME™0, MOZILLA1* FIREFOX™, APPLE1* SAFARI™, etc. Secure web browsing may be provided using Secure Hypertext Transport Protocol (HTTPS), Secure Sockets Layer (SSL), Transport Layer Security (TLS), etc. Web browsers 616 may utilize facilities such as AJAX™, DHTML™, ADOBE1* FLASH™, JAVASCRIPT™, JAVA™, Application Programming Interfaces (APIs), etc. In some embodiments, the computer system 600 may implement a mail server (not shown in Figure) stored program component. The mail server may be an Internet mail server such as Microsoft Exchange, or the like. The mail server may utilize facilities such as ASP™, ACTIVEX™, ANSI™ C++ / C#, MICROSOFT1*, .NET™, CGI SCRIPTS™, JAVA™, JAVASCRIPT™, PERL™, PHP™, PYTHON™, WEBOBJECTS™, etc. The mail server may utilize communication protocols such as Internet Message Access Protocol (IMAP), Messaging Application Programming Interface (MAPI), MICROSOFT1* exchange, Post Office Protocol (POP), Simple Mail Transfer Protocol (SMTP), or the like. In some embodiments, the computer system 600 may implement a mail client stored program component. The mail client (not shownin Figure) may be a mail viewing application, such as APPLERMAIL™, MICROS OFTRENTOURAGE™, MICROSOFT1* OUTLOOK™, MOZILLARTHUNDERBIRD™, etc.

[0059] Furthermore, one or more computer-readable storage media may be utilized in implementing embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory on which information or data readable by a processor may be stored. Thus, a computer-readable storage medium may store instructions for execution by one or more processors, including instructions for causing the processor(s) to perform steps or stages consistent with the embodiments described herein. The term “computer- readable medium” should be understood to include tangible items and exclude carrier waves and transient signals, i.e., be non-transitory. Examples include Random Access Memory (RAM), Read-Only Memory (ROM), volatile memory, non-volatile memory, hard drives, Compact Disc Read-Only Memory (CD ROMs), Digital Video Disc (DVDs), flash drives, disks, and any other known physical storage media.

[0060] The present disclosure provides a method and a system for mitigating data drifts in process plants. In the present disclosure, metadata of a process in a process plant is collected when a data drift is detected in the process due to a change in scenario in the process plant. An Al model deployed in the first process and its parameters are affected due to the data drift. The metadata is used to identify another plant with an Al model trained on a similar scenario. As the Al model of another process plant is trained on the similar scenario, parameters of the Al model of another process plant to update the parameters of the Al model of the process plant. The Al model is trained using the updated parameters. In this way, the Al model affected by the data drift is trained using the parameters of another Al model, and thereby the data drift is mitigated autonomously.

[0061] The present disclosure enables collaborative learning between Al models of different plants by mutual knowledge transfer between the plants via the system, to mitigate the data drift. In the present disclosure, the system collects only the metadata of the process affected by the data drift and identifies another plant with similar scenario to mitigate the data drift. In this way, sensitive data of the plants are not collected, thus preserving data privacy.

[0062] The terms "an embodiment", "embodiment", "embodiments", "the embodiment", "the embodiments", "one or more embodiments", "some embodiments", and "one embodiment" mean "one or more (but not all) embodiments of the invention(s)" unless expressly specified otherwise.

[0063] The terms "including", "comprising", “having” and variations thereof mean "including but not limited to", unless expressly specified otherwise.

[0064] The enumerated listing of items does not imply that any or all of the items are mutually exclusive, unless expressly specified otherwise. The terms "a", "an" and "the" mean "one or more", unless expressly specified otherwise.

[0065] A description of an embodiment with several components in communication with each other does not imply that all such components are required. On the contrary a variety of optional components are described to illustrate the wide variety of possible embodiments of the invention.

[0066] When a single device or article is described herein, it will be readily apparent that more than one device / article (whether or not they cooperate) may be used in place of a single device / article. Similarly, where more than one device or article is described herein (whether or not they cooperate), it will be readily apparent that a single device / article may be used in place of the more than one device or article, or a different number of devices / articles may be used instead of the shown number of devices or programs. The functionality and / or the features of a device may be alternatively embodied by one or more other devices which are not explicitly described as having such functionality / features. Thus, other embodiments of the invention need not include the device itself.

[0067] The illustrated operations of Figure 5 show certain events occurring in a certain order. In alternative embodiments, certain operations may be performed in a different order, modified, or removed. Moreover, steps may be added to the above-described logic and still conform to the described embodiments. Further, operations described herein may occur sequentially or certain operations may be processed in parallel. Yet further, operations may be performed by a single processing unit or by distributed processing units.

[0068] Finally, the language used in the specification has been principally selected for readability and instructional purposes, and it may not have been selected to delineate or circumscribe the inventive subject matter. It is therefore intended that the scope of the invention be limited not by this detailed description, but rather by any claims that issue on an application based here on. Accordingly, the disclosure of the embodiments of the invention is intended to be illustrative, but not limiting, of the scope of the invention, which is set forth in the following claims.

[0069] While various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art. The various aspects and embodiments disclosed herein are for purposes of illustration and are not intended to be limiting, with the true scope being indicated by the following claims.Referral Numerals:

Claims

We claim:

1. A method for mitigating data drift in process plants, the method comprising: receiving, by a processor (206), metadata of a first process in a first process plant, when a data drift is detected in the first process due to a change in an input scenario in the first process plant, wherein first model parameters associated with a first Artificial Intelligence (Al) model (106i) deployed in the first process are affected due to the data drift; identifying, by the processor (206), a second process plant configured with a second Al model (IO62) trained on the input scenario, based on the metadata; receiving, by the processor (206), second model parameters of the second Al model (IO62) from the second process plant, upon the identifying; and updating, by the processor (206), the first model parameters of the first Al model (IO61) based on the second model parameters, for mitigating the data drift in the first process plant.

2. The method as claimed in claim 1, wherein the first Al model (IO61) and the second Al model (IO62) are deployed in a first soft sensor (104i) and a second soft sensor (1042) in the first process plant and the second process plant respectively.

3. The method as claimed in claim 1, wherein the metadata corresponds to data of the first process that is affected by the data drift.

4. The method as claimed in claim 1, wherein the identifying further comprising: identifying operating conditions associated with the input scenario in the first process plant, based on the metadata; comparing the operating conditions associated with the input scenario in the first process plant with pre-stored data comprising operating conditions associated with each of a plurality of scenarios in a plurality of process plants; and identifying the second process plant configured with the second Al model (IO62) trained on the input scenario based on the comparison.

5. The method as claimed in claim 1, wherein the mitigating comprising: receiving the first model parameters of the first Al model (IO61) from the first process plant;updating the first model parameters based on the second model parameters using an aggregation strategy, wherein the first Al model (106i) is trained with updated first model parameters.

6. The method as claimed in claim 5, wherein the updating comprising: identifying a plurality of aggregation strategies to update the first model parameters with the second model parameters; and updating the first model parameters using the aggregation strategy of the plurality of aggregation strategies selected based on characteristics of the first model parameters and the second model parameters.

7. A system (102) for mitigating data drift in process plants, the system (102) comprises: a processor (206); and a memory (204), wherein the memory (204) stores processor-executable instructions, which, on execution, cause the processor (206) to: receive metadata of a first process in a first process plant, when a data drift is detected in the first process due to a change in an input scenario in the first process plant, wherein first model parameters associated with a first Artificial Intelligence (Al) model (106i) deployed in the first process are affected due to the data drift; identify a second process plant configured with a second Al model (IO62) trained on the input scenario, based on the metadata; receive second model parameters of the second Al model (IO62) from the second process plant, upon the identifying; and update the first model parameters of the first Al model (IO61) based on the second model parameters, for mitigating the data drift in the first process plant.

8. The system (102) as claimed in claim 7, wherein the first Al model (IO61) and the second Al model (IO62) are deployed in a first soft sensor (104i) and a second soft sensor (1042) in the first process plant and the second process plant respectively.

9. The system (102) as claimed in claim 7, wherein the metadata corresponds to data of the first process that is affected by the data drift. 0 The system (102) as claimed in claim 7, wherein the processor (206) is configured to identify the second process plant by:identifying operating conditions associated with the input scenario in the first process plant, based on the metadata; comparing the operating conditions associated with the input scenario in the first process plant with pre-stored data comprising operating conditions associated with each of a plurality of scenarios in a plurality of process plants; and identifying the second process plant configured with the second Al model (IO62) trained on the input scenario based on the comparison. The system (102) as claimed in claim 7, wherein the processor (206) is configured to mitigate the data drift by: receiving the first model parameters of the first Al model (IO61) from the first process plant; updating the first model parameters based on the second model parameters using an aggregation strategy, wherein the first Al model (IO61) is trained with updated first model parameters. The system (102) as claimed in claim 11, wherein the processor (206) is configured to update the first model parameters by: identifying a plurality of aggregation strategies to update the first model parameters with the second model parameters; and updating the first model parameters using the aggregation strategy of the plurality of aggregation strategies selected based on characteristics of the first model parameters and the second model parameters.

Images