System and method for development and deployment of self-organizing cyber-physical systems for manufacturing industries

A processor implemented method of monitoring an industrial plant, comprising:

• monitoring (902) a plurality of components of the industrial plant by a soft sensor which is digital twins of physical sensors;
• and performing a self-monitoring of the industrial plant, wherein the self-monitoring comprises a self-diagnosis comprising steps of:
• determining (910.a) operation of any of the plurality of components outside a pre-defined range of operation in a multi-layered interacting structure, as a deviation from a pre-defined normal range of operation of each of a plurality of industrial plant components, using one or more hardware processors, wherein determined performance is compared with an expected performance to determine the deviation, wherein when a determined deviation is below a threshold of the deviation, then working of a component is in a ‘normal range’ of operation, wherein when the determined deviation exceeds the threshold of the deviation, then working of the component is ‘outside the pre-defined normal range of operation’, and wherein the plurality of industrial plant components includes sensors, actuators, and controllers, wherein upon determining quality of plant data, soft sensors substitutes lack of quality of data in case physical sensors are found to be faulty or non-functional, wherein the soft sensors, are digital twins of the physical sensors, are monitored continuously to ascertain performance and corresponding behavioral models are updated;
• monitoring the performance separately at component level and at plant level continuously;
• and recommending (910.b) at least one action in response to the determined deviation from the pre-defined normal range of operation, via the one or more hardware processors, wherein the action is one or more of a) optimization of the industrial plant, b) predictive maintenance of at least one of the plurality of components, c) generating an alert for one or more users, and d) self-updation of one or more behavioral models that affect performance of the industrial plant, wherein the recommended at least one action is triggered, wherein the alert is generated upon identifying one or more pre-defined types of events which occurred or were predicted to occur in a prediction horizon and the alert contains information about an event and one or more recommendations to assist the one or more users or an operator, wherein the one or more behavioral models represent a prediction model, a classification model, an optimization model used to represent behavior of the industrial plant, detecting trigger conditions for the recommended at least one action in terms of values of different parameters or combination of parameters, as configured, wherein the trigger conditions comprises:
• determining whether a previously triggered self-optimization execution is in progress or being implemented;
• waiting till the previously triggered self-optimization execution is terminated in response to determining that the previously triggered self-optimization execution is in progress;
• triggering a self-optimization in response to determining that there is no ongoing previously triggered self-optimization execution;
• determining whether a time elapsed after a previous execution exceeded a threshold of time;
• triggering the self-optimization execution in response to determining that the time elapsed after the previous execution exceeded the threshold of time;
• waiting to trigger the self-optimization execution till the time elapsed exceed the threshold of time, in response to determining that the time elapsed after the previously triggered self-optimization execution is within the threshold of time;
• wherein the trigger conditions are defined in terms of model quality parameters, upon detecting that a new regime has been flagged and a change in raw materials, wherein determining a deviation and recommending at least one action comprises:
• performing analysis of the industrial plant data, comprising:
• determining whether a determined quality of the industrial plant data meets a defined threshold of quality of plant data, wherein the determined quality of the industrial plant data not meeting the defined threshold of quality of plant data is considered as an instance of bad quality of data;
• generating additional data to compensate for any determined instance of bad quality of data to improve quality of the industrial plant data, wherein the method is performed at multiple levels including a lowest level and a global level, wherein at the lowest level, the method monitors plant measurements of process variables, actuator signals, or controller outputs using their corresponding soft sensors and at the global level, the method monitors the industrial plant comprising of a plurality of assets, operations or processes that are carried out in the industrial plant, materials and their flow, an environment in which the industrial plant operates, the one or more users operating the industrial plant, and automation systems that operate at an individual asset or operation level, to optimize performance of the industrial plant at the global level, wherein the method uses the industrial plant data and predictions of the one or more behavioral models to monitor the soft sensors for degradation in their predictive capabilities which are caused due to a plurality of factors including a physical failure of a sensor, inappropriate or infrequent calibration of the sensor, degradation of the predictive capabilities of the one or more behavioral models of the sensor, wherein industrial plant operators act upon the predictions of the one or more behavioral models to adjust quality of the industrial plant data, wherein the method provides real-time decisions to the industrial plant operators for optimization of Key Performance Indicators (KPIs) related to a physical system or process under consideration by working with a physical entity including the asset, a unit operation or the complete industrial plant, wherein a self-organizing system is deployed and updated further to enhance decision making capabilities for future requirements, wherein simulated data is archived with corresponding real-time data in a knowledge-base for the industrial plant for future use.

A computer program product comprising a non-transitory computer readable medium having a computer readable program embodied therein, wherein the computer readable program, when executed, causes a computing device to perform:

• monitoring a plurality of components of an industrial plant, via one or more hardware processors;
• and performing a self-monitoring of the industrial plant, wherein the self-monitoring comprises a self-diagnosis comprising the steps of:
• determining operation of any of a plurality of industrial plant components outside a pre-defined range of operation in a multi-layered interacting structure, as a deviation from a pre-defined normal range of operation of each of the plurality of industrial plant components, using the one or more hardware processors, wherein a determined performance is compared with an expected performance to determine the deviation, wherein when the determined deviation is below a threshold of the deviation, then working of a component is in a ‘normal range’ of operation, wherein when the determined deviation exceeds the threshold of the deviation, then working of the component is ‘outside the pre-defined normal range of operation’, and wherein the plurality of industrial plant components includes sensors, actuators, and controllers, monitoring the determined performance separately at component level and at plant level continuously, wherein upon determining quality of plant data, soft sensors substitutes lack of quality of data in case physical sensors are found to be faulty or non-functional, wherein the soft sensors, are digital twins of the physical sensors, are monitored continuously to ascertain performance and corresponding behavioral models are updated;
• and recommending at least one action in response to the determined deviation from the pre-defined normal range of operation, via the one or more hardware processors, wherein the action is one or more of a) optimization of the industrial plant, b) predictive maintenance of at least one of the plurality of components, c) generating an alert for one or more users, and d) self-updation of one or more behavioral models that affect performance of the industrial plant wherein the recommended at least one action is triggered, wherein the alert is generated upon identifying one or more pre-defined types of events occurred or predicted to occur in a prediction horizon and the alert contains information about an event and one or more recommendations to assist the one or more users or an operator, wherein the one or more behavioral models represent a prediction model, a classification model, an optimization model used to represent behavior of the industrial plant, detecting trigger conditions for the at least one action in terms of values of different parameters or combination of parameters, as configured, wherein the trigger conditions comprises:
• determining whether a previously triggered self-optimization execution is in progress or being implemented;
• waiting till the previously triggered self-optimization execution is terminated in response to determining that the previously triggered self-optimization execution is in progress;
• triggering a self-optimization in response to determining that there is no ongoing execution;
• determining whether a time elapsed after a previous execution exceeded a threshold of time;
• triggering the self-optimization execution in response to determining that the time elapsed after the previous execution exceeded the threshold of time;
• waiting to trigger the self-optimization execution till the time elapsed exceed the threshold of time, in response to determining that the time elapsed after the previous execution is within the threshold of time;
• wherein the trigger conditions are defined in terms of model quality parameters, upon detecting that a new regime has been flagged and change in raw materials, wherein determining the deviation and recommending the at least one action comprises:
• performing analysis of industrial plant data, comprising:
• determining whether a determined quality of the industrial plant data meets a defined threshold of quality of plant data, wherein the determined quality of the industrial plant data not meeting the defined threshold of quality of plant data is considered as an instance of bad quality of data;
• generating additional data to compensate for any determined instance of bad quality of data to improve quality of the industrial plant data, wherein a self-organizing system is configured to function at multiple levels including a lowest level and a global level, wherein at the lowest level, the self-organizing system monitors plant measurements of process variables, actuator signals, or controller outputs using their corresponding soft sensors and at the global level, the self-organizing system monitors a complete industrial plant comprising of a plurality of assets, operations or processes that are carried out in the industrial plant, materials and their flow, an environment in which the industrial plant operates, the one or more users operating the industrial plant, and automation systems that operate at an individual asset or operation level, to optimize performance of the complete industrial plant at the global level, wherein the self-organizing system uses the industrial plant data and predictions of the one or more behavioral models to monitor the soft sensors for degradation in their predictive capabilities which are caused due to a plurality of factors including a physical failure of a sensor, inappropriate or infrequent calibration of the sensor, degradation of the predictive capabilities of the behavioral model of the sensor, wherein industrial plant operators act upon the predictions of the one or more behavioral models to adjust quality of the industrial plant data, wherein the self-organizing system provides real-time decisions to the industrial plant operators for optimization of Key Performance Indicators (KPIs) related to a physical system or process under consideration by working with a physical entity including the assets, a unit operation or the complete industrial plant, wherein the self-organizing system is deployed and updated further to enhance decision making capabilities for future requirements, wherein simulated data is archived with corresponding real-time data in a knowledge-base for the industrial plant for future use.