Method and system for optimization of agglomeration of ores

A system for optimizing the operation of an agglomeration plant, the system comprising:

• an input/output interface for receiving a plurality of data as an input from one or more data sources of the agglomeration plant at a predetermined frequency, wherein the plurality of data comprises of a real-time and non-real-time data;
• one or more hardware processors;
• a memory in communication with the one or more hardware processors, wherein the one or more first hardware processors are configured to execute programmed instructions stored in the one or more first memories, to:
• preprocess the plurality of data;
• obtain simulated data using the preprocessed data and a plurality of soft sensors, wherein the simulated data is integrated with preprocessed data to obtain integrated data;
• determine a first set of parameters using a physics-based wet agglomeration model and the integrated data;
• determine a second set of parameters using physics-based and data-driven charging models, the first set of parameters and the integrated data;
• determine a third set of parameters using a physics-based thermal agglomeration model,
• the first set of parameters, the second set of parameters and the integrated data;
• determine a final set of parameters using a plurality of data-driven models, the first, the second and the third set of parameters, and the integrated data;
• configure an optimizer using the plurality of physics-based and data-driven models of the agglomeration plant to optimize a plurality of key performance parameters of the agglomeration plant;
• generate at least one recommendation using the configured optimizer, wherein the recommendations comprise of optimal settings of a plurality of manipulated variables,
• wherein the manipulated variables includes height of the wet agglomerate in the thermal agglomeration from lower to upper limits, a strand speed from lower to higher limits, percentage of return fines or percentage of undersize agglomerate from lower to upper limits and a reduction degradation index (RDI) of the agglomerate from lower to upper limits;
• provide the generated recommendations to optimize the key performance parameters of the agglomeration plant;
• and
• monitor the key performance parameters of the agglomeration plant using self-learning physics-based and data-driven models and re-tune or re-train physics-based and data-driven models of the agglomeration plant, wherein the self-learning physics-based and data-driven models of the agglomeration plant comprising:
• computing, via the one or more hardware processors, a model performance index for each of the plurality of physics-based and data-driven models in real-time by comparing the model outputs with the corresponding measured variables, wherein the model performance index includes an error metrics, percentage of points error, a coefficient of determination and a customized performance metrics;
• determining, via the one or more hardware processors, the models responsible for model performance index exceeding the pre-determined thresholds and triggering one method out of a plurality of self-learning methods based on the model type;
• executing, via the one or more hardware processors, first type of self-learning method for drifted plant KPI models, wherein the drifted plant KPI model include productivity, fuel rate, yield, agglomerate quality, tumbler index, reduction degradation index, reducibility index, the plant KPI;
• performing self-learning of the physics-based and data-driven models of the agglomeration plant by re-training active models, wherein the active models are re-trained using recent and historical data of the agglomeration plant to obtain re-trained models;
• executing, via the one or more hardware processors, second type of self-learning method, wherein the active physics-based model of the thermal agglomeration unit is re-tuned in at least one level based on the sensitivity of tuning parameters to the model predictions, followed by the first type of self-learning method using recent and historical data of the agglomeration plant to obtain re-tuned and re-trained models;
• executing, via the one or more hardware processors, third type of self-learning method, wherein the active physics-based model of wet agglomeration and data-driven charging models are re-tuned and re-trained followed by the second type of self-learning method using recent and historical data of the agglomeration plant to obtain re-tuned and re-trained models;
• assessing, via the one or more hardware processors, the efficacy of executed self-learning methods by computing the model performance index using the predictions from the re-tuned and re-trained models with the corresponding measured variables;
• and
• recommending, via the one or more hardware processors, at least one re-tuned and re-trained model for activation in the agglomeration plant wherein the model performance index of the at least one re-tuned and re-trained model is below the pre-determined thresholds.

One or more non-transitory machine readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause:

• receiving a plurality of data as an input from one or more data sources of the agglomeration plant at a predetermined frequency, wherein the plurality of data comprises of a real-time and non-real-time data;
• preprocessing the plurality of data;
• obtaining simulated data using the preprocessed data and a plurality of soft sensors, wherein the simulated data is integrated with preprocessed data to obtain integrated data;
• determining a first set of parameters using a physics-based wet agglomeration model and the integrated data;
• determining a second set of parameters using physics-based and data-driven charging models, the first set of parameters and the integrated data;
• determining a third set of parameters using a physics-based thermal agglomeration model, the first set of parameters, the second set of parameters and the integrated data;
• determining a final set of parameters using a plurality of data-driven models, the first, the second and the third set of parameters, and the integrated data;
• configuring an optimizer using the plurality of physics-based and data-driven models of the agglomeration plant to optimize a plurality of key performance parameters of the agglomeration plant;
• generating at least one recommendation using the configured optimizer, wherein the recommendations comprise of optimal settings of a plurality of manipulated variables, wherein the manipulated variables includes height of the wet agglomerate in the thermal agglomeration from lower to upper limits, a strand speed from lower to higher limits, percentage of return fines or percentage of undersize agglomerate from lower to upper limits and a reduction degradation index (RDI) of the agglomerate from lower to upper limits;
• providing the generated recommendations to optimize the key performance parameters of the agglomeration plant;
• and
• monitoring the key performance parameters of the agglomeration plant using self-learning physics-based and data-driven models and re-tune or re-train physics-based and data-driven models of the agglomeration plant, wherein the self-learning physics-based and data-driven models of the agglomeration plant comprising:
• computing, via the one or more hardware processors, a model performance index for each of the plurality of physics-based and data-driven models in real-time by comparing the model outputs with the corresponding measured variables, wherein the model performance index includes an error metrics, percentage of points error, a coefficient of determination and a customized performance metrics;
• determining, via the one or more hardware processors, the models responsible for model performance index exceeding the pre-determined thresholds and triggering one method out of a plurality of self-learning methods based on the model type;
• executing, via the one or more hardware processors, first type of self-learning method for drifted plant KPI models, wherein the drifted plant KPI model include productivity, fuel rate, yield, agglomerate quality, tumbler index, reduction degradation index, reducibility index, the plant KPI;
• performing self-learning of the physics-based and data-driven models of the agglomeration plant by re-training active models, wherein the active models are re-trained using recent and historical data of the agglomeration plant to obtain re-trained models;
• executing, via the one or more hardware processors, second type of self-learning method, wherein the active physics-based model of the thermal agglomeration unit is re-tuned in at least one level based on the sensitivity of tuning parameters to the model predictions, followed by the first type of self-learning method using recent and historical data of the agglomeration plant to obtain re-tuned and re-trained models;
• executing, via the one or more hardware processors, third type of self-learning method, wherein the active physics-based model of wet agglomeration and data-driven charging models are re-tuned and re-trained followed by the second type of self-learning method using recent and historical data of the agglomeration plant to obtain re-tuned and re-trained models;
• assessing, via the one or more hardware processors, the efficacy of executed self-learning methods by computing the model performance index using the predictions from the re-tuned and re-trained models with the corresponding measured variables;
• and
• recommending, via the one or more hardware processors, at least one re-tuned and re-trained model for activation in the agglomeration plant wherein the model performance index of the at least one re-tuned and re-trained model is below the pre-determined thresholds.