System for action determination

A method for action determination, comprising:

• a) providing a processor with information relating to a plurality of past actions over a period of time associated with an operation of an industrial process, wherein the information relating to the plurality of past process actions comprises one or more operator actions;
• b) providing the processor with information relating to a plurality of past process events over the time period associated with the operation of the industrial process, wherein the information relating to the plurality of past process events comprises a plurality of alarms, wherein the processor is configured to ignore one or more alarms that appear uniformly over the time period, and wherein the processor is configured to cluster multiple identical alarms within a defined short time interval relative to the period of time into a single alarm;
• c) providing the processor with information relating to a new process event;
• d) determining by the processor a correlation between at least some of the plurality of past actions provided at step a) with at least some of the past process events provided at step b), wherein determination of the correlation comprises a statistical analysis to detect a relationship between alarms and actions, wherein determination of the correlation comprises a selection of alarms that occur rarely during the time period, wherein determination of the correlation comprises determination of an action to alarm matrix, wherein determination of the correlation comprises a statistical inversion of the action to alarm matrix to determine a plurality of rules and associated probabilities, and wherein step d) comprises applying a principal component analysis to filter most significant rules and wherein filtering of the most significant rules comprises removal of one or more rules that correspond to an alarm that triggers a significantly high number of rules;
• e) determining by the processor at least one recommended action from the information relating to the new process event provided at step c), the determination comprising utilization of the correlation determined at step d), and wherein step e) comprises utilizing the most significant rules ordered by their associated probabilities;
• f) outputting, by an output unit, the at least one recommended action;
• and
• g) performing automated operation of the industrial process based on the at least one recommended action, wherein performing the automated operation of the industrial process based on the at least one recommended action comprises closing and/or opening of a valve.

A non-transitory computer-readable medium having processor-executable instructions stored thereon, wherein the processor-executable instructions, when executed by a processor, facilitate:

• a) providing the processor with the information relating to the plurality of past actions over the period of time associated with the operation of the industrial process, wherein the information relating to the plurality of past process actions comprises one or more operator actions;
• b) providing the processor with the information relating to the plurality of past process events over the time period associated with the operation of the industrial process, wherein the processor is configured to ignore one or more alarms that appear uniformly over the time period, and wherein the processor is configured to cluster multiple identical alarms within a defined short time interval relative to the period of time into a single alarm;
• c) providing the processor with the information relating to the new process event;
• d) determining by the processor the correlation between the at least some of the plurality of the past actions provided at step a) with at least some of the past process events provided at step b), wherein determination of the correlation comprises a statistical analysis to detect a relationship between alarms and actions, wherein determination of the correlation comprises a utilization of natural language processing and word embeddings to obtain vectorial representations of information relating to the plurality of past actions and the information relating to the plurality of past process events, and wherein actions and events are determined to be correlated when related actions and events are mapped close together in feature space with respect to other mappings, wherein determination of the correlation comprises determination of an action to alarm matrix, wherein determination of the correlation comprises a statistical inversion of the action to alarm matrix to determine a plurality of rules and associated probabilities, and wherein step d) comprises applying a principal component analysis to filter most significant rules and wherein filtering of the most significant rules comprises removal of one or more rules that correspond to an alarm that triggers a significantly high number of rules;
• e) determining by the processor the at least one recommended action from the information relating to the new process event provided at step c), the determination comprising utilization of the correlation determined at step d), and wherein step e) comprises utilizing the most significant rules ordered by their associated probabilities;
• f) outputting by the output unit the at least one recommended action;
• and
• g) performing automated operation of the industrial process based on the at least one recommended action, wherein performing the automated operation of the industrial process based on the at least one recommended action comprises closing and/or opening of a valve.