Abstract: A method and system for removing undesirable inferences from a machine learning model include a search component configured to receive a rejected explanation of model output provided by the machine learning model, identify data samples to unlearn by selecting training samples from training data that were used to train the machine learning model, the selected training samples being associated with explanations that are similar to the rejected explanation according to a calculated similarity measure, and pass the data samples to unlearn to a machine unlearning unit.

Abstract: A method includes training a first control model by utilizing a first set of input data as first input, resulting in a trained first control model; copying the trained first control model to a second control model, wherein, after copying, the second input layer and the plurality of second hidden layers is identical to the plurality of first hidden layers, and the first output layer is replaced by the second output layer; freezing the plurality of second hidden layers; training the second control model by utilizing the first set of input data as second input, resulting in a trained second control model; and running the trained second control model by utilizing a second set of input data as second input, wherein the second output outputs the quality measure of the first control model.

Abstract: A computer-implemented method of generating a training data set for training an artificial intelligence module includes providing first and second data sets, the first data set including first data elements indicative of a first operational condition, the second data set including second data elements indicative of a second operational condition that matches the first operational condition. The method further comprises determining a data transformation for transforming the first data elements into the second data elements; applying the data transformation to the first data elements and/or to further data elements of further data sets, thereby generating a transformed data set; and generating a training data set for training the AI module based on at least a part of the transformed data set.

Abstract: A method for applying machine learning to an application includes: a) generating a candidate policy by a learner; b) executing a program in at least one simulated application based on a set of candidate parameters provided based on the candidate policy and a state of the at least one simulated application, execution of the program providing interim results of tested sets of candidate parameters based on a measured performance information of the execution of the program; c) collecting a predetermined number of interim results and providing an end result based on a combination of the candidate parameters and/or the state with the measured performances information by a trainer; and d) generating a new candidate policy by the learner based on the end result.

Abstract: An industrial plant operator intervention system for use in an industrial plant includes a processing unit configured to monitor and analyze industrial plant operation data to detect an anomaly in the industrial plant operation data that warrants initiating an operator intervention, and in response to detecting the anomaly, automatically determine a user interface configuration of a user interface to be presented to a designated operator who is to perform the operator intervention. The user interface configuration is determined on the basis of technical context data, including industrial plant operation data associated with the anomaly, and on the basis of operator data pertaining to the designated operator, in such a manner that an anomaly-related and operator-specific user interface configuration is obtained.

Abstract: A method for generating a process model modeling a manual mode procedure instance of a plant process includes providing log events of operational actions; selecting related sequences of manual mode operational actions from the log events; filtering the related sequences according to an individual plant section; identifying a sequential order from the filtered related sequences; determining statistical properties of values of related process variables and/or statistical properties of values of related set point changes to each sequential ordered manual mode operational action from the filtered related sequences; generating the process model of the manual mode procedure instance by arranging related manual mode operational actions with the sequential order of each operational action assigned with the statistical properties of the values of related process variables and/or assigned with the statistical properties of the values of the related set point changes.

Abstract: A method used for generating a zoomable diagram includes providing a topology diagram of at least a part of the industrial plant for interacting with components of the industrial plant; generating, for each zoom-level of the zoomable diagram, a zoomed diagram comprising a zoomed view of each one of the plurality of objects and connections, by applying at least following rules to each object: from a first predefined zoom-level, collapse all objects of a plant-segment to one object and keep only connections that cross the plant-segment's border, and introduce an plant-segment related alarm measure as a function of all alarm statuses of the plurality of objects of the plant-segment; and when interacting with the components of the industrial plant, determining, for each plant-segment, the plant-segment related alarm measure, and outputting the plant-segment related alarm measure for healing a cause of the alarm.

Abstract: A method for determining an appropriate sequence of actions to take during operation of an industrial plant includes obtaining values of a plurality of state variables that characterize an operational state of the plant (or a part thereof); encoding by at least one trained state encoder network the plurality of state variables into a representation of the operating state of the plant; mapping by a trained state-to-action network the representation of the operating state to a representation of a sequence of actions to take in response to the operating state; and decoding by a trained action decoder network the representation of the sequence of actions to the sought sequence of actions to take.

Abstract: A method for training a machine-learning model to assess at least one condition of industrial equipment, and/or at least one condition of a process running in an industrial plant, based on measurement data gathered by a plurality of sensors, includes: obtaining a plurality of records of measurement data that correspond to a variety of operating situations and a variety of conditions; obtaining, for each record of measurement data, a label that represents a condition in the operating situation characterized by the record of measurement data; and determining a plausibility of at least one record of measurement data, and/or a plausibility of at least one label, based at least in part on a comparison with at least one other record of measurement data, with at least one other label, and/or with additional information about the industrial equipment, and/or about the industrial plant where the industrial equipment resides, and/or about the process.

Abstract: A computer-implemented method for monitoring machine learning models in a distributed setup includes obtaining model activity data relating to activity of the machine learning models in the distributed setup; analyzing the obtained model activity data; and, based on the analysis of the model activity data, outputting model management data for managing the activity of the machine learning models in the distributed setup.

Abstract: A method for detecting an anomaly includes obtaining a time-series of historical process variables within a predefined time span; determining a cycle time of the historical process variables; clustering the historical process variables into clusters based on cycle time; arranging the clusters into a tree; storing the tree; obtaining a time-series of a plurality of current process variables, which correspond to the historic process variables; and detecting the anomaly of at least one device by identifying a cycle time of a current process variable that is longer than the cycle time of a corresponding historic variable.

Abstract: A method for interactive explanations in industrial artificial intelligence systems includes providing a machine learning model and a set of test data, a set of training data and a set of historical data simulating a piping and process equipment; predicting a result for the piping and process equipment based on the machine learning model using the set of test data and the set of training data, wherein the set of historical data is used by the machine learning model to predict at least one parameter of the piping and process equipment; and presenting the predicted at least one parameter on a piping and instrumentation diagram of the piping and process equipment.

Abstract: A method for recommending an operational command includes receiving an alarm from a sensor and/or an operator; obtaining the current state of the plant that includes a current process value and/or operational command; comparing the current state to a list of historic states, each comprising a plurality of historic process values and/or historic operational commands; when the current state matches a subset of at least one of the historic states, starting a simulation and running a plurality of simulations, each based on a variation of at least one of the historic operational commands; determining, for each simulation of the plurality of simulations, a quality value, based on at least one quality criterion; and recommending the variation of the operational command that resulted in the simulation with the highest quality value.

Abstract: A method for deciding on a machine learning model result quality based on the identification of distractive samples in the training data includes providing a first result of the model based on initial training data; determining a first performance of the first result of the model; logging input data; providing a second result of the model based on initial training data and the input data, determining a second performance of the second result of the model and thereon based identifying erroneous data within the input data and/or the training data.

Abstract: A method for explanation of machine learning results based on using a model collection includes training at least two machine learning models with at least two competing strategies for the at least one dataset; and using the least two machine learning models to yield at least two different predictions and/or at least two explanations for the at least one dataset.

Abstract: A method for enabling user feedback and summarizing return of investment for machine learning systems includes providing a training data set and an initial machine learning model; providing a result of the initial machine learning model; receiving feedback on the result of the initial machine learning model from a user enriching the training dataset based on the feedback to an enriched data set; and retraining the initial machine learning model to a retrained machine learning model based on an enriched data set.

Abstract: There is provided an explainer system for explaining an alarm raised by a machine learned model of an industrial automation system. The explainer system is configured to: receive model output from the machine learned model trained to predict anomalous behaviour in the industrial automation system and to raise the alarm; process the model output using at least one prediction explanation technique to identify at least one influential feature which contributed to the model output; use the identified at least one influential feature to extract contextual information from at least one machine-readable information source pertaining to the industrial automation system; and prepare the extracted contextual information for display to an operator of the industrial automation system, to enable the operator to select an appropriate action to take in response to the alarm for ensuring proper functioning of the industrial automation system.

Abstract: A method for determining an appropriate sequence of actions to take during operation of an industrial plant includes obtaining values of a plurality of state variables that characterize an operational state of the plant (or a part thereof); encoding by at least one trained state encoder network the plurality of state variables into a representation of the operating state of the plant; mapping by a trained state-to-action network the representation of the operating state to a representation of a sequence of actions to take in response to the operating state; and decoding by a trained action decoder network the representation of the sequence of actions to the sought sequence of actions to take.

Abstract: A method for training a prediction model includes obtaining training samples representing states of the process that do not cause the undesired event; obtaining based on a process model and a set of predetermined rules that stipulate states having an increased likelihood of the undesired event occurring; training samples representing states with an increased likelihood to cause the undesired event; providing samples to the to-be-trained prediction model to obtain a prediction of the likelihood for occurrence of the undesired event in a state of the process represented by the respective sample; rating a difference between the prediction and the label of the respective sample using a predetermined loss function; and optimizing parameters such that, when predictions are made, the rating by the loss function improves.

Abstract: A method for determining the state of health of an industrial process executed by at least one industrial plant comprising an arrangement of entities, and the state of each such entity, includes obtaining values of the entity state variables; providing the values to a model to obtain a prediction of the state of health; determining propagation paths for anomalies between said entities; determining importances of the states of health of the individual entities for the overall state of health of the process; and aggregating the individual states of health of the entities to obtain the overall state of health of the process.