Methods and systems for improving asset operations based on identifying significant changes in sensor combinations during relevant events.

Abstract

The present invention discloses a method and system for improving asset operation based on identifying significant changes in sensor combinations in relevant events. The method includes the following steps: providing an asset representation of an asset with an alarm aggregation system; creating a list of calculated indicators, which has at least one calculated indicator in the asset configured to calculate, predict or estimate the indicated asset behavior; merging the list of calculated indicators into a group of entity connections by associating list elements in the list of calculated indicators, thereby generating an interconnected representation; associating the identified asset behavior with at least one related object and / or at least one related process that damages the asset to cause the identified asset behavior to be unable to optimally perform, conduct and / or achieve; and generating at least one prompt associated with at least one related object and / or at least one related process that damages the identified asset behavior.

Description

Technical Field

[0001] The present invention relates to methods and systems for improving asset operations based on identifying significant changes in sensor combinations in related events where the asset fails to meet expected behavior. Background Art

[0002] Predictive maintenance and fault detection and analysis are critical in many industries where unpredictable problems can be costly, involving a number of adverse outcomes, including loss of money, operational downtime, equipment loss, property damage, penalties, compensation, and sometimes even human death. To prevent such damage, many industrial plants have installed sensors to help monitor plant production and its processes, generating alarms to notify operators of potential problems. Due to the nature of complex systems that include a large number of possible normal system states, and the fact that individual alarms can involve a large number of sensors, the use of machine learning algorithms can often produce valuable information as well as misleading false alarms, complicating the analysis and focus on important asset elements related to asset operation and performance.

[0003] U.S. Patent No. US10,318,364B2, having common inventorship and assigned to the same entity as the present invention, teaches a method for problem-alarm aggregation. Even in a system environment where such problem-alarm aggregation is implemented, determining which so-called "incidents" are of high concern, as well as the root causes, factors and their interrelationships of such events is a challenging task.

[0004] It would be desirable to have methods and systems for improving asset operations based on identifying significant changes in sensor combinations during relevant events.These methods and systems would overcome the various limitations described above. Summary of the invention

[0005] It is an object of the present invention to provide methods and systems for improving asset operations based on identifying significant changes in sensor combinations during relevant events.

[0006] Note that the wording "exemplary" is used herein to refer to examples of embodiments and / or implementations and is not meant to necessarily convey a more desirable use case. Similarly, the words "alternative" and "alternatively" used herein refer to examples outside of the classification of intended embodiments and / or implementations and are not meant to necessarily convey a more desirable use case. Therefore, it can be understood from the above that "exemplary" and "alternative" can be applied to multiple embodiments and / or implementations herein. Various combinations of such alternative and / or exemplary embodiments are also contemplated herein.

[0007] For the sake of clarity, several terms used herein are specifically defined below. As used herein, the term "asset" refers to an integrated physical system having a purpose or primary activity that is composed of a large number of intertwined entities, objects, parts, and processes that work together to achieve that purpose or primary activity. Examples of such assets include industrial complexes, production facilities, chemical refineries, integrated physical operations, integrated machinery, automated machinery, self-driving vehicles, locomotives, ships, aircraft, integrated computer programs, and biological systems.

[0008] The term "object" as used herein refers to any physical item (i.e., something that can be touched or accessed). The term "process" as used herein refers to a non-physical item that affects at least one object. The term "entity" as used herein refers to a broader class of objects and / or processes. The terms "feature" or "attribute" as used herein refer to a property of an entity. The term "sensor" as used herein refers to a sensor, a measurement, a user tag, or a column in a database.

[0009] The term "asset representation" as used herein refers to a model of an asset that includes not only the physical parts of the asset, but also the characteristics of the asset, such as asset processes, features, and sensors. The asset representation of a specific asset, whether a plant / factory structure, a machine, a chemical refinery, a biological system, and / or a production process, includes all or some of the sensors that generate data-driven alerts.

[0010] As used herein, the terms "identified asset behavior" or "identified behavior" refer to specific behavior of an asset, which may be favorable (i.e., content to be encouraged) or unfavorable (i.e., content to be discouraged). Such identified behavior may be used to set goals or objectives. For example, one type of identified behavior may be a specific production level. Another type of identified behavior may be a reduction in the occurrence of a particular predefined type of event (such as an identified event) (expanding from the attribution meaning in U.S. Patent No. 10,318,364 B2).

[0011] As used herein, the term "computed metric" refers to a metric that is calculated, predicted, or estimated based on data from sensors in an asset.

[0012] The term "template" as used herein refers to a framework for an asset representation of a sub-asset in an asset. Such a sub-asset may appear more than once in the asset in the same or different variations. The template may include additional information about the modeled sub-asset, such as failure analysis guidelines, preferences, priorities, rules, or recommendations.

[0013] The term "alert" as used herein refers to highlighting an abnormal sensor-measurement combination in an asset. Such an alert helps clarify the underlying anomaly in the asset and its possible cause.

[0014] As used herein, the term "graph" refers to a set of vertices connected by edges. As used herein, the term "data-driven alerts" or simply "alerts" refers to alerts generated based on real-world data measured and / or detected on assets and their entities. Alerts are generated through data mining, machine learning, or deep learning algorithms (e.g., sensor thresholds, problem-specific alert scripts, pattern detection of known problems, pattern likelihood of known problems, classification of known problems, abnormal state detection, measurement deviations from algorithmic predictions, trend detection, and / or new cluster detection).

[0015] Therefore, according to the present disclosure, there is provided for the first time a method for identifying and improving asset behavior based on an interconnected representation of comprehensive physical operations, the method comprising the following steps:

[0016] (a) providing an asset representation of an asset, wherein the asset is an integrated physical system having a primary activity; wherein an object is a physical item; wherein a process is a non-physical item that affects at least one object; wherein entities are physical objects and / or processes of the asset; and wherein the entities are adapted to act together to achieve the primary activity; and wherein the asset representation comprises: (i) a sensor list having all relevant sensors, all relevant measurement results and / or all relevant sensor data columns in a database related to real-world data measured and / or detected in the asset; (ii) an object list having at least two relevant objects in the asset; (iii) a process list having at least two relevant processes in the asset; (iv) an entity connection group obtained by associating list elements in the sensor list, the object list and the process list; wherein all relevant objects and all relevant processes in the asset are appropriately listed and correspondingly associated to the asset representation of the asset; (b) (c) merging the list of calculated indicators into the entity connection group by associating list elements in the list of calculated indicators, wherein all related objects and all related processes in the asset are appropriately listed and correspondingly associated to the asset representation of the asset, thereby producing the interconnected representation; (d) associating the identified asset behavior with at least one of the related objects and / or at least one of the related processes that impairs the asset so that the asset cannot optimally perform, conduct and / or achieve the identified asset behavior; and (e) generating at least one prompt associated with at least one of the related objects and / or at least one of the related processes that impairs the identified asset behavior.

[0017] Alternatively, said at least one calculated metric is defined as a function of a subset of said sensor list elements.

[0018] Alternatively, the method further comprises the steps of: (f) calculating scores for respective combinations of the sensor list elements using a predefined function, wherein the combinations are related to the identified asset behavior; and (g) prioritizing the at least one prompt based on the scores.

[0019] Alternatively, the step of creating includes using at least one time period during which the identified asset behavior occurred or did not occur to create the at least one calculated indicator.

[0020] Alternatively, the method further comprises the steps of: (f) comparing a first given asset representation having the identified asset behavior with a second given asset representation lacking the identified asset behavior to produce a comparative representation; and (g) associating the at least one prompt with at least one list element appearing in the comparative representation.

[0021] Alternatively, the method further comprises the following steps: (f) training a machine learning algorithm to predict the state of the identified asset behavior represented by the asset, wherein the training step also identifies the related list elements.

[0022] Alternatively, said step of merging comprises extending said evaluation of said sensor list using said asset representation as input to said at least one calculated metric.

[0023] Alternatively, the method further comprises the following steps: (f) verifying the entity connection group represented by the asset by analyzing the real-world data obtained by the relevant sensors; and (g) cleaning the entity connection group represented by the asset by retaining connections with direct influence in the entity connection group and by removing connections with remote and / or indirect influence from the entity connection group.

[0024] Alternatively, the method further comprises the following steps: (f) collecting user feedback regarding the at least one prompt; (g) learning a prompt scoring function based on the user feedback; (h) using the prompt scoring function to calculate scores for respective combinations of the sensor list elements, wherein the combinations are related to the identified asset behavior; and (i) prioritizing the at least one prompt based on the score.

[0025] Alternatively, the method further comprises the step of: (f) modifying the at least one prompt based on at least a template of the asset representation.

[0026] According to the present disclosure, a system for identifying and improving asset behavior based on an interconnected representation of integrated physical operations is provided for the first time, the system comprising: (a) a CPU for performing computing operations; (b) a memory module for storing data; (c) an alarm aggregation module configured to: (i) provide an asset representation of an asset, wherein the asset is an integrated physical system having a primary activity; wherein an object is a physical item; wherein a process is a non-physical item that affects at least one object; wherein entities are physical objects and / or processes of the asset; and wherein the entities are adapted to act together to achieve the primary activity; and Wherein, the asset representation includes: (A) a sensor list having all relevant sensors, all relevant measurement results and / or all relevant sensor data columns in a database related to real-world data measured and / or detected in the asset; (B) an object list having at least two relevant objects in the asset; (C) a process list having at least two relevant processes in the asset; (D) an entity connection group obtained by associating list elements in the sensor list, the object list and the process list; wherein all relevant objects and all relevant processes in the asset are appropriately listed and (d) a calculated indicator module configured to create a list of calculated indicators having at least one calculated indicator for the asset, wherein each calculated indicator is a metric configured to calculate, predict or estimate an indicated asset behavior based on evaluating the list of sensors in the asset, and wherein each calculated indicator is directly affected by the at least two related processes; (e) an indicator connection module configured to merge the list of calculated indicators into the entity connection group by associating list elements in the list of calculated indicators, wherein , all relevant objects and all relevant processes in the asset are appropriately listed and correspondingly associated into the asset representation of the asset, thereby generating the interconnected representation; (f) an identified behavior module, which is configured to associate the identified asset behavior with at least one of the related objects and / or at least one of the related processes that damage the asset so that the asset cannot optimally perform, conduct and / or achieve the identified asset behavior; and (g) a prompt generation module, which is configured to generate at least one prompt associated with at least one of the related objects and / or at least one of the related processes that damage the identified asset behavior.

[0027] Alternatively, said at least one calculated metric is defined as a function of a subset of said sensor list elements.

[0028] Alternatively, the system further comprises: (h) a score calculation module configured to calculate scores for respective combinations of the sensor list elements using a predefined function, wherein the combinations are related to the identified asset behavior; and (i) a prompt prioritization module configured to prioritize at least one of the prompts based on the score.

[0029] Alternatively, the calculated indicator module is further configured to create the at least one calculated indicator using at least one time period during which the identified asset behavior occurred or did not occur.

[0030] Alternatively, the system further comprises: (h) a representation comparison module configured to compare a first given asset representation having the identified asset behavior with a second given asset representation lacking the identified asset behavior to produce a comparison representation; and (i) a prompt association module configured to associate the at least one prompt with at least one list element appearing in the comparison representation.

[0031] Alternatively, the system further comprises: (h) a training module configured to train a machine learning algorithm to predict the state of the identified asset behavior represented by the asset, wherein the training also identifies the related list elements.

[0032] Alternatively, the metric connection module is further configured to extend the evaluation of the sensor list using the asset representation as an input to the at least one calculated metric.

[0033] Alternatively, the system further comprises: (h) a data verification module configured to verify the entity connection group represented by the asset by analyzing the real-world data obtained by the relevant sensors; and (i) a connection cleaning module configured to clean the entity connection group represented by the asset by retaining connections with direct influence in the entity connection group and by removing connections with remote and / or indirect influence from the entity connection group.

[0034] Alternatively, the prompt generation module is further configured to: (i) collect user feedback regarding the at least one prompt; (ii) learn a prompt scoring function based on the user feedback; (iii) use the prompt scoring function to calculate scores for respective combinations of the sensor list elements, wherein the combinations are related to the identified asset behavior; and (iv) prioritize the at least one prompt based on the score.

[0035] Alternatively, the prompt generation module is further configured to modify the at least one prompt based on at least a template of the asset representation.

[0036] According to the present disclosure, a non-transitory computer-readable storage medium is provided for the first time, which has a computer-readable code contained on the non-transitory computer-readable storage medium for identifying and improving asset behavior based on an interconnected representation of integrated physical operations, the computer-readable code comprising: (a) program code for providing an asset representation of an asset, wherein the asset is an integrated physical system with a primary activity; wherein an object is a physical item; wherein a process is a non-physical item that affects at least one object; wherein entities are physical objects and / or processes of the asset; and wherein the entities are adapted to act together to achieve the primary activity; and wherein the asset representation comprises: (i) a sensor list having all relevant sensors, all relevant measurement results, and / or all relevant sensor data columns in a database related to real-world data measured and / or detected in the asset; (ii) an object list having at least two relevant objects in the asset; (iii) a process list having at least two relevant processes in the asset; (iv) an entity connection group obtained by associating list elements in the sensor list, the object list, and the process list; wherein all relevant objects and all relevant processes in the asset (a) program code for creating a list of calculated indicators having at least one calculated indicator for the asset, wherein each calculated indicator is a metric configured to calculate, predict or estimate an indicated asset behavior based on evaluating the list of sensors in the asset, and wherein each calculated indicator is directly affected by the at least two related processes; and (c) program code for merging the list of calculated indicators into the entity representation by associating list elements in the list of calculated indicators. (d) program code for associating an identified asset behavior with at least one of the related objects and / or at least one of the related processes that damages the asset so that the asset cannot optimally perform, conduct and / or achieve the identified asset behavior; and (e) program code for generating at least one prompt associated with at least one of the related objects and / or at least one of the related processes that damages the identified asset behavior.

[0037] Alternatively, said at least one calculated metric is defined as a function of a subset of said sensor list elements.

[0038] Alternatively, the computer readable code further comprises: (f) program code for calculating scores for combinations of the sensor list elements using a predefined function, wherein the combinations are related to the identified asset behavior; and (g) program code for prioritizing the at least one prompt based on the score.

[0039] Alternatively, the creating comprises creating the at least one calculated indicator using at least one time period during which the identified asset behavior occurred or did not occur.

[0040] Alternatively, the computer-readable code also includes: (f) program code for comparing a first given asset representation having the identified asset behavior with a second given asset representation lacking the identified asset behavior to produce a comparison representation; and (g) program code for associating the at least one prompt with at least one list element appearing in the comparison representation.

[0041] Alternatively, the computer readable code further comprises: (f) program code for training a machine learning algorithm to predict states of the identified asset behavior represented by the asset, wherein the training also identifies related list elements.

[0042] Alternatively, said merging comprises extending said evaluation of said sensor list using said asset representation as an input to said at least one calculated metric.

[0043] Alternatively, the computer-readable code also includes: (f) program code for verifying the entity connection group represented by the asset by analyzing the real-world data obtained by the relevant sensors; and (g) program code for cleaning the entity connection group represented by the asset by retaining connections with direct effects in the entity connection group and by removing remote and / or indirect effects from the entity connection group.

[0044] Alternatively, the computer readable code further comprises: (f) program code for collecting user feedback regarding the at least one prompt; (g) program code for learning a prompt scoring function based on the user feedback; (h) program code for calculating scores for combinations of the sensor list elements using the prompt scoring function, wherein the combinations are related to the identified asset behavior; and (i) program code for prioritizing the at least one prompt based on the score.

[0045] Alternatively, the computer readable code further comprises: (f) program code for modifying the at least one prompt based on at least a template of the asset representation.

[0046] These and other embodiments will be apparent from the following detailed description and examples. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The present invention is herein described, by way of example only, with reference to the accompanying drawings, in which:

[0048] Figure 1 is a diagram of exemplary prompts for concept events in an asset involving temperature and vibration sensors according to an embodiment of the present invention;

[0049] Figure 2 is a diagram of an exemplary simplified graphical user interface (GUI) for collecting user tags for occurrences of specific periods of identified asset behavior in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The present invention relates to a method and system for improving asset operation based on identifying significant changes in sensor combinations during relevant events. The principles and operation of providing such a method and system according to the present invention may be better understood with reference to the accompanying description and drawings.

[0051] The asset data graph (ADG) described herein accurately describes an asset representation if the results of a set of predefined queries are the same on both the graph and the asset representation. Similarly, an asset representation describes an asset if the results of a set of predefined queries on the asset representation are the same as the results of a human expert. Thus, having a graph for the ADG enables the replacement of queries to human experts by employing a library of standard queries on the ADG derived from the asset representation.

[0052] Such a system does not know about the assets and has no access to them. Therefore, the system cannot verify that the asset representation is a faithful representation of the asset. The system is responsible for making the resulting representation valid for use in alert aggregation. Instead of the modeling language experts who communicate with "asset experts" in traditional modeling systems, the system provides asset experts with a simple and well-defined process for modeling their assets themselves.

[0053] The system's Asset Structure panel contains a hierarchical object structure of assets, where objects may be items that contain other entities. The system's Asset Behavior panel contains hierarchical processes in assets, where a root process implements the primary functionality of an asset. Processes can be connected to objects in such an Asset Structure panel, which can affect the process, be affected by the process, or both. Note that processes can be directed, which indicates the direction of the influence.

[0054] The output of such a system is a sensor-incorporated asset representation that is converted into an ADG for automatic querying using a library of standard queries on the AGD. The responses to a set of predefined queries need to match those of asset experts and be described in terms that asset experts fully understand.

[0055] All elements of the ADG, whether sensor elements, entity elements (i.e., object elements and / or process elements), or attribute elements, can include additional metadata and information, such as detailed explanations or references to user guides or problem-solving manuals. Data measurements and sensors are connected to the measured attributes, and these attributes are in turn connected to related entities in the ADG, which makes the data a local part of the ADG.

[0056] In one embodiment of the invention, the asset representation undergoes a validation and cleaning process in which asset representation links are validated through analysis of relevant sensor data. Links that imply direct causes or effects are retained, and links that imply remote or indirect effects are deleted. The weights of the links that are retained and the elements of the asset representation may be changed during the cleaning process.

[0057] Alarms are aggregated into events, which are identified and scored (e.g., reflecting event importance, relevance, or significance) based on an asset representation or ADG. This aggregation typically produces multiple events of interest. Each alarm originates from at least one sensor that is connected to the attributes describing an entity in the ADG. All alarms triggered by adjacent, subordinate, similar, or process-connected entities of the asset representation are accumulated into a single event. The resulting events are expressed in terms used to describe the relevant elements of the asset in the asset representation, which are terms and expressions used and understood by maintenance and security teams.

[0058] In some exemplary use cases, an event may be presented with a prompt that highlights an unusual combination of sensor-measurements in the event. Such prompts help clarify potential problems and their causes. Referring to the accompanying drawings, Figure 1 FIG. 1 is a diagram of an exemplary prompt of a symbolic event in an asset involving a temperature sensor and a vibration sensor according to an embodiment of the present invention. Assume that the event evolves around a compressor in the asset, such as Figure 1 As shown in . An aggregation of many evolving alarms occurs, the event may involve multiple sensors and components, and a person with technical knowledge should inspect the compressor for repair.

[0059] To aid in this task, the system emphasizes the most unusual sensor-measurement combinations in an event, which are called cues. Figure 1 Such a prompt is depicted which highlights to the technician the change in the operation of the East compressor around the "May 27" data entry in the relative temperature-vibration behavior.

[0060] In one embodiment of the invention, potential prompts are created by listing all combinations of elements in an event and their elements that are relevant to an asset representation. In another embodiment of the invention, potential prompts are created by applying template guidelines, rules, priorities, and / or suggestions to elements in an event that are also included in a predefined template. For each potential prompt, a number of attributes (including but not limited to statistical attributes about the measurement results) are generated. Such potential prompts are scored, and the highest scores are presented as prompts about the event.

[0061] In one use case, the score of a potential prompt is determined by a predefined function of its attributes. In yet another use case, feedback on the usefulness of each prompt can be provided, and machine learning functionality can be activated to learn the function of the attributes that score the potential prompts, so that high-scoring prompts are more likely to be helpful.

[0062] In yet another embodiment, the scores of potential prompts, including elements that are also part of the template, are influenced by the template definitions, rules, and / or recommendations. In yet another embodiment, the scores are influenced by the user's responses to previously presented prompts in that event or in a different event. In another use case, the prompts are presented in an order configured to improve overall event analysis, while in another use case, the prompts are presented in an order designed to speed up the resolution of a specific event.

[0063] The system enables definition of calculated indicators that are configured to calculate, predict or estimate indicated asset behavior based on sensors or measurements in / of the asset.

[0064] In one use case, the input may include a method or function for deriving a calculated metric that specifies an identified asset behavior based on measurements from the asset sensors. An example of such a designation of an identified event is designating a boiler alarm as "occurring" when the temperature measured by sensor TMP 503 is above 400°C, or when the pressure measured by sensor PRS 503 is above 10 atmospheres.

[0065] In yet another use case, the system can automatically create calculated metrics, such as by assigning identified asset behaviors to specific events. The system can automatically connect identified behavioral elements to entities that are important in the asset representation related to the event.

[0066] In yet another use case, the associated calculated metric may also include a recommendation for action in the event of such an identified behavior (e.g., in the event of a boiler alarm, reduce pressure to 8 atmospheres). In another calculated metric, the identified event is positive (e.g., the asset is in optimal production).

[0067] In another such use case, the inputs defining the identified asset behaviors may include examples of historical periods when the identified behaviors occurred and did not occur. User tags may be accumulated through interactive interactions. The automatically created calculated indicators use machine learning methods to learn based on the accumulated historical periods to predict the occurrence of the identified behaviors.

[0068] Figure 2 is a diagram of an exemplary simplified graphical user interface (GUI) for collecting user tags for occurrences of specific time periods of identified asset behavior according to an embodiment of the present invention. The system presents the selected time periods and the associated event tags generated by the system for those time periods to the asset expert, and allows the asset expert to correct, refine, change or approve the presented tags of the identified behavior. GUI 2 depicts a diagram of a track 4 showing the readings of the sensors that the user examines in order to tag the identified behavior.

[0069] The sensor is identified by index 6 at the bottom of the GUI 2 as (for example) sensor TMP 301. The user can load additional sensors for inspection by writing their names in the text boxes in the sensor panel entry 8. Panels 10, 12 and 14 indicate the relevant labels currently generated by the system for each sub-period, and panels 10 and 14 indicate that the system predicts that the sub-period will exhibit the identified behavior (depicted by a "√" in the indicator box). Panel 12 indicates that the system predicts that the identified behavior will not be exhibited in this sub-period (depicted by an "×" in the indicator box). The user can change the labels in each sub-panel to switch between identified and unidentified behaviors.

[0070] Arrows 16 enable the user to change the position of the borders of various panels, and panel buttons 18, 20, and 22 enable the user to add a panel with a "relative" label in the middle of an existing sub-panel (e.g., a sub-panel marked with a "√" can have an area labeled "×" inside it, and vice versa). When the user presses the "Approve" button 24, the interaction with the GUI 2 (i.e., the user experience) ends, and the labels in the GUI 2 are assigned to the correct user labels. The system receives the user labels and generates an improved predictor for the identified behavior. The system then presents another time period to the user for labeling. The user can ignore the request to label. However, the more precise the labeling time period provided by the user, the better the definition of the identified behavior.

[0071] In one use case, the order of time periods presented for label refinement, correction, or approval is determined randomly. In another use case, the order of the time periods presented is configured to show similar patterns (e.g., to encourage user interaction). In yet another use case, the order of the time periods presented is determined based on the uncertainty of identifying events in these time periods. In another use case, the time periods presented are selected after checking the differences in the predictions of the calculated indicators when the identified behavior labels in the time periods are assumed to show the identified events (“√” in the indication box) or when the identified event labels are assumed to not show the identified events (“×” in the indication box). Therefore, the presented time periods represent the differences in labeling that have the greatest impact on the calculated indicator results. In yet another use case, the order of the presented time periods is determined based on the likelihood that the presented time periods will result in the greatest reduction in the total uncertainty of event identification.

[0072] The system may ask the user to add an element representing the identified behavior and connect the element to the asset representation, particularly the process that directly affects the identified behavior. In yet another use case, the system may automatically connect the identified behavior element to the entity containing the sensor that is considered or observed by the user when defining the calculated metric. For example, in Figure 2 GUI 2, an asset expert can add sensors to the view via the sensor panel entry 8 to better understand the system over a given period of time, resulting in better user-label assignment of identified behaviors.

[0073] For all types of calculated sensors (e.g., function-defined, period-defined, or system-defined calculated metrics), the system can collect all indicated sensors, connect the sensors (and entities connected to the sensors) to elements of the identified behavior in the asset representation, and / or increase the weight of those links. After the elements of the identified behavior are connected to entities that have been embedded in the asset representation, the system can consider some entities connected to those entities (e.g., up to a certain distance) as affecting the calculated metric, and consider the measurements of the sensors in those entities as inputs to the calculated metric.

[0074] The system can activate the above-mentioned verification and cleaning procedures by analyzing the relevant sensor data to remove connections from non-influencing elements to the identified behavior elements, and select a subset of relevant sensors that influence the identified behavior. Ultimately, the system can add all of these elements or only the influencing elements to the input of the relevant calculated indicator, which learns to identify the identified behavior (e.g., by using a machine learning algorithm).

[0075] The calculated metrics are used to investigate and improve the operational aspects represented by the identified behaviors. The system provides hints in the form of potential ways to improve the operational aspects through the following capabilities.

[0076] A. By comparing the contributing factors generated for the calculated indicators for the occurrence of the identified behaviors with those generated for the non-occurrence. In both cases, the relevant contributing factors are derived from the asset representation, sensor measurements, and cleaning processes performed individually for each of these behaviors.

[0077] B. By operating the system during a transition period where the system changes from expected behavior to unexpected behavior or vice versa, a relevant alarm aggregation system (e.g., as described in U.S. Patent No. 10318364B2) can generate relevant events, their main elements, and possible related prompts as described above, which enables investigation, understanding and focus on how to improve the expected operating aspects and identification of the factors that cause the impact.

[0078] C. Train a machine learning algorithm to predict states where the identified behavior occurs and does not occur. The important element in the trained machine learning model is the distinction between these two states, and is the main contributing factor to the identified behavior.

[0079] When the major factors influencing the identified behavior are identified, relevant prompts are generated as described above and presented to the user.

[0080] While the invention has been described with respect to a limited number of embodiments, it will be appreciated that many variations, modifications and other applications of the invention are possible.

Claims

1. A method for identifying and improving asset behavior based on an interconnected representation of integrated physical operations, the method comprising the steps of: (a) providing an asset representation of an asset, wherein the asset is an integrated physical system having a primary activity; wherein objects are physical items; wherein processes are non-physical items that affect at least one object; wherein entities are physical objects and / or processes of the asset; and wherein the entities are adapted to act together to achieve the primary activity; and wherein the asset representation comprises: (i) a sensor list having all relevant sensors, all relevant measurements, and / or all relevant sensor data columns in a database related to real-world data measured and / or detected in the asset; (ii) an object list having at least two related objects in the asset; (iii) a process list having at least two associated processes in the asset; (iv) an entity connection group obtained by associating list elements in the sensor list, the object list and the process list; wherein all relevant objects and all relevant processes in the asset are appropriately listed and correspondingly associated to the asset representation of the asset; (b) creating a list of calculated indicators having at least one calculated indicator for the asset, wherein each calculated indicator is a metric configured to calculate, predict, or estimate an identified asset behavior based on evaluating the list of sensors in the asset, wherein the identified asset behavior is defined as a specific behavior of the asset, and wherein each calculated indicator is directly affected by the at least two related processes; (c) merging the list of calculated indicators into the entity connection group by associating list elements in the list of calculated indicators, wherein all related objects and all related processes in the asset are appropriately listed and correspondingly associated into the asset representation of the asset, thereby generating the interconnected representation; (d) associating the identified asset behavior with at least one of the related objects and / or at least one of the related processes that impairs the asset such that the asset is unable to optimally perform, conduct and / or achieve the identified asset behavior; and (e) generating at least one alert associated with at least one of the related objects and / or at least one of the related processes that compromise the identified asset behavior, wherein the at least one alert highlights an abnormal combination of sensor-measurement results in the asset.

2. The method according to claim 1, wherein: The at least one calculated metric is defined as a function of a subset of the sensor list elements.

3. The method according to claim 1, further comprising the steps of: (f) calculating a score for each combination of the sensor list elements using a predefined function, wherein the combination is related to the identified asset behavior; and (g) prioritizing the at least one prompt based on the score.

4. The method according to claim 1, wherein: The step of creating includes using at least one time period during which the identified asset behavior occurred or did not occur to create the at least one calculated indicator.

5. The method according to claim 1, further comprising the steps of: (f) comparing a first given asset representation having the identified asset behavior with a second given asset representation lacking the identified asset behavior to produce a comparative representation; as well as (g) associating the at least one prompt with at least one list element appearing in the comparison representation.

6. The method according to claim 1, further comprising the steps of: (f) training a machine learning algorithm to predict the state of the identified asset behavior represented by the asset, wherein the training step also identifies related list elements.

7. The method according to claim 1, wherein: The step of merging includes extending the evaluation of the sensor list using the asset representation as an input to the at least one calculated metric.

8. The method according to claim 1, further comprising the steps of: (f) verifying the set of physical connections represented by the asset by analyzing the real-world data obtained by the associated sensors; and (g) cleaning the entity connection set of the asset representation by retaining connections with direct influence in the entity connection set and by removing connections that are remote and / or have indirect influence from the entity connection set.

9. The method according to claim 1, further comprising the steps of: (f) collecting user feedback regarding the at least one prompt; (g) learning a prompt scoring function based on the user feedback; (h) calculating a score for each combination of the sensor list elements using the prompt scoring function, wherein the combination is related to the identified asset behavior; and (i) prioritizing the at least one prompt based on the score.

10. The method according to claim 1, further comprising the steps of: (f) modifying the at least one prompt based on at least a template of the asset representation.

11. A system for identifying and improving asset behavior based on an interconnected representation of integrated physical operations, the system comprising: (a) a CPU for performing computational operations; (b) a memory module for storing data; (c) an asset representation module configured to: (i) providing an asset representation of an asset, wherein the asset is an integrated physical system having a primary activity; wherein objects are physical items; wherein processes are non-physical items that affect at least one object; wherein entities are physical objects and / or processes of the asset; and wherein the entities are adapted to act together to achieve the primary activity; and wherein the asset representation comprises: (A) a sensor list having all relevant sensors, all relevant measurements, and / or all relevant sensor data columns in a database related to real-world data measured and / or detected at the asset; (B) a list of objects having at least two related objects in the asset; (C) a process list having at least two associated processes in the asset; (D) an entity connection group obtained by associating list elements in the sensor list, the object list, and the process list; wherein all related objects and all related processes in the asset are appropriately listed and correspondingly associated to the asset representation of the asset; (d) a calculated indicator module configured to create a list of calculated indicators, the list of calculated indicators having at least one calculated indicator for the asset, wherein each calculated indicator is a metric configured to calculate, predict or estimate an identified asset behavior based on evaluating the list of sensors in the asset, wherein the identified asset behavior is defined as a specific behavior of the asset, and wherein each calculated indicator is directly affected by the at least two related processes; (e) an indicator connection module configured to merge the list of calculated indicators into the entity connection group by associating list elements in the list of calculated indicators, wherein all related objects and all related processes in the asset are appropriately listed and correspondingly associated into the asset representation of the asset, thereby generating the interconnected representation; (f) an identified behavior module configured to associate the identified asset behavior with at least one of the related objects and / or at least one of the related processes that impairs the asset such that the asset is unable to optimally perform, conduct and / or achieve the identified asset behavior; and (g) a prompt generation module configured to generate at least one prompt associated with at least one of the related objects and / or at least one of the related processes that compromise the identified asset behavior, wherein the at least one prompt highlights an abnormal combination of sensor-measurement results in the asset.

12. The system according to claim 11, wherein: The at least one calculated metric is defined as a function of a subset of the sensor list elements.

13. The system according to claim 11, further comprising: (h) a score calculation module configured to calculate a score for each combination of the sensor list elements using a predefined function, wherein the combination is related to the identified asset behavior; and (i) A prompt prioritization module configured to prioritize the at least one prompt based on the score.

14. The system according to claim 11, wherein: The calculated indicator module is further configured to create the at least one calculated indicator using at least one time period during which the identified asset behavior occurred or did not occur.

15. The system according to claim 11, further comprising: (h) a representation comparison module configured to compare a first given asset representation having the identified asset behavior with a second given asset representation lacking the identified asset behavior to produce a comparison representation; as well as (i) A prompt correlation module configured to correlate the at least one prompt with at least one list element appearing in the comparison representation.

16. The system according to claim 11, further comprising: (h) a training module configured to train a machine learning algorithm to predict a state of the identified asset behavior represented by the asset, wherein the training also identifies related list elements.

17. The system of claim 11, wherein: The metric connection module is further configured to extend the evaluation of the sensor list using the asset representation as an input to the at least one calculated metric.

18. The system according to claim 11, further comprising: (h) a data verification module configured to verify the set of entity connections represented by the asset by analyzing the real-world data obtained by the associated sensors; and (i) a connection cleaning module configured to clean the entity connection group of the asset representation by retaining connections with direct influence in the entity connection group and by removing connections that are remote and / or have indirect influence from the entity connection group.

19. The system of claim 11, wherein the prompt generation module is further configured to: collecting user feedback regarding the at least one prompt; learning a prompt scoring function based on the user feedback; The score of each combination of the sensor list elements is calculated using the prompt scoring function, wherein, The combination is related to the identified asset behavior; and The at least one prompt is prioritized based on the score.

20. The system of claim 11, the prompt generation module further configured to modify the at least one prompt based on at least a template of the asset representation.

21. A non-transitory computer readable storage medium having computer readable code embodied thereon for identifying and improving asset behavior based on an interconnected representation of integrated physical operations, the computer readable code comprising: (a) Program code for providing an asset representation of an asset, wherein the asset is an integrated physical system having a primary activity; wherein objects are physical items; wherein processes are non-physical items that affect at least one object; wherein entities are physical objects and / or processes of the asset; and wherein the entities are adapted to act together to achieve the primary activity; and wherein the asset representation comprises: (i) a sensor list having all relevant sensors, all relevant measurements, and / or all relevant sensor data columns in a database related to real-world data measured and / or detected in the asset; (ii) an object list having at least two related objects in the asset; (iii) a process list having at least two associated processes in the asset; (iv) an entity connection group obtained by associating list elements in the sensor list, the object list and the process list; wherein all relevant objects and all relevant processes in the asset are appropriately listed and correspondingly associated to the asset representation of the asset; (b) program code for creating a list of calculated indicators, the list of calculated indicators having at least one calculated indicator for the asset, wherein each calculated indicator is a metric configured to calculate, predict, or estimate an identified asset behavior based on evaluating the list of sensors in the asset, wherein the identified asset behavior is defined as a specific behavior of the asset, and wherein each calculated indicator is directly affected by the at least two related processes; (c) program code for merging the list of calculated indicators into the entity connection group by associating list elements in the list of calculated indicators, wherein all related objects and all related processes in the asset are appropriately listed and correspondingly associated into the asset representation of the asset, thereby generating the interconnected representation; (d) program code for associating the identified asset behavior with at least one of the related objects and / or at least one of the related processes that impairs the asset so that the asset cannot optimally perform, conduct and / or achieve the identified asset behavior; and (e) program code for generating at least one alert associated with at least one of the related objects and / or at least one of the related processes that compromise the identified asset, wherein the at least one alert highlights an abnormal combination of sensor-measurement results in the asset.

22. The non-transitory computer-readable storage medium of claim 21, wherein: The at least one calculated metric is defined as a function of a subset of the sensor list elements.

23. The non-transitory computer readable storage medium of claim 21, the computer readable code further comprising: (f) program code for calculating a score for each combination of the sensor list elements using a predefined function, wherein the combination is related to the identified asset behavior; and (g) Program code for prioritizing the at least one prompt based on the score.

24. The non-transitory computer-readable storage medium of claim 21, wherein: Creating includes creating the at least one calculated indicator using at least one time period during which the identified asset behavior occurred or did not occur.

25. The non-transitory computer readable storage medium of claim 21, the computer readable code further comprising: (f) program code for comparing a first given asset representation having the identified asset behavior with a second given asset representation lacking the identified asset behavior to produce a comparison representation; as well as (g) program code for associating the at least one prompt with at least one list element appearing in the comparison representation.

26. The non-transitory computer readable storage medium of claim 21, the computer readable code further comprising: (f) Program code for training a machine learning algorithm to predict states of the identified asset behavior represented by the asset, wherein the training also identifies related list elements.

27. The non-transitory computer-readable storage medium of claim 21, wherein: Merging includes extending the evaluation of the sensor list using the asset representation as an input to the at least one calculated metric.

28. The non-transitory computer readable storage medium of claim 21, the computer readable code further comprising: (f) program code for verifying the set of physical connections represented by the asset by analyzing the real-world data obtained by the associated sensors; as well as (g) Program code for cleaning the entity connection group of the asset representation by retaining connections with direct impact in the entity connection group and by removing connections that are remote and / or have indirect impact from the entity connection group.

29. The non-transitory computer readable storage medium of claim 21, the computer readable code further comprising: (f) program code for collecting user feedback regarding the at least one prompt; (g) program code for learning a prompt scoring function based on the user feedback; (h) program code for calculating a score for each combination of the sensor list elements using the prompt scoring function, wherein the combination is related to the identified asset behavior; and (i) Program code for prioritizing the at least one prompt based on the score.

30. The non-transitory computer readable storage medium of claim 21, the computer readable code further comprising: (f) Program code for modifying the at least one prompt based on at least a template of the asset representation.

Images