Propagating authored packages to asset management platforms
Techniques for using an LLM agent to predict a state of an asset are disclosed. The LLM agent uses bitemporal modeling to track status information of an asset. The status information includes an uptime or a downtime of the asset. The LLM agent is trained to detect anomalies with respect to at least one of the uptime or the downtime of the asset and to predict uptimes and downtimes. The LLM agent generates a prediction regarding a future uptime or a future downtime of the asset. Subsequently, the LLM agent collects data reflecting an actual uptime or an actual downtime of the asset during the given range of time. The LLM agent makes a comparison between the prediction and the actual data and then further trains or instructs the LLM agent based on this comparison.
1 . A method for using a machine learning (ML) model to predict a state of an asset, the method comprising:
using bitemporal modeling to track status information of an asset, wherein the status information includes at least one of an uptime or a downtime of the asset, and wherein the bitemporal modeling includes a first record associated with a current state of the asset's status information and a second record associated with the asset's status information at a selected point in time earlier than the current state;
feeding the tracked status information used by the bitemporal modeling to an ML model, wherein the ML model, during a first training stage, is trained to detect anomalies with respect to at least one of the uptime or the downtime of the asset and to generate predictions regarding uptimes and downtimes;
causing the ML model to generate a prediction regarding a future uptime or a future downtime of the asset, the future uptime or the future downtime being predicted for a given range of time that is to subsequently transpire;
subsequently, collecting data reflecting an actual uptime or an actual downtime of the asset during the given range of time;
comparing the ML model's prediction of the future uptime or the future downtime of the asset with the data reflecting the actual uptime or the actual downtime of the asset during the given range of time;
further training the ML model based on said comparison, such that the ML model is subjected to a second training stage to correct for discrepancies between the ML model's prediction of the future uptime or the future downtime of the asset and the data reflecting the actual uptime or the actual downtime of the asset;
using the trained ML model to identify at least one of: a trend and a pattern;
generating an updated package based on the identified at least one of: a trend and a pattern, wherein the updated package includes updated information on at least one of: operation, maintenance, and use of an asset; and
causing the updated package to be installed on the asset to change operational behavior of the asset to reflect contents of the updated package.
2 . The method of claim 1 , wherein the bitemporal modeling provides an auditable log for the asset.
3 . The method of claim 1 , wherein the status information is obtained from a plurality of different sources, including at least one text source and at least one image source.
4 . The method of claim 1 , wherein the ML model determines a standard operating range for the asset, and wherein the ML model determines whether the tracked status information includes data representative of anomalous behavior for the asset.
5 . The method of claim 1 , wherein the prediction includes a recommendation generated by the ML model, the recommendation including a recommended maintenance event for the asset.
6 . The method of claim 1 , wherein the prediction includes a recommendation by the ML model, the recommendation including a recommended part replacement event.
7 . The method of claim 1 , wherein the status information includes at least one of: sensor data obtained from the asset, surface reconstruction data for the asset, LIDAR data for the asset, or image data for the asset.
8 . The method of claim 1 , further comprising rendering a custom view on a user interface, wherein the custom view is based at least in part on the identified at least one of: a trend and a pattern.
9 . A computer system that uses a machine learning (ML) model to predict a state of an asset, the computer system comprising:
a processor system; and
a storage system that stores instructions that are executable by the processor system to cause the computer system to:
use bitemporal modeling to track status information of an asset, wherein the status information includes at least one of an uptime or a downtime of the asset, and wherein the bitemporal modeling includes a first record associated with a current state of the asset's status information and a second record associated with the asset's status information at a selected point in time earlier than the current state;
feed the tracked status information used by the bitemporal modeling to the ML model, which is trained to detect anomalies with respect to at least one of the uptime or the downtime of the asset;
cause the ML model to generate a prediction regarding a future uptime or a future downtime of the asset, the future uptime or the future downtime being predicted for a given range of time that is to subsequently transpire;
subsequently, collect data reflecting an actual uptime or an actual downtime of the asset during the given range of time;
compare the ML model's prediction of the future uptime or the future downtime of the asset with the data reflecting the actual uptime or the actual downtime of the asset during the given range of time;
further train ML model based on said comparison, such that the ML model is further instructed to correct for discrepancies between the ML model's prediction of the future uptime or the future downtime of the asset and the data reflecting the actual uptime or the actual downtime of the asset;
using the trained ML model to identify at least one of: a trend and a pattern; and
generating an updated package based on the identified at least one of: a trend and a pattern, wherein the updated package includes updated information on at least one of: operation, maintenance, and use of an asset; and
causing the updated package to be installed on the asset to change operational behavior of the asset to reflect contents of the updated package.
10 . The computer system of claim 9 , wherein the ML model is further instructed in an attempt to reduce false positives as to whether operations of the asset conform with a set of governing criteria.
11 . The computer system of claim 9 , wherein the ML model is further configured to analyze multiple, disparate sets of data and to identify relationships between those multiple, disparate sets of data.
12 . The computer system of claim 9 , wherein the ML model is further configured to analyze (i) a procedure for the asset, (ii) image data of the asset, and (iii) sensor data for the asset, and wherein the ML model identifies one or more relationships between the procedure, the image data, and the sensor data.
13 . The computer system of claim 9 , wherein comparing the ML model's prediction with the data includes overlaying data corresponding to the prediction onto the data reflecting the actual uptime or the actual downtime onto a common plot and then identifying deviations.
14 . The computer system of claim 9 , wherein comparing the ML model's prediction with the data includes performing one or more data slicing actions on the ML model's prediction and/or the data, resulting in a modified data set.
15 . The computer system of claim 9 , wherein the ML model tracks when the asset operates outside of an established range, and wherein the ML model determines a reason as to why the asset is operating outside of the established range.
16 . The computer system of claim 9 , wherein the status information of the asset is obtained from one or more of the following: image data, computer vision data, graph data, or telemetry data.
17 . The computer system of claim 9 , wherein the ML model is further configured to:
perform optical character recognition (OCR) on a procedure for the asset, wherein the OCR identifies a particular characteristic for the asset;
acquire sensor data for the asset, wherein the sensor data relates to the particular characteristic of the asset; and
generate a relationship between the particular characteristic identified from performing the OCR and the acquired sensor data.
18 . A method for using a machine learning (ML) model to predict a state of an asset, the method comprising:
using bitemporal modeling to track status information of an asset, wherein the status information includes at least one of an uptime or a downtime of the asset, and wherein the bitemporal modeling includes a first record associated with a current state of the asset's status information and a second record associated with the asset's status information at a selected point in time earlier than the current state;
feeding the tracked status information used by the bitemporal modeling to an ML model, wherein the ML model, during a first training stage, is trained to detect anomalies with respect to at least one of the uptime or the downtime of the asset;
causing the ML model to generate a prediction regarding a future uptime or a future downtime of the asset, the future uptime or the future downtime being predicted for a given range of time that is to subsequently transpire;
subsequently during the given range of time, collecting data reflecting an actual uptime or an actual downtime of the asset during the given range of time, wherein the collected data includes sensor data obtained for the asset;
comparing the ML model's prediction of the future uptime or the future downtime of the asset with the data reflecting the actual uptime or the actual downtime of the asset during the given range of time, wherein said comparing is performed using a chart that identifies deviations between the actual uptime or the actual downtime and the predicted further uptime or the predicted further downtime;
further training the ML model based on said comparison, such that the ML model is subjected to a second training stage to correct for the deviations between the ML model's prediction of the future uptime or the future downtime of the asset and the data reflecting the actual uptime or the actual downtime of the asset;
using the trained ML model to identify at least one of: a trend and a pattern; and
generating an updated package based on the identified at least one of: a trend and a pattern, wherein the updated package includes updated information on at least one of: operation, maintenance, and use of an asset; and
causing the updated package to be installed on the asset to change operational behavior of the asset to reflect contents of the updated package.
19 . The method of claim 18 , wherein the bitemporal modeling provides an auditable log for the asset, and wherein the status information is obtained from a plurality of different sources.
20 . The method of claim 18 , wherein the ML model determines a standard operating range for the asset, and wherein the ML model determines whether the tracked status information includes data representative of anomalous behavior for the asset.
21 . The method of claim 18 , wherein the prediction includes a recommendation generated by the ML model, the recommendation including a recommended maintenance event for the asset.