Systems and methods for providing dynamic representation of assets in a facility

A system comprising:

• a processor;
• and
• a memory communicatively coupled to the processor, wherein the memory comprises one or more instructions which when executed by the processor cause the processor to:
• determine a first vector for an asset from amongst a plurality of assets in a facility, wherein the first vector is derived from one or more static parameters associated with the asset, and wherein to determine the first vector for the asset, the processor is configured to:
• identify a hierarchical representation of the asset in the facility, wherein the hierarchical representation comprises the one or more static parameters associated with the asset, and wherein the one or more static parameters comprises:
• a location of the asset in the facility, a category to which the asset belongs in the facility, and one or more parameters measured by the asset in the facility;
• convert the hierarchical representation into one or more sequential representations, each of the one or more sequential representations comprise a relation of each of the one or more parameters with the location and the category of the asset;
• and
• convert each of the one or more sequential representations into a vector using a natural language processing (NLP) algorithm, wherein the vector comprises the one or more static parameters represented as one or more values;
• receive telemetry data from the asset, wherein the telemetry data comprises a plurality of readings from the asset at a plurality of timestamps, and wherein each reading from amongst the plurality of readings is received at a respective timestamp;
• determine a second vector for the asset at each of the plurality of timestamps, wherein the second vector comprises a product of the first vector and a corresponding reading at the respective timestamp;
• convolute the second vector at each of the plurality of timestamps with a machine learning algorithm to obtain a dynamic representation of the asset, wherein the convolution of the second vector comprises:
• determining a weight matrix based on the second vector for the asset at each of the plurality of timestamps;
• and
• training the machine learning model with the weight matrix;
• and
• predict by the machine learning algorithm, a maintenance requirement for the asset based on the dynamic representation.

A non-transitory, computer-readable storage medium having program instructions embodied thereon, the program instructions executable by a processor to cause the processor to:

• determine a first vector for an asset from amongst a plurality of assets in a facility, wherein the first vector is derived from one or more static parameters associated with the asset, and wherein to determine the first vector for the asset, the processor is configured to:
• identify a hierarchical representation of the asset in the facility, wherein the hierarchical representation comprises the one or more static parameters associated with the asset, and wherein the one or more static parameters comprises:
• a location of the asset in the facility, a category to which the asset belongs in the facility, and one or more parameters measured by the asset in the facility;
• convert the hierarchical representation into one or more sequential representations, each of the one or more sequential representations comprise a relation of each of the one or more parameters with the location and the category of the asset;
• and
• convert each of the one or more sequential representations into a vector using a natural language processing (NLP) algorithm, wherein the vector comprises the one or more static parameters represented as one or more values;
• receive telemetry data from the asset, wherein the telemetry data comprises a plurality of readings from the asset at a plurality of timestamps, and wherein each reading from amongst the plurality of readings is received at a respective timestamp;
• determine a second vector for the asset at each of the plurality of timestamps, wherein the second vector comprises a product of the first vector and a corresponding reading at the respective timestamp;
• convolute by a machine learning algorithm, the second vector at each of the plurality of timestamps with a machine learning algorithm to obtain a dynamic representation of the asset, wherein the convolution of the second vector comprises:
• determining a weight matrix based on the second vector for the asset at each of the plurality of timestamps;
• and
• training a machine learning algorithm with the weight matrix;
• and
• predict by the machine learning algorithm, a maintenance requirement for the asset based on the dynamic representation.