Granted Patent B2

US 12,367,417 · App. 16/921,951 · Granted Jul 22, 2025

Inventors: Subanatarajan Subbiah (Neulussheim, DE); Benjamin Kloepper (Mannheim, DE)

Assignee: ABB Schweiz AG

G06N20/00G05B23/024

Abstract

Claims (23)

1. A computer-implemented method for assessing 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, the method comprising:

obtaining at least one record of measurement data that describes an operating situation of the equipment, and/or of the process;

obtaining a plurality of trained machine-learning models that are each configured to map the record of measurement data to an indication of the condition;

determining, for each of the trained machine-learning models, a figure of fitness to assess a situation described by the record of measurement data by determining a measure for a type and/or a strength of any disturbance contained in the measurement data and determining a fitness of the trained machine-learning model to process measurement data that contain disturbances of the type and/or strength;

selecting from the trained machine-learning models whose figure of fitness meets a predetermined criterion a simple model and a computationally intensive model, wherein the selection of the trained machine-learning models is further based on one or more of a requirement for accuracy and for a confidence of the assessment of the condition; and

mapping the record of measurement data to a sought condition using the selected trained machine-learning models by cascading the plurality of trained machine-learning models in that:

in an initial screening of the plant, a fast cursory assessment of the record of measurement data is performed by the simple model; and

in response to the simple model indicating that the condition may be abnormal, the computationally intensive model that works on complex measurement data is applied and used to further investigate the condition, and

in response to the abnormal condition meeting a predetermined criterion, taking one or more corrective actions including:

flagging the equipment for maintenance and/or repair, and/or for future more frequent inspection;

changing operating parameters of the equipment to ease a load on the equipment;

changing operating parameters of the process to move the process to a stable state;

and/or shutting down the equipment, and/or the process, to prevent further damage.

2. The method of claim 1 , wherein the determining the measure for the quality of the measurement data comprises:

supplying the measurement data to a further trained machine-learning model that is configured to produce a classification value, and/or a regression value, as a measure for the quality.

3. The method of claim 1 , further comprising:

before obtaining the record of measurement data using a plurality of sensors, selecting a subset of the plurality of sensors according to a database that stores associations between equipment whose condition is to be assessed and sensors that are suitable for assessing the condition of this equipment.

4. The method of claim 3 , wherein selecting of the subset of the plurality of sensors is dependent on the sensors operating on a finite supply of a consumable.

5. The method of claim 1 , wherein an interval at which the obtaining of records of measurement data is repeated is determined based on a previous history of the assessed condition.

6. The method of claim 1 , further comprising: selecting at least one sensor from which to obtain further measurement data, and/or a piece of equipment to which further measurement data are to relate, based at least in part on an already determined plausibility, and/or on an already assessed condition.

7. The method of claim 1 , wherein the plurality of sensors comprises sensors mounted on a service robot that is configured to move between different pieces of equipment, and/or between different processes, in the industrial plant.

8. The method of claim 7 , further comprising:

determining a location of the service robot in the industrial plant, and/or an identification of the equipment that is being sensed by the sensors of the service robot, based on a match between measurement data captured by the sensors and pre-stored information about a layout of the industrial plant, and/or about an inventory of the equipment installed in the plant.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (50)