Fluid inspection using machine learning

Patent metadata

Field	Value
Publication number	US-12461087-B2
Application number	US-202217893834-A
Country	US
Kind code	B2
Filing date	Aug 23, 2022
Priority date	Aug 23, 2022
Publication date	Nov 4, 2025
Grant date	Nov 4, 2025

Abstract

A method includes determining, using a processing device, a set of observations from coolant data, the coolant data being received from one or more sensors in an environment associated with a coolant. The method further includes determining, using a machine learning model and the set of observations, a contamination level of the coolant. The method also includes initiating an operation, using the processing device, responsive to determining the coolant contamination level.

First claim

What is claimed is:

1 . A method, comprising:
determining, using a processing device, a set of observations from coolant data received from one or more sensors in an environment associated with a coolant in a datacenter cooling system, the set of observations comprising at least one of: a fluid turbidity measurement, a pressure measurement, a conductivity measurement, or a potential hydrogen (pH) level measurement; determining, using the processing device, performance data including at least one of power consumption measurements, temperature measurements, or clock frequency measurements of one or more computing devices; processing the set of observations with the performance data using a machine learning model that determines whether the set of observations matches a contaminated coolant profile or an uncontaminated coolant profile and outputs a contamination level of the coolant based on a result of the processing; and initiating predictive maintenance of the datacenter cooling system, using the processing device, responsive to determining the coolant contamination level and that the coolant data matches a contaminated coolant profile.

2 . The method of claim 1 , further comprising:
receiving, using the processing device, second coolant data from the one or more sensors; determining, using the processing device, a second set of observations from the second coolant data; determining, using the machine learning model and the second set of observations, a second contamination level of the coolant; and refraining from initiating the predictive maintenance responsive to detecting the coolant is uncontaminated.

3 . The method of claim 1 , wherein the one or more sensors include at least one of a light spectroscopy sensor, a fluid turbidity sensor, a pressure sensor, a potential hydrogen (pH) sensor, or a conductivity sensor.

4 . The method of claim 1 , further comprising:
determining, using the processing device, a second set of observations from a second set of data, the second set of data including network information wherein detecting whether the coolant is contaminated or uncontaminated is based on at least in part using the second set of observations.

5 . The method of claim 1 , wherein the machine learning model comprises one of a classification model, a feature detection model, an anomaly detection model, or a pattern recognition model trained to detect whether the coolant is contaminated or uncontaminated.

6 . The method of claim 1 , wherein the predictive maintenance comprises one of transmitting an alert the coolant is contaminated, initiating a contamination analysis, determining a temperature of one or more devices of the environment, initiating a fluid analysis, or determining a thermal resistance of one or more devices associated with the coolant.

7 . The method of claim 1 , further comprising:
determining, using the machine learning model and the set of observations, that the coolant will become contaminated responsive to determining the set of observations.

8 . The method of claim 1 , wherein the coolant data collected by the one or more sensors is received by the processing device remotely.

9 . A system comprising:
a processing device to: determine a set of observations from coolant data being received from one or more sensors in a computing environment associated with a coolant in a datacenter cooling system, the set of observations comprising at least one of: a fluid turbidity measurement, a pressure measurement, a conductivity measurement, or a potential hydrogen (pH) level measurement; determine performance data including at least one of power consumption measurements, temperature measurements, or clock frequency measurements of one or more computing devices; process the set of observations with the performance data using a machine learning model that determines whether the set of observations matches a contaminated coolant profile or an uncontaminated coolant profile and outputs a contamination level of the coolant based on a result of processing the set of observations; and initiate predictive maintenance of the datacenter cooling system, using the processing device, responsive to determining the coolant contamination level and that the coolant data matches a contaminated coolant profile.

10 . The system of claim 9 , wherein the processing device is further to:
receive second coolant data from the one or more sensors; determine a second set of observations from the second coolant data; determine, using the machine learning model and the second set of observations, a second contamination level of the coolant; and refrain from initiating the predictive maintenance responsive to determining the coolant is uncontaminated.

11 . The system of claim 9 , wherein the one or more sensors include at least one of a light spectroscopy sensor, a fluid turbidity sensor, a pressure sensor, a potential hydrogen (pH) sensor, or a conductivity sensor.

12 . The system of claim 9 , wherein the predictive maintenance comprises one of transmitting an alert the coolant is contaminated, initiating a contamination analysis, determining a temperature of one or more devices of the computing environment, initiating a fluid analysis, or determining a thermal resistance of one or more devices associated with the coolant.

13 . The system of claim 9 , wherein the coolant data collected by the one or more sensors is received by the processing device remotely.

14 . A method, comprising:
determining, using a processing device, a set of observations from coolant data received from one or more sensors associated with a first portion of a computing environment associated with a coolant in a datacenter cooling system, the set of observations comprising at least one of: a fluid turbidity measurement, a pressure measurement, a conductivity measurement, or a potential Hydrogen (pH) level measurement; determining, using the processing device, performance data including at least one of power consumption measurements, temperature measurements, or clock frequency measurements of one or more computing devices; training, using the set of observations and the performance data, a machine learning (ML) model to determine a cooling efficiency and power efficiency associated with the first portion; determining, using the ML model, the cooling efficiency and power efficiency associated with a second portion of the computing environment; and determining whether to initiate predictive maintenance of the datacenter cooling system at the first portion or the second portion responsive to determining the cooling efficiency and power efficiency associated with the second portion.

15 . The method of claim 14 , wherein the computing environment comprises a plurality of portions, including the first portion and the second portion, and the method further comprises:
determining, using the ML model, a second cooling efficiency and second power efficiency associated with each portion of the plurality of portions of the computing environment; and determining a third portion to perform a second operation responsive to determining the second cooling efficiency and power efficiency associated with each portion of the plurality of portions.

16 . The method of claim 14 , further comprising;
determining, using the processing device, a second set of observations from a second coolant data, the second coolant data being received from a second set of one or more sensors associated with the second portion of the computing environment, wherein determining the cooling efficiency and power efficiency

Assignees

• Nvidia Corp

Inventors

• Albright Ryan Kelsey
• Mecham William Andrew
• Cader Tahir
• Thompson Michael Scott
• Carkin Aaron Richard
• Weese William Ryan
• Goska Benjamin Joseph
• Davis Marc

Classifications

• G06N20/00

  Machine learning · CPC title

• G01N21/9072

  with illumination or detection from inside the container · CPC title

• G01N11/02

  by measuring flow of the material · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title