Predicting internal corrosion in gas flow lines using machine learning
In an example method, a system obtains first data indicating a plurality of properties of a plurality of gas flow lines. The properties include, for each of the gas flow lines (i) data representing a flow rate of a gas through that gas flow line, (ii) data representing a pressure of the gas in that gas flow line, and (iii) data representing an additive included in the gas in that gas flow line, such as a substance for inhibiting corrosion. For each of the gas flow lines, the system uses a computerized neural network to determine a risk of corrosion associated with that gas flow line based on the properties of that gas flow line, determines whether the metric for that gas flow line is greater than a threshold level, and if so, generates a notification for presentation to a user.
1. A method comprising:
obtaining, using one or more processors, a plurality of sets of training data, wherein each of the plurality of sets of training data comprises first data indicating a plurality of first properties of a plurality of first gas flow lines, wherein the plurality of first properties comprise, for each of the first gas flow lines:
data representing a flow rate of a gas through that first gas flow line during a respective observation period,
data representing a pressure of the gas in that first gas flow line during the observation period,
data representing a first additive included in the gas in that first gas flow line during the observation period, wherein the first additive comprises a substance for inhibiting corrosion in that first gas flow line, and
data representing whether that first gas flow line became corroded subsequent to the observation period;
training, using the one or more processors, a computerized neural network based on the plurality of sets of training data;
obtaining, using one or more processors, second data indicating a plurality of second properties of a plurality of second gas flow lines, wherein the plurality of second properties comprise, for each of the plurality of second gas flow lines:
data representing a flow rate of a gas through that second gas flow line,
data representing a pressure of the gas in that second gas flow line, and
data representing a second additive included in the gas in that second gas flow line, wherein the second additive comprises a substance for inhibiting corrosion in that second gas flow line;
for each of the plurality of second gas flow lines:
determining, using the one or more processors and the computerized neural network, a first metric representing a risk of corrosion associated with that second gas flow line based on the second properties of that second gas flow line,
determining, using the one or more processors, whether the first metric for that second gas flow line is greater than a threshold level, and
responsive to determining that the first metric for that second gas flow line is greater than the threshold level, generating, using the one or more processors, a notification indicating the first metric for that second gas flow line for presentation to a user to facilitate at least one of a repair or maintenance of that second gas flow line.
2. The method of claim 1 , wherein the plurality of second properties comprise, for each of the plurality of second gas flow lines:
a composition of the gas in that second gas flow line.
3. The method of claim 1 , wherein the plurality of second properties comprise, for each of the plurality of second gas flow lines:
a composition of that second gas flow line.
4. The method of claim 1 , further comprising:
determining that the second data comprises incomplete data regarding one or more of the plurality of second gas flow lines, and
in response to determining that the second data comprises incomplete data regarding the one or more of the plurality of second gas flow lines, modifying the second data to include additional data regarding the one or more of the plurality of second gas flow lines.
5. The method of claim 1 , further comprising:
determining that the second data comprises inaccurate data regarding one or more of the plurality of second gas flow lines, and
in response to determining that the second data comprises inaccurate data regarding the one or more of the plurality of second gas flow lines, modifying the second data to correct the inaccurate data regarding the one or more of the plurality of second gas flow lines.
6. The method of claim 1 ,
wherein the plurality of first properties comprise, for each of the plurality of first gas flow lines:
data representing an extent of corrosion in that first gas line.
7. The method of claim 1 , further comprising:
for each of the plurality of second gas flow lines, determining a degree of severity of the risk of corrosion associated with that second gas flow line.
8. The method of claim 7 , wherein the plurality of second gas flow lines comprise one or more hydrocarbon gas flow lines.
9. A system comprising:
one or more processors; and
one or more non-transitory computer readable media storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
obtaining a plurality of sets of training data, wherein each of the plurality of sets of training data comprises first data indicating a plurality of first properties of a plurality of first gas flow lines, wherein the plurality of first properties comprise, for each of the first gas flow lines:
data representing a flow rate of a gas through that first gas flow line during a respective observation period,
data representing a pressure of the gas in that first gas flow line during the observation period,
data representing a first additive included in the gas in that first gas flow line during the observation period, wherein the first additive comprises a substance for inhibiting corrosion in that first gas flow line, and
data representing whether that first gas flow line became corroded subsequent to the observation period;
training a computerized neural network based on the plurality of sets of training data:
obtaining second data indicating a plurality of second properties of a plurality of second gas flow lines, wherein the plurality of second properties comprise, for each of the plurality of second gas flow lines:
data representing a flow rate of a gas through that second gas flow line,
data representing a pressure of the gas in that second gas flow line, and
data representing a second additive included in the gas in that second gas flow line, wherein the additive comprises a substance for inhibiting corrosion in that second gas flow line; and
for each of the plurality of second gas flow lines:
determining, using the computerized neural network, a first metric representing a risk of corrosion associated with that second gas flow line based on the second properties of that second gas flow line,
determining whether the first metric for that second gas flow line is greater than a threshold level, and
responsive to determining that the first metric for that second gas flow line is greater than the threshold level, generating a notification indicating the first metric for that second gas flow line for presentation to a user to facilitate at least one of a repair or maintenance of that second gas flow line.
10. The system of claim 9 , wherein the plurality of second properties comprise, for each of the plurality of second gas flow lines:
a composition of the gas in that second gas flow line.
11. The system of claim 9 , wherein the plurality of second properties comprise, for each of the plurality of second gas flow lines:
a composition of that second gas flow line.
12. The system of claim 9 , the operations further comprising:
determining that the second data comprises incomplete data regarding one or more of the plurality of second gas flow lines, and
in response to determining that the second data comprises incomplete data regarding the one or more of the plurality of second gas flow lines, modifying the second data to include additional data regarding the one or more of the plurality of second gas flow lines.
13. The system of claim 9 , further comprising:
determining that the second data comprises inaccurate data regarding one or more of the plurality of second gas flow lines, and
in response to determining that the second data comprises inaccurate data regarding the one or more of the plurality of second gas flow lines, modifying the second data to correct the inaccurate data regarding the one or more of the plurality of second gas flow lines.
14. The system of claim 9 ,
wherein the plurality of first properties comprise, for each of the plurality of first gas flow lines:
data representing an extent of corrosion in that first gas line.
15. The system of claim 9 , the operations further comprising:
for each of the plurality of second gas flow lines, determining a degree of severity of the risk of corrosion associated with that second gas flow line.
16. The system of claim 15 , wherein the plurality of second gas flow lines comprise one or more hydrocarbon gas flow lines.
17. One or more non-transitory computer readable media storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
obtaining a plurality of sets of training data, wherein each of the plurality of sets of training data comprises first data indicating a plurality of first properties of a plurality of first gas flow lines, wherein the plurality of first properties comprise, for each of the first gas flow lines:
data representing a flow rate of a gas through that first gas flow line during a respective observation period,
data representing a pressure of the gas in that first gas flow line during the observation period,
data representing a first additive included in the gas in that first gas flow line during the observation period, wherein the first additive comprises a substance for inhibiting corrosion in that first gas flow line, and
data representing whether that first gas flow line became corroded subsequent to the observation period;
training a computerized neural network based on the plurality of sets of training data;
obtaining second data indicating a plurality of second properties of a plurality of second gas flow lines, wherein the plurality of second properties comprise, for each of the plurality of second gas flow lines:
data representing a flow rate of a gas through that second gas flow line,
data representing a pressure of the gas in that second gas flow line, and
data representing a second additive included in the gas in that second gas flow line, wherein the additive comprises a substance for inhibiting corrosion in that second gas flow line; and
for each of the plurality of second gas flow lines:
determining, using the computerized neural network, a first metric representing a risk of corrosion associated with that second gas flow line based on the second properties of that second gas flow line,
determining whether the first metric for that second gas flow line is greater than a threshold level, and
responsive to determining that the first metric for that second gas flow line is greater than the threshold level, generating a notification indicating the first metric for that second gas flow line for presentation to a user to facilitate at least one of a repair or maintenance of that second gas flow line.