Early warning and automated detection for lost circulation in wellbore drilling

1. A system comprising:

a processing device; and

a memory device that includes instructions executable by the processing device for causing the processing device to:

generate a first machine-learning model trained using historic drilling operation data for monitoring for a lost circulation event; and

receive real-time data associated with a drilling operation that is concurrently occurring with receiving the real-time data;

apply the first machine-learning model to the real-time data to identify a lost circulation event that is occurring for a predetermined period of time for the drilling operation; and

display, via a user interface and based on determining that the lost circulation event is occurring for the predetermined period of time for the drilling operation, a user interface feature indicating the lost circulation event, wherein the user interface further comprises a mud loss indicator.

2. The system of claim 1 , wherein the memory device further includes instructions executable by the processing device for causing the processing device to:

generate a second machine-learning model using the historic drilling operation data for detecting at least one precursor indicating a likelihood of the lost circulation event occurring;

apply the second machine-learning model to the real-time data to detect the at least one precursor above a pre-set threshold indicating that the lost circulation event is likely to occur for the drilling operation; and

output an indicator based on detecting the at least one precursor above the pre-set threshold.

3. The system of claim 2 , wherein the indicator is an alarm or the user interface feature.

4. The system of claim 1 , wherein the memory device further includes instructions executable by the processing device for causing the processing device to:

pre-process the historic drilling operation data for training the first machine-learning model by performing operations comprising:

resampling and cleaning the historic drilling operation data to remove data that is unrelated to detecting the lost circulation event;

performing a feature selection process on the historic drilling operation data using time series analysis;

detecting and removing outlier data in the historic drilling operation data using density-based spatial clustering of applications with noise and a pre-set parameter for determining the outlier data; and

at least one of: (i) removing the outlier data and data that is not a number or (ii) replacing, with an average of a plurality of data values that surround a data value to be replaced; and

pre-process the real-time data prior to applying the real-time data to the first machine-learning model by performing operations comprising:

temporary caching to normalize the real-time data;

resampling and cleaning the real-time data to remove data that is unrelated to detecting the lost circulation event;

performing the feature selection process on the real-time data using time series analysis;

detecting and removing outlier data in the real-time data using density-based spatial clustering of applications with noise and the pre-set parameter for determining the outlier data; and

at least one of: (i) removing the outlier data and data that is not a number or (ii) replacing, with an average of a plurality of data values that surround a data value to be replaced.

5. The system of claim 2 , wherein the memory device further includes instructions executable by the processing device for causing the processing device to:

train the first machine-learning model using the historical drilling operation data and unsupervised spectral clustering for automated mud loss zone learning to enable the first machine-learning model to segment the real-time data into a normal zone and a lost circulation zone,

wherein the normal zone indicates the lost circulation event is not occurring and the lost circulation zone indicates the lost circulation event is occurring; and

train the second machine-learning model using the historic drilling operation data and unsupervised spectral clustering for precursor learning to enable the second machine-learning model to detect the at least one precursor for the lost circulation event in the normal zone of the real-time data.

6. The system of claim 2 , wherein the memory device includes instructions executable by the processing device for causing the processing device to apply the first machine-learning model to the real-time data to identify any lost circulation event that is occurring for the drilling operation and apply the second machine-learning model to detect the precursor above the pre-set threshold indicating that the lost circulation event is to occur for the drilling operation by performing operations comprising:

determining whether the lost circulation event is occurring for the drilling operation;

in response to determining that the lost circulation event is occurring, output the alarm representing that the lost circulation event is detected;

in response to determining that the lost circulation event is not occurring, apply the second machine-learning model to the real-time data to detect the precursor above the pre-set threshold indicating that the lost circulation event is to occur for the drilling operation; and

in response to detecting the precursor above the pre-set threshold, output the indicator representing that the lost circulation event is to occur for the drilling operation.

7. The system of claim 1 , wherein the memory device includes instructions executable by the processing device for causing the processing device to identify the lost circulation event is occurring for the drilling operation by performing operations comprising:

receiving the real-time data in a plurality of consecutive rows with each of the plurality of consecutive rows having a set number of entries and the set number of entries correspond to a known time period; and

detecting the real-time data is indicating the lost circulation event is occurring for a predetermined number of the plurality of consecutive rows.

8. A method, comprising:

generating, by a computing device, a first machine-learning model trained by using historic drilling operation data for detecting that a lost circulation event is occurring;

receiving real-time data associated with a drilling operation that is concurrently occurring with receiving the real-time data;

applying the first machine-learning model to the real-time data to identify a lost circulation event that is occurring for a predetermined period of time for the drilling operation; and

displaying, via a user interface and based on determining that the lost circulation event is occurring for the predetermined period of time for the drilling operation, a user interface feature indicating the lost circulation event, wherein the user interface further comprises a mud loss indicator.

9. The method of claim 8 , further comprising:

generating, by a computing device, a second machine-learning model trained by using historic drilling operation data for predicting that a lost circulation event is likely to occur;

receiving real-time data associated with the drilling operation that is concurrently occurring with receiving the real-time data;

applying the second machine-learning model to the real-time data to detect a precursor above a pre-set threshold indicating that the lost circulation event is to occur for the drilling operation; and

outputting an indicator based on detecting the precursor above the pre-set threshold.

10. The method of claim 8 , further comprising:

pre-processing the historic drilling operation data for training the first machine-learning model by performing operations comprising:

resampling and cleaning the historic drilling operation data to remove data that is unrelated to detecting the lost circulation event;

performing a feature selection process on the historic drilling operation data using time series analysis;

detecting and removing outlier data in the historic drilling operation data using density-based spatial clustering of applications with noise and a pre-set parameter for determining the outlier data; and

at least one of: (i) removing the outlier data and data that is not a number or (ii) replacing, with an average of a plurality of data values that surround a data value to be replaced; and

pre-processing the real-time data prior to applying the real-time data to the first machine-learning model by performing operations comprising:

temporary caching to normalize the real-time data;

resampling and cleaning the real-time data to remove data that is unrelated to detecting the lost circulation event;

performing the feature selection process on the real-time data using time series analysis;

detecting and removing outlier data in the real-time data using density-based spatial clustering of applications with noise and the pre-set parameter for determining the outlier data; and

at least one of: (i) removing the outlier data and data that is not a number or (ii) replacing, with an average of a plurality of data values that surround a data value to be replaced.

11. The method of claim 9 , further comprising:

training the first machine-learning model using the historical drilling operation data and unsupervised spectral clustering for automated mud loss zone learning to enable the first machine-learning model to segment the real-time data into a normal zone and a lost circulation zone,

wherein the normal zone indicates the lost circulation event is not occurring and the lost circulation zone indicates the lost circulation event is occurring; and

training the second machine-learning model using the historic drilling operation data and unsupervised spectral clustering for precursor learning to enable the second machine-learning model to detect the at least one precursor for the lost circulation event in the normal zone of the real-time data.

12. The method of claim 9 , wherein applying the first machine-learning model to the real-time data to identify any lost circulation event that is occurring for the drilling operation and applying the second machine-learning model to detect the precursor above the pre-set threshold indicating that the lost circulation event is to occur for the drilling operation comprises:

determining whether the lost circulation event is occurring for the drilling operation;

in response to determining that the lost circulation event is occurring, outputting the alarm representing that the lost circulation event is detected;

in response to determining that the lost circulation event is not occurring, applying the second machine-learning model to the real-time data to detect the precursor above the pre-set threshold indicating that the lost circulation event is to occur for the drilling operation; and

in response to detecting the precursor above the pre-set threshold, outputting the indicator representing that the lost circulation event is to occur for the drilling operation.

13. The method of claim 8 , wherein identifying the lost circulation event is occurring for the drilling operation comprises:

receiving the real-time data in a plurality of consecutive rows with each of the plurality of consecutive rows having a set number of entries and the set number of entries correspond to a known time period; and

detecting the real-time data is indicating the lost circulation event is occurring for a predetermined number of the plurality of consecutive rows.

14. A non-transitory computer-readable medium comprising instructions that are executable by a processing device for causing the processing device to perform operations comprising:

generating a first machine-learning model trained by using historic drilling operation data for detecting that a lost circulation event is occurring;

receiving real-time data associated with a drilling operation that is concurrently occurring with receiving the real-time data;

applying the first machine-learning model to the real-time data to identify a lost circulation event that is occurring for a predetermined period of time for the drilling operation; and

displaying, via a user interface and based on determining that the lost circulation event is occurring for the predetermined period of time for the drilling operation, a user interface feature indicating the lost circulation event, wherein the user interface further comprises a mud loss indicator.

15. The non-transitory computer-readable medium of claim 14 , further comprising instructions executable by the processing device for causing the processing device to:

generate a second machine-learning model using the historic drilling operation data for detecting at least one precursor indicating a likelihood of the lost circulation event occurring;

apply the second machine-learning model to the real-time data to detect the at least one precursor above a pre-set threshold indicating that the lost circulation event is likely to occur for the drilling operation; and

output an indicator based on detecting the at least one precursor above the pre-set threshold.

16. The non-transitory computer-readable medium of claim 15 , wherein the indicator is an alarm or the user interface feature.

17. The non-transitory computer-readable medium of claim 14 , further comprising instructions executable by the processing device for causing the processing device to:

pre-process the historic drilling operation data for training the first machine-learning model by performing operations comprising:

resampling and cleaning the historic drilling operation data to remove data that is unrelated to detecting the lost circulation event;

performing a feature selection process on the historic drilling operation data using time series analysis;

detecting and removing outlier data in the historic drilling operation data using density-based spatial clustering of applications with noise and a pre-set parameter for determining the outlier data; and

at least one of: (i) removing the outlier data and data that is not a number or (ii) replacing, with an average of a plurality of data values that surround a data value to be replaced; and

pre-process the real-time data prior to applying the real-time data to the first machine-learning model by performing operations comprising:

temporary caching to normalize the real-time data;

resampling and cleaning the real-time data to remove data that is unrelated to detecting the lost circulation event;

performing the feature selection process on the real-time data using time series analysis;

detecting and removing outlier data in the real-time data using density-based spatial clustering of applications with noise and the pre-set parameter for determining the outlier data; and

at least one of: (i) removing the outlier data and data that is not a number or (ii) replacing, with an average of a plurality of data values that surround a data value to be replaced.

18. The non-transitory computer-readable medium of claim 15 , further comprising instructions executable by the processing device for causing the processing device to:

train the first machine-learning model using the historical drilling operation data and unsupervised spectral clustering for automated mud loss zone learning to enable the first machine-learning model to segment the real-time data into a normal zone and a lost circulation zone,

wherein the normal zone indicates the lost circulation event is not occurring and the lost circulation zone indicates the lost circulation event is occurring; and

train the second machine-learning model using the historic drilling operation data and unsupervised spectral clustering for precursor learning to enable the second machine-learning model to detect the at least one precursor for the lost circulation event in the normal zone of the real-time data.

19. The non-transitory computer-readable medium of claim 15 , further comprising instructions executable by the processing device for causing the processing device to apply the first machine-learning model to the real-time data to identify any lost circulation event that is occurring for the drilling operation and apply the second machine-learning model to detect the precursor above the pre-set threshold indicating that the lost circulation event is to occur for the drilling operation by performing operations comprising:

determining whether the lost circulation event is occurring for the drilling operation;

in response to determining that the lost circulation event is occurring, output the alarm representing that the lost circulation event is detected;

in response to determining that the lost circulation event is not occurring, apply the second machine-learning model to the real-time data to detect the precursor above the pre-set threshold indicating that the lost circulation event is to occur for the drilling operation; and

in response to detecting the precursor above the pre-set threshold, output the indicator representing that the lost circulation event is to occur for the drilling operation.

20. The non-transitory computer-readable medium of claim 14 , further comprising instructions executable by the processing device for causing the processing device to identify the lost circulation event is occurring for the drilling operation by performing operations comprising:

receiving the real-time data in a plurality of consecutive rows with each of the plurality of consecutive rows having a set number of entries and the set number of entries correspond to a known time period; and

detecting the real-time data is indicating the lost circulation event is occurring for a predetermined number of the plurality of consecutive rows.