Value balancing for oil or gas drilling and recovery equipment using machine learning models

A method comprising:

• processing time series data related to operation of a piece of oil or gas drilling or recovery equipment using a first sliding window to generate processed data including features relevant to failure prediction;
• applying the processed data to an equipment failure prediction model;
• determining a threshold cutoff value for a feature relevant to failure prediction for the piece of oil or gas drilling or recovery equipment by extracting a metric using a second sliding window to smooth an output of an equipment failure prediction model, wherein the threshold cutoff value is determined based on the metric;
• applying the threshold cutoff value to the output of the equipment failure prediction model to train the equipment failure prediction model;
• deriving a recall value and an average hour-loss value from the trained equipment failure prediction model;
• cross validating the trained equipment failure prediction model with an embedded value optimization using a cost associated with running the piece of oil or gas drilling or recovery equipment to failure and a cost of predicting a failure too early that leads to unused life of piece of oil or gas drilling or recovery equipment;
• identifying a predictive maintenance to be performed and a time for performing the predictive maintenance based on a prediction of equipment failure;
• and
• performing the identified predictive maintenance for the piece of oil or gas drilling or recovery equipment and the time to perform the identified predictive maintenance based on the prediction of equipment failure by the cross validated and trained equipment failure prediction model.

A non-transitory computer-readable medium having stored therein instructions that are executable by a processor for performing operations comprising:

• processing time series data related to operation of a piece of oil or gas drilling or recovery equipment using a first sliding window to generate processed data including features relevant to failure prediction;
• applying the processed time series data to an equipment failure prediction model;
• determining a threshold cutoff value for a feature relevant to failure prediction for the piece of oil or gas drilling or recovery equipment by extracting a metric using a second sliding window to smooth an output of an equipment failure prediction model, wherein the threshold cutoff value is determined based on the metric;
• applying the threshold cutoff value to the output of the equipment failure prediction model to train the equipment failure prediction model;
• deriving a recall value and an average hour-loss value from the trained equipment failure prediction model;
• cross validating the trained equipment failure prediction model with an embedded value optimization using a cost associated with running the piece of oil or gas drilling or recovery equipment to failure and a cost of predicting a failure too early that leads to unused life of piece of oil or gas drilling or recovery equipment, and
• identifying a predictive maintenance to be performed and a time for performing the predictive maintenance based on a prediction of equipment failure, wherein the identified predictive maintenance for the piece of oil or gas drilling or recovery equipment and the time to perform the identified predictive maintenance is based on a prediction of equipment failure by the cross validated and trained equipment failure prediction model.