Devices, systems, and methods for forecasting device failures

A method comprising:

• obtaining sensor data that were generated by a plurality of sensors;
• selecting a first event code, corresponding first characteristics, and a corresponding first machine-learning model, wherein the first event code corresponds to a first event, and wherein the corresponding first machine-learning model was trained based, at least in part, on first extracted sensor data that were generated within a temporal range of a time of a first occurrence of the first event and on second extracted sensor data that were generated within the temporal range of a time of a second occurrence of the first event, and wherein the first extracted sensor data and the second extracted sensor data were extracted from sensor data that included additional sensor data that were not generated within the temporal range of a time of any occurrence of the first event;
• normalizing at least some of the sensor data based on the corresponding first characteristics;
• inputting the normalized sensor data into the corresponding first machine-learning model, which produces a first output;
• and
• in a case where the first output indicates a forecasted occurrence of the first event, calculating a future time of the forecasted occurrence of the first event and outputting the future time and an indication of the forecasted occurrence of the first event.

One or more non-transitory computer-readable media storing instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations comprising:

• obtaining sensor data that were generated by a plurality of sensors;
• selecting a first event code, corresponding first characteristics, and a corresponding first machine-learning model, wherein the first event code corresponds to a first event, and wherein the corresponding first machine-learning model was trained based, at least in part, on first extracted sensor data that were generated within a temporal range of a time of a first occurrence of the first event and on second extracted sensor data that were generated within the temporal range of a time of a second occurrence of the first event, and wherein the first extracted sensor data and the second extracted sensor data were extracted from sensor data that included additional sensor data that were not generated within the temporal range of a time of any occurrence of the first event;
• normalizing at least some of the sensor data based on the corresponding first characteristics;
• inputting the normalized sensor data into the corresponding first machine-learning model, which produces a first output;
• and
• in a case where the first output indicates a forecasted occurrence of the first event, calculating a future time of the forecasted occurrence of the first event and outputting the future time and an indication of the forecasted occurrence of the first event.