Drift detection in dynamic processes

A system comprising:

• a communication interface configured to receive a first plurality of predictor values for a mechanical process occurring during or before a first time interval, a designated outcome value occurring during a second time interval, and a second plurality of predictor values for the mechanical process occurring during or before the second time interval, the first time interval preceding the second time interval;
• a processor configured to determine a drift value for the second time interval by determining one or more derivatives of a polynomial function fitted to the second plurality of predictor values, to compare the drift value to a designated drift threshold, to compare the error value to a designated error threshold, and to retrain the prediction model when it is determined that the drift value exceeds the designated drift threshold;
• and
• a storage device configured to store the retrained prediction model, wherein an estimated outcome value is determined based on the retrained prediction model, wherein the communication interface is further configured to halt the mechanical process based on the estimated outcome value meeting a designated condition.

One or more non-transitory computer readable media having instructions stored thereon for performing a method, the method comprising:

• receiving via a communication interface a first plurality of predictor values occurring during or before a first time interval for a mechanical process;
• determining an estimated outcome value for a second time interval by applying a prediction model via a processor to the first plurality of predictor values;
• receiving via a communication interface a designated outcome value detected by a sensor and occurring during the second time interval and a second plurality of predictor values for the mechanical process occurring during or before the second time interval, the first time interval preceding the second time interval;
• determining an error value via the processor based on the estimated outcome value and the designated outcome value;
• determining a drift value for the second time interval by determining one or more derivatives of a polynomial function fitted to the second plurality of predictor values;
• comparing the drift value to a designated drift threshold and comparing the error value to a designated error threshold;
• retraining the prediction model when it is determined that the drift value exceeds the designated drift threshold;
• storing the retrained prediction model on a storage device;
• determining an estimated outcome value for the mechanical process based on the retrained prediction model;
• and
• halting the mechanical process based on the estimated outcome value meeting a designated condition.