Intelligent system for identifying sensor drift
A method for identifying sensor drift can include: setting an autocorrelation threshold for a sensor in a long-short term memory (LSTM) model developed based on historical process measurements from an analogous sensor to a sensor; collecting measured data from the sensor; applying the LSTM model to the measured data from the sensor, wherein applying the LSTM model comprises: applying the LSTM model to the measured data from the sensor to yield LSTM predicted data; calculating key performance indicators (KPIs) of the LSTM data based on an accumulated slow drift error (ASDE) model, wherein the KPIs comprise an error, an accumulated prediction error, an accumulated slow-drift error, and an estimated autocorrelation; and identifying sensor drift when the estimated autocorrelation violates the autocorrelation threshold.
1. A method comprising:
collecting historical process measurements from an analogous sensor to a sensor;
developing a long-short term memory (LSTM) model, wherein developing comprises:
selecting training data from the historical process measurements from the analogous sensor;
creating lookback dataset from the training data with an interval of 5 to 100;
defining a topology of the LSTM model; and
training a LSTM network of the LSTM model with the lookback dataset and the topology; and
setting an autocorrelation threshold for the sensor in the LSTM model.
2. The method of claim 1 , wherein developing the LSTM model further comprises normalizing at least one of: the training data, the historical process measurements, and the lookback dataset.
3. The method of claim 1 , wherein developing the LSTM model further comprises reshaping the training data into a 3-dimensional tensor.
4. The method of claim 1 , wherein the topology has parameters that comprise a loss function and a number of layers.
5. The method of claim 4 , wherein the loss function is selected from the group consisting of a mean-square root, a mean absolute error, a hinge, a binary cross-entropy, and a mean-squared logarithmic error.
6. The method of claim 1 further comprising:
running the LSTM model to predict measurements in validation data.
7. The method of claim 6 further comprising:
redeveloping the LSTM model with additional historic process measurements until a R 2 value of greater than 0.80 for a predicted value versus true value cross-plot is achieved.
8. A method comprising: setting an autocorrelation threshold for a sensor in a long-short term memory (LSTM) model developed based on historical process measurements from an analogous sensor to a sensor; collecting measured data from the sensor; applying the LSTM model to the measured data from the sensor, wherein applying the LSTM model comprises: applying the LSTM model to the measured data from the sensor to yield LSTM predicted data; calculating key performance indicators (KPIs) of the LSTM data based on an accumulated slow drift error (ASDE) model, wherein the KPIs comprise an error, an accumulated prediction error, an accumulated slow-drift error, and an estimated autocorrelation; and identifying sensor drift when the estimated autocorrelation violates the autocorrelation threshold.
9. The method of claim 8 , wherein the KPIs are based on parameters comprising a mean factor for a fraction of rolling standard deviation from a mean, a threshold for an estimated autocorrelation of accumulated slow drift prediction error, and predefined number of samples for a rolling mean and a rolling standard deviation.
10. The method of claim 8 further comprising:
triggering an alarm when the sensor drift is identified.
11. The method of claim 8 further comprising:
recalibrating the sensor when the sensor drift is identified.
12. The method of claim 8 further comprising:
replacing the sensor when the sensor drift is identified.
13. The method of claim 8 further comprising:
collecting the historical process measurements from the analogous sensor;
developing the LSTM model, wherein developing comprises:
selecting training data from the historical process measurements from the analogous sensor;
creating lookback dataset from the training data with an interval of 5 to 100;
defining a topology of the LSTM model; and
training a LSTM network of the LSTM model with the lookback dataset and the topology; and
setting an autocorrelation threshold for the sensor in the LSTM model.
14. The method of claim 13 , wherein developing the LSTM model further comprises normalizing at least one of: the training data, the historical process measurements, and the lookback dataset.
15. The method of claim 13 , wherein developing the LSTM model further comprises reshaping the training data into a 3-dimensional tensor.
16. The method of claim 13 further comprising:
running the LSTM model to predict measurements in validation data.
17. The method of claim 13 further comprising:
redeveloping the LSTM model with additional historic process measurements until a R 2 value of greater than 0.80 for a predicted value versus true value cross-plot is achieved.