The main issue with FOS was coming up with an algorithm that determines the important deliverables of thermocouple systems by using time history data generated from a first-order system. Deliverables include: our MATLAB coded algorithm -which outputs the parameters yL, yH, ts, and � - an executive MATLAB file -which outputs SEE and standard deviation values- and a regression plot which compares the prices of FOS models to �. Determining these four parameters within 20.0% of the calibrated data points merits a fairly successful model. We had to deal with one important constraint: we are working with data specific to a first-order system, and must use equations that appropriately match.
Our algorithm inputs data representing a cooling or heating curve and outputs four parameters: yL, yH, ts, and � (calculated by the prior 3 parameters). We compare � of each FOS system to find the most effective system.
The first design choice we made was how we determined ts. ts is best represented when there is a huge spike in the temperature data. Before ts, the difference between two points should be zero; ts is the value where the first non-zero value difference occurs. While making our algorithm we had to adjust for noise, as a minute nonzero change would occur before ts. We first checked …show more content…
We should never disregard outliers, but we should only compare the FOS modelled points which are closer together as those are most representative. Furthermore, when looking at the values we calculated when compared to the calibrated data, there is never more than a 20.0% error, fulfilling our original goal as well as representing the data in an accurate matter. Even so, the highest error percentage we calculated was 16.0% from the � of the clean data from the heating set. Overall, our models are pretty accurate and provide trends that can be reasonably
Our algorithm inputs data representing a cooling or heating curve and outputs four parameters: yL, yH, ts, and � (calculated by the prior 3 parameters). We compare � of each FOS system to find the most effective system.
The first design choice we made was how we determined ts. ts is best represented when there is a huge spike in the temperature data. Before ts, the difference between two points should be zero; ts is the value where the first non-zero value difference occurs. While making our algorithm we had to adjust for noise, as a minute nonzero change would occur before ts. We first checked …show more content…
We should never disregard outliers, but we should only compare the FOS modelled points which are closer together as those are most representative. Furthermore, when looking at the values we calculated when compared to the calibrated data, there is never more than a 20.0% error, fulfilling our original goal as well as representing the data in an accurate matter. Even so, the highest error percentage we calculated was 16.0% from the � of the clean data from the heating set. Overall, our models are pretty accurate and provide trends that can be reasonably