ways to build in sensor and system uncertainty

We have a control system which provides a bang-bang action on temperature sources. There are two sensors, one with a higher frequency response and less accurate and one with a slow freq response but more accurate. We are using these to react and turn off loads as not to build too high of case temp.

There are 8 units we have in a temp chamber, each unit also has 5 external thermocouples that are used to monitor the case temp, along with these two sensors above within each unit for the control.

We see measurement noise in all the sensor readings, when looking through the data for the eight units.

Can this noise be modeled in Simulink or matlab as a random variable input, so as for each simulation one would get the variance seen in the real world test data in the simulation? This input would then approximate the variance actually measured.

Is this easily done?

Answer 1

[Sebastian Castro]

Yes, you can do this relatively easily.

On the MATLAB side, you can always create a vector of normally distributed values with the randn function.

On the Simulink side, the equivalent would be to use the Random Number block.

- Sebastian