Standard Deviation - For Confidence in Your Application Measurements
Avoiding Measurement Errors
It is always disheartening to find out after hours of data logging and subsequent statistical analysis, that your measurements are not good enough. For example, an average temperature for a PID set point may have an error that causes an oven to melt numerous parts in a batch. How can these types of errors be avoided? This is where the critically acclaimed standard deviation comes into consideration.
How far a data point is from the average is considered the deviation of that point. The squared root of the average of all the squared deviations is defined as standard deviation or sigma (σ). Taking the standard deviation into consideration (from the feedback data from your thermocouples logged during a benchmark test) will help alleviate these problems.
Values In Standard Deviations
The values referenced to the standard deviation are significant and reliable enough to be used in your process’s algorithms. A data point could be only a few standard deviations away, but too far away may not be consistent with the model.
It is well documented that 3 standard deviations reflect that the data set with 99.7% confidence in the model. Current PLCs guarantee an efficient way to collect and to process data, therefore, consider using a small ladder or structure text program to calculate the standard deviation of your data set in order to minimize errors that are normally found when using average values.