Analyser Update of Inferential

Automatic updating using an on-stream analyser measurement is quite different from updating with laboratory results. Analysers can have a reputation of poor reliability but we describe later in the chapter techniques that prevent spurious measurements from disturbing the process or being used to update an inferential. With this measurement validation in place analysers are far less prone to random errors than the laboratory. Secondly analysers provide measurements far more frequently and so the delay introduced by filtering will be far less. 

We apply dynamic compensation in the form of a deadtime/lead-lag algorithm. This is tuned in exactly the same way as described in Chapter 6 covering bias feedforward. By performing open loop steps on the MV we obtain the dynamics of both the inferential and 

The process gain of the analyser and the inferential should be the same and so K should be 1. If the test shows that this is not the case the problem should be resolved before commissioning analyser update - indeed before using the inferential in a controller. From Equation (6.21)) 

The analyser deadtime should be significantly larger than that of the inferential -otherwise the inferential serves little purpose - except perhaps as a back-up in the event of analyser failure. So B will be positive. If not the case, the dynamic compensation should be applied to the analyser measurement. 

If the analyser is discontinuous and its sample interval greater than the time it takes the process to reach steady state, then it may not show significant lag. As T2 should not be set to zero (because of the effect on the TUTl ratio) then it is wise only to include the deadtime compensation - by removing the lead-lag or setting T equal to T2. 

---