Model acceptance for transfer-function-based technique predictability

4 Model acceptance for transfer-function-based technique predictabilitY 

Explainability is only the first criterion for model acceptance. To be useful, the model must be able to predict the behaviour one or more key variables to a reasonable level of accuracy. Supposing the model is to be used to predict the behaviour of a certain important variable, 0, (e.g. the peak value of reactor exit temperature during a transient) then clearly the uncertainty in the nominally constant parameters will affect the tolerance to be placed on the calculated value of 0. Providing equation (24.61) or the more general equation (24.63) is satisfied, then a conservative estimate of the variance, o, to be associated with the variable, 0, is 

On the other hand, we may give full weight to knowledge derived from the results of the transient by replacing Ty in equation (24.64) by the weighted contribution 

Naturally, for us to have high confidence in the model s predictions, the plant transient against which the model has been tested has to be sufficiently wide-ranging to traverse all the regions of the model upon which later predictions will be based. 

---