Methods of quantitative validation for dynamic simulation models

Abstract

The paper presents a number of techniques developed at AEE Winfrith for assessing and improving dynamic simulation models. A method for selecting best-fit parameters for control engineering applications is discussed first. Then, in the main body of the report, the concept of model-distortion is developed, whereby a model's validity may be assessed quantitatively by the extent of time-variation or distortion needed in its nominally constant parameters to match a recorded transient perfectly.

Model validation of discrete transfer functions using the distortion method

“…This obviously depends on the number of parameters and data points, and the choice of significance level for F. Table 1 shows values for a for a selection of circumstances. Given the approximations made in the evaluation of a by Butterfield and Thomas [3] and the uncertainties in the choice of p and q in the Table, this range is not at odds with Butterfield's suggested value of a % 0.8. 3 AN EXAMPLE…”

“…The statistics of these distortions can be examined to see how they compare with a priori estimates of their variability. Butterfield [3] suggests two methods of comparison: 0 comparison against known uncertainties such as manufacturing tolerances;…”

“…An estimate for the variability is then a x A ,. The factor, a x 0.8, is determined approximately by making various assumptions about the behaviour of the model [3]. Another way of deriving suitable values of a using the concept of confidence limits will be described below.…”

Modeling the antigen recognition by B-cell and T-cell receptors through thermostatted kinetic theory methods

“…The reader interested in mathematical and in silico models that are quantitatively in agreement with biological data and based on a similar framework of the kinetic theory for active particles, is referred [10,27]. It is worth stressing that the validation and tuning with data is still a very hard problem and new effective methods or the proper combination of existing methods (see [3,15]) should be developed to reach this goal.…”

Mechatronic Systems

“…So proving the correctness of a model is a matter of importance. If we now interpret -as did Butterfield in [55] -a model as a scientific theory, then the validation of the model must be placed within narrow boundaries. According to Popper [338] the following is true for the validation of a theory:…”