| 摘要 |
<p>The presented invention is an algorithm for producing analytic process model that serves for discovery of one or more hidden process variables. This is an interactive procedure that from input measurement process variables (1), by the knowledge of very limited number of reference or goal function data, selects those process models that point to dominant process variables by means of testing their compatibility with the hidden process variable (10). Omitting of these dominant process variables by renormalizing process input and goal data the complex recombination models are obtained (11) that possess a better qualitative correlation with the hidden variable (12) with as more expressive disturbance function. In case of known disturbance function it is possible to directly renormalize data (9A) to make the model by recombination of normalized data (11) and to obtain hidden process variables (12).</p> |
