UNAMBIGUOUSNESS-AMBIGUITY OF PARAMETRIC OPTIMIZATION OF MOTOR-CAR SYSTEMS IN THE CONDITIONS OF CRITERIAL UNCERTAINTY
Annotation. The general methodology of parametric optimization of systems is considered for two arbitrary cri-teria simultaneously. The so-called principle of expanding an optimization problem is proposed, which creates the basis for finding guaranteed unambiguous solutions, without resorting to artificial formal means of «collapse» of the two cri-teria into one. It turns out that a very common multiplicative criterion for so-called fair trade-off actually expresses the average geometric basic criteria. It is easy to reduce (lead down) it to additive. Therefore, it is certainly not known, why he should give preference to the arithmetic mean (after the appropriate coordinate) of the dimensions of the primary criteria. There are more subjective and far-fetched than objective and truthful in the criterion of a fair compromise.
Perfection is a permanent process — it has a beginning but has no end. In that the new" perfections arise from time to time and each of them definitely use a certain time, then, of course, the process of perfection is a step-by-step process, an endless step to an unattainable ideal. This particular circumstance should be taken into account.
Described algorithms for optimal search formally reproduce on a primitive model plane the real process of step-by-step improvement of all man-made - from acceptable to better... There are no examples when something was created immediately unconditionally optimally (and the ideal — at all not recognizable and therefore not embodied). At each step, one of the algorithms regulates minimizing the value of a single criterion, without affecting it, without changing the other. That is why there are no conflicts outside the attractor. Only within the attractor, for which the line (which is a one-dimensional attractor) rules on the model plane, the consistency disappears. Another algorithm combines a series of steps in each of which only one parameter varies, and the gain at the same time has both supporters of one perfection, and supporters of some other perfection. Consequently, there are no conflicts, until the algorithm does not attract the attractor, which this time is an area on a model plane, that is a two-dimensional attractor.
Within the attractor, all solutions to the optimization problem is appropriate without a doubt, even advisable to consider completely equivalent. However, in fact, insurmountable subjectivism does not allow us to adhere to this idea (let's say, without the participation of any dictator).