Multivarijabilni modeli upravljanja u rudarstvu

Abstract

osetljivost i rizik. Više pobuda generiše ovakvu atribuciju. Prva je, multivarijabilnost ambijentalnih uslova i funkcionalna i strukturna složenost rudarskih sistema (čovek - priroda (radna sredina) - mašina – okruženje); Druga se odnosi na ne postojanje opšte naučne saglasnosti oko ocene pogodnosti modela za podršku odlučivanju i upravljanje. Treća, ako bi saglasnost i postojala, nije sigurno da bi raspoloživi modeli, metode i taktike, dale ekvivalentne rezultate za iste rudničke uslove. Četvrta, sistemske nauke još uvek nemaju opšte prihvaćen i u praksi primenljiv algoritam izbora najboljeg rešenja u slučaju kolebanja multimodelskih poredaka alternativa. Prisutni su različiti pristupi u premoštavanju problema, oni se međusobno ne isključuju, ali ponuđena rešenja nemaju verifikaciju opšte i univerzalne primenljivosti. Ovo je uticalo da istraživanja u okviru disertacije, budu usmerena kritičkim stavom prema pristupima zasnovanim na izboru najboljeg modela za analizu i donošenje upravljačkih odluka u rudničkim uslovima, drugačijim od uobičajenih industrijskih uslova. U disertaciji se umesto izbora najboljeg, najpogodnijeg, ili najprikladnijeg modela za podršku odlučivanju, predlaže proceduralni prilaz. Ovakav prilaz podrazumeva istovremeno uključivanje u analizu više modela sa korektnom aproksimacijom multivarijabilnih rudničkih uslova. Pošto cilj nije izbor najboljeg modela već najboljeg rešenja zadatog problema, prema postavljenom algoritmu postupak izbora najbolje rangirane alternative ili najboljeg poretka alternativa, zavisi od kolebanja multimodelskih rangova. U slučaju ekvivalencije multimodelskih rangova, formirani poredak alternativa prihvata se kao definitivan, u suprotnom konačni poredak alternativa definiše se ponderisanjem. Ako je cilj najbolja (prvorangirana) alternativa, analiza je opciona zavisno od stepena ekvivalencije multimodelskih rangova. Primenljivost i operativnu korisnost postavljenog pristupa, odnosno algoritma „proceduralnog postupka’’, eksperimentalno je testirana sa četiri modela na šest rudničkih problema, različite strateške, taktičke ili operativne važnosti. Prosečna korelativnost multimodelskih i ponderisanih poredaka u 50% testova je u opsegu visoke – jake veze (0,70-0,89), a u preostalih 50% u opsegu veoma visoke-veoma jake veze (0,90-0,99). Prosečna korelativnost ponderisanih i multimodelskih poredaka (0,813) veća je za 5,58% od prosečne korelativnosti (0,77) multimodelskih poredaka. Primenjeni pristup u istraživanjima i metrični elementi eksperimentalnih rezultata, obezbeđuju objektivnost sagledavanja i vrednovanja predloženog algoritma za podršku odlučivanju u multivarijabilnim uslovima. Rezultujući ishodi ovih saznanja, potvrđuju validnost i praktičnu primenljivost postavljenog algoritma...

The decision making and management in mining engineering are characterized by complexity, variability, ambiguity, sensibility and risk. There are several incentives for such attribution. The first is multi-variability of environmental conditions and the functional and structural complexity of mining production systems (Human – nature (working environment) – machine – narrower and wider environment). The second is related to the non-existing scientific consensus regarding the assessment on the suitability of decision making and management model. The third, even if such consensus existed, it is not sure that available models, methods and tactics would give equivalent results for different mining conditions. The fourth is that system sciences still do not have generally accepted and practically applicable algorithm for the selection of the best solution in case of instability of the multi-model order of alternatives. There are several approaches in overcoming this problem that are not mutually exclusive, but the offered solutions do not have verification of the general and universal applicability. This was a determining point that the research within the scope of this dissertation should be forwarded by a critical standpoint toward approaches based on the selection of the best model for analysis and management decision making in mining conditions, with major differences from conditions in other industries. In the dissertation, opposed to the selection of the best, the most adequate or the most suitable model for the support to decision making, the procedural approach is recommended. This approach means to include simultaneous models in the analysis, with accurate approximation of the multi-variable conditions in mines. Since the goal is not to select the best model, but rather the best solution for the task in question, the procedure of selection of the best ranked alternative, or the best order of alternatives is accomplished depends on the instabilities of multi-model ranks according to the algorithm set. In case of multi-model ranks equivalence, the order of alternatives is accepted as the final solution, otherwise the process of defining the final order of alternatives is achieved by pondering. If the goal is the best (the first ranked) alternative, the procedure is optional, depending on the multi-model ranks equivalence degree. The applicability and the operative usability of the approach set, i.e. the algorithm of the “procedural method” developed in this dissertation on four models, is tested with six mining engineering problems with different strategic, tactical or operational importance. The average correlation of multi-model and pondered orders for 50% of tests lies within the class of high – strong connection (0.70-0.89), and for other 50% within the class of very high – very strong connections (0.90-0.99). The average correlation of the pondered and multi-model orders (0.813) is larger by 5.58 % than the average correlation (0.77) of the multi-model orders. The applied approach in the research, and metric elements of the experimental tests results are ensuring the objective assessment and valuation of the suggested algorithm for the support to decision making in multi-variable conditions. The outcome of these findings are confirming the validity and practical applicability of the suggested approach...