Detekcija i analiza grešaka u implementaciji dinamičkih diskretnih modela upravljanja zalihama

Social sciences / Operations research
Predmet istraţivanja ove doktorske disertacije je pristup za detekciju i analizu grešaka u dinamiĉkim diskretnim spredšit modelima upravljanja zalihama, zasnovan na karakteristikama problema i naĉinu modeliranja. Postojeći pristupi za obezbeĊenje kvaliteta spredšit modela pokazali su se kao perspektivni, ali su nedovoljno ispitani i prilagoĊeni konkretnim problemima. Spredšit modeli korišćeni za evaluaciju ovih pristupa su najĉešće namenski kreirani, nisu realni primeri, relativno su mali i nisu prilagoĊeni kompleksnim problemima sa velikim brojem zavisnosti, tako da njihova primena u realnim okolnostima, kao i skalabilnost nisu potvrĊeni. Ovi pristupi, u najvećem broju sluĉajeva, zasnovani su na osobinama spredšit aplikacija, ali i idejama iz razliĉitih oblasti kao što su: softversko inţenjerstvo, operaciona istraţivanja i druge. Postojeći pristupi ne uzimaju u obzir karakteristike problema i naĉin modeliranja, koji znaĉajno utiĉu na nastanak, ali i mogućnost otkrivanja grešaka. Oni podrazumevaju da su izlazne vrednosti modela unapred poznate ili da korisnik moţe da obezbedi sve informacije o strukturi i ograniĉenima modela, što veoma ĉesto nije moguće. U skladu sa navedenim, opravdan je i neophodan razvoj novog pristupa, koji bi omogućio unapreĊenje kvaliteta spredšit modela za upravljanja zalihama, odnosno detekciju i analizu grešaka u njima. Osnovni cilj ove disertacije je kreiranje novog pristupa za obezbeĊenje višeg kvaliteta dinamiĉkih diskretnih modela upravljaĉkih problema operacionog menadţmenta, konkretno upravljanja zalihama, razvijenih u spredšitovima, razvojem algoritma za detekciju i analizu grešaka u navedenim modelima i utvrĊivanje njihovih uzroka na brz i efikasan naĉin.
The subject of this doctoral thesis refers to error detection and debugging approach for dynamic discrete spreadsheet models of inventory control problems, based on problem characteristics and modelling method. Existing quality assurance approaches are very perspective, but insufficiently tested and adapted to actual problems. Spreadsheet models used for evaluation of those approaches are usually created just in that purpose and aren’t real practical examples. Experimental spreadsheet models are generally small and aren’t adjusted to complex problems with many dependencies. Applicability and scalability of existing quality assurance approaches has not been proven in wider context. In most cases, approaches are based on spreadsheet application characteristics and ideas from different scientific areas, such as: software engineering, operations reaserch and others. Existing approaches do not consider problem characteristics and modelling method, which significantly influence error occurrence and error detection. Those approaches consider that model output values are known in advance or that all informations about model structure and constraints are provided by user. Very often, both presumptions are unacceptable for models used in practice. In accordance with aforementioned, development of new improved quality assurance approach for inventory control spreadsheet models is justified and necessary. Main goal of this dissertation is to create new spreadsheet quality assurance approach for dynamic discrete models of operations management problems, specifically inventory control problems, developed in spreadsheets, by developing algorithm for error detection and quick and efficient debugging for mentioned models. Newly developed error detection and debugging approach for dynamic discrete spreadsheet models of inventory control problems, presented in this dissertation, is adapted to users and allows quality improvement of inventory control spreadsheet models.