Razvoj sinteznog modela upravljanja rizikom kod rotornih bagera

Abstract

Procena rizika tehničkih sistema predstavlja neizostavni segment upotrebnog kvaliteta i upravljanja imovinom. Neželjeni događaji sa rizikom od velikih posledica su havarije i zastoji. Havarije bagera mogu da ugroze živote i zdravlje ljudi i stvore ekonomske posledice. Otkazi elemenata stvaraju dugotrajne zastoje odnosno gubitak planirane proizvodnje. U postojećim metodama za ocenu rizika prisutno je više nedostataka, koji su koncentrisani u postavkama jednakosti elemenata rizika i načinu kalkulacije rizika. Pored toga, predvidljivost rezultata je izražena, a preciznost modela nije razmatrana. Dizajn modela koji se prezentuje u disertaciji, usmeren je ka smanjenju uticaja nedostataka na analize rizika. Inovativni model bazira se na fazi logici i ekspertnim sistemima gde su implementirane dve višekriterijumske metode. Fazi ekspertni sistemi kao domen veštačke inteligencije, su optimalni za analize fenomena gde dominira neodređenost, višeznačnost, neizvesnost i subjektivnost. Algoritamska struktura modela treba da umanji uticaj subjektivnosti, doprinese vrednovanju parcijalnih indikatora u odnosu na okolnosti, ukloni predvidljivost ishoda, poveća preciznost rezultata i unapredi sisteme održavanja. Univerzalnost fazi ekspertskog modela ogleda se kroz mogućnost korišćenja različitih tipova podataka: lingvističkih, numeričkih ili funkcija. Ulazne veličine se mapiraju na fazi skupove. Metoda AHP je primenjena za definisanje koeficijenata značajnosti parcijalnih indikatora. TOPSIS metoda koristi se u propoziciji modela kao deo fazi zaključivanja. Rezultati modela iskazan su kao procentualni odnos intenziteta rizika uz lingvistički opis. Koncizan oblik krajnjeg rezultata daje jednu numeričku vrednost u odnosu na različite skale. Primenom programskog jezika Python i softverskog okvira Streamlit, model je dizajniran u aplikativnoj i vizuelnoj formi. Prezentovani model je verifikovan na rotornom bageru SRs2000. Analizirane su dve forme rizika: po strukturnu stabilnost mašine i po prekid proizvodnog procesa. Rezultati prvog modela su prikazani kao rang slabih mesta bagera sa preporukama aktivnosti održavanja. Rezultati drugog model definišu trenutno stanje čitavog bagera. U obe forme rizika modela disperzija je manja u odnosu na najčešće korišćene modela. Ostvareni doprinosi su i smanjenje subjektivnosti i predvidljivosti ishoda, pomeranje težišta u skladu sa okolnostima analize

Risk assessment of technical systems is an indispensable part of the quality of service and asset management. Adverse events with the risk of major consequences are breakdowns and downtime. Excavator breakdowns can endanger people's lives and health and have great economic consequences. Failures of excavator elements cause long-term downtime i.e. a loss from the planned production. There are several shortcomings in the existing methods used for risk assessment. They are concentrated in the settings of the equality of risk elements and the method of risk calculation. In addition, the predictability of results is pronounced, whereas the accuracy of the model has not been considered. The design of the synthetic model presented in the dissertation is aimed at reducing the impact of the shortcomings on risk analyses. The innovative model is based on fuzzy logic and expert systems where two multi-criteria methods are implemented. Fuzzy expert systems, as a domain of artificial intelligence, are optimal for the analyses of the phenomena in which uncertainty, ambiguity, and subjectivity dominate. The algorithmic structure of the model should reduce the impact of subjectivity, contribute to the aggravation of partial indicators in relation to the circumstances, remove predictability of outcomes, increase the accuracy of results, and improve the existing maintenance systems. The universality of the fuzzy expert model is reflected in the possibility of using different types of data: linguistic, numerical, or functions. The input values are mapped to fuzzy sets. The AHP method has been applied to define the weight coefficients of partial indicators in risk. The TOPSIS method is used in the model composition as part of the fuzzy screening. The results of the model are expressed as a percentage of risk intensity with a linguistic description. The concise shape of the final result is defined through a single numerical value concerning different scales. By applying the Python programming language and the Streamlit framework, the model is designed in the application and visual form. The presented model was verified on a bucket-wheel excavator with the label SRs2000. Two forms of risk were analysed: risk of the machine structural stability and risk of the production process interruption. The results of the first model are presented through the ranking of excavator weak points with recommendations for maintenance activities. The results of the second model define the current state of the entire excavator. In both forms of risk, the model proved to be more precise (less dispersion) than the most used other models. The achieved contributions are also the reduction of subjectivity and predictability of outcomes, shifting the model center of gravity according to the analysis circumstances.