prikaz prve stranice dokumenta Monitoring the state of health of the pipes of a water distribution network using machine learning algorithms
Access restricted to authenticated users
master's thesis
Monitoring the state of health of the pipes of a water distribution network using machine learning algorithms
2022. urn:nbn:hr:179:602176
Čmrlec, Marta
University of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
Department of Electronics and Computing

Cite this document

Čmrlec, M. (2022). Monitoring the state of health of the pipes of a water distribution network using machine learning algorithms (Master's thesis). Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture. Retrieved from https://urn.nsk.hr/urn:nbn:hr:179:602176

Čmrlec, Marta. "Monitoring the state of health of the pipes of a water distribution network using machine learning algorithms." Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2022. https://urn.nsk.hr/urn:nbn:hr:179:602176

Čmrlec, Marta. "Monitoring the state of health of the pipes of a water distribution network using machine learning algorithms." Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2022. https://urn.nsk.hr/urn:nbn:hr:179:602176

Čmrlec, M. (2022). 'Monitoring the state of health of the pipes of a water distribution network using machine learning algorithms', Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, accessed 24 April 2024, https://urn.nsk.hr/urn:nbn:hr:179:602176

Čmrlec M. Monitoring the state of health of the pipes of a water distribution network using machine learning algorithms [Master's thesis]. Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture; 2022 [cited 2024 April 24] Available at: https://urn.nsk.hr/urn:nbn:hr:179:602176

M. Čmrlec, "Monitoring the state of health of the pipes of a water distribution network using machine learning algorithms", Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Split, 2022. Available at: https://urn.nsk.hr/urn:nbn:hr:179:602176

Please login to the repository to save this object to your list.
Metadata
TitleMonitoring the state of health of the pipes of a water distribution network using machine learning algorithms
Title (croatian)Promatranje stanja cijevi u vodovodnoj mreži korištenjem algoritma strojnog učenja
AuthorMarta Čmrlec
MentorEugen Mudnić (mentor)
GranterUniversity of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
(Department of Electronics and Computing)
Split
Defense date and country2022-09-14, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
In this paper, monitoring the state of health of the pipes of a water distribution network (WDN) is implemented using a machine learning algorithm. This is done using a data set collected by a WDN company in Barcelona, Spain. The company's goal is to be able to predict when a pipe in the network will fail based on the characteristics of that pipe. If the predictions are accurate, the pipe in question can be preventively replaced without causing any damage. This strategy is called predictive maintenance and is the best maintenance management strategy available. Predictive maintenance is implemented using machine learning algorithms, in this case supervised learning and Python programming language. Since the data set consists mostly of data on pipes that have not failed by the end of this study (right-censored data), a lot of information would be lost if the usual regression or classification algorithms were used. For this reason, a reliability or survival analysis algorithm called the random survival forest (RSF), is used in this study. The RSF algorithm is an extension of the random forest algorithm that is able to consider data where the event of interest has not yet occurred by the end of the study, censored data. Its goal is to predict when a pipe is likely to fail based on the information obtained from the survival data set. The survival data set consists of feature values, the survival or censoring time t, and an event indicator that has a value of 0 if the data sample is still censored at time t, or a value of 1 if the data sample has failed at time t. Numerous survival trees are created using the variations of the original survival data set, each of which provides its own prediction. The RSF averages all the predictions to make the final decision, the estimate of the failure time of a pipe, i.e. remaining useful life of a pipe.
Abstract (croatian)
U svrhu ovog diplomskog rada promatrano je stanje cijevi u sustavu za opskrbu vodom korištenjem algoritma strojnog učenja. Algoritam je treniran na bazi podataka prikupljenoj od strane firme za upravljanje vodovodnom mrežom u Barceloni. Cilj ovog projekta je omogućiti firmi da predvidi kada će se pojaviti kvar na cijevi u vodovodnom sustavu na temelju karakteristika te cijevi. Ukoliko su dobivena predviđanja točna, moguće je preventivno popraviti promatranu cijev i time izbjeći eventualnu štetu. Ovakav se pristup naziva prediktivno održavanje i pokazalo se kao najbolja dostupna strategija održavanja sustava. Implementirana je korištenjem nadziranog strojnog učenja i programskog jezika Python. Kako se baza podataka sastoji pretežito od podataka o cijevima koje do kraja studije nisu doživjele kvar, velika bi količina informacija bila izgubljena kod implementacije uobičajenih regresijskih i klasifikacijskih algoritama. Iz tog je razloga u ovom radu korišten jedan od algoritama analize preživljavanja ili pouzdanosti, nasumična šuma preživljavanja (engl. random survival forest). Algoritam je nastao kao proširenje algoritma nasumičnih šuma, uvođenjem mogućnosti razmatranja podataka koji nisu doživjeli promatrani događaj, u ovom slučaju kvar, do kraja provođenja studije. Cilj algoritma je predvidjeti kada je vjerojatno da će cijev doživjeti kvar na temelju informacija prikupljenih iz baze podataka o preživljavanju cijevi u mreži. Baza podataka sastoji se od karakteristika cijevi, vremena preživljavanja t, te indikatora događaja čija je vrijednost 0 u slučaju da promatrana cijev još uvijek nije doživjela kvar do trenutka t, ili 1 u slučaju da je doživjela kvar u trenutku t. Brojna stabla preživljavanja su izgrađena temeljem varijacija na originalnu bazu podataka, od kojih svako pruža vlastito predviđanje. Nasumična šuma preživljavanja računa prosječnu vrijednost svih predviđanja stabala u svrhu dobivanja konačne procjene vremena kvara cijevi, tj. preostalog vijeka trajanja cijevi.
Keywords
Keywords (croatian)
Languageenglish
URN:NBNurn:nbn:hr:179:602176
Study programmeTitle: Computing
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsAccess restricted to authenticated users
Terms of useLicence In copyright icon
Created on2023-03-24 08:40:43