TY - THES

T1 - Application of machine learning algorithms in assessing the technical condition of pipeline transport facilities

AU - Siraeva, Aliia

N1 - no embargo

PY - 2023

Y1 - 2023

N2 - Today, the problem of industrial equipment diagnostics is actively discussed by scientists. First of all, timely diagnosis and early prediction of developing defects directly affect the efficiency of enterprises. For this purpose, enterprises actively introduce new developing technologies, methods and systems, aimed at reducing downtime, energy and material losses and significant increase in socio-economic welfare, as traditional approaches to monitoring the technical condition of equipment and diagnostics, based on periodic inspection does not allow to determine reliably accurate information about the technical condition. One of the innovations are predictive diagnostic systems based on machine learning methods. To date, there are already works and experimental studies of such methods, but further research and validation of such models is needed. The main purpose of the work is to study the application of machine learning methods to assess the technical condition of objects of the main pipeline transport, on this basis it is planned to develop an algorithm based on data taken from the electric drive of oil pumping equipment.A block based on current and voltage sensors and an Arduino microcontroller for digitization of the obtained data were developed to take the electric characteristics data. To assess the possibility of machine learning algorithms in the work were used such machine learning models as Logistic regression models, and Decision Tree Classificator model. Through experimentation and application of the developed algorithm, it was found that this algorithm detects the presence of defects with an accuracy of more than 80%, which confirms the potential of integrating physical and digital systems of the production environment.

AB - Today, the problem of industrial equipment diagnostics is actively discussed by scientists. First of all, timely diagnosis and early prediction of developing defects directly affect the efficiency of enterprises. For this purpose, enterprises actively introduce new developing technologies, methods and systems, aimed at reducing downtime, energy and material losses and significant increase in socio-economic welfare, as traditional approaches to monitoring the technical condition of equipment and diagnostics, based on periodic inspection does not allow to determine reliably accurate information about the technical condition. One of the innovations are predictive diagnostic systems based on machine learning methods. To date, there are already works and experimental studies of such methods, but further research and validation of such models is needed. The main purpose of the work is to study the application of machine learning methods to assess the technical condition of objects of the main pipeline transport, on this basis it is planned to develop an algorithm based on data taken from the electric drive of oil pumping equipment.A block based on current and voltage sensors and an Arduino microcontroller for digitization of the obtained data were developed to take the electric characteristics data. To assess the possibility of machine learning algorithms in the work were used such machine learning models as Logistic regression models, and Decision Tree Classificator model. Through experimentation and application of the developed algorithm, it was found that this algorithm detects the presence of defects with an accuracy of more than 80%, which confirms the potential of integrating physical and digital systems of the production environment.

KW - Gerätediagnose

KW - maschinelles Lernen

KW - elektrischer Antrieb

KW - technischer Zustand

KW - Entscheidungsbaum

KW - Clustering

KW - logistische Regression

KW - binäre Klassifikation

KW - maschinelle Lernalgorithmen

KW - Equipment diagnostics

KW - machine learning

KW - electrical drive

KW - technical condition

KW - decision tree

KW - clustering

KW - logistical regression

KW - binary classification

KW - machine learning algorithms

U2 - 10.34901/mul.pub.2023.210

DO - 10.34901/mul.pub.2023.210

M3 - Master's Thesis

ER -