Predictive Fault Detection in Elevators from Change Points analysis
Saha, Arindam (2022)
2022
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202304135205
https://urn.fi/URN:NBN:fi:amk-202304135205
Tiivistelmä
Time series analysis can be a crucial and valuable form of data source to track changes over time. It has a variety of applications such as gaining business insights, weather forecasts, stock prices predictions, and application performance. It can also be used as a tool for anticipating potential device malfunctions. This incentive is the motivation for this thesis with the goal of identifying probable issues in the elevator machines in advance. The technicians can be alerted and remedy the problem before it materializes, leading to overall improved performance.
This thesis aims to predict possible breakpoints or maintenance needs for elevators based on change points data generated from internet of things (IoT) devices connected to the elevators.
This work analyzes the statistics in the time series change points data, such as high peaks, and correlate them with previous historical maintenance needs of the elevators, which are recorded as service orders. The data is visualized to determine if there may be a possible breakpoint in the future.
Change point data collected over 1 year for 10 elevators was used as a test set. It was found that the peak values of change points correlate with maintenance service needs reported within a certain time frame after the peak. There were various sets of parameters which were tested against the model. The results have been measured with metrices such as accuracy, precision, f1Score, recall, specificity. The metric values differed with different sets of parameters.
Based on this work it can be concluded that change points data is a valuable source of information when used in predicting breakdown of an elevator. The data can be further refined to have more accurate results and other algorithms can also be investigated for the same. The current study serves as a proof of concept to determine whether such data can be used for further research on a larger set of elevators.
This thesis aims to predict possible breakpoints or maintenance needs for elevators based on change points data generated from internet of things (IoT) devices connected to the elevators.
This work analyzes the statistics in the time series change points data, such as high peaks, and correlate them with previous historical maintenance needs of the elevators, which are recorded as service orders. The data is visualized to determine if there may be a possible breakpoint in the future.
Change point data collected over 1 year for 10 elevators was used as a test set. It was found that the peak values of change points correlate with maintenance service needs reported within a certain time frame after the peak. There were various sets of parameters which were tested against the model. The results have been measured with metrices such as accuracy, precision, f1Score, recall, specificity. The metric values differed with different sets of parameters.
Based on this work it can be concluded that change points data is a valuable source of information when used in predicting breakdown of an elevator. The data can be further refined to have more accurate results and other algorithms can also be investigated for the same. The current study serves as a proof of concept to determine whether such data can be used for further research on a larger set of elevators.