The method developed by scientists from Novosibirsk State Technical University (NSTU) is based on statistical processing of the main drive activation protocols.
The result of the work is the development and scientific substantiation of the elevator reliability coefficient. The work was carried out within the framework of the study "Detection of elevator malfunctions by detecting the breakdown of an inhomogeneous Poisson process."
"The task set by the customer, the founder of the Novosibirsk company LIFT-COMPLEX DS LLC, Konstantin Kolupaev, a graduate of NETI, was to analyze data on the operation of elevators. They are recorded discretely in the protocol, and it is impossible to directly understand from them why the elevator was not working: because it was broken or because there were no calls. Therefore, we applied statistical analysis, which allowed us to assess the reliability of each elevator," said Artem Kovalevsky, Professor of the Department of Higher Mathematics at NSTU-NETI, Doctor of Physico—Mathematical Sciences.
The analysis method used a mathematical model from the theory of random processes — a Poisson flow with variable intensity. The fact is that the use of the elevator is not uniform throughout the day. For example, in residential buildings, the most intense period is in the morning, while in shopping malls, peak loads occur in the early evening hours. There are also differences between weekdays, holidays, and pre-holidays.
"We have developed and justified an algorithm that allows us to determine with high accuracy why the elevator did not turn on at each specific time interval: due to the absence of people in it or due to a malfunction. The algorithm is based on calculating the probability that no one wanted to use the elevator in a given period of time. If this probability is very low, then a malfunction of the elevator is recorded. The practical algorithm was implemented by Natalia Zakrevskaya, a senior lecturer at the Department of Higher Mathematics, and tested on both factual and synthetic data," explained Artyom Kovalevsky.
As part of the research work, an algorithm for calculating the elevator reliability coefficient, the results of its calculation for real elevators were prepared, and the accuracy of the method was checked on synthetic (artificially generated) data.
The developed algorithm for estimating the reliability coefficient of the elevator is provided to the partner company for use. The elevator reliability coefficient will make it possible to classify elevators by reliability and make scientifically sound decisions about choosing a specific elevator model even at the construction stage of new buildings.