Tezislər / Theses

THE 3
rd
 INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS 
dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
108
The equipment operation periods of time with and without such defects 
are called abnormal and normal, respectively. Learning algorithms are 
determined by the predictive analytics methods. For example, the regression 
model method uses learning intervals for establishing the regression 
coefficients and the calculated criterion threshold (the model output 
parameter) classifying the equipment operation time (normal/abnormal). The 
method of artificial neural networks tunes the weighing coefficients of 
neurons on learning intervals. The plant operation at each moment of time t 
is described by the vector of. In carrying out its measurements by means of 
a monitoring system at a certain step Δt (e.g., 5 min), a sequence of vectors 
united into the matrix is obtained. Each moment of time t, is related by an 
expert to one of two classes, one of which corresponds to normal operation, 
and the other to abnormal (pre-emergency or emergency) state of the plant. 
Denoting the class tag at the time moment t as y(t) (y(t) = 0 for the normal 
state and y(t) = 1 for the emergency state), we write the vector of tags as 
follows: Now, to develop a model from the composition of available 
measurements, it is necessary to determine the output variable and the input 
variables that “explain” it.
As a result of the calculations, there will be some 
collection of data from the algorithm and inputs for the output with the 
graphical form.

---

THE 3
rd
 INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS 
dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
109
References: 
1. [Li, S.; Nie, Y.; Li, J. Condition monitoring and diagnosis of power equipment: Review and 
prospective. High Volt. 2017, 2,p 82–91] 
2. [Carvalho, T.; Soares, F.; Vita, R.; Francisco, R.; Basto, J.; Alcala, S. A systematic 
literature review of machine learning methods applied to predictive maintenance p 137-
140] 
3. [S. Cheng and M. Pecht, “Multivariate state estimation technique for remaining useful life 
prediction of electronic products,” in Proc. AAAI Fall Symposium on Artificial Intelligence 
for Prognostics, Arlington, VA, Nov. 9– 11, 2007 (AAAI, Menlo Park, CA, 2007) p 26-32] 
4. [L. He, R. A. Levine, J. Fan, J. Beemer, and J. Stronach, “Random forest as a predictive 
analytics alternative to regression in institutional research,” Pract. Assess., Res. Eval. P 
1-23] 
5. [Z. Cha and Y. Ma, Ensemble Machine Learning: Methods and Applications]