FORECASTING THE VOLUME PRODUCTIVITY OF A SUPPLIER USING CLUSTERIZATION METHODS
DOI:
https://doi.org/10.31471/1993-9981-2022-1(48)-109-114Keywords:
centrifugal supercharger, volumetric productivity, clustering, forecasting, Fuzzy C-Means algorithm, subtractive fuzzy clustering algorithm.Abstract
Methods for predicting the volumetric productivity of a centrifugal supercharger, which is based on the clustering approach, are performed in this paper. Their advantage over existing forecasting methods is that they take into account the chaotic component of experimental data and include modern information technology (fuzzy logic, neural networks, genetic algorithms). Due to the large number of iterative approaches, the value of the error of the objective function is gradually reduced to the desired value, which provides a fairly good quality of forecasting.
The prediction method based on the Fuzzy C-Means algorithm, which is based on solving fuzzy c-means, is implemented in the Matlab software environment. For this method, the number of clusters is twenty, and the number of iterative approaches ranges from one hundred. The value of the standard deviation equal to 0.036816 is calculated.
Another method for predicting the volumetric performance of a centrifugal supercharger, which is based on the subtractive fuzzy clustering algorithm, is also considered. Fuzzy rules were synthesized autonomously from experimental data, and a Sugeno-type model was used to construct a fuzzy rule base. For each point, the value of the potential is calculated, which shows the possibility of forming a cluster in its vicinity. The centers of the clusters are selected iteratively from the points with maximum potentials. It is established that the value of the standard deviation for this method is 0.024289. The obtained results confirm a good degree of convergence of the experimental values to the predicted ones.
Experimental data used for testing of the proposed methods were obtained during the operation of the centrifugal supercharger type 16 ГЦ2-395/53-76С of Dolyna linear production administration of gas transmittal pipelines in November 2014.
Downloads
References
Wang J., Chi D., Wu J., Lu H. Chaotic time series method combined with particle swarm optimization and trend adjustment for electricity demand forecasting. Expert Systems with Applications. 2011. № 38. P. 8419-8429.
Hromov A.V., Voronyn Y.M., Hatylo V.R., Prokopalo E.T. Otsennka hyperparametrov v zadachakh prohnozyrovanyya na osnove klasteryzatsyy. Pytannya prykladnoyi matematyky i matematychnoho modelyuvannya. 2016. № 16. S. 25-46.
Widiputra H., Kho H., Pears R., Kasabov N. A novel evolving clustering algorithm with polynomial regression for chaotic time-series prediction. Neural Inf Process. 2009. Vol. 5864. P. 114-121.
Pakhomova V.M., Dmitriyev S.YU. Rozrobka pidsystemy operatyvnoho prohnozuvannya prostoyiv prybuvayuchykh poyizdiv na osnovi ANFIS-systemy. Informatsiyno-keruyuchi systemy na zaliznychnomu transporti. 2013. №4. S. 46-55.
Chatfield C. The Analysis of Time Series : An Introduction, Sixth Edition. CRC Press. 2009. 352 p.
Phu L., Anh D.T. Mostif-based method for initialization the K-means clustering for time series data. Springer. 2011. Vol. 7106. P. 11-20.
Dovidnyk inzhenera dyspetchersʹkoyi sluzhby / za zah. red. kand. tekhn. nauk YU.V. Ponomarʹova ta M.P. Khymka. Kyyiv : UTSEBOPnaftohaz, 2009. 245 s.
Remneva VV Features of application of partition quality evaluation criteria in Fuzzy C-Means clustering algorithm. Modern information and innovative technologies in transport. 2013. № 1. pp. 157–158.
.png)
