To read the full version of this content please select one of the options below:

Non-equidistant GM(1,1) model based on GCHM_WBO and its application in corrosion rate prediction

Yuanjie Zhi (School of Automation and Electrical Engineering, University of Science and Technology Beijing (USTB), Beijing, China)
Dongmei Fu (School of Automation and Electrical Engineering, University of Science and Technology Beijing (USTB), Beijing, China)
Hanling Wang (Unit 96421 of Chinese People’s Liberation Army, Baoji, China)

Grey Systems: Theory and Application

ISSN: 2043-9377

Article publication date: 7 November 2016

Abstract

Purpose

The purpose of this paper is to present a new model which combines the non-equidistant GM(1,1) model with GCHM_WBO (generalized contra-harmonic mean (GCHM); weakening buffer operator (WBO)). The authors use the model to solve the deadlock that for a large number of non-equidistant corrosion rate, it is difficult to establish a reasonable prediction model and improve the prediction accuracy.

Design/methodology/approach

This research consists of three parts: non-equidistant GM(1,1) model, GCHM_WBO operator, and the optimization of morphing parameter (contained in GCHM, control the intensity of the weakening operator). The methodology is explained as follows. First, the authors built a non-equidistant GM(1,1) model with GCHM_WBO weakened data, of which morphing parameter was randomly selected. Next, the authors calculated the error between prediction data of model and the real data, and adjusted the morphing parameter according to the error and property of GCHM. Then, the authors generated a new non-equidistant GM(1,1) based on new morphing parameter, and repeated the previous step until the termination condition was satisfied. Finally, the model with appropriate morphing parameter was used to implement the prediction of new data.

Findings

This paper finds a property of GCHM, which is a monotonic increasing function of morphing parameter in some specific conditions. Based on the property and the fixed point axiom of WBO, an algorithm was designed to search an appropriate morphing parameter. The appropriate morphing parameter was implemented for the purpose of improving the accuracy of the model. The model was applied to predict the corrosion rate of six steels at Guangzhou experimental station. The results showed that the proposed method can get more accuracy in prediction capability compared to the models with the original data and AWBO weakened data. The method is applicable to long-term forecasts in case of data scarcity.

Practical implications

Corrosion will cause huge economic loss to a country; therefore, it is important to judge the remaining useful life of a material or equipment; the foundation for judgement of which is the prediction of material corrosion rate. However, the prediction of corrosion rate is very difficult because of corrosion data’s features, such as small sample size, non-equidistant, etc. The proposed method can be used to implement long-term forecast of corrosion data with only one sample and non-equidistant samples.

Originality/value

This paper presented a model which combines the non-equidistant GM(1,1) model with GCHM_WBO to handle the problem of long-term forecasting of corrosion data. In the modelling process, the proposed morphing parameter searched through algorithm can improve the prediction accuracy of the model. Therefore, the model can provide effective and reliable result when data are of a small sample size and non-equidistant.

Keywords

Acknowledgements

The project is supported by key project of the National Natural Science Foundation of China (No. 51131001), National Environmental Corrosion Platform (No. 41500009).

Citation

Zhi, Y., Fu, D. and Wang, H. (2016), "Non-equidistant GM(1,1) model based on GCHM_WBO and its application in corrosion rate prediction", Grey Systems: Theory and Application, Vol. 6 No. 3, pp. 365-374. https://doi.org/10.1108/GS-09-2015-0061

Publisher

:

Emerald Group Publishing Limited

Copyright © 2016, Emerald Group Publishing Limited