A robust prediction method based on Kriging method and fuzzy c-means algorithm with application to a combine harvester

• Published in: Structural and multidisciplinary optimization Vol. 65; no. 9
• Main Authors: Shi, Maolin, Liang, Zhenwei, Zhang, Jian, Xu, Lizhang, Song, Xueguan
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.09.2022; Springer Nature B.V
• Subjects: Algorithms; Cleaning; Combine harvesters; Computational Mathematics and Numerical Analysis; Data analysis; Engineering; Engineering Design; Mathematical models; Measurement techniques; Optimization; Outliers (statistics); Parameters; Prediction models; Research Paper; Robustness (mathematics); Theoretical and Applied Mechanics; Training; Outlier detection; Robust prediction; Fuzzy; Combine harvester; Kriging method; means algorithm
• ISSN: 1615-147X; 1615-1488
• DOI: 10.1007/s00158-022-03364-0

In real-world problems, engineering data often suffer from outliers due to deficiencies in measurement techniques, recording errors, or other reasons. In this work, a robust prediction method is proposed based on the Kriging method and fuzzy c-means algorithm. An outlier detection strategy is designed based on the fuzzy c-means algorithm, in which Kriging method is used to evaluate the relationship between the input and output. The membership of each training sample is calculated based on the prediction error of the Kriging model of each cluster and is used to judge whether the training sample is an outlier. The detected outliers are removed from the training samples, and then the remaining training samples are used to construct the final prediction model. The effect of the parameters of the proposed method on its performance is studied through an one-dimensional and a four-dimensional numerical problem, and ten benchmark functions are used to test its performance thoroughly. The results indicate that the proposed method produces much better performance in terms of outlier detection accuracy and prediction accuracy than the conventional outlier detection method and the Kriging method. Similar results can be found in the experiments on engineering problems. The proposed method is applied to model and analyze the operation data of the cleaning device of a combine harvester. The effect of the operation parameters of the cleaning device on the grain impurity ratio is studied, and the operation parameters of the cleaning device are optimized based on the prediction model of the proposed method. The analysis and optimization results provide a reference for the operation and control of the cleaning device of a combine harvester.