The Accuracy Comparison of Assessing the Amount of Steel Corrosion in Concrete with Several Different Algorithms

Abstract:

Many before studies showed that it was difficult to ensure the accuracy of assessing the amount of steel corrosion in the cracking concrete with artificial neural network [3] method while the study sample size was small. This paper introduces several different algorithms to assess the amount of steel corrosion in concrete. The experimental results show that compared with other algorithms, the predictive value of the support vector machine algorithm is the closest to the measured value.

• Periodical: Applied Mechanics and Materials (Volumes 166-169)
• Pages: 1958-1962
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.166-169.1958
• Online since: May 2012
• Authors: Ping Jie Cheng
• Keywords: Accuracy, Assessing, Steel Corrosion

References

[1] Weiping Zhang. The corrosion damage of reinforced prediction in concrete structure and durability assessment (PhD thesis). Tongji University, 1999. (In Chinese)

[2] Yunling Hui. The extent of corrosion of reinforced in concrete structures and prediction of experimental studies .Industrial buildings, 1996,26 (4): 11-13. (In Chinese)

[3] Haibin Chen, Ditao Niu, Jinxiu Pu. Assessment the amount of reinforced corrosion in concrete using artificial neural network technology, 1999 (2): 51-55. (In Chinese)

[4] Hui Dong, Helin Fu, etc. Predicted landslide displacement based on Takens theory and SVM, China Journal of Highway, 2007,9 (20 Volume 5): 13-18. (In Chinese)

[5] Ditao Niu. Durability and lifetime prediction of concrete structures, Science Press, 2002,10. (In Chinese) Table 1 Sample (learning set) Sample No. w[mm] fcuk[Mpa] c[mm] d[mm] ηc ηj ηc/ηj 1 0.180 21.21 15.0 22 6.84 6.89 　 2 0.250 21.63 17.0 20 14.51 14.46 　 3 0.333 26.58 30.0 18 24.53 24.48 　 4 0.667 29.86 30.0 22 35.64 35.59 　 5 0.500 22.34 55.0 30 15.70 15.65 　 6 0.750 31.76 14.0 30 4.00 4.05 　 7 0.667 21.21 30.0 25 35.30 35.25 　 8 1.100 25.68 15.0 20 14.80 14.75 　 9 0.667 20.31 12.0 25 18.60 18.55 　 10 0.667 23.88 16.0 25 21.10 21.05 　 11 0.850 18.00 17.3 12 7.04 7.09 　 12 2.000 18.00 15.4 12 9.60 9.65 　 13 1.800 18.00 12.1 12 4.20 4.25 　 14 0.500 18.00 14.8 12 10.92 10.97 　 15 0.500 20.00 20.0 8 9.70 9.75 　 16 0.225 20.00 20.0 8 14.45 14.4 　 17 0.300 20.00 40.0 12 7.33 7.38 　 18 0.350 10.00 20.0 12 8.52 8.57 　 19 0.250 20.00 10.0 12 6.37 6.42 　 20 1.150 18.00 15.7 12 6.72 6.77 　 21 0.300 20.00 20.0 12 4.88 4.93 　 Table 2 Support vector machine regression test results Sample No. W[mm] fcuk[Mpa] c[mm] d[mm] ηc ηj ηc/ηj 1 0.700 28.31 30.0 25 21.2 21.15 1.002 2 0.500 22.31 33.5 30 13.1 13.15 0.996 3 0.200 21.21 20.0 20 7.44 7.49 0.993 4 1.000 21.64 30.0 28 20.6 20.55 1.002 5 0.600 30.54 18.0 12 14.05 14 1.004 6 0.900 18.00 13.1 12 7.68 7.73 0.994 7 1.500 18.00 11.7 12 8.32 8.37 0.994 8 0.300 20.00 20.0 16 5.85 5.9 0.992 9 0.500 20.00 40.0 12 6.39 6.44 0.992 10 0.400 20.00 10.0 12 7.9 7.85 1.006 Mean 　 　 　 　 　 　 0.998 Table 3 Neural network test results Sample No. w/mm fcuk[Mpa] c[mm] d[mm] ηc ηj ηc/ηj 1 0.700 28.31 30.0 25 21.2 26.35 0.805 2 0.500 22.31 33.5 30 13.1 15.39 0.851 3 0.200 21.21 20.0 20 7.44 5.52 1.348 4 1.000 21.64 30.0 28 20.6 21.72 0.948 5 0.600 30.54 18.0 12 14.05 23.95 0.587 6 0.900 18.00 13.1 12 7.68 7.91 0.971 7 1.500 18.00 11.7 12 8.32 7.13 1.167 8 0.300 20.00 20.0 16 5.85 5.58 1.048 9 0.500 20.00 40.0 12 6.39 10.56 0.605 10 0.400 20.00 10.0 12 7.9 7.74 1.021 Mean 　 　 　 　 　 　 0.935 Table 4 Linear regression test results Sample No. w/mm fcuk[Mpa] c[mm] d[mm] ηc ηj ηc/ηj 1 0.700 28.31 30.0 25 21.2 19.98 1.061 2 0.500 22.31 33.5 30 13.1 18.96 0.691 3 0.200 21.21 20.0 20 7.44 12.70 0.586 4 1.000 21.64 30.0 28 20.6 17.87 1.153 5 0.600 30.54 18.0 12 14.05 14.88 0.944 6 0.900 18.00 13.1 12 7.68 8.07 0.952 7 1.500 18.00 11.7 12 8.32 8.39 0.992 8 0.300 20.00 20.0 16 5.85 10.94 0.535 9 0.500 20.00 40.0 12 6.39 13.28 0.481 10 0.400 20.00 10.0 12 7.9 8.07 0.979 Mean 　 　 　 　 　 　 0.837