American Society of Civil Engineers


On-line Monitoring of Wastewater True Color Using Digital Image Analysis and Artificial Neural Network


by Ruey-Fang Yu, (corresponding author), (Associate Professor, Dept. of Safety, Health and Environmental Engineering, National United Univ., Miao-Li 360, Taiwan, R.O.C. E-mail: rfyu@nuu.edu.tw), Wen-Po Cheng, (Professor, Dept. of Safety, Health and Environmental Engineering, National United Univ., Miao-Li 360, Taiwan, R.O.C.), and Mei-Ling Chu, (Research Assistant, Dept. of Safety, Health and Environmental Engineering, National United Univ., Miao-Li 360, Taiwan, R.O.C.)

Journal of Environmental Engineering, Vol. 131, No. 1, January 2005, pp. 71-79, (doi:  http://dx.doi.org/10.1061/(ASCE)0733-9372(2005)131:1(71))

     Access full text
     Purchase Subscription
     Permissions for Reuse  

Document type: Journal Paper
Abstract: The American Dye Manufactures’ Institute (ADMI) 3 and 31 wavelength (WL) methods are the most well-known analytical methods for measuring wastewater true color. However, these two methods use a spectrophotometer as the measurement device. The measurements thus must be conducted in the laboratory and cannot provide on-line monitoring data for process control. This study applies the digital image analysis (DIA) method to develop an effective and economical method for on-line monitoring of the wastewater true color. The DIA method was used to measure three types of colored samples including the platinum-cobalt color standard solutions, synthetic colored samples and real colored samples. Experimental results show that the DIA method performed not only with good accuracy and precision but also high sensitivity in the measurement of true color. However, the DIA values present different linear relationship with the ADMI 3 and 31 WL values for different sample hues. Direct substitutions of DIA, ADMI 3 and 31 WL values may be inappropriate. Finally, a back-propagation neural network (BPNN) was proposed to calibrate the DIA, ADMI 31 and 3 WL values. The BPNN models were proven to be highly effective for the calibrations among the ADMI 3, 31 WL, and DIA values.


ASCE Subject Headings:
Color
Digital techniques
Imaging techniques
Neural networks
Wastewater management