A new method of COD testing with BP neural network

A new optical method of testing the chemical oxygen demand (COD) of wastewater with multi-wavelength at UV band is reported. The normal UV254 is based on the UV light absorption by wastewater at 254nm according to the Beer-Lambert law. But not all of the wastewater has obviously absorption at 254nm. The new method use back-propagation artificial neural network(BP network) to determine COD of the samples. After the BP network is training by a series wastewater's samples which have similar ingredient but different COD, the sample's absorbance at UV band(between 240~400nm) is analyzed, and the specific wavelength which is associative with sample's COD is picked out. These wavelength can be used to calculated the COD value, and the result is close to standard chemical method. Comparing with UV254, the new method is more stability and precise, and better adaptability to varied wastewater.

I. Introduction

Chemical oxygen demand (COD) is one of the significant parameters of water quality. It reflects the level of organic compounds in water pollution. Different from the traditional chemical measurement method, the ultraviolet spectrometry method can quickly predict COD and does not need any reagents, in another words, it avoid secondary pollution.