Application Of Back-propagation Artificial Neural Network In Speciation Of Plumbum And Cadmium In City Soil

Urban soil is an important component in urban ecosystems, which plays an important role for the quality of living environment and ecological function. Lead and cadmium are the major pollution element in the soil, which also are toxic heavy metals for the human body. They can enter the body through the food chain. Lead pollution is the most common and most serious in the soil of cities. The main source of lesd pollution are from traffic and transport, atmospheric deposition and industrial activities. Lead poisoning harms all the body systems and organs, especially for the nervous system, hematopoietic system, circulatory system and digestive system. The superfluous cadmium in body can induces the function change of kidney, lung, liver, testis, brain and blood systems, which have carcinogenic and some mutagenicity. In addition too much cadmium can lead to osteoporosis, and fractures symptoms.The heavy metals in the soil have the peculiarities of long accumulate time, complexity of existent state and toxicity. The study found that the capacity of soil accumulation, biotoxicity and its influence on plants, humans and the environment, which is not only related to the total concentration of heavy metals, but also depends on the distribution of chemical species of heavy metals. There are five chemical species of heavy metals in soil, namely exchangeable species, carbonate species, Fe-Mn oxide species, organic species and residue species. Usually, The exchangeable species is the primary one which can harm plants and being by food chain, the carbonate species and Fe-Mn oxide species are potential dangerous species for plants and being. If the soil environmental is changed (for example, pH), the two species will be released to environment. The residue species is the most stable species. So it is more meaningful for evaluating the harm of the human health and environment conservation by with heavy metal speciation. Thus, exchangeable, carbonate bound, Fe-Mn oxides and organic-bound and is called the "non-residue". The greater "non-residue" content proportion of metal in the environment , the greater harm intense the activity the environmental and biological hazards. So, it is significance that evaluating the harm of heavy metals for human and environment by heavy metals speciation.In heavy metals speciation, there are many types apparatuses for testing the concentration of heavy metals species. The appropriate apparatuses can be chosen according by the types of sample. This paper chooses graphite furnace atomic absorption spectrometry by researching the speciation of Pb and Cd in urban soil, but the main problem of determination is that the relatively pretreatment of urban soil sample is cumbersome. The Tessier sequential extraction method is one effective method for speciation, which is through the method of selective dissolution or selective retention by certain chemical reagents. Each specie of heavy metals separation and test in turn by sequential extraction method. The inadequacies of this method is the complex and long time operation. For the leaching agent, it is difficult to find the high selective extractants and it is unavoidable that various species interfere with each other in the sequential extraction method. Therefore the establishment of a simple, rapid and sensitive method for heavy metals speciation in soil is necessary.This paper founds a method for exchangeable species, carbonate species, Fe-Mn oxide species, organic species and residue species analysis of cadmium and plumbum by combining the back-propagation artificial neural network with graphite furnace atomic absorption spectrometry point to the question of cockamamie pretreatment processes and complex determination processes of determination of heavy metals specie content in soil. This is a simple, rapid and sensitive analysis method. In the application of artificial neural network, the network learing function is mostly characteristic of artificial neural network. The characteristics of learning set and unknown set has favorable comparability for ependability of prediction results, Building corresponding and dependable learning set by experimentation (Tessier sequence extraction method) for characteristics of unknown samples (species kinds and species concentration). Discussing the influence of chosen of learning set for the prediction results. The experimentation indicate that the method of building learning set in this paper is no prominent influence for prediction results. Establishing reasonable network structure by discussing of every parameters of artificial neural network. Selection of the number of hidden layer nodes: If the number of hidden layer nodes is increased, the system of ANN is stable and not easily to cause oscillation, but learning velocity get slowly; If the number of hidden layer nodes is decreased, which results that he network system instable and prone to oscillation. Selection of learning step length (η): learning step length is a infinite small amount, which means that a long time to learning. However, in practice. in order to speed up the convergence speed, learning step length from (0,1] between the values. Learning step length is increased, which can cause the convergence rate speed up, but the system of ANN unstably and oscillationally; learning step length decreases, which can cause that the convergence rate slowed down and network system is stable, but the learning time is prolonged. In addition, there are the selections of the initial value, learning steps and learning function and training goal in this paper.This paper describes a method of chemical speciation analysis about lead and cadmium in soil by combining the back-propagation artificial neural network with graphite furnace atomic absorption spectrometry. Determination the total concentrations of Pb and Cd in soil samples by GF-AAS after the soil sample by digestion of HNO3-HF-H2O2 system. Putting the concentrations value in the trained BP-ANN, then the network can predict synchronously the five species concentrations value of heavy metals in soil samples. Using this method for Pb and Cd species concentrations in the soil of Changchun yan'an road. The result indicates that the method of the technology of combining the BP-ANN and GF-AAS in this paper has determination determine Pb and Cd content in exchangeable species, carbonate species, Fe-Mn oxide species, organic species and residue species is feasibility, which inaugurates a new approach for chemometrics application for speciesion, which also provide dependable evidence for detection and treatment of Pb and Cd contamination in urban soil.Comparing the prediction result of BP-ANN-GF-AAS and the experimental result of Tessier extraction method compare, the result indicates that there is no systemic error between the two methods, which validates feasibility and veracity of this method. Determination the samples needs 6-8 days by Tessier method and needs one day by this method in this paper, which save manpower and material resources. Furthermore, escaping the errors by "various species interfere with each other", which proves that this analysis method is simple, rapid, accurate and practical comparing traditional methods (Tessier sequential extraction method).