1-9 Al-Khwarizmi Engineering Journal,Vol. 12, No. 3, P.P. Modeling the removal of Olive Pips Ayad A.H. Faisal *,** Environmental Engineering Department *Email: (Received Abstract The uptake of Cd(II) ions from simulated wastewater onto olive pips was modeled using artificial neural network (ANN) which consisted of three layers. Based on 112 batch experiments, pH (2-6), initial concentration (25-250 mg/l), biosorbent dosage (0.05 temperature (20-60ºC) were studied. The maximum uptake (=9 60 min, 6, 50 mg/l, 1 g/100 ml, 250 rpm and 25ºC respectively. Tangent sigmoid and linear transfer functions of ANN for hidden and output layers respectively with 7 neurons were sufficient to present good predictions for cadmium removal efficiency with coefficient of correlation The sensitivity analysis for outputs of ANN signified that the relative importance of initial pH equal to 38 % and it is the influential parameter in the treatment process, followed by initial concentration, agitation speed, biosorbent dosage, time and temperature. Keywords: Neural network, Adsorption, 1. Introduction One of the most widely and rapidly spreading problems in the field of conservation and protection of water resources is water pollution with heavy metals. It is closely associated with several human activities such as mining, the metal processing industry, petroleum industry, power industry and on a smaller scale by electroplating wastes, metal-based pigments and numerous other industrial wastes, besides the high exhaust emissions in urban regions from car engines, burning of hospital wastes and domestic solid waste, as well as wasteland-landfills several methods to treat the metal contaminated effluent, but the selection of the adequate method is based on the concentration of pollutant and the cost of treatment [2]. These method chemical precipitation, electroplating, ion exchange, and reverses osmosis which are Khwarizmi Engineering Journal,Vol. 12, No. 3, P.P. 1- 9 (2016) of Cadmium Ions from Aqueous Olive Pips Using Neural Network Technique Ayad A.H. Faisal* Zahraa Saud Nassir* Environmental Engineering Department / University of Baghdad Email: ayadabedalhamzafaisal@yahoo.com **Email: Zahraa_saud@yahoo.com (Received 20 December 2015; accepted 17 April 2016)` The uptake of Cd(II) ions from simulated wastewater onto olive pips was modeled using artificial neural network consisted of three layers. Based on 112 batch experiments, the effect of contact time (10 250 mg/l), biosorbent dosage (0.05-2 g/100 ml), agitation speed (0 ed. The maximum uptake (=92 %) of Cd(II) was achieved at optimum parameters of 60 min, 6, 50 mg/l, 1 g/100 ml, 250 rpm and 25ºC respectively. Tangent sigmoid and linear transfer functions of ANN for hidden and output layers respectively with 7 neurons were sufficient to present good predictions for cadmium removal efficiency with coefficient of correlation sis for outputs of ANN signified that the relative importance of initial pH equal to 38 % and it is the influential parameter in the treatment process, followed by initial concentration, agitation speed, biosorbent dosage, , Olive pips, Modeling, Equilibrium. One of the most widely and rapidly spreading problems in the field of conservation and protection of water resources is water pollution with heavy metals. It is closely associated with several human activities such as mining, the metal sing industry, petroleum industry, power industry and on a smaller scale by electroplating based pigments and numerous other industrial wastes, besides the high exhaust emissions in urban regions from car engines, nd domestic solid landfills [1]. There are several methods to treat the metal contaminated effluent, but the selection of the adequate method is based on the concentration of pollutant and the . These methods include chemical precipitation, electroplating, ion exchange, and reverses osmosis which are expensive and inefficient especially for low metal concentration as cited by many studies Adsorption technique is considered a promising method and this can be attributed to its simplicity in design and operation, low cost, and insensitivity to toxic materials. Investigation the usage of low cost sorbents for treating large quantities of wastewater has become a substantial issue. concern, olive pips as an example for wastes generated from forestry activities and industrial agriculture acquired a great importance in the removing of metal ions [6] Modeling of adsorption process is a topic of interest for the prediction of the metal partitioning between the aqueous solution and the solid surface, and its subsequent application to the design of adsorption treatment units Adsorption isotherm model Freundlich, Elovich, Temkin and others are used mostly for description the equilibrium relationship between the adsorbate concentration in the fluid Al-Khwarizmi Engineering Journal (2016) Aqueous Solutions onto Using Neural Network Technique ** University of Baghdad The uptake of Cd(II) ions from simulated wastewater onto olive pips was modeled using artificial neural network the effect of contact time (10-240 min), initial 2 g/100 ml), agitation speed (0-250 rpm) and %) of Cd(II) was achieved at optimum parameters of Tangent sigmoid and linear transfer functions of ANN for hidden and output layers respectively with 7 neurons were sufficient to present good predictions for cadmium removal efficiency with coefficient of correlation equal to 0.99798. sis for outputs of ANN signified that the relative importance of initial pH equal to 38 % and it is the influential parameter in the treatment process, followed by initial concentration, agitation speed, biosorbent dosage, expensive and inefficient especially for low metal concentration as cited by many studies [3-5]. Adsorption technique is considered a promising can be attributed to its simplicity in design and operation, low cost, and insensitivity to toxic materials. Investigation the usage of low cost sorbents for treating large quantities of wastewater has become a substantial issue. In this s as an example for wastes generated from forestry activities and industrial agriculture acquired a great importance in the [6]. Modeling of adsorption process is a topic of interest for the prediction of the metal partitioning between the aqueous solution and the solid surface, and its subsequent application to the design of adsorption treatment units [7]. Adsorption isotherm models such as Langmuir, Freundlich, Elovich, Temkin and others are used mostly for description the equilibrium relationship between the adsorbate concentration in the fluid Ayad A.H. Faisal Al-Khwarizmi Engineering Journal, Vol. 12, No. 3, P.P. 1- 9 (2016) 2 phase and the adsorbate concentration in the adsorbent particles at a given temperature [8]. Hence, there is a need to use a more representative model that can identify the equilibrium/ non- equilibrium biosorption process for different values of temperature. Consequently, ANN technique has drawn great attention as an alternative approach in the determination of complex relationship between operating parameters. ANN based predictive models are powerful in terms of learning the nonlinear relationships to understand and solve and thereby achieving ability to predict accurately [9]. So, the present study aimed to characterize the non- equilibrium/equilibrium non-isotherm cadmium removal from aqueous solutions onto olive pips using ANN model in comparison with batch experimental results for different operational conditions. 2. Description of ANN The working basis of ANN is inspired from the biological neurons of natural network. Neuron, or node, is represented the principle core of this network. The neuron impulse or the output of a node is determined as weighted sum of the input signals from the proceeding neuron, altered by the transfer function. The learning capability of a neuron is accomplished by adjusting the weights in conformity to chosen learning algorithm. Basically, the topology of ANN is consisted of input, hidden and output layers and the number of neurons is depended on the number of input and output parameters. The significant step in the solution procedure using ANN is finding the number of neuron at a hidden layer and this can be achieved by trial and error [17]. 3. Experimental Work 3.1. Materials The olive pips were collected from the olive oil industry and the dirt particles were removed from these pips by washing with deionized water. They were then air-dried, ground in a mechanical grinder and sieved to obtain the desired size between 125-710 µ m size particles. The simulated water was contaminated with Cd(II) by dissolved Cd(NO3)2 (manufactured by BDH, England) and kept at temperature of the room (25ºC). The pH of the prepared solution was adjusted by using 0.1 M NaOH or 0.1 M HNO3 as required and this stock solution was used to prepare any required concentration of Cd(II). 3.2. Batch Tests The effect of contact time (t), initial pH, initial concentration (Co) of Cd(II), olive pips dosage, agitation speed and temperature (T) on the performance of sorbent material was investigated based on batch biosorption tests. Adsorbent dosages varied from 0.05 to 2 g were introduced into 250 ml flasks with 100 ml solution containing 50 mg/l of metal ions. These flasks were agitated using an orbital shaker (Edmund Buhler SM25, German) with t equal to 60 min and agitation speed of 250 rpm. Samples were withdrawn at specified time intervals ranged from 10 to 240 min and then passed through filter paper. Cd(II) concentrations in the samples were analyzed by atomic absorption spectrophotometer (AAS) (Shimadzu, Japan). Tests were carried out in a pH range of 2-6 to determine the effect of initial pH on biosorption. While effects of operating temperatures ranged from 20 to 60 ºC were investigated. Removal efficiency (R) of Cd(II), which is represented the predicted value of ANN model, was calculated as follows: � = ��������� × 100 …(1) where Ce is the equilibrium concentration of cadmium remaining in the solution at the end of the test. 4. Results and Discussion 4.1. Developing and Optimization of the ANN Model ANN model with Levenberg–Marquardt backpropagation (LMA) training algorithm for correlating the removal efficiency of cadmium ions from aqueous solution by biosorption method was developed. This algorithm was calculated using Matlab program version 7.10.0.499 (R2010a). The experimental data was divided into training, validation and test subsets with corresponding proportions of 60, 20 and 20% respectively. This step is very important in the development of optimized topology for ANN. Ayad A.H. Faisal The training data is the biggest set and is used by neural network to learn pattern presented in the data by updating the network weights. The testing data is used to evaluate the quality of the network. The final check on the perfor generalization ability of the trained network is made using validation data. Transfer functions of tangent sigmoid (tansig) at hidden layer and linear (purelin) at output layer were applied in the present study. The input variables to the feed forward neural network were; biosorbent dosage, t, initial pH, Co, agitation speed and T. In addition, the removal efficiency was chosen as output variable. Figure 1 illustrates best topology for the variation of parameters was depended on the minimum mean square error (MSE) of the training and prediction set solution procedure was beginning with neurons in the hidden layer for optimization of the network. Table 1 illustrates the dependence between the neuron number and MSE for th LMA algorithm. It seems from Table 1 that the MSE of the network was the highest (= 0.0015) for 2 hidden neurons and it is decreased significantly to minimum value of 0.000335 with Hence, the 7 hidden neurons were best case. When the neurons changed 16, MSE was increased and this behavior attributed to the properties of perform and the input parameters [9]. The training was stopped after 30 epochs for the LMA because the differences between training error and validation error started to increase. Figure 2 presents the MSE for subsets described previously with LMA and the best regression was set in Figure 3. It can be seen that the correlation coefficient for training, validation, testing and all data was 0.99795, 0.99874, 0.99793 and 0.99798 respectively. Al-Khwarizmi Engineering Journal, Vol. 12, No. 3 The training data is the biggest set and is used by neural network to learn pattern presented in the data by updating the network weights. The testing data is used to evaluate the quality of the network. The final check on the performance and generalization ability of the trained network is made using validation data. Transfer functions of tangent sigmoid (tansig) at hidden layer and linear (purelin) at output layer were applied in the present study. The input variables to the feed rward neural network were; biosorbent dosage, t, initial pH, Co, agitation speed and T. In addition, the removal efficiency was chosen as illustrates best topology for ANN and of parameters was calculated the minimum mean square error (MSE) of the training and prediction sets. The solution procedure was beginning with two for optimization of the illustrates the dependence between the neuron number and MSE for the Table 1 that the MSE of the network was the highest (= 0.0015) for 2 hidden and it is decreased significantly to ith 7 neurons. adopted as the changed from 7 to behavior can be of performance index The training was stopped after 30 epochs for the LMA because the error and validation Figure 2 presents the MSE for subsets described previously with LMA and the best regression was set in Figure 3. It can be seen that the correlation coefficient for training, validation, was 0.99795, 0.99874, Fig. 1. The optimal architecture of ANN Table 1, MSE values of the training set for Cd(II) on the olive pips. MSE (×10No. of neurons 1.5002 0.9633 0.8384 0.9725 0.5526 0.3357 0.7118 0.5609 0.84010 0.67811 0.43112 0.45313 0.34914 0.45215 0.78816 Fig. 2. MSE of training, validation and test for the LMA algorithm. Khwarizmi Engineering Journal, Vol. 12, No. 3, P.P. 1- 9 (2016) The optimal architecture of ANN. MSE values of the training set for Cd(II) on the MSE (×10-3) 1.500 0.963 0.838 0.972 0.552 0.335 0.711 0.560 0.840 0.678 0.431 0.453 0.349 0.452 0.788 raining, validation and test for the Ayad A.H. Faisal Fig. 3. Training, validation and testing regression for the LMA algorithm. 4.2. Biosorbent Dosage One of the important parameters that strongly affect the sorption capacity is the biosorbent dosage. This effect was studied different boisorbent dosages into solution contaminated with 50 mg/l pH of 6 as shown in the Figure 4. These conducted with agitation speed of 250 rpm, min and T of 25ºC. This figure signified that efficiency was changed from 70 to 92 % response to the variation of olive pips 0.05 to 1 g/ 100 ml at conditions described previously. This can be resulted from the increasing of biosorption sites, i.e. prepared for contact with dissolved contaminant due to increase of biosorbent dosage reaching the equilibrium sorption increment in the biosorbent dosage do not cause any significant change in the removal efficiency and, so, 1 g/ 100 ml was considered choice for further experiments to investigat influences of other parameters. Fig good agreement between the predictions and experimental data with correlation coefficient 0.994. Al-Khwarizmi Engineering Journal, Vol. 12, No. 4 Training, validation and testing regression One of the important parameters that strongly affect the sorption capacity is the biosorbent studied by adding different boisorbent dosages into 100 ml of 50 mg/l of Cd(II) and These tests were with agitation speed of 250 rpm, t of 60 that the removal efficiency was changed from 70 to 92 % in olive pips dosage from / 100 ml at conditions described resulted from the i.e. surface area, prepared for contact with dissolved contaminant dosage [10]. Due to ion, additional the biosorbent dosage do not cause any significant change in the removal efficiency considered the best investigate the Figure 4 states a good agreement between the predictions and experimental data with correlation coefficient of Fig. 4. Variation of ANN outputs and experimental results with biosorbent dosage ( pH= 6, Co =50 mg/l, agitation speed = 250 rpm and T = 25 °C). 4.3. Contact time and initial pH of solution The sorption uptake is first minutes and, then, remains approximately stabilized as a function achieved the equilibrium Figure 5. This can be explained on the basis of availability large number of binding sites for metal ions which are decreas decrease can be caused slowed down sorption uptake due to generatio forces [11]. In addition, results proved that the removal efficiencies of metal ions w at pH of 2 and this may be due to high concentration of H+ ions which are competed with Cd(II) ions for binding sites. The pH means low concentrations of hydrogen ions and this reduces the competition between these ions and ions of pollutant, i.e. reduce of positive surface charge, resulting in the of removal efficiency from 11 to 92 % as pH changed from 2 to 5 [12]. The results showed the efficiency was decreas further increase of pH and this generation of soluble hydroxyl complexes which are precipitated from the studies of true sorption experimental results signified that the equilibrium can be achieved at t of 60 min. Also, explains that obtained results from the proposed ANN model and experimental data agreement with correlation coefficient not than 0.985. Khwarizmi Engineering Journal, Vol. 12, No. 3, P.P. 1- 9 (2016) ANN outputs and experimental sorbent dosage (t= 60 min, initial =50 mg/l, agitation speed = 250 rpm and ontact time and initial pH of solution is varied rapidly in the , then, remains approximately stabilized as a function of contact time until the equilibrium state as illustrated in can be explained on the basis of large number of binding sites for metal ions which are decreased with time. This decrease can be caused slowed down in the o generation of repulsive results proved that the removal ions were approximately low and this may be due to high ions which are competed with Cd(II) ions for binding sites. The higher values of means low concentrations of hydrogen ions this reduces the competition between these ions and ions of pollutant, i.e. reduce of positive surface charge, resulting in the rapidly increment of removal efficiency from 11 to 92 % as pH . The results showed that efficiency was decreased in response to any and this may be due to of soluble hydroxyl complexes which are precipitated from the liquid phase making true sorption impossible [13]. The signified that the equilibrium of 60 min. Also, Figure 5 explains that obtained results from the proposed and experimental data are in good agreement with correlation coefficient not less Ayad A.H. Faisal Fig. 5. Variation of ANN outputs and experimental results with t for different values of (dosage =1g/100 ml min, Co =50 mg/l, agitation speed = 250 rpm and T = 25 °C). 4.4. Initial concentration Figure 6 shows that the removal efficiency of Cd(II) onto the olive pips decreased from higher values (≈ 97 %) to lower values ( function of metal concentration. This logical because of the lack of sufficient sites that are required for sorption Cd(II) ions present in the liquid phase Conversely, the results explained that the removal efficiencies were higher with lower concentrations because all ions in aqueous phase with the binding sites. As a result, the tr yield can be increased by diluting the wastewaters containing high metal ion concentrations clear that there is a good agreement between outputs of ANN and experimental data with correlation coefficient of 0.997. Al-Khwarizmi Engineering Journal, Vol. 12, No. 5 Variation of ANN outputs and experimental for different values of initial pH =50 mg/l, agitation 6 shows that the removal efficiency of creased from higher %) to lower values (≈ 38 %) as a function of metal concentration. This trend is sufficient binding sorption much more liquid phase. he results explained that the removal lower concentrations aqueous phase can interact with the binding sites. As a result, the treatment yield can be increased by diluting the wastewaters containing high metal ion concentrations [14]. It is re is a good agreement between experimental data with Fig. 6. Variation of ANN outputs and experimental results with Co (dosage =1g/100 ml initial pH= 6, agitation speed = 250 rpm and °C). 4.5. Agitation speed Approximately 10 % of the Cd(II) was onto olive pips at agitation speed uptake increases with the increase of agitation speed up to 250 rpm at which maximum contaminant removal can be achieved illustrated in Figure 7. speed can be improved the diffusion of towards the biosorbent and can be developed between sites [15]. Figure 7 shows rpm is sufficient and there is no substantial change in removal efficiency beyond this value. However, this figure stated that the ANN model presents a good prediction for the correlation coefficient of 0.996 Khwarizmi Engineering Journal, Vol. 12, No. 3, P.P. 1- 9 (2016) Variation of ANN outputs and experimental dosage =1g/100 ml min, t =60 min, initial pH= 6, agitation speed = 250 rpm and T = 25 % of the Cd(II) was sorbed agitation speed of zero and the uptake increases with the increase of agitation to 250 rpm at which maximum contaminant removal can be achieved as . Increasing of agitation the diffusion of pollutants and the sufficient contact between the solutes and active shows that agitation with 150 and there is no substantial change in removal efficiency beyond this value. However, that the ANN model presents a good prediction for the experimental data with 0.996. Ayad A.H. Faisal Fig. 7. Variation of ANN outputs and experimental results with agitation speed (dosage =1g/100 t =60 min, initial pH= 6, Co= 50 mg/l and 4.6. Temperature Table 2 signifies that the sorption Cd(II) ions onto olive pips was varied from 85 to 96 %, when the corresponding temperature was changed from 20 to 60 ºC. This may be due to increasing the diffusion rate of the contaminant ions across the external boundary layer and the internal pores of the sorbent particles. that the adsorption process is an endothermic process [12]. However, optimization between cost of heating energy and the achieved increase of biosorption efficiency at high temperatures, temperature with value of 25 ºC was for batch tests. The results of Table 2 that the predictions are approaching from measured values with correlation co Table 2, Experimental and ANN output values of Cd(II) ions removal efficiency as a function of temperature (dosage = 1 g/100ml, t= 60 min, initial mg/l, and agitation speed= 150 rpm). Removal efficiency (%)Temp. (ºC) Exp. 85.84 20 92 25 94.3 30 95.12 40 95.52 50 96.3 60 Al-Khwarizmi Engineering Journal, Vol. 12, No. 6 Variation of ANN outputs and experimental dosage =1g/100 ml min, = 50 mg/l and T = 25 °C). he sorption uptake of varied from 85 to temperature was ºC. This may be due to the diffusion rate of the contaminant ions across the external boundary layer and the internal pores of the sorbent particles. This shows that the adsorption process is an endothermic optimization between the the achieved increase at high temperatures, ºC was very suitable Table 2 signified approaching from the correlation coefficient 0.9. Experimental and ANN output values of Cd(II) ions removal efficiency as a function of temperature initial pH= 6, Co= 50 Removal efficiency (%) ANN 87.54 90.77 92.5 92.47 94.62 95.75 4.7. Biosorption characteristics The biosorption data described by ANN model equilibrium concentration ( corresponding equilibrium sorption capacity ( A good agreement can be recognized predicted and measured sorption capacity equal to Fig. 8. Comparison of the experimental results with the qe values obtained by ANN model. 4.8. Sensitivity analysis The sensitivity analysis was the relative importance of the input based on the neural net weight matrix and Garson equation. Garson (1991) proposed an equation based on the partitioning of connection weights as follows (Faisal, 2015): � = ∑ �� �� ��� �∑ ������ ������ �×������� ∑ �∑ � ������ �∑ ������ �������������������� where Ij is the relative importance of the parameter on the output parameter the numbers of input and hidden neurons respectively, W's are connection weights, the superscripts i, h and o refer to input, hidden and output layers, respectively, and subscripts and n refer to input, hidden and output neurons, respectively. The results proved that the most influential parameter is initial pH of relative importance equal Khwarizmi Engineering Journal, Vol. 12, No. 3, P.P. 1- 9 (2016) haracteristics sorption data (Figure 8) were also described by ANN model in the terms of equilibrium concentration (Ce) and the corresponding equilibrium sorption capacity (qe). can be recognized between the values with maximum 10 mg/g. Comparison of the experimental results with values obtained by ANN model. nalysis he sensitivity analysis was aimed to calculate the relative importance of the input parameters based on the neural net weight matrix and Garson equation. Garson (1991) proposed an equation on the partitioning of connection weights as � ��×������ � � � �!×������ �" …(2) is the relative importance of the jth input parameter, Ni and Nh are the numbers of input and hidden neurons are connection weights, the refer to input, hidden and output layers, respectively, and subscripts k, m refer to input, hidden and output neurons, The results proved that the most influential initial pH of aqueous phase with to 38% for biosorption Ayad A.H. Faisal process under consideration (Fig parameter is followed by Co (17%), agitation speed (15%), dosage (14%), t (8 %), and the rank is temperature with relative (7%). However, the experimental ranges used for fitting ANN model, as proved by many researchers, are specified the influential variable and influence of each variable [16]. Fig. 9. Sensitivity analysis using artificial neural network. 5. Conclusions Best operating parameters for biosorption of Cd(II) ions from simulated wastewater onto olive pips were specified depended on batch tests and these values are; t of 60 min, initial pH of 6, biosorbent dosage of 1 g/ 100 ml, C agitation speed of 150 rpm and T results proved that the achieved maximum uptake and biosorbent capacity of 92 % and respectively. The ANN of three layers with transfer functions consisted of tangent sigmoid at the hidden layer and linear at the output layer is very efficient in the description of Cd(II) biosorption process onto olive pips with coefficient of correlation equal to 0.99798 and MSE of 0.00033. Finally, the results of sensitivity showed that the initial pH of aqueous phase is the most inf parameter governed this process with relative importance of 38 %. 6. References [1] M. A. Hashim, S. Mukhopadhyay, J. N. Sahu and B. Sengupta,"Remediation technologies for heavy metal contaminated groundwater". J. of Environmental Management, vol. pp. 2355-2388, 2011. Al-Khwarizmi Engineering Journal, Vol. 12, No. 7 (Figure 9). This (17%), agitation %), and the last importance of the experimental ranges used for as proved by many the influential variable Sensitivity analysis using artificial neural Best operating parameters for biosorption of Cd(II) ions from simulated wastewater onto olive pips were specified depended on batch tests and of 60 min, initial pH of 6, Co of 50 mg/l, of 25 ºC. The results proved that the achieved maximum uptake and biosorbent capacity of 92 % and 10 mg/g The ANN of three layers with transfer functions consisted of tangent sigmoid at the ut layer is very efficient in the description of Cd(II) biosorption process onto olive pips with coefficient of MSE of 0.00033. Finally, the results of sensitivity showed that the initial pH of aqueous phase is the most influential parameter governed this process with relative M. A. Hashim, S. Mukhopadhyay, J. N. Sahu and B. Sengupta,"Remediation technologies for heavy metal contaminated groundwater". J. of Environmental Management, vol. 92, [2] N. M. Hilal, I. A. Ahmed and R. E. El Activated and Nonactivated Date Pits Adsorbents for the Removal of Copper(II) and Cadmium(II) from Aqueous Solutions. International Scholarly Research Network ISRN Physical Chemistry, [3] A. Dabrowski, Z. Hubicki, P. Podkoscielny and E. Robens," Selective removal of the heavy metal ions from waters and industrial wastewaters by ion Chemosphere, vol. 56, pp. 91 [4] E. S. Z. El-Ashtoukhy, N. K. Amin, and O. Abdelwahab," Removal of lead(II) and copper(II) from aqueous solution using pomegranate peel as a new adsorbent", Desalination, vol. 223, pp. 162 [5] B. Volesky, and A. Leusch,"The influence of film diffusion on cadmium bio marine biomass. J. Biotechnology", vol. 43, pp. 1-10, 1995. [6] S. E. Samra, B. Jeragh, A. M. EL and A. A. El-Asmy,"Biosorption of Pb2+ from Natural Water using Date Pits: A Green Chemistry Approach", Mod. Chem. Appl., vol. 2, no. 13, 2014. [7] A. A. Faisal,"Modeling simulated wastewater by adsorption on eggshell using artificial neural network", Journal of the Association of Arab Universities for Basic and Applied Sciences, vol. 21, no. 2, pp. 1-20, 2015. [8] O. Hamdaoui and E. Naffrechoux,"Modelin of adsorption isotherms of phenol and chlorophenols onto granular activated carbon", J. Hazard. Mater., vol. 147, pp. 381 2007. [9] K. Yetilmezsoy and S. Demirel,"Artificial neural network (ANN) approach for modeling of Pb(II) adsorption from aqueous s Antep pistachio (Pistacia Vera L.) shells", J. Hazard. Mater., vol. 153, pp. 1288 2008. [10] B. M. W. P. K. Amarasinghe, R. A. Williams, "Tea waste as a low cost adsorbent for the removal of Cu and Pb from wastewater", Chem. Eng. J., vol. 132, 299–309, 2007. [11] G. O. El-Sayed, H. A. Dessouki and S. S. Ibrahim,"Biosorption of Ni(II) and Cd(II) ions from aqueous solutions onto rice straw", Chem. Sci. J., vol. 9, 1 [12] M. Alkan, B. Kalay, M. Do Demirbaş,"Removal of copper ions from aqueous solutions by kaolinite and batch design", J. Hazard. Mater., vol. 153, pp. 867 876, 2008. Khwarizmi Engineering Journal, Vol. 12, No. 3, P.P. 1- 9 (2016) N. M. Hilal, I. A. Ahmed and R. E. El-Sayed," Activated and Nonactivated Date Pits Adsorbents for the Removal of Copper(II) and Cadmium(II) from Aqueous Solutions. International Scholarly Research Network ISRN Physical Chemistry, 2012. A. Dabrowski, Z. Hubicki, P. Podkoscielny and E. Robens," Selective removal of the heavy metal ions from waters and industrial wastewaters by ion-exchange method", Chemosphere, vol. 56, pp. 91-106, 2004. Ashtoukhy, N. K. Amin, and O. Abdelwahab," Removal of lead(II) and copper(II) from aqueous solution using pomegranate peel as a new adsorbent", Desalination, vol. 223, pp. 162-173, 2008. B. Volesky, and A. Leusch,"The influence of film diffusion on cadmium bio-sorption by ss. J. Biotechnology", vol. 43, S. E. Samra, B. Jeragh, A. M. EL-Nokrashy Asmy,"Biosorption of Pb2+ from Natural Water using Date Pits: A Green Chemistry Approach", Mod. Chem. Appl., vol. A. A. Faisal,"Modeling of Cd(II) removal from simulated wastewater by adsorption on eggshell using artificial neural network", Journal of the Association of Arab Universities for Basic and Applied Sciences, vol. 21, no. 2, O. Hamdaoui and E. Naffrechoux,"Modeling of adsorption isotherms of phenol and chlorophenols onto granular activated carbon", J. Hazard. Mater., vol. 147, pp. 381–394, K. Yetilmezsoy and S. Demirel,"Artificial neural network (ANN) approach for modeling of Pb(II) adsorption from aqueous solution by Antep pistachio (Pistacia Vera L.) shells", J. Hazard. Mater., vol. 153, pp. 1288–1300, B. M. W. P. K. Amarasinghe, R. A. Williams, "Tea waste as a low cost adsorbent for the removal of Cu and Pb from wastewater", Chem. Eng. J., vol. 132, pp. Sayed, H. A. Dessouki and S. S. Ibrahim,"Biosorption of Ni(II) and Cd(II) ions from aqueous solutions onto rice straw", Chem. Sci. J., vol. 9, 1–11, 2010. M. Alkan, B. Kalay, M. Doğan, Ő. ,"Removal of copper ions from aqueous solutions by kaolinite and batch design", J. Hazard. Mater., vol. 153, pp. 867– Ayad A.H. Faisal Al-Khwarizmi Engineering Journal, Vol. 12, No. 3, P.P. 1- 9 (2016) 8 [13] C. Raji, G. N. Manju, T. S. Anirudhan, "Removal of heavy metal ions from water using sawdust-based activated carbon", Indian J. Eng. Mater., vol. 4, pp. 254-260, 1997. [14] H. J. Rao, G. Kalyani, K. V. Rao, T. A. Kumar, K. Mariadas, Y. P. Kumar, P. Vijetha, P. Pallavi, B. Sumalatha and K. Kumaraswamy, "Kinetic studies on bio- sorption of lead from aqueous solutions using eggshell powder", International Journal of Biotechnology and Biochemistry, vol. 6, no. 6, pp. 957-968, 2010. [15] J. Anwar, W. Zaman, U. Shafique, M. Salman, A. Dar and S. Anwar, "Removal of Pb(II) and Cd(II) from water by adsorption on peels of banana", Bio-resource Technology, vol. 101, pp. 1752-1755, 2010. [16] Y. Yang, X. Lin, B. Wei, Y. Zhao, and J. Wang, "Evaluation of adsorption potential of bamboo biochar for metal-complex dye: equilibrium, kinetics and artificial neural network modeling", Int. J. Environ. Sci. Technol., 2013. [17] J. Kabuba and A. M. Bafubiandi, "Comparison of equilibrium study of binary system Co-Cu ions using adsorption isotherm models and neural network", International Conference on Mining, Mineral Processing and Metallurgical Engineering, Johannesburg (South Africa). April 15-16, 2013. �ارزم� ا ������ ا ���ما��د � � ا ��ة ���� � ��د12���� ا )2016( 1- 9، ! � 3، ا 9 &% �$�#"� ازا' �ن ,��+��ام ()��% ا+�� �ى ا$ /�� %�0�� �� ا� � ا��$�ت ا &�دم��م م1 ا �% ا4!%����3�� ا �د $�!6 *ا��د � � ا ��ه ����� **زھ6اء � ���� ا����� **،* ����ا�� / �� ا������ !/ � )'�اد %$#" ayadabedalhamzafaisal@yahoo.com : ا��,�0 ا/�.-,و*( * Zahraa_saud@yahoo.com : ا��,�0 ا/�.-,و*( ** لخالصـةا � ا#-9از ا80*$ت ا�.$د#�8م > ; *8ى ا�09-8ن )�6-�5ام ا�3�.1 ا�"2��1 ا/%?@* �Aت$B�ث طEF G# 1*8.-@1 ا��<$�HI . ; < ١١٢ا>-@$دآ (,MA� "Nا�� �B0,H(��, ز#G ا�-@$س ، F6A ��) ٢٤٠-١٠(�A درا�B�� � @T 8ل ، د��U#$Tا� �� ا�@$دة ، �-, /# '�) ٢٥٠- ٥٠(ا�-,!�9 ا/)-�اW( ، ) ٦- ٢(ا��ا��@! � و در%� ا�T,ارة /دوره) ٢٥٠- ٠(�,>� ا/ھ-9از، #\ ١٠٠/]�) ١٥- ٠.٠١(ا�@$زه �*$F)٦٠-٢٠ ( �09� �(= �A ا�82Tل > ; ا> ; ا#-9از � .$د#�8م. � ٩٢(% �A/وف ا�@_ ; ا,aل ا�Eb G#�:٦٠ �B��/دورة ٢٥٠،#\١٠٠/]�١،�-,/# '�٥٠،د�B�09�٢٥،د�� �ودوال Tangent sigmoid و�� و%� ان .� c# )ا�-8ا� ; < ��� وا�5$ر%�d5@ا� �B�Hا� G# \.� ��<$�HI/1 ا�� � 3�.1 ا�"2��H508\ ا�T-ءة از ٧ا�$d.� ة��� /M0$د �8A"$ت %�N$! $ت�8م >2�ا�� ا�.$د# � و)���� 2A\ ا�; ٠.٩٩٧٩٨#c #"$#\ ارA�$ط �0$وي M�$"@ا� ���, ا/!�, > ; >@ F$-ا� i -@A ��Wا/)-�ا ��U#$Tا� �� أن ا��ا���$�Tا� \� TA ,٣٨أظ� % �,ا در%� ا�T,ارةbوا m�8ا� �F ا�@$ده ا�@$زه ��@! �F ا/ھ-9از �<,� �F )W9 ا/)-�ا���$ ا�-,! 0.