MATERIALS AND METHODS

The rice straw biomass was collected from rice fields in the rural areas outside the city of Banda Aceh, Indonesia. Prior to reducing its size, the rice straw was dried in the open air for a week. The dried rice straw was then ground in a milling machine and results were screened into three particle size categories of 20, 40 and 60 mesh. The moisture content of the samples was adjusted to 5, 10 and 15% by weight, respectively. A 10% starch was introduced into the sample to promote the binding among particles.

Proximate and ultimate analysis of the rice straw was performed at the Research and Development Center for Mineral and Coal Technology, Bandung, Indonesia. Proximate analysis provides information on the weight percent of the moisture, Volatile Matter (VM) when heated up to 950°C, Fixed Carbon (FC) and ash in a biomass material, whereas the ultimate analysis presents the weight percent of major elements such as carbon (C), hydrogen (H) and oxygen (O), as well as other elements of sulphur (S) and nitrogen (N). Both proximate and ultimate analysis was conducted in accordance with the recommendation of American Society for Testing Materials (ASTM). Determination of moisture content in dry sample was carried out consistent with ASTM D. 3173. Ash and volatile matter were analyzed in line with ASTM D.3174 and ISO 562, respectively. The fixed carbon was obtained through the following calculation: 100-total percentage of moisture content, ash and volatile matter. With respect to the ultimate analysis, elements of C, H and O were determined using ASTM standard method D. 5373 while total sulphur was performed along with ASTM standard D. 4239. Oxygen was calculated following Eq. 1 (Zhen, 1993):

(1)

where, C, H, N, S and Ash are carbon, hydrogen, nitrogen, sulphur and ash percentages in the biomass, respectively. Gross calorific value was measured using adiabatic bomb calorimeter following the ASTM standard D. 5685.

The independent variables being studied were densifying pressure X₁, particle size X₂ and moisture content X₃, keeping a 10% starch binder as a constant variable. The dependent variables analyzed were density, relaxation and durability of densified biomass produced. The densification experiments were conducted using a bench type manually operated laboratory hydraulic press having a capacity of 20 ton (Hydraulic Press Shop CMC ISO9002) and a densification die. The densification die was constructed from stainless steel cylinder of 30 mm in internal diameter and 250 mm in length, equipped with a stamp of 30 mm in external diameter. A Bob-Behnken design with three levels, low, medium and high coded as -1, 0 and +1 was applied to this study. The level values of each variable and code investigated in this study is presented in Table 1.

The density of the produced briquettes was found by a simple method as a ratio of weight and volume determined from the briquette geometric shape.

Table 1:	Experimental range and levels of independent variables

Table 2:	Bob-behnken design matrix along with experimental data predicted results

Relaxation in volume was measured after the briquettes were stored for a week, utilizing a vernier calliper. The durability of the produced solid fuels was measured in accordance with ASAE S269 (ASAE, 1996).

A number of 17 runs were randomly performed to optimize the process variable, as shown in Table 2 together with the experimental and predicted results of the dependent variables: the density, relaxation and durability of the densified rice straw biomass. The experimental data were analyzed by RSM using Design Expert software (Version 6.06, State-Ease Inc. Minneapolis, USA) to fit the second order polynomial relationship, as shown in Eq. 2:

(2)

where, Y is the predicted response and X₁, X₂ and X₃ are coded independent variables corresponding to the pressure, particle size and moisture content, respectively. Constants β_o, β_i and β_ij are linear term, quadratic term and cross product term coefficients, respectively. The coded values are related to the real values through Eq. 3:

(3)

where, Z is the coded value (-1, 0 or +1) and X is the corresponding original un-coded value, while X_o the mid value of the domain, ΔX represents as the increment of X for every unit of Z.

For the purpose of optimizing multiple response variables, it is necessary to establish the optimum criteria in accordance to the Desirability Function (DF) approach, as proposed by Derringer and Suich (1980). The maximum or minimum value of the variable response is determined on the basis of technical and/or economical considerations. The general approach is to first convert each response Y_k into an individual desirability function d_k = h (Y_k) that may vary over the range of 0≤d_k =1. If the response Y_k meets the goal or target value, then d_k = 1 and if the response falls beyond the acceptable limit, then d_k = 0. The individual desirability functions are then combined into a single composite response, the so-called Desirability Function (DF), defined in Eq. 4 as the geometric mean of the different d_k-values:

(4)

It is clear from Eq. 4 that DF will be close to 1.0 if all individual desirability functions are also close to 1.0. Therefore, DF = 1.0 implies that all response variables are at their respective optimum or target value condition. This type of methodology has been successfully applied for the optimization of mechanical densification process of agricultural crop residues for the production cattle feed (Munoz-Hernandez et al., 2006).

RESULTS AND DISCUSSION

Table 3 presented the results of proximate and ultimate analysis of rice straw from this study and compared with the one of Calvo et al. (2004). With the exception to the volatile matter in the proximate analysis, other data in both analyses of the two studies are comparable. With regard to HHV, the result obtained from Calvo et al. (2004) is slightly higher than that of this study, due to its higher carbon content. However, both HHVs are in the range of normal values, as shown by Huang et al. (2008) who studied 172 rice straw samples from around China and found out that the minimum and maximum values of HHV are 3051 cal g^-1 and 4000 cal g^-1, respectively.

Table 3:	Properties of rice straw biomass
aThis study, bCalvo et al. (2004)

Table 2 presented the design matrix in the coded units in conjunction with the experimental data and the predicted values of three response variables, the density, relaxation and durability of densified rice straw. The predicted values of the response were calculated from quadratic model fitting techniques utilizing Design Expert software. The experimental data, the density, relaxation and durability of solid fuel produced were utilized to develop the statistical model using multiple regression analysis method to fit the response function in accordance to Eq. 2. The resulted relationships between each response variable and independent variables of pressure, particle size and moisture content are presented in Table 4, where Y₁, Y₂ and Y₃ are density, relaxation and durability of the solid fuel produced, respectively.

The significance of the statistical model shown in Table 4 was evaluated by the F-test analysis of variance (ANOVA) presented in Table 5. In Table 4, the value of “Prob>F” less than 0.0500 revealed that the quadratic model of the response variable is significant at 95% confidence level. From inspection of the value of “Prob>F”, it can be concluded that all models representing the response variables are significant. The model showed a relatively high determination coefficient, R² and low the coefficient of variation C.V. These values are obtained as follows: R² = 0.8371 and C.V. = 3.12 for Y₁, R₂ = 0.9931 and C.V. = 4.82 for Y₂ and R² = 0.9697 and C.V. = 24.21 for Y₃. The closer the determination coefficient to unity, the better agreement of the model suits the experimental data, showing less the difference between the calculated and measured values.

Myers and Montgomery (2002) also suggested that the model adequacy can be evaluated not only from R² but also from adjusted R², predicted R² and prediction error sum of squares (PRESS). A good model is indicated by a large R² and a low PRESS. In this case, R² = 0.8371; adjusted-R² = 0.6276; predicted-R² = -1.6068; adeq precision = 7.761 and PRESS = 0.14 for Y₁. A negative "Pred R-Squared" implies that the overall mean is a better predictor of the response than the current model.

Table 4:	The fitted model equations
Y1 = Density, Y2 = Relaxation and Y3 = Dlurability of solid fuel

Table 5:	Analysis of variance (ANOVA) of the fitted models
Y1 = Density, Y2 = Relaxation and Y3 = Dlurability of solid fuel

"Adeq Precision" measures the signal to noise ratio. A ratio greater than 4 is desirable. The current study showed ratios of 4.801 for Y₁, 37.289 for Y₂ and 16.763 for Y₃, respectively which indicate an adequate signal, confirming each model can be used to navigate the design space.

It is clearly seen that the residual values are normally distributed on both sides of the line indicating that the experimental data are in excellent agreement with the predicted values. The above findings indicate outstanding adequacy of the proposed quadratic model to represent the variable responses of density, relaxation and durability of the densified rice straw in the range of pressure: 3000-7000 psi, particle size: 20-60 mesh and moisture content: 5-15%, respectively.

The normal probability plots showing the distribution of residual value defined as the difference between the predicted and experimental data for all response variables of density, relaxation and durability of the solid fuel are forming a straight line, as shown in Fig. 1a-c, respectively.

Fig. 1:	Normal probability plot of residuals for (a) density, (b) relaxation and (c) durability of rice straw briquette

OPTIMIZATION OF THE RESPONSE VARIABLE

The determination of optimum operating conditions for the density of densified rice straw is aimed at obtaining high quality of solid fuel and minimum operating cost of production. One feature of high quality densified solid fuel is indicated by its density of higher than 1.0 g.cm^-3. Minimum operating cost can be achieved when the pressure is at minimum (X1<5000 psi), particle size is at maximum (X₂ >40 mesh) and moisture content is at maximum (X₃>10%) levels. The second order polynomial model to represent all response variables, namely the density, relaxation and durability of densified rice straw, were utilized to optimize the operating conditions of the independent variables. This model is merely valid in the selected experimental domain. In this study, pressure, particle size and moisture content were chosen in the range of 3000-7000 psi, 20-60 mesh and 5-15%, respectively.

Applying the desirability function (DF) method, the Design Expert software produced a number of 10 solutions of which each has a DF = 1, as shown in Table 6. However, among 10 solutions only three meet the predetermined criteria. The first is 3244.0 psi for pressure, 58.08 mesh for particle size and 14.63% for moisture content. The second is 3002.8 psi for pressure, 34.90 mesh for particle size and 7.77% for moisture content and the last is 3912.4 psi for pressure, 48.39 mesh for particle size and 12.77% for moisture content.

The first, second and the third acceptable solutions yield the density of densified rice straw of 1.1856, 1.0390 and 1.1378 g cm^-3, respectively. As the pressure of the third solution is the highest among the available solutions, coupled with somewhat higher in moisture content, this solution yields the highest density of the product and the highest operational cost. The first solution has a higher pressure and finer particle size in comparison to the second solution. As a consequence, the first solution requires higher operational cost than that of the second solution.

Table 6:	Alternative solutions that meet DF=1 for optimization of process parameter

Although the second solution produces the lowest briquette density, it gives the lowest operational cost. In addition, the density produced by the second solution is still higher than the minimum density of 1.0 g cm^-3 required for producing a solid fuel briquette. It is therefore appropriate to select the second acceptable solution as the optimum.

CONCLUSION

A desirability function approach has been utilized to optimize the process variables of pressure, particle size and moisture content on the multi-response variables of density, relaxation and durability of solid fuel produced through a mechanical densification of rice straw biomass. The optimum conditions to produce solid fuels from rice straw biomass were obtained at a pressure of 3002.8 psi, the particle size of 34.90 mesh and raw material water content of 7.77%. With a minimum number of experimental runs, this technique is an efficient way for solution of optimization problems.

ACKNOWLEDGMENT

The authors wish to express their gratitude to Directorate General of Higher Education (DGHE), Ministry of National Education (MONE) of Republic of Indonesia for providing travel grant under PAR-C project to the first author for the preparation of the manuscript at the University of Leeds, UK.