Artificial Intelligence in Agriculture 3 (2019) 52–61

Contents lists available at ScienceDirect

Artificial Intelligence in Agriculture

journal homepage: http://www.keaipublishing.com/en/journals/artificial-
intelligence-in-agriculture/

Survey on solid wastes management by composting: Optimization of key

process parameters for biofertilizer synthesis from agro wastes using
response surface methodology (RSM)
Christian O. Asadu a,⁎, Samuel O. Egbuna a, Thompson O. Chime a, Chibuzor N. Eze b, Dibia Kevin c,
Gordian O. Mbah a, Anthony C. Ezema b
a
Department of Chemical Engineering, Enugu State University of Science and Technology, P.M.B 01660, Enugu State, Nigeria
b
Department of Microbiology, University of Nigeria, Nsukka, Enugu State, Nigeria
c
Department of Chemical Engineering, Nnamdi Azikiwe University, P.M.B 5025 Awka, Anambra State, Nigeria

a r t i c l e i n f o a b s t r a c t

Article history: The optimization of key process parameters for the transformation of agro wastes into biofertilizer has been dem-
Received 10 October 2019 onstrated using response surface methodology (RSM). Biofertilizer was produced by composting using 120 L ca-
Received in revised form 14 December 2019 pacity drum made of polyethylene as the composter. Composting time (X1), dosage ratio (X2) and moisture
Accepted 16 December 2019
content (X3) were the independent factors while percentage nitrogen, phosphorus and potassium (N.P.K)
Available online 20 December 2019
were the response factors. The outcomes exhibited that composting time, dosage ratio and moisture content
Keywords:
all significantly affects the mineralization of N.P.K at probability value of 0.0001. The coefficients of determination
Optimization also called regression coefficients of 98.60%, 99.79% and 97.80% for nitrogen, phosphorus and potassium observed
Composting between the predicted and the real value are obvious that the developed regression models can fit the experi-
Sawdust mental data well. It was seen from the optimization studies that the pinnacle value of N.P.K from the ideal con-
Sewage sludge ditions are 9.62%, 8.97% and 5.62. Characterization of the composite uncovered that biofertilizer produced has a
Vegetable wastes high potential for commercial application on agricultural land. It can be concluded that combination of sawdust,
Response surface methodology sewage sludge and vegetable waste is a good mixture for biofertilizer synthesis. Also, the nutrients release by the
compost materials during the process of composting may be maximized when process conditions are circum-
spectly managed within the reported optimal value.
© 2019 The Authors. Production and hosting by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction been a major issue to most urban areas in Nigeria. Aside Pyrolysis, burn-
ing in incinerator and storage, a large portion of the natural wastes from
In any emerging or modern society, waste management and control human, creature, rural and mechanical foundations presenting genuine
is dependably a characteristic piece of such society. Wellsprings of resi- ecological and medical issues can be overseen by anaerobic/aerobic fer-
dential and commercial wastes have developed widely in Nigeria over mentation producing biogas and biofertilizer individually through
the previous decade. The by-product of agricultural exercises is gener- proper composting. These procedures are exceptionally worthwhile re-
ally alluded to as “Agricultural Wastes” (Westerman and Bicudo, membering that, they limit harm to the earth and create monetarily sig-
2005). These wastes essentially appear as harvest deposits (leftover nificant items from wastes (Ogazi and Omueti, 2000; stentiford, 1996).
stalks, straw, leaves, roots, husks, shells etcetera) and animal waste (ex- As indicated by (Stentiford, 1996; Dumitrescu et al., 2009), Composting
crement). Agricultural wastes can be overseen by changing over to al- is defined as biological oxidative degradation of organic matter in
ternative manures (bio-fertilizer) through proper composting (Oltjen wastes under controlled conditions which allows the growth of micro-
and Beckett, 2006). The management of solid wastes has dependably organisms that convert biodegradable natural wastes into an end prod-
uct that is adequately stable for application in agricultural land without
Abbreviations: RSM, response surface methodology; CCD, central composite design; N, antagonistic ecological impacts. The end product of composting has
nitrogen; P, phosphorus; K, potassium; ASTM, American Society for Testing and Materials; been observed to be more valuable to plants and soil biodiversity
ANOVA, analysis of variance; CV, coefficient of variation; TOC, total organic carbon; R2, co- Rasapoor et al.,(2009). Carbon dioxide, water, mineral particles and
efficient of regression or coefficient of determination.; AAS, atomic absorption
spectrometer.
humus are the primary products of aerobic composting. The procedure
⁎ Corresponding author. decimates pathogens due the accelerated temperature by microbial ex-
E-mail addresses: [email protected], [email protected] (C.O. Asadu). ercises. Likewise, nitrogen fixing bacteria converts nitrogen from

https://doi.org/10.1016/j.aiia.2019.12.002
2589-7217/© 2019 The Authors. Production and hosting by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61 53

unstable ammonia to stable natural structures. The procedure for the were changed by giving 10 mm equidistant gaps in six layers on the pe-
most part decreases the volume of waste. For the reason that riphery of the drums utilizing a hand driller to encourage the air dissem-
composting is an effective strategy for reusing waste, this investigation ination inside the drums. Two inspecting windows (one each at center
has an extraordinary importance for most developed and developing and base part) were given in the drums to gather the intermittent sam-
nation that needs to deal with, consistently, bigger amounts of biode- ples for investigation. The arrangements for the leachate collection from
gradable waste, for example, a waste administration organization the base of the drums were likewise given.
(Enugu State Waste Management Agency) in Enugu State Nigeria over-
saw both as local waste (86,883 tons in 2010) and sewage sludge 2.3. Composting operations
(53,750 tons in 2010). Sludge rich in natural and mineral mixes, for ex-
ample, (nitrogen and phosphorus) and in lipids is created by treating The composting process was carried out in open space to allow the
modern effluents. At the point when sludge is arranged without treat- natural aeration. The drum was supported on the bricks and the plastic
ment, it turns into a wellspring of contamination (Ilegbune, 2006). tray was kept below the drum for the collection of leachate. For the
With the end goal to avert genuine natural issues, for example, sullying study, 90 kg of the homogenized wastes samples at different dosage ra-
of groundwater by leachate, contamination of the air with foul gases tios as shown in Table 3 were added into the drums. Aeration was
and so forth, Sludge must be stabilized by chemical, organic and bio- achieved by manual turning of the composts once daily. Various operat-
chemical methods (Mowoe, 2001; Jimoh, 2005). ing and product quality parameters such as pH, temperature, organic
Optimization is a strong decision making which is the act of produc- matter and total carbon were monitored as the compost last. Samples
ing the best results under certain conditions (Datta, 2011). The main were taken from the drum after every 5 days for laboratory analysis.
components of optimization are; (i) The objective function, (ii) The var- The moisture content was monitored and maintained using electronic
iables and (iii) The constraints. Optimization problems can be stated to moisture meter (Reotemp 648(800) San Diego CA).
maximize or minimize that is subject to the constraints. Optimization
problems could be constrained optimization problems which are sub- 2.4. Physiochemical characterization of the samples
ject to constraints or unconstrained optimization problems which is
subject to no constraints considering the nature of the equations for The ASTM D2974-07 was used in the analysis of the percentage com-
the objective functions and constrains. Optimization problems can be position of organic matter, ash content and moisture content (ASTM,
linear, non linear, geometric and quadratic programming problems. 2007). The nitrogen content was estimated by Kjeldahl's method
Essential of process streamlining of key process parameters involved Sobiecka et al. (2007). Phosphorus and potassium contents were ana-
in the conversion of agro wastes cannot be overemphasized. The reason lyzed using Atomic Absorption spectrophotometer (AAS) (Model
for statistically designing an experiment is to gather regular connection 2010, VGP manufacturer USA). pH was measured in the filtrate solution
between different components influencing the procedure towards find- using pH-meter 340I/SET (Texcare Instrument, New Delhi India, Preci-
ing the most appropriate conditions. Process optimization was done in sion /sensitivity 0.01/−59.16 mV/pH @25°C). Method for the determi-
this work using response surface methodology (RSM). It is a statistical nation of total organic carbon (TOC) in soil and sediment was used for
tool used mainly for optimization. RSM is for the most part utilized to organic carbon content determination (Schumacher, 2003).
study about the impact of independent factors on the response(s). It is
additionally used to ponder the impacts of multiple variables and their 2.5. Statistical analysis and mathematical modeling
cooperation. These different elements are the independent factors
while the response(s) are the dependent variables (Datta, 2011). RSM In order to examine if there is a relationship between the dependent
relates item properties by utilizing regression equation that portrays in- and independent variables, the data gathered were subjected to regres-
terrelations between information factors and item properties (Adeyanju sion analysis utilizing response surface methodology of Design expert
et al., 2016). The most prevalent and regularly utilized type of RSM is the version 8.0.7.1. Regression analysis was utilized to show a response
central composite design (CCD) and Box Behnken design. In this work, (Yi) as a scientific capacity of a couple of consistent elements. Every re-
central composite design was utilized for the process optimization. sponse (Yi) was represented by mathematical equation that relates the
These designs are rotatable or near rotatable. The factors studied were response surfaces. The response was represented as second-order poly-
composting time, moisture content (water solid ratio) and dosage nomial equation as indicated by Eq. (1).
ratio. Therefore, this research aims at contributing to waste manage-
ment by converting some agro waste into biofertilizer through k k k k
Y i ¼ f ðyÞ ¼ β0 þ ∑i¼1 βi X i þ ∑i¼1 βii X 2i þ ∑i¼1 ∑i¼1 βij X i X j þ ε ð1Þ
composting and afterward, optimize the key process parameters in
composting using response surface methodology.
where Yi is the predicted response used to relate the independent vari-
2. Materials and method ables, k is the number of independent variablesXi(i = 1, 2,3); while β is
a constant coefficient and βi, βij and βiiis the linear, interaction and
2.1. Materials square terms respectively and ε is the random error term. Multivariant
regression analysis with model Eq. (1) was carried out on data using de-
Raw materials used in this research were; sawdust, dewatered sew- sign expert 8.0.7.1 software to yield Eq. (2) which was used to optimize
age sludge and vegetable wastes (comprising of different left over of
fresh green leaves, watermelon, cabbage, lettuce, cucumber etc). The Table 1
sawdust was sourced from local sawmill, the vegetable wastes was Coded and un-coded values of the independent factors.
sourced from the local market and were shredded and homogenized Code Actual value of
with a cutter to improve decomposition during composting. Dewatered independent
sewage sludge was sourced from waste water treatment plant of Wilson variable
Industry Nigeria Limited Nsukka. −α Xmin
−1 ðα−1ÞX max þ ðα þ 1ÞX min
2.2. Design of composting drums 2
0 X max þ X min
2
The composting was done in polyethylene drums of 120 L capacity. +1 ðα−1ÞX min þ ðα þ 1ÞX max
The drums were reasonably changed for air dissemination. The drums 2
54 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61

Table 2
Factor levels of independent variables for the synthesis of biofertilizer.

Independent factors −α (Axial) Low level Medium level High level +α (Axial)

−1.68 −1 0 +1 + 1.68

Process duration X1 21.59 25 30 35 38.41

Dosage ratio X2 2.32 (1:2:1) 3 (1:1:1) 4 (2:1:1) 5 (2:2:1) 5.68 (3:1:1)
Moisture content X3 46.59 50 55 60 63.41

the product responses. moisture content (50–60%). Every single other parameter were kept
constant. The working extents and five institutionalized dimensions
Y ¼ β0 þ β1 X 1 þ β2 X 2 þ β3 X 3 þ β11 X 21 þ β22 X22 þ β33 X 23 þ β12 X 1 X 2 were built up after a few fundamental runs. In light of CCD, the test
þ β13 X 1 X 3 þ β23 X 2 X 3 þ ε ð2Þ runs contain 20 trials (8 factorial points, 6 center points and 6 axial
points). Every one of the treatment were perfumed in randomized
The model developed for every determination was then inspected order. RSM and second order of three factors (composting time (X1),
for significance and lack of fit, while response surface plot was struc- dosage ratio (X2) and moisture content (X3), five level combination
tured after removal of the non-significance terms with the same soft- coded as −1.68, −1, 0, +1, and +1.68 as demonstrated by
ware. RSM was utilized in enhancing the procedure parameters for (Nahemiah et al., 2015; Snedecor and Cochran, 2008), was adopted to
composting. The composting times (X1), dose ratio (X2) and moisture decide the impacts of the independent factors on the responses (N.P.
content (X3) were the independent variables studied to optimize the K). Utilizing the coded dimensions, the typical dimensions were com-
yield (Y) of nitrogen, phosphorus and potassium. The independent fac- puted and delineated in Table 4, including 20 trial runs and diverse for-
tors were coded to lie at ±1 for the factorial point, 0 for the center point mulation compositions. The independent factors having the least P-
and ± balphaN for the axial points as demonstrated in Table 1. The fac- value (or the most noteworthy F-ratio) demonstrates the most critical
tors were signified by Eq. (3); (P b 0.05) impact on the reliant factors (Samaram et al., 2015; Yolmeh
et al., 2014). The non-significant terms (P N 0.05) were expelled from
X i −X j the created model in this investigation, with the exception of in a cir-
Zj ¼ ; i ¼ 1; 2; 3 ð3Þ
Δj cumstance whereby a quadratic or collaboration impact including that
factor would be critical (Samaram et al., 2015). Examination on the pro-
where Zj are the coded values of the independent variables; while Xi and ductivity of the model was finished by the assurance of the number of
Xj are its real values and real value at the central point respectively. Δj is significant terms, regression equation P-value, lack of fit P-value and co-
the step change of the variable Xi. The coded levels of the independent efficient of regression (R2) Yolmeh et al., 2014.
variables used in the RSM design were as shown in Table 2. The outcome with R2 values near 1 demonstrates that the model is
more exact. The high value of adjusted and predicted coefficient of de-
2.6. Statistical design and data interpretation termination is additionally a sign of the sufficiency of the model fits
for the experimental data, Yolmeh et al. (2014). The process optimiza-
The central composite design (CCD) was utilized to contemplate the tion was done utilizing graphical and numerical advancement methods
impacts of the factors towards their responses and subsequently in op- to decide the optimum composting conditions. The interaction impact
timization studies. This technique is appropriate for fitting a quadratic was likewise considered utilizing the three dimensional (3D) surface
surface and it improves the viable parameters with a minimal number plots obtained from the final model (Samaram et al., 2015). The yield
of experiments, and in addition to investigate the association between of N.P.K would be compared with the predicted response values got
the parameters. In order to describe the impacts of composting time, from the final reduced model in order to demonstrate both the viability
dosage ratio, and moisture content on the synthesis of biofertilizer,
batch experiment were performed which was dependent on the CCD. Table 4
In order to define the experimental range, preliminary experiment Real value of independent factors with responses.
were first performed. As the structure ranges were built up, they were
Std Run Factor Factor 2 Factor 3 Response Response 2 Response3
coded to lie at ±1α for the factorial point, 0 for the center point 1 X2: X3 : 1 Phosphorus Potassium
and ± 1α for the axial points. The codes were ascertained as an element X1: Dosage Moisture Nitrogen (%) (%)
of the scope of enthusiasm of each factor as appeared in Table 1. In this Time ratio (w: content (%)
investigation, a small scale composting was conducted utilizing a mix- (days) w) (%)

ture of sawdust, dewatered sewage sludge and vegetable wastes to cre- 9 1 21.59 4.00 55.00 9.62 8.01 4.64
ate biofertilizer. The plan depended on five dimensions of the three 3 2 25.00 5.00 50.00 9.45 8.27 5.08
14 3 30.00 4.00 63.41 8.61 7.28 4.49
factors as appeared in Table 2. The independent factors considered
1 4 25.00 3.00 50.00 9.01 7.43 3.95
were composting time (25-35 days), dosage ratio (3–5 w:w) and 17 5 30.00 4.00 55.00 9.31 8.63 4.45
12 6 30.00 5.68 55.00 9.21 7.87 5.62
Table 3 5 7 25.00 3.00 60.00 9.24 8.05 5.41
Physiochemical properties of the raw materials before composting. 16 8 30.00 4.00 55.00 9.31 8.61 4.45
6 9 35.00 3.00 60.00 9.38 8.97 5.13
Parameters Organic waste 4 10 35.00 5.00 50.00 8.23 7.08 4.97
20 11 30.00 4.00 55.00 9.31 8.62 4.45
Sawdust Vegetable waste Sewage sludge
2 12 35.00 3.00 50.00 8.74 7.73 4.54
Moisture content (%) 23.3±6.5 37.8±6.5 33.6±6.5 13 13 30.00 4.00 46.59 8.03 6.91 4.18
Total Organic carbon (%) 58.2±0.5 23.6±0.5 24.2±0.5 11 14 30.00 2.32 55.00 9.42 8.9 5.27
Ash content (%) 24.3±0.1 17.8±0.1 22.8±0.1 19 15 30.00 4.00 55.00 9.31 8.64 4.45
Nitrogen (N) (%) 0.9±0.1 4.34±0.1 3.36±0.1 10 16 38.41 4.00 55.00 8.72 7.78 4.36
Phosphorus (P) (%) 1.04±0.1 6.4±0.1 3.2±0.1 15 17 30.00 4.00 55.00 9.31 8.7 4.45
Potassium (K) (%) 0.6±0.1 4.7±0.1 1.89±0.1 18 18 30.00 4.00 55.00 9.31 8.63 4.45
pH 5.87±1.0 7.15±1.0 7.23±1.0 7 19 25.00 5.00 60.00 9.03 7.49 5.37
Organic matter (%) 65±1.0 77±1.0 86.4±1.0 8 20 35.00 5.00 60.00 8.6 6.7 4.47
 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61 55

and unwavering quality of the regression fitted for the expected re- the Predicted R-squared value. The synopsis of P-value shows that a
sponses (Samaram et al., 2015). quadratic model fitted the ANOVA examination and subsequently it
was recommended (Table 5). The linear and 2FI models were not pro-
3. Results and discussion posed. The Cubic model is constantly associated on the grounds that
the CCD does not contain enough runs to help a full cubic model (Filli
3.1. Nutrient composition in the raw materials et al., 2010; Trinh and Kang, 2010). An significant level of 95% was uti-
lized henceforth all terms whose P-value are b0.05 are viewed as signif-
Percentage composition of nitrogen in sawdust as shown in Table 3 icant terms.
showed that sawdust must be blended with other nitrogen rich organic The F-value tests were performed utilizing the ANOVA to ascertain
wastes before composting to maintain some level of nitrogen in the the significance of each sort of model. Besides evaluating the signifi-
compost which would sustain the organisms responsible for biodegra- cance, the adequacy of the models was evaluated by applying the
dation. Sewage sludge and vegetable wastes contained reasonable per- lack-of-fit test. This test is utilized in the numerator in an F-trial of the
centage nitrogen 4.34% and 3.36% as shown in Table 3. The values agree null hypothesis and shows that a proposed model fits well or not. The
closely with the work done by Okon (2000), which reported that the test for lack of-fit contrasts the variation around the model with pure
value of nitrogen in dewatered sewage sludge and vegetable wastes be- variation within the replicated observations. This test estimated the am-
fore aerobic composting are 3.85% and 4.79%. These results showed that pleness of the diverse models dependent on response surface investiga-
these wastes can supply the amount of nitrogen which can initiate the tion (Manpreet et al., 2011). Henceforth, the quadratic model with the
growth of microbes and enhance biodegradation. The organic matter most reduced insignificant model lack of fit was proposed.
was observed to be high in all the raw materials as presented in Table 6 for N.P.K demonstrates the regression coefficients of the in-
Table 3. The results agree with the report by Gajalakshmi and Abbasi tercept, linear, quadratic and interactive terms of the models. The out-
(2008); which states that high percentage of organic matter in waste comes demonstrated that over 98% of the general framework factors
samples is an indication that they are good substrates for bio-fertilizer can be explained by the quadratic model equations (Table 4). The signif-
production. icance of every coefficient in the models was checked from the P-value
(P b 0.05) of the terms. The lower the models P-value (higher F-value)
3.2. Experimental design and model formulations the better the significance of the input variable effect on the responses
(Shrivastsvs et al., 2008).
Table 2 shows the experimental parameters, ranges and level of in- From Table 6, the P-value for the models were (b0.0001) for nitro-
dependent variables examined in this work and the results are shown gen and additionally phosphorus and potassium which meant high sig-
in Table 4. As regards to regression analysis, model fitting is the process nificance in the prediction of the response factors and also the model
of developing a probabilistic model that best describes the relationships appropriateness. The F-value was 59.37, 536.89 and 49.31 for nitrogen,
between the dependent and independent variables. RSM was applied in phosphorus and potassium individually. These qualities were moder-
developing the model and optimization of the process by first ately high, accordingly showing that the models were exceptionally sig-
performing series of experimental runs (Table 4) to adequately and re- nificant at above 95% confidence level. Their P-value built up the
liably measure the variables response before developing mathematical importance of the considerable number of coefficients as appeared in
model of second order response surface best fit, and finally determine Table 6. From Table 6, every single liner term of time (X1), dose ratio
the optimal set of experimental parameters producing the optimal re- (X2) and moisture content (X3), the quadratic term of moisture content
sponse value Damirel and Kayan (2012). In this study, effect of (X23) and in addition the interactive term among time and dosage ratio
composting time (X1), dosage ratio (X2), and moisture content (X3) (X1X2), time and moisture content (X1X3) and between dose ratio and
and their interactions each at three levels on the yield of N.P.K were in- moisture content were all significant with P-value b 0.05 for nitrogen.
vestigated. Observed response data (in triplicates) from experimental From Table 6 additionally, all linear terms of time (X1), dose ratio (X2)
runs (Table 4) were used to develop models (Table 5) using least square and moisture content (X3), the quadratic term of time (X21), dosage
techniques as described by Filli et al., (2010). The three (3) response ratio (X22), moisture content (X23) and the interactive term among time
variables (nitrogen N, phosphorus P and potassium P) were correlated and dose proportion (X1X2), time and moisture content (X1X3) and be-
with the independent variables using the second order polynomial as tween dose ratio and moisture content were all significant with P-
represented by Eq. (2). X1, X2, and X3 represents the composting time, values b 0.05 for potassium, while every single liner term of time (X1),
dosage ratio and moisture content respectively. The coefficient with dose ratio (X2) and moisture content (X3), the quadratic term of dose
one factor (X1, X2, and X3) represents the sole effects of that particular proportion (X22) and also the interactive term among time and dosage
factor, while coefficients with two factors (X1X2, X1X3 and X2X3) and ratio (X1X2), time and moisture content (X1X3) and between dosage
those with second order terms (X11, X22, and X33) represents the inter- ratio and moisture content (X2X3) were all significant with P-value
action between the three factors and the squared effect respectively. A under 0.05 for potassium. The values of the coefficient of determination
positive value of the regression terms indicates a synergistic effect, (R2) were 98.16% for nitrogen, 99.79% for phosphorus and 97.80% for
while negative sign indicates an antagonistic effect Filli et al., (2010). potassium, accordingly a sign that the models fit the experimental
data. Likewise, the values of the adjusted coefficient of regression (Adj
3.2.1. ANOVA analysis and model fitting R2) and predicted coefficient of regression (Pred R2) were 96.51% and
The Analysis of Variance (ANOVA) was utilized to translate the cen- 85.24% for nitrogen, 99.61% and 98.67% for phosphorus and 95.81%
tral composite design. The nitty gritty table of insights looks at the Se- and 83.26% for potassium individually which indicates the model's sig-
quential P-value, the Lack of fit P-value, the adjusted R-squared and nificance and sensible accuracy of the fitted models respectively.

Table 5
Second order polynomial equation obtained for the three responses.

Response variables Second order polynomial models Regression coefficient

R2 R2adj

N +9.31-0.24 * X1-0.10 * X2+0.13*X3-0.19*X1*X2+0.15*X1*X3-0.11*X2*X3-0-0.34*X23 98.16 96.51

P +8.64-0.084*X1-0.32*X2+0.097*X3-0.40*X1*X2+0.13*X1*X3-0.38*X2*X3-0.27*X21-0.095*X22-0.55*X23 99.79 99.61
K +4.45-0.086*X1+0.11*X2+0.17*X3-0.16*X1*X2-0.21*X1*X3-0.28*X2*X3+0.37*X23 97.80 95.81
56 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61

Table 6 The tests for adequacy of the regression models, significance of indi-
Analysis of variance (ANOVA) for full quadratic model for the response variables. vidual of model coefficients and the lack of fit test were performed uti-
Source Sum of df Mean F value P-value lizing the same statistical package. The P-value were utilized as an
squares square Prob N F apparatus to check the essentialness of every one of the coefficients,
Nitrogen which thus are important to comprehend the example of the common
Model 3.45 9 0.38 59.37 b 0.0001 collaborations between the test factors (Shrivastsvs et al., 2008). Higher
X1-Time 0.79 1 0.79 123.08 b 0.0001 the F-test value and small P-value indicates high significance of the re-
X2-Dosage ratio 0.15 1 0.15 22.66 0.0008
lating coefficient (Taran and Aghaie, 2015). The adequate precision
X3-Moisture content 1 0.24 36.57 0.0001
0.24 measures the signal to noise ratio and compares the range of the pre-
X1 X 2 0.29 1 0.29 44.75 b 0.0001 dicted value at the design points to the average prediction error. The
X1 X3 0.18 1 0.18 27.89 0.0004 adquate predicion ratio above 4 indicates adequate model efficacy
X2 X3 0.11 1 0.11 16.39 0.0023 (Taran and Aghaie, 2015). Hence, the adquate precision ratios of
X21 0.022 1 0.022 3.47 0.0921
X22 2.022E-003 1 2.022E-003 0.31 0.5880
25.996, 71.078 and 25.534 for N.P.K indicate adquate signal. This indi-
X23 1.67 1 1.67 258.05 b 0.0001 cates that an adequate relationship of signal to noise ratio exists. The
Residual 0.065 10 6.454E-003 C.V called coefficient of variation which is defined as the ratio of the
Lack of Fit 0.28 10 0.028 0.19 0.9920 standard deviation of estimate to the mean value of the observed re-
Pure Error 0.000 5 0.000
sponse is independent of the unit. It is also a measure of reproducibility
Cor Total 3.51 19
Std. Dev. 0.080 R-Squared 0.9816 and repeatability of the models (Chen et al., 2011). The calculations in-
Mean 9.06 Adj R-Squared 0.9651 dicated the C.V value of 0.89% for nitrogen, 0.54% for phosphorus and
C.V. % 0.89 Pred R-Squared 0.8524 1.99% for potassium which showed that the model can be considered
PRESS 0.52 Adeq Precision 25.996 reasonably reproducible (because its CV was not N10%) (Chen et al.,
Phoshorus
2011). The response values obtained by inserting the independent
Model 9.22 9 1.02 536.89 b 0.0001
X1-Time 0.096 1 0.096 50.48 b 0.0001 values are the predicted values of the model. These values are compared
X2-Dosage ratio 1.40 1 1.40 733.69 b 0.0001 to the actual experimental values. The result of this comparison is
X3-Moisture content 1 0.13 67.10 b shown in the Table 7. From the table, it is seen that there is a close cor-
0.13 0.0001
relation between the actual experimental response and the predicted
X1 X 2 1.28 1 1.28 670.91 b 0.0001
X1 X3 0.13 1 0.13 68.17 b 0.0001
response. This confirms the effectiveness of the process for biofertilizer
X2 X3 1.14 1 1.14 597.55 b 0.0001 synthesis.
X21 1.04 1 1.04 543.13 b 0.0001
X22 0.13 1 0.13 68.04 b 0.0001 3.2.2. Model adequacy check
X23 4.38 1 4.38 2293.21 b 0.0001
It is very important in RSM that the developed models (Table 5) pro-
Residual 0.019 10 1.908E-003
Lack of Fit 0.014 5 2.799E-003 2.75 0.1453 vide an adequate approximation for application in real system, and
Pure Error 5.083E-003 5 1.017E-003 there are principally two methods used for this check, these are graph-
Cor Total 9.24 19 ical and numerical method Filli et al. (2010). The graphical technique
Std. Dev. 0.044 R-Square 0.9979
considers the idea of the nature of the residuals (distinction between
Mean 8.02 Adj R-Squared 0.9961
C.V. % 0.54 Pred R-Squared 0.9867
the observed values and its fitted) of the model while the numerical
PRESS 0.12 Adeq Precision 71.078 method utilizes the coefficient of determination (R2) and adjusted R2
Potassium (R2adj). For the most part, it is imperative to check the fitted model to
Model 3.89 9 0.43 49.31 b 0.0001 guarantee that it gives the estimate to the genuine framework. On the
X1-Time 0.10 1 0.10 11.46 0.0069
off chance that the model does not demonstrate a sufficient fit, further
X2-Dosage ratio 0.15 1 0.15 17.54 0.0019
X3-Moisture content 1 0.41 46.62 b examination and improvement of the fitted response surface may give
0.41 0.0001
X1 X2 0.22 1 0.22 24.87 0.0005
X1 X2 0.34 1 0.34 39.33 b 0.0001 Table 7
X2 X3 0.64 1 0.64 72.90 b 0.0001 Responses with predicted values of nitrogen for biofertilizer synthesis.
X21 0.019 1 0.019 2.12 0.1764
X22 1.97 1 1.97 225.23 b 0.0001 Standard Nitrogen Phosphorus Potassium
X23 7.276E-003 1 7.276E-003 0.83 0.3835 order
Residual 0.088 10 8.758E-003 Actual Predicted Actual Predicted Actual Predicted
Lack of Fit 0.019 5 3.787E-003 2.96 0.1295 value value value value value value
Pure Error 0.000 5 0.000
Cor Total 3.97 19 1 9.01 9.00 7.43 7.38 3.95 3.98
Std. Dev. 0.094 R-Squared 0.9780 2 8.74 8.60 7.73 7.76 4.54 4.56
Mean 4.71 Adj 0.9581 3 9.45 9.40 8.27 8.30 5.62 5.67
R-Squared 4 8.23 8.24 8.05 7.07 4.97 5.00
C.V. % 1.99 Pred 0.8326 5 9.24 9.19 7.08 8.08 5.41 5.31
R-Squared 6 9.38 9.39 8.97 8.96 5.13 5.05
PRESS 0.67 Adeq 25.534 7 9.03 9.14 7.49 7.48 5.37 5.29
Precision 8 8.62 8.57 6.70 6.77 4.47 4.37
9 9.62 9.60 8.01 8.02 4.64 4.69
10 8.72 8.79 7.78 7.74 4.36 4.40
11 9.21 9.52 8.90 8.91 5.27 5.32
12 8.03 8.06 7.87 7.83 5.08 5.09
Consequently, from these outcomes, it could be recommend that both 13 8.61 8.13 6.91 6.92 4.18 4.09
14 9.31 8.57 7.28 7.24 4.49 4.60
liner and quadratic terms were the primary deciding factors for the
15 9.28 9.31 8.70 8.64 4.45 4.45
yield of N.P.K. Adjusted R2 is a measure of the variation around the 15 9.33 9.31 8.61 8.64 4.44 4.45
mean clarified by the model, balanced for the quantity of terms in the 17 9.31 9.31 8.63 8.64 4.43 4.45
model (Taran and Aghaie, 2015). The Adjusted R2 diminished as the 18 9.27 9.31 8.63 8.64 4.46 4.45
quantity of terms in the model increments if those extra terms do not 19 9.32 9.31 8.64 8.64 4.45 4.45
20 9.31 9.31 8.62 8.64 4.43 4.45
increase the value of the model (Taran and Aghaie, 2015).
 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61 57

Design-Expert® Software
Design-Expert® Software Phosphorus Predicted vs. Actual
Nitrogen Predicted vs. Actual
Color points by value of 9.00
Color points by value of 10.00 Phosphorus:
Nitrogen: 8.97
9.62
6.7 2
8.03
8.50

9.50

6
8.00

P r e d ic te d
P re d ic te d

9.00

7.50

8.50
7.00

8.00
6.50

8.00 8.50 9.00 9.50 10.00

6.50 7.00 7.50 8.00 8.50 9.00

Actual
Actual

a b
Design-Expert® Software
Design-Expert® Software
Potassium
Nitrogen Normal Plot of Residuals
Color points by value of
Potassium:
Predicted vs. Actual Color points by value of
Nitrogen:
5.62 9.62
6.00
99
3.95
8.03

95
5.50
90

N o r m a l % P r o b a b ilit y
80
5.00 70
Predicted

50

4.50 30
6
20

10
4.00
5

3.50 1

3.50 4.00 4.50 5.00 5.50 6.00

-3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00

Actual

Internally Studentized Residuals

c
d
Design-Expert® Software Design-Expert® Software
Phosphorus Normal Plot of Residuals Potassium Normal Plot of Residuals
Color points by value of Color points by value of
Phosphorus: Potassium:
8.97 5.62
99 99
6.7 3.95
95 95
90 90
Normal % Probability
N o r m a l % P r o b a b ility

80 80
70 70

50 50

30 30
20 20

10 10
5
5

1
1

-4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00

-3.00 -2.00 -1.00 0.00 1.00 2.00

Internally Studentized Residuals Internally Studentized Residuals

e f
58 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61

poor or deluding results as stated by Li and Fu (2005). The residuals mineralization of critical soil supplement amid composting of agricul-
from least square fits to assume a fundamental job in making a decision tural wastes for biofertilizer synthesis.
about model adequacy (Myers and Montgomery, 2002). Fig. 1A (a, b
and c), shows the distribution of the predicted value against the actual 3.2.3. Factors and interactive effects on the mineralization of nitrogen dur-
experimental values for nitrogen, phosphorus and potassium. From ing composting
the plots, each of the observed values was compared to the predicted The interactive relationship between the independent and depen-
values calculated from the models. dent variables are depicted by plotting 3-D graphs known as response
The regression coefficients of 98.16%, 99.79% and 97.80% observed surface graphs generated by the models. These plots were used to
between the predicted and real values for the response factors are show the effect of process parameters on the yield of nitrogen, phospho-
proof that that the regression model can represent to the experimental rus and potassium (Lee and Wang, 1997). The linear, quadratic and in-
data well. It could be seen that that the points on the diagrams were teraction terms of the models (Table 5) were applied to create 3-D
sensibly dispersed almost a straight line demonstrating that the funda- response surface graphs. Every margin on the graph denotes a specific
mental supposition of typicality in this examination was proper and value for the heights of the surface above the plane define for the com-
along these lines approve the models developed. The Normal plot of re- bination of the levels of the independent variables (Liu et al., 2011). The
siduals as appeared in Fig. 1B (d, e and f), was utilized to check whether 3-D response surfaces were generated by keeping one variables at its
the points will pursue a straight line in which we presume that the re- zero level (null point or midpoint) and carefully varying the other two
siduals pursue a typical dissemination. It was seen that the points variables within the experimental range.
were firmly conveyed to the straight line of the plot. This affirms the Table 6 showed that the interaction of dosage ratio and time when
great connection between the trial values and the predicted values of the moisture content were kept constant at 55% were statistically signif-
the response, however some little disperse like a “S” shape is constantly icant as evident from the P-values (P b 0.0001 for nitrogen and phos-
anticipated. This observation shows that the central composite design is phorus and P b 00005 for potassium). It was observed from Fig. 2A.(a-
well fitted into the model and thus can be used to perform the optimi- c), that as the dosage ratio were varied at different fractions, percentage
zation operation for the process. Also, the straight line formed by the nitrogen, phosphorus and potassium increases with increase in number
data points is an indication that neither response transformation is re- of days (time) until it reaches an optimal point. Increase in both variable
quired nor there was any apparent problem with normality assumption beyond the optimal point resulted to decrease in the percentage yield.
of the regression model equations. This is in harmony with the report by At dosage ratio slightly above 5w:w and time above 33 days, the yield
Damirel and Kayan, (2012). starts to decrease. The observation shows that dosage ratio and time
The R2 measures how much of the observed variability in the exper- has a significant effect on the mineralization of important soil nutrients
imental data could be accounted for by the models; while R2adj on the from agro wastes during composting which is in harmony with the re-
other hand modifies R2 by taking into account the number of predictors port by Gajalakshmi and Abbasi (2008). The shape of the contour lines
in the model. R2 and R2adj are calculated using Eqs. (4) and (5). in Fig. 2A.(a-c), is also an indication of strong interaction effect between
dosage and time. The contours are somewhat curve which depicts the
Sum of square residual fact that the line will meet at a certain point and interaction would
R2 ¼ ð4Þ take place.
Model sum of square þ Sum of square residual
The interaction effect between moisture content and time were ob-
served to be statistically significant at P-value (P b 0.0004 for nitrogen
n−1  
and P b 0.0005 for phosphorus) but not significant for potassium and
R2adj ¼ 1− 1−R2 ð5Þ
n−p was removed as shown in Table 6. It was observed from Fig. 2B. (d-f),
that keeping the dosage ratio constant at dosage ratio of 4w: w, the
where n is the number of experimental runs, and p is the quantity of in- yield of nitrogen and phosphorus increased by varying the moisture
dicators in the model, not including the steady term. Kooche et al., content between 50 and 60% with increase in number of days (time).
(2009), recommended that for a decent fitted model, R2 ought not be The curve nature of the contour lines in Fig. 2B.(d-f), shows that interac-
under 80%, while Chauhan and Gupta (2004), announced R2 more note- tion of the two factors is imminent and nutrient release certain. The per-
worthy than 78% as worthy for fitting a model. In this examination, the centage nitrogen and phosphorus was observed to have decreased at
models created showed R2 going somewhere in the range of 97.80% and moisture content above 60% and moisture content below 50%. This
99.79% while R2adj extends somewhere in the range of 95.81 and trend is in harmony with the report by (Rasapoor et al., 2009;
99.61% connoting fitness of the developed model equations in anticipat- Stentiford, 1996) which states that excess moisture content impedes
ing nutrient release in the compost during composting when the three the entrance of oxygen to the compost system thereby changing the
independent factors are mathematically combined. system from the desired aerobic co-fermentation to anaerobic co- fer-
The R2 and R2adj values are near unity. (Lee and Wang, 1997; mentation; also low moisture content inhibits the growth and multipli-
Zaibunnisa et al., 2009), detailed that when R2 is nearer to solidarity, cation of microorganisms and hence prolongs the rate of nutrient
the better the exact model fit the experimental data. It is not any release during composting.
more news that adding extra factor to the model will dependably Furthermore, the interaction effect of moisture content and dosage
build R2, not considering of whether the extra factor is statistically sig- ratio were statistically significant at P b 0.05 (0.0023 for nitrogen,
nificant or not. Consequently, a large R2 does not always necessitate ad- 0.0001 for phosphorus and potassium) as shown in Table 6. It was ob-
equacy of the model. For this reason, Koocheki et al. (2009) declare that served from Fig. 2C. (g), that the contour lines are somewhat parallel,
it is more fitting to utilize R2adj of over 90% to assess the model adequacy. which signifies poor interaction between moisture content and dosage
Higher R2adj demonstrated that non-critical terms have not been incor- ratio. From Fig. 2C.(h and i), the contour lines are mostly curve and
porated into the model as obvious in this investigation. The general im- not parallel to each other which is an indication of good interaction
pression is that the residuals (Fig. 1) diffuse haphazardly in plain view, and positive influence on the yield. Keeping the time constant at
proposing that the fluctuation of the first perception is steady for all 30 days, the yield of nitrogen, phosphorus and potassium increased at
value of responses(Y). Since the plots in Fig. 1 are agreeable, it very verifying moisture content and dosage ratio within the experimental
well may be reasoned that the model is satisfactory to depict the range. The increase in the release of nitrogen at varying dosage ratio

Fig. 1. A: (a, b, c) Plots of predicted value vs experimental value for nitrogen, phosphorus and potassium. B: (d, e and f) Plots of normal probability plot of residual for nitrogen, phosphorus
and potassium.
 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61 59

Design-Expert® Software Design-Expert® Software

Factor Coding: Actual Factor Coding: Actual
Nitrogen Phosphorus
Design points above predicted value Design points above predicted value
9.62 Design points below predicted value
9.8 9
8.97
8.03
9.6 6.7
X1 = A: Time 8.5
X2 = B: Dosage ratio 9.4 X1 = A: Time
X2 = B: Dosage ratio
Actual Factor

P h o s p h o ru s
C: Moisture content = 55.00
N it r o g e n 9.2 Actual Factor 8
C: Moisture content = 55.00
9
7.5
8.8

8.6 7

5.00 35.00 5.00 35.00

4.50 33.00 4.50 33.00
31.00 31.00
4.00 4.00
29.00 29.00
3.50 3.50
B: Dosage ratio 27.00 A: Time B: Dosage ratio 27.00 A: Time
3.00 25.00 3.00 25.00

a b
Design-Expert® Software Design-Expert® Software
Factor Coding: Actual Factor Coding: Actual
Potassium Nitrogen
Design points above predicted value Design points above predicted value
5.62 5.8 9.62 9.8
3.95 5.6 8.03 9.6
X1 = A: Time 5.4 X1 = A: Time
X2 = B: Dosage ratio 9.4
5.2 X2 = C: Moisture content
Actual Factor 9.2
5 Actual Factor
P o ta s s iu m

C: Moisture content = 55.00 B: Dosage ratio = 4.00

4.8 N it r o g e n 9
4.6 8.8
4.4
8.6
4.2
8.4

5.00 35.00
4.50 33.00 60.00 35.00
31.00 58.00 33.00
4.00
29.00 56.00 31.00
3.50
B: Dosage ratio 27.00 A: Time 54.00 29.00
3.00 25.00
C: Moisture content 52.00 27.00 A: Time
50.00 25.00

c
Design-Expert® Software
Factor Coding: Actual Design-Expert® Software
Factor Coding: Actual
Phosphorus
Potassium
Design points above predicted value
Design points above predicted value
Design points below predicted value 9 5.62 6
8.97
3.95
6.7
8.5 X1 = A: Time 5.5
X1 = A: Time X2 = C: Moisture content
X2 = C: Moisture content
P h o s p h o ru s

8 Actual Factor 5
Actual Factor
P o ta s s iu m

B: Dosage ratio = 4.00

B: Dosage ratio = 4.00

7.5 4.5

7 4

60.00 35.00 60.00 35.00

58.00 33.00 58.00 33.00
56.00 31.00 56.00 31.00
54.00 29.00 54.00 29.00

C: Moisture content 52.00 27.00 A: Time C: Moisture content 52.00 27.00 A: Time
50.00 25.00 50.00 25.00

f
60 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61

could be attributed to the component of the organic mixtures which has outcomes from the validation experiment showed a rate yield of phos-
a great influence in the range of carbon to nitrogen ratio, which is the phorus was 8.91%, which moderately concurs with the model antici-
one of the key performance indicator in composting. Agro wastes pated value. The ideal condition for potassium was: Composting time
must be blended at a certain dosage to maintain the carbon to nitrogen of 25 days, dosage ratio of 5w:w and moisture content of 50%. The antic-
ratio from 15:1 to 35:1 for proper composting. López et al. (2010), sug- ipated rate yield for potassium at this condition was 5.67%. Utilizing the
gested that C/N of around 20:1–35:1 are normally advisable, but good ideal state of 25 days, 5 w:w and 50%, the rate yield of potassium was
results have been obtained with values out of this range. In general, at tentatively validated. The consequence of the validation demonstrates
higher C/N ratios (above 35:1), the composting process is thought to that the rate yield was 5.69%, which was moderately near the model's
be delayed through lack of nitrogen, whereas smaller C/N ratios anticipated value. The closeness of the validated value and predicted
(b15:10) leads to excessive nitrogen loss and eventually to microbial or anticipated rate yield of N.P.K shows the authenticity of the models.
toxicity due to high level of ammonia. Mixing of different materials is Additionally, a comparatively high R2 value (98.16% for N, 99.79% for P
commonly necessary to achieve adequate C/N ratio for composting and 97.80% for K) demonstrated the closeness between the tentatively
through proper dosage ratio (López et al., 2010). validated experimental values and the anticipated or predicted values
as appeared in Table 7. This further checks the accuracy of the proposed
3.2.4. Process optimization and validation of the composting process model.
As quoted by Josh et al. (2014), it is not possible to define a single op-
timum for a process since it can change depending on the level of other 3.3. Comparison of the results with prior results
factors; however one of the optimum solutions was selected at desir-
ability of 1 for the process composting. The consequences of the confir- The yield of nitrogen, phosphorus and potassium from the afore-
mation of the ideal numerical arrangement demonstrated that the ideal mentioned examination was observed to be 9.62%, 8.97% and 5.62% re-
arrangement anticipated the genuine arrangement intently. The devia- spectively. The result of the percentage nitrogen was higher when
tion of the test yield of nitrogen, phosphorus and potassium from the compared with 4.7% nitrogen reported for the composting of mixture
anticipated yield was a nearby match. The diagram of the predicted of sawdust and dewatered sewage sludge by Bazrafshan et al. (2006).
values against experimental (real values) as appeared in Fig. 1 likewise The results was analyzed using Box Behnken Design and Validated at
affirmed the closeness of the predicted and real value. The data points of predicted value of percentage nitrogen 4.75% with regression coefficient
the optimization runs falls along the inclining of the squared chart dem- of 95.6%. Also, 4.9%, 5.6% and 2.3% for nitrogen, phosphorus and potas-
onstrating the cozy relationship of the predicted and real points and all sium reported for the composting of mixture of sawdust plus chicken
things considered, the quadratic model was satisfactory for the litter by Egbuna et al. (2016), which was the optimized results using
examination. CCD method of analysis and validated at predicted value of 4.95, 5.8,
The developed model for the process parameters was streamlined and 2.4 with regression coefficients of 92.5%, 94.2% and 91.7%. More-
utilizing response optimizer software that is available in design expert over, 5.5%, 6.1% and 3.4% for nitrogen, phosphorus and potassium re-
version 8.7.0.1. The software provides ideal answers for the input vari- ported by Dumitrescu et al., (2009), after composting a mixture of
able combinations. The optimization is additionally interactive and con- sawdust plus sewage sludge plus dry leaves and analyzed with CCD
siders bargain among the different independent factors and the and validated at the predicted values of 5.6%, 6.3% and 3.9% with regres-
response(s) (Agu et al., 2015). The RSM was utilized to portray the con- sion cofficients of 91.8%, 93.6% and 90.8%. Nevertheless, after analysis
nection between the process parameters and response factors (% yield) with BBD, 7.84% nitrogen was reported by Olayinka and Adebayo
for the composting procedure. These process parameters are (1989) after composting a mixture of sawdust plus cowdung and vali-
composting time, dosage ratio and moisture content while N.P.K is the dated at the predicted value of 7.95% with regression coefficient of
response. The ideal setting is determined by characterizing the con- 96.7%, thus indicating the potential of the biofertilizer produced in this
straints and the objective function of the dependent factors. This way, contest for commercial application. The difference in N.P.K yield ob-
the best values for both the input and response parameters are deter- tained by prior researchers and that obtained in this research work
mined (Myers and Montgomery, 2002; Agu et al., 2015). In this research could be attributed to factors such as substrate type (nature of the nitro-
(work), the objective function of the response factors expands the rate gen rich agro-wastes), carbon to nitrogen ratio of the composting mix-
yield of N.P.K. This is subject to the accompanying constraints: ture, compsting time (process duration), degradability of the substrates
Composting time (time) (21.59 ≥ 38.41 days), dosage ratio and the nature of autochtonous microbes that aids the decomposition
(2.32 ≥ 5.68 w:w) and moisture content (46.59 ≥ 63.41%). The devel- (Haug, 2009). Other perceived factors could be aeration rate (depending
oped models were utilized by the response optimizer to provide the on rate of compost agitation), compost temperature and compost pH
ideal outcomes for the responses and the independents factors. (Trinh and Kang, 2010).
The ideal conditions for the for maximum yield of nitrogen, phos-
phorus and potassium from the compost (mixtures of sawdust + sew- 4. Conclusion
age sludge + vegetable waste) concerning the proposed second order
polynomial equations were: Composting time of 22 days, dosage ratio The use of response surface methodology and central composite de-
of 4 w:w and moisture content of 55% for nitrogen. At this condition, sign was helpful in determination of the ideal working conditions for
the predicted yield of nitrogen was 9.60%. Utilizing the ideal states of composting of organic wastes for biofertilizer synthesis. It was built up
22 days, 4w:w and 55%, the rate yield of nitrogen was tentatively vali- that the second order polynomial model was adequate to define and an-
dated. The outcome from the validation studies demonstrated a rate ticipate the process responses to variation of input variables within the
yield of 9.58%, which is relatively near the model's anticipated value. experimental range. The validity of the models was demonstrated by
For phosphorus, the ideal condition was: Composting time of 35 days, fitting the estimations of the factors to the model equations and carry-
dosage ratio of 3 w:w and moisture content of 60%. The anticipated ing out experiments utilizing the same values. The graphical optimiza-
rate yield of phosphorus was 8.96%. Utilizing the ideal state of 35 days, tion utilized to locate the ideal conditions for the composting of agro
3 w:w and 60%, the rate yield of phosphorus was validated. The wastes was characterized by the composting time of 22 days, dosage

Fig. 2. A: (a, b and c), 3-D response surface plots showing the relationship between independent variables (time and dosage ratio) and dependent variables (nitrogen, phosphorus and
potassium). B: (d, e and f) 3-D response surface plots showing the relationship between independent variables (moisture content and time) and dependent variables (nitrogen,
phosphorus and potassium). C: (h, i and j) 3-D response surface plots showing the relationship between independent variable (moisture content and dosage ratio) and dependent
variables (nitrogen, phosphorus and potassium).
 C.O. Asadu et al. / Artificial Intelligence in Agriculture 3 (2019) 52–61 61

ratio of 4w:w and moisture content of 55% with 9.62% yield of nitrogen, Gajalakshmi, S., Abbasi, S.A., 2008. Solid waste management by composting. Environ. Sci.
Technol. 38, 311–400.
composting time of 35 days, dosage ratio of 3 w:w and moisture content Haug, R.T., 2009. The Practical Handbook of Compost Engineering. Boca Raton, Florida.
of 60% with 8.97% yield of phosphorus and composting time of 30 days, Ilegbune, T.O., 2006. Legal regulation of industrial waste management. In: Ajomo, M.A.,
dosage ratio of 6 w:w and moisture content of 55% with 5.62% yield of Omobolaji, A. (Eds.), Selected Essays in Environmental Law and Sustainable Develop-
ment. Lagos, NIALS, p. 88.
potassium. Characterization of the composite demonstrated its possibil- Jimoh, I.A., 2005. A new approach to municipal waste management in Nigeria. Interna-
ities for commercial application on agricultural land. tional Conference on Energy, Environment and Disasters- INCEED, Charlotte N.C,
USA - July 24–30, He Says that the Around Growth Rate Is About 2.7–3 Percent.
Josh, L.P., Chris, P., Rachel, L.G., 2014. Comparison of response surface methodology (RSM)
Authors' contributions and artificial neural networks (ANN) towards efficient extraction of artemisinin from
artemisiaannua. Ind. Crop. Prod. 58, 15–24.
Koocheki, A., Taherian, A.R., Razavi, S.M.A., Bostan, A., 2009. Response surface methodol-
ogy for optimization of extraction yield, viscosity, and hue and emulsion stability of
mucilage extracted from Lepidiumperfoliatum seeds. Food Hydrocoll. 23,
Author Contribution 2369–2379. https://doi.org/10.1016/j.foodhyd.06.014.
Lee, C.L., Wang, W.L., 1997. Biological Statistics. Science press, Beijing, Peoples Republic of
Christian O. Provided study materials, reagents, writing the initial draft, China.
Asadu coordinated research activities and analyses of the data, review Li, Q.H., Fu, C.L., 2005. Application of response surface methodology for extraction optimi-
and editing zation of germinant pumpkin seeds protein. Food Chem. 92, 701–706.
Samuel O. Writing the initial draft, provided funds and supervised the data Liu, S., Yang, F., Zhang, C., Ji, H., Hong, H., Deng, C., 2011. Optimization of process param-
Egbuna analysis, reviewing and editing the manuscript eters for supercritical carbon dioxide extraction of passiflora seed oil by response sur-
Thompson O. Provided funding, Coordinated research activities and supervised face methodology. J. Supercrit. Fluids 48, 9–14.
Chime data analysis, reviewing and editing the manuscript López, M., Huerta-Pujol, O., Martínez-Farré, F.X., Soliva, M., 2010. Approaching compost
Chibuzor N. Eze Provided funding, writing the initial draft and data analysis, stability from Klason lignin modified method: chemical stability degree for OM and
reviewing and editing N quality assessment. Resour. Conserv. Recycl. 55, 171–181.
Gordian O. Provided funding and supervised the draft and data analysis Manpreet, S. Bhatti, DhritiKapoor, Rajeev K. Kalia, Reddy, Akepati S., Thukral, Ashwani K.,
2011. RSM and ANN modeling for electrocoagulation of copper from simulated
Mbah
wastewater: multi objective optimization using genetic algorithm approach. Desali-
Dibia Kevin Provided study materials, reagent, equipments, review and
nation 274, 74–80.
editing
Mowoe, K.M., 2001. Quality of life and environmental pollution and protection in
Anthony C. Provided reagents, study materials, data analysis Omotola. In: A., J. (Ed.), Environmental Law in Nigeria. 177. Faculty of Law, University
Ezema of Lagos (Municipal Solid Waste Landfills, Final Section – Supplement E, November).
Myers, R.H., Montgomery, R.C., 2002. Response Surface Methodology, Process and Product
Optimization Using Design Experiment. 2nd ed. Wiley 0471412554.
Acknowledgments Nahemiah, D., Nkama, I., Badau, M. halide, 2015. Application of Respponse surface meth-
odology (RSM) and Central Composite Design (CCD) to optimize minerals composi-
tion of rice-cowpea composite blends during extrusion cooking. Int’l J. food science
The authors wish to thank in a special way, Center for Energy re- and nutrition engineering 5 (1), 40–52.
search, University of Nigeria Nsukka for making their laboratory and Ogazi, J.N. and J.A.I. Omueti (2000), Waste utilization through organo-mineral production
equipments available for the work. Also, our appreciation goes to Pro- in South Western Nigeria. Proc. 18th Int. Symp. Anim., Agric. Food Process, Des-
Moines, Iowa, USA.
fessor Gibson Nebechukwu Aneke of the department of Chemical Engi-
Okon, Y., 2000. Response of cereal and forage grasses to inoculation with N2 fixing bacte-
neering, Enugu State University of Science and Technology and ria. In: Veeger, C., Newton, W.E. (Eds.), Advances in the Nitrogen Fixation Researc
Professor Philomena Kanwulia Igbokwe of the department of Chemical Nijoff/Junk. The Hague, pp. 30–39.
Engineering, Nnamdi Azikiwe University, Awka, Nigeria for their im- Olayinka, A., Adebayo, A., 1989. Effect of pre- incubated sawdust based cow dung on the
growth and nutrient uptake of Zea mays L. and on soil chemical properties. Biol. Fert.
measurable contributions towards the success of this research. Soil 7, 176–179.
Oltjen, J.W., Beckett, J.L., 2006. Role of ruminant livestock in sustainable agricultural sys-
References tems. J. Anim. Sci. 74, 1406–1409.
Rasapoor, M., Nasrabadi, T., Kamali,M., Hoveidi, H (2009), The effects of aeration rate on
Adeyanju, J.A., Ogunlakin, G.O., Adekunle, A.A., Alawode, G.E., Majekolagbe, O.S., 2016. generated compost quality, using aerated static pile. In: Waste Management 9, p.
Optimiza-tion of oil extracted from coconut using response surface methodology. 570–573.
J. Chem. Pharm. Res. 8 (1), 374–381. Samaram, S., Mirhosseini, H., Tan, C.P., Ghazali, H.M., Bordbar, S., Serjouie, A., 2015. Opti-
Agu, C., Menkiti, M., Kadurumba, C., Menkiti, N., 2015. Process parameter optimisation for misation of ultrasound-assisted extraction of oil from papaya seed by response sur-
transformer oil extraction from Terminalia catappa seed using response surface face methodology: oil recovery, radical scavenging antioxidant activity, and
methodology. J. Chin. Adv. Mater. Soc 3 (4), 328–344. oxidation stability. Food Chem. 172, 7–17.
ASTM D2974-07, 2007. Standard Test Methods for Moisture, Ash, and Organic Matter of Schumacher, B.A., 2003. Methods for the determination of total organic carbon (TOC) in
Peat and Other Organic Soils. ASTM International, West Conshohocken, PA www. soils and sediments. U.S. Environmental protection agency, Washington, DC. 2002
astm.org. EPA/600/r-02/069 (Ntis PB-100822).
Bazrafshan, Edriss, Zazoul, Mohammd Ali, Bazraafshan, Javad, Mohsen, Anoushivan, 2006. Shrivastsvs, A., Sandagar, P., Baja, I., Singhal, R., 2008. Media optimization for the produc-
Evaluation of microbial and chemical parameters during wasteswater sludge and tion of U-linolenic acid by cunninghamella echinulata varielegans MTCC 522 using
sawdust co-composting. J.App.Sci.Environ. 10 (2), 115–119. response surface methodology. Int. J. Food Eng. 4 (2), 1–32.
Chauhan, B., Gupta, R., 2004. Application of statistical experimental design for optimiza- Snedecor, G.W., Cochran, W.G., 2008. Statistical Methods. 7th ed. Iowa State University
tion of alkaline protease production from Bacillus sp RGR-14. Process Biochem. 39, Press, Ames, IA.
2115–2122. Sobiecka, E., van der Sloot, H., Janssen, E., Gawlik, B.M., 2007. Project HORIZONTAL Valida-
Chen, G., Chen, J., Srinivasakannan, C., Peng, J., 2011. Application of response surface tion Report on Kjeldahl Nitrogen. Validation of a Horizontal Standard for the Determi-
methodology for optimization of the synthesis of synthetic rutile from titania slag. nation of Kjeldahl nitrogen in Soil Sludges and Treated Bio Waste in a European Inter-
Appl. Surf. Sci. 258 (7), 3068–3073. comparison Exercise, (Luxembourg, European Commissions-Joint Research Centre-
Damirel, M., Kayan, B., 2012. Application of response surface methodology and central Institute for Environmental and Sustainability-European Communities).
composite design for the optimization of textile dye degradation by wet air oxidation. Stentiford, E.T., 1996. Composting control: principles and practice. In: DeBertoldi, M., Sequi, P.,
International Journal of Industrial Chemistry 3, 24. https://doi.org/10.1186/2228- Lemmes, B., Papi, T. (Eds.), The Science of Composting. Chapman abd Hall, pp. 49–59.
5547-3-24. Taran, M., Aghaie, E., 2015. Designing and optimization of separation process of iron im-
Datta, S., 2011. Adsorption removal of chromium (VI) from aqueous solution over pow- purities from kaolin by oxalic acid in bench-scale stirred-tank reactor. Applied Blay.
dered activated carbon: optimization through response surface methodology. Trinh, T.K., Kang, L.S., 2010. Application of response surface method as experimental de-
J. Hazard. Mat 173, 135–143. sign to optimize coagulation tests. Environ. Eng. Res. 15 (2), 63–70. https://doi.org/
Dumitrescu, L., Manciulea, I., Sauciuc, A., Zaha, C., 2009. Obtaining Biofertilizer by 10.4491/ear.21010.15.2.063.
Composting Vegetable Waste, Sewage Sludge and Sawdust. Vol. 2(51). Buletin of Westerman, P.W., Bicudo, J.R., 2005. Management considerations for organic waste use in
Translivana university of Brasov. agriculture. Bioresour. Technol. 96, 215–221.
Egbuna, S.O., Asadu, C.O., Onoh, M.I., 2016. Production of biofertilizer by composting saw- Yolmeh, M., Habibi, M., Najafi, R., Farhoosh, B., 2014. Optimisation of ultrasound-assisted
dust, sewage sludge and succulent tissue of green plants using an accelerator. Int’l extraction of natural pigment from annatto seeds by response surface methodology
Journal of Novel Research in Engineering and Science. 3, 62–72. (RSM). Food Chem. 155, 319–324.
Filli, K.B., Nkama, I., Abubakar, U.M., Jideani, V.A., 2010. Influence of extrusion variables on Zaibunnisa, A.H., Norasikin, S., Mamot, S., Osman, H., 2009. An experimental design ap-
some functional properties of extruded millet-soybean for the manufacture of ‘fura’: a proach for the extraction of volatile compounds from turmeric leaves (Curcuma
Nigerian traditional food. Afr. J. Food Sci. 4 (6), 342–352. http://www. domestica) using Pressurized Liquid Extraction (PLE). Journal of Food Science and
academicjournals.org/ajfs. Technology 42, 233–238.