RESULTS AND DISCUSSION

Roasted gram was coated with tomato sauce in a laboratory scale spouted bed. Central Composite Rotatable Design with 4 variables and 5 levels was used to decide the combination of experiments. In all 31 experiments were conducted, 16 of them at first order and 8 at the second order with 7 replications at the center point. The independent variables used in the design of experiments were inlet air temperature (X₁), rate of addition of tomato sauce (X₂), sample size of roasted gram (X₃) and air velocity (X₄).

The responses studied were coating thickness, percentage gain in weight of the product, moisture content of the product, weight of product and the weight loss during the process. Hedonic scale was used for the sensory evaluation of the product. Sensory ratings of flavour, taste, appearance, hardness and overall acceptability of the product were assessed.

Back to the top

4.1 Coating process

The coating of tomato sauce on roasted gram was done in a spouted bed (laboratory scale). The entire process could be understood to be taking place in two different phases, the coating phase and the drying phase. The coating phase was the one in which the tomato sauce was added to the spouted sample of roasted gram. The spouting helped inproper mixing of particles resulting in a uniform coating on the sample. There was simultaneous drying of the sauce during this phase. The coating phase was followed by the drying phase in which no more sauce was added and the particles were allowed to dry in the spouted bed. Though there was drying of sauce even during the coating phase, it was more dominant during the drying phase. The drying phase lasted for a period of 30 minutes.

It was observed that during the coating phase that the bed collapsed and was transformed into a fluidised bed. This was due to the fact that the bed characteristics of system changed in this phase as the rate of removal of moisture was less compared to the gain in weight due to absorption of moisture from tomato sauce. During the drying phase the moisture was removed rapidly and the bed again regained its spouting nature. It was also found during the preliminary studies that prolonging drying phase resulted in the loss of dry matter. This loss was due to the attrition between the particles, thus the sauce deposited on the surface of the particles got rubbed-off and escaped in the form of powder. This was in conformation with the experimental findings of Szentmarjay et al. (1991).

Back to the top

4.2 Data analysis

The data (Table A1 and A2) obtained were analysed using statistical software packages. As the basic idea was to optimise the processing parameters for coating, a complete second order model (Eq. 4.1) was used to fit the data.

A second order response function in 4 independent variables can be written as :

Y = a_o + a_iX_i + a_ij X_iX_j -------- 4.1

Where,

a₀, a_i, a_ij are constant.

X_i, X_j are Variables

The equation (4.1) was fitted to the experimental data using statistical software package, MINITAB.

The program provided the values of the coefficients of the model and the related statistics. This programme also identified unusual observations having a deviation of more than 2s from the value predicted by the model. Any such observation was removed by assigning it to zero weight and repeating the regression.

Back to the top

4.3 Model selection

The final model was selected after eliminating one or 2 unusual observations, wherever necessary. The final model was selected on the basis of F-value. If the calculated F-value was greater than tabulated value at 5% or 1% level of significance, the model was considered to be acceptable. The unusual observations that were removed, could be attributed to the experimental error or due to lack of fit in the model. In general, one or two observations from the set of 31 data had to be removed which showed deviation greater than 2s .

Back to the top

4.4 Process optimization

The second order models were used to optimize the process parameters. The optimum conditions were represented by the stationary points. The stationary point of a response surface y, could be obtained by partially differentiating it with respect to the independent variables (X₁, X₂, X₃ and X₄) and equating them to zero.

= a₁ + 2a₁₁X₁ + a₁₂X₂ + a₁₃X₃ + a₁₄X₄ = 0

= a₂ + a₂₁X₁ + 2a₂₂X₂ + a₂₃X₃ + a₂₄X₄ = 0

= a₃ + a₃₁X₁ + a₃₂X₂ + 2a₃₃X₃ + a₃₄X₄ = 0

= a₄ + a₄₁X₁ + a₄₂X₂ + a₄₃X₃ + 2a₄₄X₄ = 0

The above set of linear equations were solved simultaneously to yield the stationary points (). The stationary points obtained from the above equations may represent a maxima, minima or a saddle point. This characteristics of the stationary point could be found by evaluating the second partial derivatives.

= 2a₁₁

= 2a₂₂

= 2a₃₃

= 2a₄₄

If all the second partial derivatives at stationary point are negative, the stationary point represents a maxima, if all are positive it represents a minima and if it has mixed signs, a saddle point is indicated.

The response surface around the stationary points were examined using contour plotting. For this contours of constant response were plotted taking 2 independent variables at a time, keeping the remaining variables at their stationary point values.

Finally the multiple response optimization package was used to improve the optimum process conditions.

Back to the top

4.3 Model selection

The final model was selected after eliminating one or 2 unusual observations, wherever necessary. The final model was selected on the basis of F-value. If the calculated F-value was greater than tabulated value at 5% or 1% level of significance, the model was considered to be acceptable. The unusual observations that were removed, could be attributed to the experimental error or due to lack of fit in the model. In general, one or two observations from the set of 31 data had to be removed which showed deviation greater than 2s .

Back to the top

4.5 Coating thickness

During the coating process tomato sauce was deposited on the surface of the particles and was dried by the action of hot air. As the process continued, layers of dried sauce developed on the particle. The thickness of coat obtained varied with the change in the process conditions. Data in the Table A1 shows that the variation of coating thickness was between 0.0899 and 0.2846 mm for different process conditions.

The data was fitted in Equation 4.1 and a second order model was obtained. The regression results of the above equation are shown in Table 4.1. The model had a value of = 0.8947 and a F-value of 9.71 which was greater than the Table value of F (3,45) at 1% level of significance. Thus, the model was accepted. The parameters which significantly affected the process are indicated in Table 4.1.

Table 4.1 : Regression results for coating thickness

*** Significant at 1% ** Significant at 5% * Significant at 10%

The quadratic model was used to see the effect of each independent variables on coating thickness by keeping other independent variables at the centre point. The variation of coating thickness with inlet air temperature, rate of addition of sauce, sample size and air velocity as predicted by the model is shown in Fig. 4.1.

The model was then used to optimize the parameters for coating thickness. The stationary point was obtained as :

X₁ = – 3.27995

X₂ = – 0.82959

X₃ = – 11.5262

X₄ = – 8.5164

The predicted coating thickness at stationary point was found to be 0.568 mm.

The second derivatives at the stationary points are :

= – 0.019942

= – 0.031552

= – 0.007838

= – 0.016484

This stationary point represents a maxima as all the signs of second derivatives are negative.

The response surface around the stationary point was examined using contour plotting. For this, contours of constant response were plotted taking two independent variables and keeping the remaining variables at their stationary points. Fig. 4.2 shows the contour plots for coating thickness.

It was noted that contours represent a maxima with respect to X₁, X₂, X₃, X₄. It can be seen from the contours b, c, d and e that after a certain level negative coating thickness is indicated. This shows that beyond a particular level, there would be considerable loss of surface material and the size of the particle would reduce due to attrition between the particles. It is clear from a, that at lower temperatures the coating thickness would be greater.

The optimum process conditions for coating thickness as shown by the stationary points lied away from the range of experimental conditions. Thus, further analysis was done to obtain compromised optima which lay in the range of actual experimental conditions.

Coded (Uncoded)

X₁ = – 0.66203 66.68�C

X₂ = 0.326127 158.153 gm/hr

X₃ = – 1.23896 688.052 gm

X₄ = – 1.38281 86.172 m/min

The predicted coating thickness at this optimum condition was found to be 0.320 mm.

Back to the top

4.6 Percent weight gain

Deposition of tomato sauce on roasted gram and its subsequent drying causes a change in the weight of the individual particles. This change in weight was calculated and expressed in the form of percentage gain in weight of the particles. The gain in weight was found to be dependent on the process conditions. Data in Table A1 shows the percentage gain in weight of the particles for various combination of the processing parameters.

The second order model was developed for the percentage gain in weight of the particlesThe regression results of the above model are shown in Table 4.2 The model had a value of r² = 0.94552 and the F-value was 17.35 which was greater than the tabulated value of F (3.45) at 1% level of significance. The 17th and 24th value was found to be unusual as its deviation was greater than 2s , hence these observations were removed and the regression was done. Model was found to be acceptable at 1% level of significance. The factors which significantly affected the percent gain in weight are indicated in the Table 4.2.

Table 4.2 : Regression results for percent gain in weight

*** Significant at 1% ** Significant at 5% * Significant at 10%

The above model was used to examine the effect of each independent variable while other independent variables were kept at centre point. The variation of percent gain in weight with inlet air temperature, rate of addition of sauce, sample size and air velocity as predicted by the model are shown in Fig. 4.3.

The model was then used to optimise the parameters for the percentage gain in weight of the particles. The stationary points obtained were

X₁ = – 0.436723

X₂ = – 1.69429

X₃ = – 0.584796

X₄ = – 0.066705

The predicted percentage of gain in weight at the stationary point was found to be 3.209%.

The second derivatives at the stationary points was

= – 0.3725

= + 0.12288

= + 0.04124

= + 0.42104

The mixed signs of the second derivatives show a saddle point.

The response surface around the stationary point was then observed using contour plots. Contours of constant responses were plotted around the stationary points. Contour for percentage gain in weight are shown in Fig. 4.4. It may be noted that contours at d, e, and f represent a minima with respect to X₁ X₃ and X_4. The range of parameter which would give desirable gain in weight is shown by the shaded portion in the contours. The contours a, b and c show a saddle point when X₁ was plotted with X₄X₃ and X₂, respectively.

The optimum process conditions for percentage gain in weight as shown by the stationary points lied away from the range of experimental conditions. Thus, further analysis was done to obtain compromised optima which lay in the range of actual experimental conditions.

Coded (Uncoded)

X₁ = – 0.088576 69.557

X₂ = + 0.628609 165.715

X₃ = – 0.067623 746.619

X₄ = – 1.89933 81.0067

The predicted maximum percentage weight gain at these condition was found to be 4.39%.

Back to the top

4.7 Final moisture content

The final moisture content of the product decides the shelf life of the product. Higher moisture content causes rapid deterioration of the product. Addition of tomato sauce during the coating process raises the moisture content of the product. Thus, it was dried. The final moisture content of the product after 30 minutes of drying were recorded and is illustrated in the Table. A-1.

The second order model was developed to show the influence of the processing parameters on the final moisture content. The regression results of the model are shown in Table .4.3. The regression coefficient (r²) was found to be 0.6979. Though the r² value was less, the model was accepted as it had a F-value of 2.67 which was greater than the tabulated F-value (2.37) at 5% level of significance. The parameters significantly affecting the process are indicated in the Table.

Table 4.3 : Regression results for final moisture content of the product

*** Significant at 1% ** Significant at 5% * Significant at 10%

The above model was used to examine the effect of independent variable on the final moisture content of the product. Fig. 4.5 shows the effect of inlet air temperature, rate of addition of sauce, sample size and air velocity on final moisture content, as predicted by the model. Effect of individual variable is shown while the other variables were kept at the centre point.

The model was then used to optimise the processing parameters for the final moisture content of the product. The stationary points that were obtained were :

X₁ = – 2.19619

X₂ = 0.146306

X₃ = – 2.04495

X₄ = – 3.28091

The predicted final moisture content that was obtained at the stationary point was 4.09726% (w.b.).

The second derivatives at the stationary points were

= – 2.80

= + 6.88

= – 4.02

= – 1.64

The stationary points indicated a saddle point as the second derivatives had both positive and negative signs.

The response around the stationary points were plotted using contour plotting and the contours of constant responses are shown in Fig. 4.6. It was found that contours plotted between X₁X₃ and X₄ represent a maxima. From the contours a, b and c it can be seen that at lower temperature of inlet air the final moisture content was higher. The plot between X₂ and X₃ shows that moisture content was higher at lower values of addition of sauce and sample size.

The optimum processing conditions as shown by the stationary points lied away from the range of experimental conditions. Thus, further analysis was done to obtain compromised optima which lay in the range of actual experimental conditions.

Coded (Uncoded)

X₁ = 0.134768 70.674

X₂ = 1.98109 199.527

X₃ = – 0.169976 741.501

X₄ = 0.165172 101.652

The predicted maximum moisture at this condition was found to be 6.41129% (w.b.).

Back to the top

4.8 Product Output

At the end of the process the final product was weighed and its dry matter content was determined. It was observed that the product output varied with the process conditions. The output obtained under the actual conditions was found to differ from the expected output. The data is listed in Table A1.To the above data a second order model was fitted Regression analysis for the above model was carried out and the results of the regression are shown in Table 4.4. The regression coefficient (r²) was to be 0.89713. The model was accepted as its F-value was 9.97 which was greater than its tabulated F-value of 3.45 at 1% level of significance. The individual parameters which significantly affected the output are indicated in Table 4.4.

Table 4.4: Regression results of final output of the product

*** Significant at 1% ** Significant at 5% * Significant at 10%

The above model was used to examine the effect of each independent variable on the output. The effect of inlet air temperature, rate of addition of sauce, sample size and air velocity as predicted by the model is shown in Fig. 4.7. Here the effect of each variable was observed while the other parameters were kept at the centre point.

The model was then used to optimize the processing parameters for the output. The stationary points obtained were :

X₁ = – 1.6256

X₂ = 65.1645

X₃ = 95.5564

X₄ = 109.832

The predicted final output at the above stationary points was 2407.77 gm. As was clear from the above stationary points, that the processing conditions as indicated, would be difficult to obtain on the set-up. Thus, further optimization was essential.

The second derivatives of the stationary points were obtained as

= – 1.1

= + 3.186

= – 8.174

= – 7.32

The stationary points had mixed signs thus indicating a saddle point.

Further analysis was done to obtain compromised optima which lay in the range of actual experimental conditions.

Coded (Uncoded)

X₁ = 0.271753 71.359

X₂ = – 0.319017 142.025

X₃ = 1.8558 842.79

X₄ = 0.585505 105.855

The predicted maximum output at these conditions was 691.357 gms.

Back to the top

4.9 Attrition Loss

It was observed during the process that there was considerable loss of material in the form of powders. These powders were formed due to the rubbing of the particles against each other during spouting. As the product dried the attrition losses increased. The total loss of material in the form of powders were recorded in the Table A-1.

The second order model was developed to show the influence of the processing parameters on the loss due to attrition The regression results of the model are shown in Table 4.5. The regression coefficient (r²) was found to be 0.7573. Here the 17th data was found to have a deviation greater than 2s so it was removed. The F-value of the above model was found to be 3.34 which was greater than the tabulated value of F (2.43) at 5% level of significance. Hence the model was accepted. The parameters which significantly affected the loss are indicated in the Table 4.5.

Table 4.5 : Regression results of attrition loss

*** Significant at 1% ** Significant at 5% * Significant at 10%

The above model was used to examine the effect of each independent variable for evaluating the loss due to attrition keeping other variables at their centre point. Fig. 4.8 shows the effect of inlet air temperature rate of addition of sauce, sample size and air velocity as predicted by the model.

The model was then used to optimize the processing parameters for the loss due to attrition. The stationary points were obtained as :

X₁ = 0.336353

X₂ = – 2.15518

X₃ = – 0.19561

X₄ = – 0.345656

The predicted losses as was obtained at the stationary point was 29.7054 gm.

The stationary points indicated a saddle point at the second derivatives contained both positive and negative signs. The second derivatives of the stationary points were as follows :

= – 16.632

= 1.466

= 11.428

= 10.55

The response surface around the stationary point were plotted as contours of constant responses. The contours for losses are shown in Fig. 4.9. It can be seen that the contours plotted between X₂X₃ and X₄ show a minima. Plots of X₁ with other variables show a saddle point. The range of variable which would give the minimum loss is shown by the shaded portion.

The optimum processing conditions as shown by the stationary points lied away from the range of experimental conditions. Thus, further analysis was done to obtain compromised optima which lay in the range of actual experimental conditions.

Coded (Uncoded)

X₁ = – 0.029457 69.853

X₂ = 0.375777 159.394

X₃ = 1.56264 828.132

X₄ = – 1.19105 80.0895

The predicted maximum loss that would occur under these conditions would be 78.6323 gms.

Back to the top

4.10 Sensory evaluation

The sensory evaluation of the roasted gram coated with tomato sauce was conducted using the method described in chapter 3. The means of scores obtained from different panellists were calculated. The mean ratings based on flavour, taste, appearance, hardness and overall acceptability are given in Table A-2.

Second order models were fitted to the ratings on flavour, taste, appearance, hardness and overall acceptability. The regression results are summarised in Table 4.6 to 4.10

Table 4.6 : Regression results of ratings for flavour of the product

Table 4.7 : Regression results of ratings for taste of the product

Table 4.8 : Regression results for ratings of appearance of the product

Table 4.9 : Regression results for ratings of hardness of the product

Table 4. 10 : Regression results for the ratings of overall acceptability of the product.

*** Significant at 1%, ** Significant at 5%, * Significant at 10%

For the above models their calculated f value was found to be greater than their tabulated f value at 5% level of significance, therefore, the models were selected. Their ‘F’ values were 3.63, 3.20, 4.42, 4.01 and 3.49 for flavour taste, appearance hardness and overall acceptability, respectively. The unusual data points, having a value of deviation greater than 2 s were removed while developing the model from sensory ratings of taste (22nd data), appearance (10th data), hardness (4th and 12th data) and overall acceptability (14th data).

The variation of flavour, taste, appearance, hardness and overall acceptability with inlet air temperature, rate of addition of sauce, sample size and air velocity as described by the above models are shown in Fig. 4.10 to 4.14, respectively.

The model for overall acceptability was used to optimize the process parameter for making roasted gram coated with tomato sauce. The stationary point obtained was as follows :

X₁ = – 0.283585

X₂ = – 0.19958

X₃ = – 0.50842

X₄ = – 0.141158

The predicted overall sensory rating of the stationary point was 7.78 which on hedonic scale was rated as in between liked moderately to liked very much.

The second derivatives at the stationary point are

= – 0.30338

= – 0.43482

= – 0.3308

= – 0.47108

Therefore, this stationary point represents a maxima as all the second derivatives are negative.

The response surface around the stationary point was examined using contour plotting. For this contours of constant response were plotted taking two independent variables at a time while the remaining variables were at their stationary points, the contour plots are shown in Fig. 4.15. It can be seen that the contours represent a maxima for all combinations of X₁X₂X₃ and X₄. The liking ratings shows that the product around the centre point was liked more. The product was thus liked moderately around the centre point. The acceptable range of parameters based on overall acceptability of the product is shown by the shaded portion.

Multiple response optimization package was used to improve the optimum process conditions. It was found that the ridge analysis was not required, and the optimum conditions found by the programme was same as the stationary points.

Coded (Uncoded)

X₁ = – 0.283585 68.582

X₂ = – 0.19958 145.011

X₃ = – 0.50842 724.579

X₄ = – 0.141158 98.588

And the predicted individual sensory ratings at these conditions were :

Sensory rating based on flavour = 7.63765777

Sensory rating based on taste = 7.73085681

Sensory rating based on appearance = 7.637812926

Sensory rating based on hardness = 7.618488128

Sensory rating based on overall acceptability = 7.732352609

Back to the top

Introduction Review material and methods results and discussion summary