Abstract

Introduction

Thermal processing is one of the major techniques used for producing packaged shelf-stable food products. Retort processing of foods in semi-rigid, flexible laminates is an advanced form of food preservation by canning. It is a thermal process which imparts increased shelf life with good retention of nutrients and sensory parameters. Although different food processing technologies like high pressure processing, ohmic heating, etc. have been developed for preservation purpose, retort processing and canning are still the most used techniques in the food industry (Chen and Ramaswamy 2002a, b).

Changing lifestyle, increased work pressure and nuclear families are leading to exponential increase in the demand of ready to eat processed food in developing countries like India. A consumer would prefer his traditional food on a daily basis, provided a safe, tasty and processed option is available. In-spite of old notions and psychological reservations against canned food, Indian market has accepted many retorted and pouch packed products. Traditional Indian vegetarian foods are mostly heterogeneous systems. Typically the recipe incorporates two or more vegetables, lentils, grains and cereals; cooked with spices and condiments. Since fat, spices and protein content of these raw materials is low, one expects the thermal properties to depend only on total solids and water content. Due to cultural diversity, there are over 2,000 different recipes of staple and snack food. As a result, optimized retorting conditions are developed for a specific product. There is a need to have an unified approach of mathematical modelling so that most of the products could be categorized in terms of total solids (plant origin) and water content. Several papers have been reported on product and process development of Indian foods, with most work on non-vegetarian foods such as Biryani(Rice with red meat or chicken) and fish curry etc.  (Dileep and Sudhakara 2007; Jayakumar et al 2007; Mallick et al. 2006; Ravi Shankar et al. 2002). Non-vegetarian products are of a different class as they contain more proteins and fats, where as vegetarian products mainly consists of fibres (soluble/insoluble) and starch. Among vegetarian products, rice and curries have been mainly studied. (Abbatemarco and Ramaswamy 1994; Bindu et al. 2007; Chandrasekar et al. 2004; Chandrasekar and Srinivasa Gopal 2008; Geetha and Jayaraj Rao 2008; Prakash et al 2005; Rodríguez et al. 2003). Many of the reports on modelling of vegetarian foods have used Balls model and Stumbo model which has few limitations (Sablani and Shayya 2001). Further, focus of these papers was more on microbial destruction, lethality and heat transfer aspects as very little has been reported on the same. Some recent papers on non-Indian foods report heat transfer studies by using Artificial Neural Network and Genetic Algorithm (Chen and Ramaswamy 2002a, b; 2003; Gonçalves et al. 2005; Sablani and Shayya 2001). Others have used classical modelling as a tool (Alonso et al. 1997; Baucour et al. 2003; Chen and Ramaswamy 2007; Miri et al. 2008; Simpson et al. 2004). These models were however product specific as they were based on thermal properties of that particular product. This work was therefore undertaken with the objective to develop a semi empirical unified model for prediction of time-temperature profile during retorting of Indian vegetarian products. Temperature of the retort vessel, initial product temperature and solid content were the three independent variables considered. Cold point temperature of the retort pouch was estimated as a dependent parameter. The predictive model developed was validated using four different types of traditional Indian foods viz. Pulav, a rice product; Sambar, a curry product; Gajar-Halawa, a sweet product and Upama, a snack in a 100 l laboratory retort equipment.

Materials and methods

Retortable pouches (150 × 200 mm) made with four layers (PET 12/Al 9/NY 15/CPP 70 Microns) were procured from Joeltech Inc., Cochin, India. Pouches were filled with 250 g of material prescribed later. A specially designed stainless steel thermo-well was inserted in the pouch so that its tip was in the geometric centre of pouch. It consisted of two rubber O rings, nuts and a space bar having hole in the centre through which a thermocouple could be inserted. A thermocouple (Ellab SSA- 12050-G700-TS stainless steel, Ellab Co. Reodovre, Denmark), capable of measuring temperature in the range of −45 to 135 °C with an accuracy of 0.1 °C and response time of 0.8 s, was placed inside the thermo-well after filling and sealing the pouches. These pouches were then thermally processed in a stainless steel still retort processing unit (Laxmi Engineering, Chennai, India). 36 steam nozzles equally distributed along the walls of retort maintains an even temperature inside the vessel. Residual air in the retort vessel was steam exhausted and pouches were processed at required temperature and pressure. Cold point temperature was monitored in three pouches using a data recorder (Ellab CTF 9008 data recorder, Ellab A/S, Roedvere, Denmark). After processing the pouches at required temperature, they were cooled rapidly by pumping pressurized water into the retort till product temperature reached 50 °C.

Model validation was done by using four traditional Indian vegetarian products: Pulav (steamed rice with vegetables), Sambar (a south Indian style curry), Gajar Halawa (carrot based sweet product), Upama (wheat based snack product). Ingredient composition used for method validation for different products were as follows,

• Pulav: 260 g Basmati rice, 40 g Green peas, 40 g Carrot, 30 g Corn, 10 g Green chilli, 4 g Clove, 4 g Black pepper, 4 g Salt, 2 g Lemon juice, 42 g Vegetable oil, 564 g externally added water
• Sambar: 65 g Toor dal (lentils), 46 g Tomato, 90 g Onion, 46 g Drum sticks, 90 g Ash gourd, 46 g Bottle gourd, 11 g Tamarind extract, 8 g Salt, 16 g Spices mix, 5 g Curry leaves, 16 g Groundnut oil, 5 g Mustard, 556 g externally added water
• Gajar Halawa: 646 g Carrot, 162 g Sugar, 162 g Mawa (milk solids), 10 g Cardamom, 30 g Cashew nut
• Upama: 228 g Sooji (a coarse, sandy fraction of wheat flour obtained in milling), 65 g Onion, 25 g Coriander, 10 g Curry leaves, 25 g Ginger, 12 g Green chilli, 25 g Cashew nuts, 5 g Sugar, 15 g Lemon juice, 25 g Ground nut oil, 5 g Salt, 5 g Mustard, 560 g externally added water

All the ingredients were procured from local market of Mumbai. For Pulav, rice, green peas, carrot and corn whereas for Sambar, Toor dal and drum sticks were half cooked for 5 min in boiling water. All the products were processed in the retort processing unit as per the procedure described above. Time-temperature profile was modelled using the proposed model and the constants of the same were used to validate the unified model.

Model development

First step in model development was to test the Ball’s model (Eq. 1). Significant literature has reported the use of Ball’s model to predict time required for thermal processing of food (Abbatemarco and Ramaswamy 1994; Bindu et al. 2007; Chandrasekar et al. 2004; Chandrasekar and Srinivasa Gopal 2008; Dileep and Sudhakara 2007; Geetha and Jayaraj Rao 2008; Jayakumar et al 2007; Mallick et al. 2006; Prakash et al 2005; Ravi Shankar et al. 2002; Rodríguez et al. 2003). Ball’ model is as follow,

1

Where, Ta is ambient (retort) temperature, TP is product temperature (°C), t is time (min), fh is heating rate constant, jc is lag constant, Ti is initial product temperature (°C). Rearranging Eq. 1 to give product temperature,

2

Correction for cum up time (tcut) may be applied (Paul Singh and Heldman 2004) as,

3

The Ball’s model was applied to the retort pouches containing different products like Pulav, Sambar, Gajar Halwa and Upama. As a sample, results for Sambar are presented in Fig. 1. From Fig. 1, it can be seen that even though it gives sufficiently accurate prediction for total thermal processing time, it fails to give continuous time-temperature profile. The Ball’s model primarily applies to a cold body immersed in hot environment i.e. it was assumed that environment is at high temperature at time t = 0. But retort starts getting heated at t = 0. Although correction for cum up time (Eq. 3) was applied, it was useful only to predict final time required for processing and not for the prediction of entire time-temperature profile and hence later it affected lethality calculation (Eqs. 13, 14), which is function of product temperature.

Fig. 1

Experimental and Ball’s model predicted data for Sambar processed in retort

Since Ball’s model did not give satisfactory results, some of the other mathematical correlations were tested to fit the data (Data not reported here). A new semi-empirical model was further proposed to describe time-temperature profile (Eq. 5). The difference between pouch temperature and retort temperature is the driving force for increase in product temperature. Hence ratio of product temperature to retort temperature was considered while developing the model. In differential form, the model is,

4

Integrating Eq. 4 with initial condition of TP = TP0 at t = 0 gave,

5

Where TP is product temperature (°C) at any time t (min). TP0 is initial product temperature (°C), TR is desired retort temperature (°C) and k is empirical model constant. For heating and cooling parts, two model constants kh and kc were used hereafter. The analysis of experimental data for proposed model is given in section Model development for simulated system later.

Next objective was to develop a unified mathematical model that could predict time-temperature profile of all vegetarian products processed in retort. Typical traditional Indian foods essentially consist of: vegetables/cereals, water, salt and/or sugar, spices, fats (oil or ghee). To simulate typical food, these ingredients were grouped in four parts: (i) water: externally added or contained by vegetables (ii) dry solids in vegetables and spices (iii) salt and sugar (iv) fats. Out of which only combination of total solids and water was considered at this stage and salt, sugar and fats were neglected. Thus, major part of any Indian vegetarian food was assumed to be consisting of dry vegetable solids and water. To see whether different vegetable solids have same heating rates or different, an experiment was performed wherein 10 g of dry vegetable powder was added to 90 g water and kept in stainless steel tubes in oil bath at 100 °C. Ball’s model was applied as environment is at high temperature at time t = 0. Values of fh and jc were estimated using Microsoft© Excel 2007 SOLVER tool. From Table 1, it was seen that if solid content is same, mixture of any vegetable powder and water had similar heating rates. Thus now the simulated system consisted of (i) different combinations of vegetable solids and water (ii) initial product temperature and (iii) retort temperature. From Eq. 5, it was seen that, new proposed model already had terms of initial product temperature and retort temperature. Hence it was tried to relate empirical model constant k with the solid content.

Table 1

Ball’s model coefficients for different materials in retort pouch

To arrive at the results in less number of experiments, Response Surface Method was used with Box Behnken orthogonal design of experiments. Three factor- three level Box Behnken design (Box and Behnken 1960) involved total 17 experiments (12 factorial points plus 5 centre points to eliminate experimental errors). The design of experiments in coded form is shown in Table 2. Decoded values for the independent parameters are given in Table 3. Retort temperature was varied from 115 °C to 125 °C; initial product temperature was varied from 35 °C to 45 °C whereas solid content in the pouch was varied from 0 % to 30 %. For cooling model, initial product temperature was replaced with cooling water temperature. Experiments were performed in random order and new proposed time-temperature profile model was fitted to each retort run. Statistical parameters such as sum of squares of errors (Eq. 6), root mean square error (Eq. 7), chi-square (Eq. 8) and correlation regression coefficient (Eq. 9) were used to assess the accuracy of models.

Table 2

Box Behnken design of experiments for three factors

Table 3

Coded and actual values of independent variables

6

7

8

9

where k, k’ and kav are experimental, predicted and average values of model constant k, SS is sum of squares of errors, RSME is root mean square error, N is number of data points, z is number of constants in the model tested, R2 is correlation regression coefficient. Modelling was carried out using the least square method and CoStat version 6.4 (CoHort Software, USA) was used to perform this task.

Model constants, kh and kc were treated as response to the Box-Behnken design and quadratic model was fitted to relate kh, kc and independent process parameters i.e. retort temperature, initial product temperature and solid content in the pouch for kh whereas retort temperature, cooling water temperature and solid content in pouch for kc. Design expert (Version 8.0) software was used for analysis of RSM data. Second order polynomial (SOP) model was used to fit the data.

10

Where, Y is the predicted transform of dependent variable; β0 is a constant; βi, βii and βij are linear, squared and cross-product coefficients respectively, k is number of factors. Tools provided by the Design Expert were used to calculate model constants and coefficients. Analysis of Variance (ANOVA) was performed for the experimental and predicted responses by these models. Contour and Response surface plots were obtained for the data. Finally the data were also checked for microbial lethality as indicated by predicted F value.

Results and discussion

The present work was undertaken with the objective of developing mathematical expression for time-temperature profile of a retort pouch, developing a unified model and its application to the traditional Indian vegetarian food. The Lump parameter approach with simulated system was used for model development.

Model development for simulated system

In the view of simulated system described in section Model development, the independent process parameters were retort temperature (115–125 °C), initial product temperature (35–45 °C) and solid content (0–30 %). Experimental design was based on Table 2 and experiments were conducted randomly. Each experiment was a separate retort run and the data of the runs was checked if it fitted the newly proposed model (separately for heating and cooling section). It can be seen from Table 4 that new proposed model gave good fit to all the experimental runs. New proposed model gave prediction in 95 % confidence level and R2 > 0.97 for all the runs. To see effect of solid content on heating rate constant, a graph of kh vs. solid content was plotted (Fig. 2). But it was seen that all the points were not coinciding for solid content of 0, 15 and 30 % which infers that the other two parameters i.e. retort temperature and initial product temperature also affects kh. Hence relationship between kh and kc and all independent variables were studied by ANOVA. Model was fitted to determine the significance of all terms (linear, two factor interaction, quadratic and cubic) on kh and kc(Table 5). It can be seen that p-value for quadratic model was least for both the responses kh and kc hence it was selected.

Table 4

The Box Benken design matrix of independent variables with their corresponding response

Fig. 2

kh as a function of solid content of product in a pouch

Table 5

Model fitting summary

Table 6 is an ANOVA table showing the significance of the independent variables on each of the response variables. Lack of fit test is a measure of failure of a model to represent data in the experimental domain (Montgomery 1984). Analysis of lack of fit was performed for both the dependent variables. Experimental data fitted well to the response surface models as indicated by sufficiently good regression coefficients (R2 = 0.9). Model terms having p-value near to one were neglected from the model as they inflated the system. From Table 6 it can be seen that solid content was affecting kh on 99 % level whereas linear terms for TR and TP0 were not significant and hence neglected. Second order terms were significant at more than 90 % level. Interacting terms of TR - TP0 and TR - S were not significant whereas TP0 - S was significant at level more that 95 %. Similarly for kc linear and second order terms of S were significant at 99 % level whereas linear and second order terms of TR and Tcw were significant at level >90 %. Interacting terms were not much significant. Lack of fit was not significant for both the responses.

Table 6

ANNOVA table

Surface response plots were generated for both the responses (Fig. 3). Surface plots for TR and TP0 were saddle shape whereas for solid content it passed through minima. This may be due to the fact that initially when solid content in the pouch was low; heat transfer was due to convection. Increase in solid content decreases heating rate upto a certain extent after which a further increase in solid content results in absence of free water in pouch and there was only moist solid. Hence heat transfer was due to conduction in later case and hence kh or kc increases again. The SOP model relating kh and kc are,

11

12

 

Fig. 3

kh & kc as a function of independent parameters for the simulated system: (a) kh as a function of initial product temperature and retort temperature (b) kh as a function of solid content and initial product temperature (c) kc as a function of ...

Figure 4 shows parity plots for kh, kc and overall parity plot showing product temperature prediction which shows that model is giving good prediction of time-temperature profile.

Fig. 4

Parity plots (a) model predicted vs. experimental kh for simulated system (b) model predicted vs. experimental kc for simulated system (c) model predicted vs. experimental values of product temperature for simulated system

Model validation using traditional Indian foods

In the view of above sections, if retort temperature, initial product temperature and solid content of any product are known, temperature profile followed by the product may be predicted using the unified model developed. If a product has moisture content ≥95, prediction is very easy since the food system will behave similar to water. The main deviation in product properties will be seen when the solid content is 15 % or more. Preliminary investigations on RTE recipes revealed that most of the prepared vegetarian food has maximum solid content of 30 %. Hence for validation of the model four traditional Indian foods, namely Pulav, Sambar, Gajar Halawa and Upama were selected. Further, these products also had incorporation of multiple vegetables and in various proportions. Table 7 shows operating conditions of retort and initial conditions of products. Figure 5 shows the plots of predicted and observed values of the cold point temperature over the time for the four products. It may be observed that there was a high correlation between the experimental values and the predicted ones over the whole process. Thus the new proposed model had overcome the disadvantage of the Ball’s model. The new proposed model satisfactorily predicted the process behaviour in heating region; however prediction of cooling rate constant (kc) was not up to the mark (relative error < 10 %) (Table 8).

Table 7

Independent parameters for validation products

Fig. 5

Experimental and predicted temperature profiles for (a) Pulav (b) Sambar (c) Gajar Halawa (d) Upama: (white triangle) Retort temperature (white diamond) Product temperature-experimental (dash) Product temperature-predicted

Table 8

Predicted and observed values of kh, kc and F0

The microorganism lethality (L) which represents the lethal effect of the thermal treatment at any temperature was calculated (Tref = 121.1 °C and ‘z’ value = 10 °C). The lethality expression is given by,

13

 

And the F value was calculated by integrating the lethality L over the time t = 0 to tf,

14

 

F0 value prediction was in 90 % confidence as relative error was less than 10 % for all the products (Table 8). Although there was a small deviation between the experimental and predicted temperature in that stage of the process, there was a pronounced effect on the F values due to its exponential dependence to temperature (Fig. 6).

Fig. 6

Accumulated lethality over the time for (a) Pulav (b) Sambar (c) Gajar Halawa (d) Upama: (white diamond) experimental (dash) predicted

In the view of the above discussion, the new proposed model has advantages of continuous time-temperature prediction for the product and sufficiently good prediction of the time required for the processing. Considering the small error obtained, the new model developed is suitable to use as a predictive model for food thermal processing. This model can be further fine tuned to reduce the 10 % error occurred by considering the size and shape of the solids, percentage of salt and sugar, percentage of fats, etc.

Conclusion

The proposed semi-empirical unified model developed could be successfully extrapolated to predict product temperature over a range of process conditions for various products (relative error < 10 %). Validation with four recipes of Indian traditional food (total solids 10 to 30 %) showed a satisfactory match. Retort temperature, initial product temperature and solid content of the product were found to be significantly important parameters affecting the cold point temperature of the retort pouch. This model is simple to use, does not need exact data on thermo-physical properties of the product and hence will be useful to many other vegetarian products to predict time-temperature profile during retort processing.

Acknowledgment

Nomenclature

Greek letters

Contributor Information

S. V. Gokhale, Email: moc.liamg@ragaselahkog.

S. S. Lele, Phone: +91-22-33611111, Fax: +91-22-24145614, Email: moc.liamg@elel.atims.rd.

References