Empirical Modelling of Pasteurization Process Using Plate Heat Exchanger

DOI : 10.17577/IJERTV1IS3111

Download Full-Text PDF Cite this Publication

Text Only Version

Empirical Modelling of Pasteurization Process Using Plate Heat Exchanger

W.M.F. Wan Mokhtar

Academic staff on Faculty of Food Technology, Universiti Sultan Zainal Abidin,Terengganu, Malaysia


Pasteurization process is become the most important process especially in dairy and beverage industries. The aim of the process is to preserve the product quality and to extend the shelf life of product. The product temperature is a k ey parameter that must be controlled in order to maintain at desired value. Hence, in order to control process properly, an accurate process model needs to be developed. The objective of this study is to develop an empirical model of coconut milk pasteurization process using plate heat exchanger. The model was developed based on experimental data and represented in first order plus time delay (FOPTD) model. Overall, the obtained model gave good agreement with experimental data in validation result with maximum error is ±5%. The validation result showed that this model is suitable for use in control strategy for the further study.

Keywords: Pasteurization, Empirical modelling, FOPTD model, process reaction curve

  1. Introduction

    Coconut milk is the white liquid and contains of oil in water e mu lsion. It can be obtained by mechanical e xtraction of grated coconut meat with added water. It is very important ingredient for many food recipes of Asian and Pacific regions such as curry and deserts. It also become as a main ingredient in making of traditional cake. According to [1], coconut milk is comple x biologica l fluid, typically contained about 54% mo isture, 35% fat and 11% solid non-fat. In order to preserve the raw coconut milk, it must be pasteurized through heating process.

    Pasteurization process is one of the food preservation techniques using therma l treat ment and commonly applied in food product industry. The purpose of pasteurization is to eliminate pathogenic bacteria and vegetative organisms and as a result, the product can be extending their shelf life [2]-[3]. There are two methods of pasteurizat ion currently used in industry namely batch and continuous process es. The batch process is a traditional method and involves

    many processes such as filling, heating, holding, cooling, e mptying and cleaning. It is simp le operation, fle xib le and easy to operate. However, the long period of heating process and small p roduction has become the ma jor dra wback of this process [4]. For h igh production, the continuous process which is high temperature short time (HT ST) is recommended because of more rapidly in heating process and can be minimized the energy consumption [5]. The HTST pasteurization consists of three stages viz. regeneration, heating and cooling sections. The crucial stage is heating process using heat exchanger to ensure unpasteurized product achieve desired pasteurizat ion temperature before pass through holding tube and cooling sections. Prior to pasteurize food sample, the operation of equipment must be adequate to control the outlet food temperature in order to maintain at standard value.

    In food processing industry, the process control

    plays an important part as it is vital to ensure several require ments are met such as product quality, energy consumption, production rate and safety the equipment and the process [6]. The important process parameter needs to be controlled in order to meet the desired require ment. For instance, in pasteurization process, the control of outlet pasteurization temperature fro m heat e xchanger in heating section is a key ele ment to avoid substantial change in properties product and to meet the industrial standard [7].

    In order to imp le ment a good performance of control strategy, it high ly depends on dynamic process model to predict process behaviour [8]-[9]. Thus, it requires developing an accurately dynamic process model in order to us e in process control strategy [10]. The process model can be developed by theoretical or emp irical approach. Ho wever, theoretical approach is very comple x due to all para meters need to be considered. Hence, the purpose of this paper is to develop empirica l mode l of coconut milk pasteurizat ion process. This obtained model will be used in process control strategy at next stage of study.

  2. Materials and Methods

    2.1 Experime ntal Works

    The raw coconut milk purchased fro m local ma rket was used in this study. The experiment was carried out using laboratory pasteurizer unit (Model Kho lle r Fe 10, Malaysia) in order to determine the process model. The process diagram of this equipment is shown in Fig. 1. It was designed and fabricated by Noble Palms Sdn. Bhd., Ma laysia. It utilises a three stages of plate heat e xchanger for heating, regeneration and cooling processes. However, this study only focuses on heating process because it is a key section to ensure the product temperature achieved at desired value. The water is used as a heating mediu m. The flow rate of product was fixed at 1.5 ml/s.

    Heater Unit

    Hot water

    TI 02

    Pasteurized product

    TI 01





    TI Temperature indicator

    Manual valve

    Peristaltic pump


    constant, respectively. Therefore, the predicted process model was fitted with FOPTD model. The e xperimental data were used in e mpirica l process model develop ment by plotted using cftool fro m MATLA B R2009a. The process model para meters we re calcu lated using three methods namely M 1, M 2 and M3. The M1 is obtained fro m c ftool result, while M 2 and M 3 are tangent method and two point method, respectively [12]-[13].

    2.3 Diagnostic Evaluation and Model Validation

    Prior to study in process control strategy, the obtained process model needs to undergo diagnostic evaluation. The purpose of diagnostic evaluation is to determine the capability of process model fitted with e xperimental data. Finally, the model was validated with new set of data in order to ensure variation in operation does not significantly degrade model



    Chilled water



  3. Results and Discussion

    1. Process Modelling

      Three e xperiments were conducted for obtain process reaction curve by introduced step change in hot

      Chiller Unit

      Fig. 1. Process diagram of laboratory pasteurizer unit

      In this study, the hot water temperature and product temperature was chosen as manipulated and controlled variables, respectively. In order to determine the process model, a step change in hot water temperature was introduced from 70oC to 80oC at time = 5min. The response of product temperature was recorded until achieved at final steady state value. Three repetitions (E1, E2 and E3) were made for each e xperiment.

      2.2 Process Modelling and Parameter Estimations

      Generally, first order plus time delay (FOPTD) model is adequate to describe many dynamic processes accurately [11]. In control system design usually used FOPTD model in transfer function form as shown in

      (1) [6]. This model is represented in Laplace transform and can be converted in time doma in equation as shown in (2). The symbols of Kp, and in this equation mean of process gain, time de lay and time

      water temperature. In order to develop process model, the experimental data were plotted using cftool fro m MATLAB R2009a. Then, the predicted model was developed empirically based on these data as shown in Fig. 2. This figure shows the dynamic product temperature after the hot water temperature was changed from 70oC to 80oC at time=5 min for all e xperiments. The trend of dynamic product temperature is similar for a ll replications of e xperiments. All dynamic responses of product temperature exh ibit appropriate under FOPTD model. Fro m Fig. 2, the final steady state value of product temperature is achieved the desired pasteurizat ion temperature of coconut milk [14].

      The predicted model needs to undergo diagnostic

      evaluation before it can be used in process control strategy. This evaluation is conducted by plot the e xperimental data and predicted model on the same graph. The comparison between e xperimental data and predicted model was evaluated by coeffic ient of determination (R2). Findings exhibited that the predicted process model gives good agreement with e xperiment data for all replications with R2 is more than

      0.9. It means that the model could exp lain about 90% of the total variability with the experimental data. The predicted model is accepted for process control because it was valid model with R2 larger than 0.6 [15]. The ability of predicted model to represent dynamic behaviour of this process is interesting in order to use this model in control strategy.

      Fig. 3 shows the correlation between experimental and predicted data of product temperature through linear relat ions. This correlation indicates the measure ment of the relationship between the responses of e xperimental data and predicted model. A straight line in Figure 3 is the regression line that describes how a response of experimental data (y-a xis) changes as a predicted value (x-a xis) of mode l change. The linear relationship between these two data sets was estimated by correlation coeffic ient (R). The results showed that R values for all replications are 0.9579, 0.9516 and 0.9535, respectively. It means that relationship between these responses was good.

      The parameters of model were estimated using three methods and listed in Table 1. Fro m this table, all methods produce like ly simila r results in determination of model para meters. Therefo re, any method can apply in model para meter estimat ion. Meanwhile, co mparison between model para meters for a ll rep licat ions also e xhibited quite simila r value with sma ll value of standard deviation (SD). A low value of standard deviation signifies small variability between model parameters for each replication, with values close to the average. Hence, the average values of model parameters were chosen to represent these models. In this study, average values from M3 we re chosen due to their simple ca lculat ion [12]. These model para meters were replaced into (1) and (2) in order to get predicted model as presented in (3) and (4) for transfer function and time doma in fo rms, respectively.

      Fig. 2. Expe rimental data and predicted model for (a) E1; (b) E2; (c) E3

      Fig. 3. Corre lation of e xperimental data versus predicted data for (a) E1; (b) E2; (c) E3

      Table 1. Estimated process parameters for all methods

      M1 M2 M3

      K K K

    2. Model Validation

      Prior to used the obtain model in process control strategy, the predicted model must be validated. This validation process is a fina l stage and most important in model building sequence. The aim is to exa mine the capability of obtained models in predicting the dynamic response of product temperature. The obtained models have been validated by estimated ma ximu m error bound with new set of e xperiment data. The ma ximu m error bound is determined by plotted one set of e xperimental data versus predicted model as demonstrated in Fig. 4. This figure shows the obtained models give a good agreement with e xperimental data within ±5% of ma ximu m error bound. Thus, it can be concluded that the obtained models are acceptable to use in control strategy for the next phase of study s ince the error less than 10% [16].

      Fig. 4. Va lidation of predicted model

  4. Conclusion

    The process model of pasteurization process of coconut milk was successfully developed based on process reaction curve. The model was represented in

    p p p

    E1 0.70 1.34 7.29 0.70 1.25 7.11 0.70 1.30 7.29

    E2 0.75 0.87 6.66 0.75 0.88 6.54 0.75 0.85 6.66

    Ave 0.70 1.19



    1.15 7.08 0.70 1.17 7.20

    SD 0.45 0.27



    0.24 0.52 0.45 0.28 0.50

    E3 0.66 1.36 7.69 0.66 1.33 7.58 0.66 1.36 7.65

    first order plus time delay (FOPTD) model with R2 more than 0.9. The predicted model was validated using new set of expe rimental data with ma ximu m error is ±5%. This model will be used in the process control strategy in ne xt stage of study.

  5. Acknowledgement

    The authors would like to acknowledge Universiti Sultan Zainal Abidin for the financial support.


  6. Refe re nces

  1. J. Simuang, N. Chiewchan, and A. Tansakul, Effect of fat content and temperature on the apparent viscosity of coconut milk, Journal of Food Engineering, vol. 64, pp. 193-197, 2003.

  2. A. P. M . Hasting, Practical considerations in the design, operation and control of food pasteurization processes, Food Control, vol. 3, pp. 27 32, 1992.

  3. G. D. Saravacos, and A. E. Kostaropoulos, Handbook of Food Processing Equipment, Kluwer Academic / Plenum Publishers, New York, 2002.

  4. J. G. Brennan and A. S. Grandison, Food Processing

    Handbook, 2nd et. 2012, Wiley -VCH Verlag & Co., Weinheim, Germany , 2012.

  5. M . Lewis, and N. Heppel, Continuous Thermal Processing of Foods Pasteurization and UHT Sterilization, M aryland: Aspen Publication, 2000.

  6. D. E. Seborg, T. F. Edgar, and D. A. M ellichamp, Process Dynamics and Control, 2nd ed., United State of America: John Wiley & Sons, Inc, 2004.

  7. J. Franco, L. Saravia, V. Javi, R. Caso, and C. Fernandez, Pasteurization of goat milk using a low cost solar concentrator, Solar Energy, vol. 82, pp. 1088 1094, 2008.

  8. C. Riverol, and J. Cooney, Assessing control strategies for the supercritical extraction from coffee beans: process

    based control versus proportional integral derivative,

    Journal of Food Process Engineering, vol. 28, pp. 494 505, 2005.

  9. A. K. Jana, Chemical Process Modeling and Computer Simulation, New Delhi: Prentice-Hall, 2008.

  10. L. W. Tan, F. S. Taip, and N. A. Aziz, Simulation and control of spray drying using nozzle atomizer spray dryer, International Journal of Engineering and Technology, vol. 9, pp. 12 17, 2009.

  11. A. Negiz, P. Ramanauskas, A. Cinar, J. E. Schlesser, and

    D. J. Amstrong, M odeling, monitoring and control strategies for HTST pasteurization systems 1. Empirical model development, Food Control, vol. 9, pp. 1 15, 1998.

  12. T. E. M arlin, Process Control: Designing Processes and Control Systems for Dynamic Performance, 2nd ed., United State of America: M cGraw Hill, 2000.

  13. C. A. Smith, and A. B. Corropio, Principle and Practice of Automatic Control, 3rd ed., New York: John Wiley, 2006.

  14. H. Y. Law, C. I. Ong, N. A. Aziz, F. S. Taip, and N. M uda, Preliminary work on coconut milk fouling deposits study, International Journal of Engineering & Technology, vol. 9, pp. 18 23, 2009.

  15. W. J. Gong, Y. P. Zhang, Y. J. Zhang, G. Xu, X. Wei, and K. Lee, Optimization strategies for separation of sulfadiazines using Box-Behnken design by liquid chromatography and capillary electrophoresis, Journal of Central South University of Technology, vol. 14, pp. 196 201, 2007.

  16. S. R. H. M . Khan, Development of regression models for predicting properties of high strength concret using non-destructive test, Ph.D. dissertation, Dept. of M echanical Engineering, Universiti Putra M alaysia, M alaysia, 2007.

Leave a Reply