- Email: [email protected]
Fuzzy concepts applied to food product quality control: A review
Fuzzy Sets and Systems 157 (2006) 1145 – 1154 www.elsevier.com/locate/fss
Fuzzy concepts applied to food product quality control: A review N. Perrota,∗ , I. Ioannoub , I. Allaisc , C. Curtc , J. Hossenloppc , G. Trystramc a UMR GMPA, INRA, France b INPL Nancy, France c REQUALA, UMR Génie Industriel Alimentaire—Automatic applied to food processes, France
Available online 20 January 2006
Abstract Fuzzy logic is now a wide field of study and different tools have been developed over the last 10 years. Its implementation in food quality control for the food industry has been highlighted by several authors that have focused on different applications designed specifically for this task. This is especially true in the case of taking into account the reasoning process, expressed in linguistic terms, of operators and experts. Nevertheless, applications are still limited and few reviews on this topic are available. Consequently, the aim of this paper is to provide an overview of the application of fuzzy concepts to the control of the product quality in the food industry over the past 10 years. Future interesting developments and trends in this area are also emphasized. © 2006 Elsevier B.V. All rights reserved. Keywords: Food processes; Fuzzy logic; Control; Quality; Review
1. Introduction In the food industry, end-products must achieve a compromise between several properties, including sensory, sanitary and technological properties. Among the latter, sensory and sanitary properties are essential because they influence consumer choice and preference. Nevertheless, managing these properties right from the fabrication stage with the aim of controlling them is no easy task for several reasons: • The food industry works with many parameters that must be taken into account in parallel. A single sensory property like colour or texture can be linked individually to several dimensions registered by the human brain. • The food industry works with non-uniform, variable raw materials that, when processed, should lead to a product that satisfies a fixed standard. • The phenomena involved in the processing are highly non-linear and variables are coupled. • The food industry operates with very diverse processes and products and has requirements in terms of the portability and adaptability of the systems developed. • Little data are available in traditional manufacturing plants that produce, for example, sausage or cheese and this situation is general throughout the food industry. Furthermore, even when databases do exist, it is not always possible to use them for controlling food product quality. ∗ Corresponding author. INRA, UMR GMPA, 78850 Thiverval-Grignon, France. Tel.: +33 1 30 81 53 79.
E-mail addresses: [email protected] (N. Perrot), [email protected] (I. Ioannou), [email protected] (G. Trystram). 0165-0114/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.fss.2005.12.013
1146
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
number of articles in the field food and fuzzy
16 14 12 10 8 6 4 2 0 1991 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 year Fig. 1. Number of articles published in the field crossing the fuzzy logic and the control of the food processes.
In this context, despite the fact that the design of standards and reliable procedures for controlling the quality of products is a major objective for the food industry, automation is limited: (i) Few sensors are available to carry out such measurements. Although new sensors have been developed such as artificial noses, the road is difficult and long and inaccessible for SMEs. (ii) For several processes, it is difficult to established models sufficiently representative of the phenomenon involved, even for control purposes. (iii) Classical automated approaches are limited for the reasons mentioned below. At present many production processes rely to a great extent on the skill and experience of the operator, something that no system will be capable of replacing in the foreseeable future. Consequently, in practice, operators often play an important role and cooperate with automation so as to (1) make on-line evaluations of the sensory properties of the product and/or (2) adjust the on-line process. Moreover, experienced operators make macroscopic interpretations of the physicochemical phenomena that appear during processing, which can act in synergy with classical engineering knowledge on the process. Integrating operator and expert skill in a control framework is a relevant direction, especially for traditional processes. Nevertheless, it leads to designing mathematical tools that have to integrate (i) reasoning based on the use of linguistic symbols such as “over-coated”, “good colour”, etc., expressed not on a numerical scale but on a discontinuous graduated scale and referring to an evaluation of a deviation in comparison to a set point; (ii) an uncertainty on these symbols that is translated after fusion in a specific action; and (iii) an action that is the result of an implicit or explicit interpolation between two specific state recorded by the operator over time. Fuzzy sets and possibility theories were introduced by Zadeh in 1965 [92] as an extension of the set theory by the replacement of the characteristic function of a set by a membership function whose values range from 0 to 1. It is now a wide field of study that has seen the development of different tools over the last 10 years. Applied to the control of product quality in the food industry, it has been considered as pertinent by several authors for different applications and especially for taking into account the reasoning process, expressed in linguistic terms, of operators and experts [18,24,48,66,77,94]. Nevertheless, applications are still limited and few reviews on this topic are available. In this framework, the aim of this paper is to provide an overview of the application of fuzzy concepts for controlling product quality in the food industry over the last 10 years. The first papers on this topic appeared 15 years ago although the volume of literature really began to increase from 1996 (Fig. 1). All in all, 78 applications have been dedicated to this topic over the last 12 years. This topic involves different subjects: (1) representation of the descriptive sensory evaluation performed by a quality team, an operator, or a consumer; (2) indirect measurement of the properties of a food product; (3) diagnosis, supervision, and control of food quality. The proportion of papers dedicated to each of these research fields is illustrated in Fig. 2, which shows more than 80% of papers being dedicated to fields (2) and (3), thus they are well represented in comparison
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
1147
contribution to sensory analysis 14%
contribution to food process control 51%
contribution to indirect measurement of the food quality 35%
Percentage of assessors
Fig. 2. Proportion of the four typical fields encountered in the topic fuzzy and food processes in the whole papers.
55% 25% 15% 5%
0
a
b
c
d
e
Colour Fig. 3. Distribution of probability obtained from the assessors on a given descriptor of a chocolate: the colour upon [63].
to the others. Nevertheless, even though fewer papers are currently dedicated to the field of sensory analysis and process modelling, the emergence of fuzzy concepts is apparent in 75% of the papers dealing with this field published over the last 6 years.
2. Representation of descriptive sensory evaluation performed by a quality team, an operator and a consumer The application of fuzzy concepts is quite recent in this field. Works are more especially focused on the power of fuzzy logic to represent the semantics of human assessment in the field of sensory analysis. For example, Davidson et al. [20] suggests a linguistic format for the sensory assessment of foods as well as processing methods to analyse taste panel opinions within the framework of the fuzzy set theory. Tan et al. [82] describes an initial attempt to examine fuzzy formalism for sensory analysis and demonstrates how fuzzy sets may lead to a natural way of interpreting sensory data. They both establish that analogies exist between fuzzy entities and sensory entities, the same hierarchy being present in sensory analysis as in fuzzy logic, i.e. sensory scales as fuzzy sets, sensory attributes as fuzzy variables, and sensory answers as membership grades. Omri et al. [62] uses fuzzy models and software that replaces numeric scales with semantic ones. For example, the colour (Fig. 3) and the corresponding distribution of possibility are built on the basis of a distribution of probability obtained from assessors for a given descriptor of a chocolate. Fuzzy mathematical t-norm and t-conorm are thus handled in the place of the statistical tools used classically by the sensory analysis community.
1148
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
3. Indirect measurement or estimation of foodstuff quality Indirect measurement can be used alone for inspection, as in Chao et al. [12] for poultry postmortem classification of colour images of viscera to detect airsacculitis livers from normal livers or as in Nielsen and Paul [59] for tomato quality grading. It is also used coupled with a controller or embedded in a decision help system to control the quality of the food product as in Davidson et al. [21] to control peanut roasting (Fig. 4) or as in Perrot et al. [66] to control the quality of biscuits in an industrial tunnel oven. Requirements in terms of measurement precision and reliability are therefore different. In both cases, it is most often a “feature level fusion” as defined in Valet et al. [87]. The fusion is performed on specific features extracted from raw information generally provided by sensors such as cameras, electronic noses, NMR sensors, etc. For example, Yea et al. [91] used three commercial gas sensors to detect four kinds of fragrant smells with a discrimination rate of 99.2%. In the first step, fuzzy reasoning detects the fragrant smells followed by odour discrimination done by a neural network. De Silva et al. [23] used features, extracted by PCA, from acoustic video images to determine the firmness of herring roe. A fuzzy decision-making system test on 160 samples gave results with about 84% good discrimination. Sundic et al. [79] uses features extracted from an electronic tongue coupled with an electronic nose to mimic the human sensory perception of potato chips and cream. In some applications, when no sensors are available for technical or economic reasons, the only inputs of the indirect measurement module are human operators. In this case, the symbolic space handled by the operators to evaluate the product is used directly by the system as in Ioannou et al. [38]. In this paper, the operator sensory perception of the product is evaluated through three variables expressed on a semantic scale directly computed by the module of indirect measurement to evaluate the degree of browning. Results show about 92.5% good discrimination. This application is representative of situations often encountered in the food industry, where it is important to propose mathematical approaches capable of coping directly with symbolic data handled by the operators. The fuzzy mathematical tool used can be a fuzzy classifier like the fuzzy C means, as in Loufti et al. [50], and the fuzzy k-nearest neighbour algorithm as in Raptis et al. [69]. In this application, for example, it is used to compute the aroma and taste of wine to discriminate its age. The training set is constituted by about 140 data. Globally, the size of the training set required in the papers concerned with these types of approach is from 50 to 150 data. The fuzzy mathematical tool used can also be a fuzzy rule basis. The classical Mamdani [51] format and Takagi–Sugeno [81] formats are often handled, as in [20,23,31,76]. For example, in Harris [31], different types of milk are classified on the basis of two biochemical measurements and two microbiological measurements. For eight out of 13 papers, the parameters of the fuzzy rule basis are calibrated using a data driven approach, as in Shahin et al. [76] for sorting apples and Perrot et al. [67] for biscuit quality evaluation. For the other applications (five papers out of 13), the fuzzy rule bases are built from expert knowledge as in Nielsen and Paul [59] for tomato grading, Mauris et al. [55] for sausage crusting
Colour Defects
Nut Size (fuzzy) Fuzzy Logic Controller
Crisp Temperature
Gp/Ga (fuzzy)
Set Point Crisp Residence Time
PID Heater Control
Feedback
Feedforward
Colour Setpoint (fuzzy)
Roasting Process
Fig. 4. Indirect measurement of the residence time to control the peanut roasting upon [22].
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
Color measurement by the operator
1149
Classes of color Diagnostic and action – Fuzzy symbolic approach
Thickness measurement by a sensor
Fuzzy classification
Moisture content measurement by a sensor
Fuzzy classification
Temperature set points for the continuous oven
INDUSTRIAL TUNNEL OVEN Biscuits Fig. 5. Estimation of the state of the biscuits using both the operators and the sensors measurements in [68].
evaluation, and Da-Ven-Sun and Brosnan [17] for pizza quality evaluation. In certain examples, fuzzy rule bases are sometimes combined with other tools, such as neural networks. For example, Yea et al. [91] uses this configuration to establish discrimination of gas odours. Nevertheless, few papers deal with the problem of fusing data of different natures (numeric and linguistic), although this is a key problem in the food industry. In these papers, the ability of fuzzy logic theory to cope with such problems seems highly significant and interesting for several reasons: (i) It is a simple way of unifying different scales of measurement (linguistic and numeric scales); (ii) The mathematical formalism is easy to understand by the experts; (iii) It allows capturing different types of imprecision from sensors and human perceptions. For example, Perrot et al. [66] proposes estimating the state of a biscuit in an industrial tunnel oven by using human and sensor information (Fig. 5). Biscuit colour is evaluated by the expert, whereas information on moisture and thickness comes from sensors.
4. Diagnosis, supervision, and control of food quality This subject represents a fairly large number of published papers in the field “fuzzy and food” (51% of the total number). Most of them are classical applications of the Takagi–Sugeno controller, such as Linko et al. [47] for extrusion cooking, Zhang and Litchfield [94] for drying, Norback [60] for cheese-making, Alvarez et al. [4] for controlling isomerised hop pellet production, Honda et al. [32] for controlling the sake brewing process, and O’connor et al. [61] for controlling the brewing process. Two specific approaches are used by the authors to develop the diagnosis, control, and supervision modules: (i) data-driven approaches and (ii) “expert knowledge”-driven approaches. The first case includes 11 out of 40 papers written on this topic. We have chosen to include the fuzzy PID implementations in this category. Identification of the system is done automatically using a data basis and an optimisation algorithm. For example, Honda et al. [32] developed a fuzzy neural network to control the temperature of the Ginjo sake mashing process, O’connor et al. [61] developed a fuzzy PID to control the brewing process, Guillaume et al. [29] optimised a fuzzy rule basis using a genetic algorithm to establish a decision-aid system for the cheese-making process. In this case, the risk can be to develop models composed with a large number of rules (from 50 to 100 rules or more) and as a consequence a large number of parameters to identify. For example in Honda et al. [32], 283 data points are used to identify the model. This can lead to difficulties in handling such quantities of data to set all the parameters and is a major limitation for applications used in food processes. Moreover, the tools built can only be used to interpolate
1150
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
and not extrapolate new cases. Nevertheless, the added value brought about by using fuzzy logic in this case relies on (i) a friendly operator interface that handles data in a space close to the symbolic space handled by the operators; (ii) the possibility of automatically adapting the module to new processing conditions (such as a new formula for a biscuit) if a model or a set with enough data to characterize the process are available. In the “expert knowledge”-driven approaches, fuzzy logic is used as a way of providing a mathematical formalization of expert knowledge and embedding it in decision-aid help algorithms and controllers. For example, Davidson et al. [21] develops a fuzzy control system for continuous crossflow peanut roasting, Perrot et al. [66] proposes a fuzzy logic approach to control the quality of the biscuits in an industrial tunnel oven, Voos et al. [89] develops a fuzzy control of a drying process in the sugar industry based on operator experience, and Curt et al. [16] develops five Takagi–Sugeno modules to control the quality of the sausage during ripening. It is also used to perform supervisory tasks such as Acosta-Lazo et al. [1] for the supervision of a sugar factory. In all, 19 out of 40 papers have been written on this topic. In this case, modules are used directly without adaptation using a database while the data sets are only used to validate the approaches. The rules bases are generally compact with a limited number of rules (around 20 rules on average). For example, 11 rules are implemented to control the quality of the biscuits in Perrot et al. [66]. According to the authors Davidson [18] or O’connor et al. [61], fuzzy logic formalism is particularly well adapted for capturing how operators think and for capitalising human knowledge. Moreover, the robustness of certain specific models are underlined in some studies (for example Ioannou et al. [37]). Nevertheless, these properties depend on each specific model built and no theoretical demonstration on the robustness generally provided by fuzzy formalism is available. Adaptation of such models is generally achieved by hand although in some studies (seven papers out of 40), an initial structure of the fuzzy model is defined with the experts and an optimisation tool is used to optimise it or identify the fuzzy model’s parameters as in Perrot et al. [64] on a filtration process or Davidson et al. [21] on a peanut roasting process. The difficulty in this case is to keep the fundamental knowledge brought by the experts. Last but not least, we have chosen to highlight the papers focusing on the semantic scope of fuzzy logic and the ability of such methodology to capture expert reasoning and knowledge, and its uncertainty. Thus, Davidson et al. [21] used a fuzzy arithmetic that estimates peanut eating time and browning to control peanut roasting, Perrot et al. [63] developed a decision help system to control the cheese ripening process, integrating the uncertainty of human measurements. Petermeier et al. [68] used a hybrid approach to develop a model of the fouling behaviour of an arbitrary heat treatment device for milk. This is developed by combining deterministic differential equations with cognitive elements for the unknown parts of the knowledge model. These authors emphasize the relevance of this open field of research in the context of food processes and the interest of fuzzy symbolic representation of expert reasoning. Nevertheless, they call into question the optimality of the approaches developed on the basis of imperfect and incomplete expert knowledge. 5. Conclusion As we have seen, fuzzy logic is used in food applications to (i) capture and formalise the descriptive sensory evaluation performed by a quality team, an operator, or a consumer, (ii) develop an indirect measurement of the properties of a food product, and (iii) control food processes. Fig. 6 presents a classification of the different papers written in this area on the different research topics. If we focus on the type of approaches developed (Fig. 7), on the one hand, 33 papers deal exclusively with data-driven approaches including fuzzy PID. A total of 66% of these papers are above all dedicated to indirect measurement tasks, while 34% are dedicated to the control and modelling of processes. This category represents only a small percentage out of the total, which can be explained by a key difficulty in the food industry, especially in traditional manufacturing plants: that of constituting databases that can be easily used for control purposes. On the other hand, “expert knowledge”-driven approaches are dealt with in 35 papers. Mixed approaches are encountered in seven papers. If we focus on the type of fuzzy concepts applied (Fig. 8), 74% of the applications dealt with stem from the fuzzy set theory and most of them implement classical fuzzy logical functions (Mamdani type). On the contrary, the theory of possibility is hardly used, being dedicated to a restricted task in only three of the total of 78 papers [39,37,23]. Nevertheless, the authors underline a really interesting and open field of research. For example, in De Silva et al. [23], a firmness sensor for an automated herring roe grader is developed using fuzzy concepts. The theory of possibility is used in this case to estimate the fuzzy membership functions of the fuzzy decision-making system. The results show that this approach is more efficient than the use of classical trapezoidal membership functions.
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
1151
Food and fuzzy 1993-2004 78 papers 11 papers
40 papers
27 papers
Sensory analysis
Indirect measurement
13 papers
Supervision Decision help system Control
Process knowledge improvement
14 papers
Fuzzy rule basis
4 papers
Classification
Knowledge driven approaches
Data driven approaches 8 papers
5 papers
Data driven approaches
Knowledge driven approaches
Mixed or unknown approach
11 papers
15 papers
10 papers
Fig. 6. Classification of the papers written in the topic “fuzzy logic and the quality control of the food products”.
Mixed approaches 8%
Data driven approaches 42 %
Knowledge driven approaches 45 % Fig. 7. Proportion of the three different types of identification of the fuzzy models in the whole papers.
fuzzy adaptation of classification tools 18% Fuzzy-neural tools 8%
53% 21% Takagi-Sugeno
Mamdani or logical operators associated MaxMin
Fig. 8. Proportion of the different fuzzy concepts manipulated in the whole papers.
1152
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
List of the applications on the four topics Representation of the descriptive sensory evaluation [20,22,28,31,39,57,62,82–84,86] Indirect measurement Fuzzy rule basis Data-driven approaches: [12,13,17,23,49,76,79,91] Knowledge driven approaches: [11,36,55,59,93] Classification [2,5–7,14,33,41,46,50,52,67,69,73,75] Diagnosis, supervision or control Data-driven approaches [4,25,29,32,43–45,70,72,88,90] Knowledge-driven approaches [1,8,16,26,35,37,38,42,47,53,54,65,66,71,94] Mixed or unknown approaches Mixed: [10,21,60,61,64,80,89] Unknown: [3,40,74] Process knowledge improvement [21,30,63,68]
References [1] G.G. Acosta-Lazo, C.J. Alonso-Gonzales, B. Pulido-Junquera, Knowledge based diagnosis of a sugar process with teknolid, International Sugar J. 103 (2001) 44–51. [2] G. Adorni, D. Bianchi, S. Cagnoni, Ham quality control by means of fuzzy decision trees: a case study, IEEE International Conference on Fuzzy Systems, Proc. IEEE World Congress on Computational Intelligence, vol. 2, 1998, pp. 1583–1588. [3] L. Aldini, Utilization of fuzzy logic in diffusers, Ind. Sacc. Italiana 91 (1998) 83–89. [4] E. Alvarez, M.A. Cancela, J.M. Correa, J.M. Navaza, C. Riverol, Fuzzy logic control for the isomerized hop pellets production, J. Food Engrg. 39 (1999) 145–150. [5] C. Amza, Bones detection from chicken breast meat using a competitive Hopfield neural network and fuzzy reasoning, in: Scientific Bulletin Series C: Electrical Engineering, University Politehnica of Bucharest, 2001, pp. 53–64. [6] G. Andreotti, R. Lamanna, E. Trivellone, A. Motta, 13C NMR spectra: an easy way to distinguish milks from different animal species, J. Amer. Oil Chemist’s Soc. 79 (2002). [7] L. Ballerini, A. Hoegberg, K. Lundstroem, G. Borgefors, Color image analysis technique for measuring of fat in meat: an application for the meat industry, Proc. SPIE—Internat. Soc. Optical Engrg. 4301 (2001) 113–124. [8] R.B. Brown, T.M. Rothwell, V.J. Davidson, A fuzzy controller for infrared roasting of cereal grain, Canadian Biosyst. Engrg. 43 (2001) 3.9–3.15. [9] P. Buche, C. Dervin, A. Brouillaud-Delattre, N. Gnanou-Besse, Combining fuzzy querying of imprecise data and predictive microbiology using category-based reasoning for prediction of the possible microbial spoilage in foods: application to Listeria Monocytogenes, Internat. J. Food Microbiol. 73 (2002) 171–185. [10] D.A. Campbell, M.A. Pecar, M.J. Lees, Intelligently controlled beer filtration, Joint Conf. on Intelligent Systems, vol. 1, 1998, pp. 313–316. [11] O. Castillo, P. Melin, Intelligent quality control for manufacturing in the food industry using a new fuzzy-fractal approach, Proc. 2nd Internat. Conf. on Intelligent Processing and Manufacturing of Materials, vol. 1, 1999, pp. 151–156. [12] K. Chao, Y.R. Chen, H. Early, B. Park, Color image classification systems for poultry viscera inspection, Appl. Eng. Agric. 15 (1999) 363–369. [13] W. Chong-Yaw, R. Paramesran, F. Takeda, T. Tsuzuki, H. Kadota, S. Shimanouchi, Classification of rice grains using fuzzy artmap neural network, Proc. APCCAS. Asia-Pacific Conf. Circuits Syst. 2 (2002) 223–226. [14] Y. Chtioui, S. Panigrahi, L.F. Backer, Self-organizing map combined with a fuzzy clustering for color image segmentation of edible beans, Trans. ASAE 46 (2003) 831–838. [15] D. Corney, Food bytes: intelligent systems in the food industry, British Food J. 104 (2002) 787–805. [16] C. Curt, J. Hossenlopp, N. Perrot, G. Trystram, Dry sausage ripening control. Integration of sensory related properties, Food Control 13 (2002) 151–159. [17] Da-Ven-Sun, T. Brosnan, Pizza quality evaluation using computer vision—Part 1: pizza base and sauce spread, J. Food Engrg. 57 (2003) 81–89. [18] V. Davidson, Fuzzy control of food processes, in: G. Mittal (Ed.), Computerized Control Systems in the Food Industry, New York, Basel, Hong Kong, 1996, pp. 179–205.
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
1153
[19] V. Davidson, J. Ryks, Comparison of Monte Carlo and fuzzy math simulation methods for quantitative risk assessment, J. Food Protection 10 (2003) 1900–1910. [20] V. Davidson, J. Ryks, T. Chu, Fuzzy models to predict consumer ratings for biscuits based on digital image features, IEEE Trans. Fuzzy Systems 9 (2001). [21] V.J. Davidson, R.B. Brown, J.J. Landman, Fuzzy control system for peanut roasting, J. Food Engrg. 41 (1999) 141–146. [22] V.J. Davidson, W. Sun, A linguistic method for sensory assessment, J. Sensory Stud. (1998) 315–330. [23] C.W. De Silva, L.B. Gamage, R.G. Gosine, An intelligent firmness sensor for an automated herring roe grader, Intell. Automat. Soft Comput. 1 (1996) 99–114. [24] D. Dubois, L. Foulloy, S. Galichet, H. Prade, Performing approximate reasoning with words, in: J. Kacprzyk, L. Zadeh (Eds.), Computing with Words in Information Intelligent Systems 1, Physica-Verlag, 1999, pp. 24–29. [25] M. Estaben, M. Polit, J.P. Steyer, Fuzzy control for an anaerobic digester, Control Engrg. Practice 5 (1997) 1303–1310. [26] E.R. Fowler, New techniques for commercial bread dough mixing, IEEE Instrum. Measurement Magazine 3 (2000) 21–25. [27] F. Goyache, A. Bahamonde, J. Alonso, S. Lopez, J.J. Del Coz, J.R. Quevedo, J. Ranilla, O. Luaces, I. Alvarez, L.J. Royo, J. Diez, The usefulness of artificial intelligence techniques to assess subjective quality of products in the food industry, Food Sci. Tech. 12 (2001) 370–381. [28] A.S. Guillard, F. Blon, J.L. Vendeuvre, Influence de la matière première et du procédé, VPC 20 (1999). [29] S. Guillaume, B. Charnomordic, Knowledge discovery for control purposes in food industry databases, Fuzzy Sets and Systems 122 (2001) 487–497. [30] T. Hanai, N. Ueda, H. Honda, H. Tohyama, T. Kobayashi, Deciding the temperature course during sake mashing using a GA-FNN for quality control of sake, Seibutsu Kogaku Kaishi 76 (1998) 331–337. [31] J. Harris, Raw milk grading using fuzzy logic, Soc. Dairy Tech. 51 (1998) 52–56. [32] H. Honda, T. Hanai, A. Katayama, T.H.T. Kobayashi, Temperature control of Ginjo sake mashing process by automatic fuzzy modeling using fuzzy neural networks, J. Fermentation Bioengrg. 85 (1998) 107–112. [33] B.-G. Hu, R.G. Gosine, L.X. Cao, C.W. de Silva, Application of a fuzzy classification technique in computer grading of fish products, IEEE Trans. Fuzzy Systems 6 (1998) 144–152. [34] S.V. Iiyukhin, T.A. Haley, R.K. Singh, A survey of automation practices in the food industry, Food Control 12 (2001) 285–296. [35] I. Ioannou, N. Perrot, C. Curt, G. Mauris, G. Trystram, Development of a control system using the fuzzy set theory applied to a browning process—A fuzzy symbolic approach for the measurement of product browning: development of a diagnosis model—Part I, J. Food Engrg. 64 (2004) 497–506. [36] I. Ioannou, N. Perrot, J. Hossenlopp, G. Mauris, G. Trystram, The fuzzy set theory: A helpful tool for the estimation of sensory properties of crusting sausage appearance by a single expert, Food Quality Preference 13 (2002) 589–595. [37] I. Ioannou, N. Perrot, G. Mauris, G. Trystram, Experimental analysis of sensory measurement imperfections impact for a cheese ripening fuzzy model, IFSA Fuzzy logic, soft computing and computational intelligence theories and applications congress, Turquie (2003) 595–602. [38] I. Ioannou, N. Perrot, G. Mauris, G. Trystram, Development of a control system using the fuzzy set theory applied to a browning process—Towards a control system of the browning process combining a diagnosis model and a decision model—Part II, J Food Engrg. 64 (2004) 507–514. [39] S. Jaya, H. Das, Sensory evaluation of mango drinks using fuzzy logic, J. Sensory Stud. 18 (2003) 163–176. [40] B. Kett, Optimising recipes for price and performance, Innovations Food Tech. 18 (2003) 59–61. [41] L. Khodja, L. Foulloy, E. Benoit, T. Talou, Fuzzy techniques for coffee flavour classification, Inform. Process. Management Uncertainty Knowledge-Based Systems 1 (1996) 709–719. [42] S. Kim, S.I. Cho, Neural network modeling and fuzzy control simulation for bread-baking process, Trans. ASAE 40 (1997) 671–676. [43] A.B. Koc, P.H. Heinemann, G.R. Ziegler, W.B. Roush, Fuzzy logic control of whole milk powder processing, Trans. ASAE 45 (2002) 153–163. [44] T. Kurz, M. Fellner, T. Becker, A. Delgado, Observation and control of the beer fermentation using cognitive methods, J. Inst. Brewing 107 (2001) 241–252. [45] Lahtinen, Identification of fuzzy controller for use with a falling film evaporator, Food Control 12 (2001) 175–180. [46] M. Leunissen, V. Davidson, Y. Kakuda, Analysis of volatile flavour components roasted peanuts using supercritical fluid extraction and gaz chromatography mass spectrometry, J. Agric. Food Chem. 44 (1996) 2694–2699. [47] P. Linko, K. Uemura, Y. Zhu, T. Eerikainen, Application of neural modeling in fuzzy extrusion control, Trans. I. Chem. E 70 (1992) 131–137. [48] S. Linko, Expert systems—What can they do for the food industry, Trends Food Sci. Tech. 9 (1998) 3–12. [49] E. Llobet, E.L. Hines, J.W. Gardner, S. Franco, Non-destructive banana ripeness determination using a neural network-based electronic nose, Measurement Sci. Tech. 10 (1999) 538–548. [50] A. Loufti, P. Wide, Symbolic estimation of food odors using fuzzy techniques, IPMU Information Processing and management of uncertainty in knowledge based systems, Annecy (France) 2 (2002) 919–925. [51] E. Mamdani, Applications of fuzzy algorithms for control of simple dynamic plant, IEE 121 (12) (1974) 1585–1588. [52] F. Marcelloni, Recognition of olfactory signals based on supervised fuzzy C-means and k-NN algorithms, Pattern Recognition Lett. 22 (2001) 1007–1019. [53] G. Martinez, A. Lopez, A. Esnoz, P. Virseda, J. Ibarrola, A new fuzzy control system for white wine fermentation, Food Control 10 (1999) 175–180. [54] G. Mauris, N. Perrot, L. Berrah, S. Hamlaoui, Determining a confidence index of cheese ripening prediction by fuzzy trend sensory indicators, IEEE Instrum. Measurement Tech. Conf. 2 (2003) 1005–1008. [55] G. Mauris, N. Perrot, P. Lambert, J. Philippe, Fuzzy techniques evaluate sausage quality, IEEE Instrum. Measurement 3 (2000) 14–17. [56] M.J. McGrath, J.F. O’Connor, S. Cummins, Implementing a process control strategy for the food processing industry, J. Food Engrg. 35 (1998) 313–321. [57] T. Naes, E. Kubberod, H. Sivertsen, Identifying and interpreting market segments using conjoint analysis, Food Quality Preference 12 (2001) 133–143.
1154
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
[58] C. Nicolas, P. Carel, J. Hossenlopp, G. Trystram, D.N. Rutledge, C. Emonet, Correlation between sensory data, instrumental data (gas sensors, physico-chemical analysis) and palatability measurements of twelve moist foods for cats, Sci. Aliments 19 (1999) 135–141. [59] H.M. Nielsen, W. Paul, Modelling image processing parameters and consumer aspects for tomato quality grading, IFAC Mathematical and Control Appl. Agric. Horticulture (1997) 141–146. [60] J.P. Norback, Natural language computer control of crucial steps in cheese making, Artificial Neural Networks Dairy Industry 49 (2) (1994) 119–122. [61] B. O’Connor, C. Riverol, P. Kelleher, N. Plant, R. Bevan, E. Hinchy, J. D’Arcy, Integration of fuzzy logic based control procedures in brewing, Food Control 13 (2002) 23–31. [62] M.N. Omri, I. Urdapilleta, J. Barthelemy, B. Bouchon-Meunier, C.A. Tijus, Semantic scales and fuzzy processing for sensorial evaluation studies, IPMU Information Processing and Management of Uncertainty in Knowledge-Based Systems, Spain II (1996) 715–719. [63] N. Perrot, L. Agioux, I. Ioannou, G. Mauris, G. Corrieu, G. Trystram, Decision support system design using the operator skill to control cheese ripening—Application of the fuzzy symbolic approach, J. Food Engrg. 64 (2004) 321–333. [64] N. Perrot, L. Mé, G. Trystram, J. Trichard, M. Decloux, Optimal control of the microfiltration of sugar product using a controller combining fuzzy and genetic approaches, Fuzzy Sets and Systems 94 (1998) 309–322. [65] N. Perrot, G. Trystram, F. Guely, Food product quality feed-back control and fuzzy sets—Application to an industrial baking process, in: Second European Congress of Chemical Engineering, 7ème congrès Français de Génie des Procédés-ECCE2, Montpellier, France, 1999. [66] N. Perrot, G. Trystram, F. Guely, F. Chevrie, N. Schoesetters, E. Dugre, Feed-back quality control in the baking industry using fuzzy sets, J. Food Process Engrg. 23 (2000) 249–279. [67] N. Perrot, G. Trystram, D. Le Gennec, F. Guely, Sensor fusion for real time quality evaluation of biscuit during baking. Comparison between Bayesian and Fuzzy approaches, J. Food Engrg. 29 (1996) 301–315. [68] H. Petermeier, R. Benning, A. Delgado, U. Kulozik, J. Hinrichs, T. Becker, Hybrid model of the fouling process in tubular heat exchangers for the dairy industry, J. Food Engrg. 55 (2002) 9–17. [69] C.G. Raptis, C.I. Siettos, C.T. Kiranoudis, G.V. Bafas, Classification of aged wine distillates using fuzzy and neural network systems, J. Food Engrg. (2000) 267–275. [70] Riza, D. Priyasta, F.R. Triputra, I. Ismail, PLC-based fuzzy controller for sterilizing process in crude palm oil mill, intelligent control for agricultural applications 2001, Proceedings volume from the 2nd IFAC/CIGR Workshop, 2002, pp. 247–252. [71] J. Ruger, B. Langhans, S. Alender, Fuzzy control in pulp drying, Zuckerindustrie 120 (1995) 387–398. [72] R. Rywotycki, Food frying process control system, J. Food Engrg. 4 (2003) 339–342. [73] C. Sarbu, H.W. Zwanziger, Fuzzy classification and comparison of some Romanian and German mineral waters, Analytical Lett. 34 (2001) 1541–1552. [74] Seung-Ju-Lee, Chang-Gi-Hong, Tack-Su-Han, Jun-Young-Kang, Young-An-Kwon, Application of fuzzy control to start-up of twin screw extruder, Food Control 13 (2002) 301–306. [75] M.A. Shahin, E.W. Tollner, R.W. Mcclendon, Artificial intelligence classifiers for sorting apples based on watercore, J. Agric. Engrg. Res. 73 (2001) 265–274. [76] M.A. Shahin, B.P. Verma, E.W. Tollner, Fuzzy logic model for predicting peanut maturity, Trans. ASAE 43 (2000) 483–490. [77] R. Singh, F. Ou-Yang, Neuro-fuzzy technology for computerized automation, in: G. Mittal (Ed.), Computerized Control Systems in the Food Industry, New York, Basel, Hong Kong, 1996, pp. 119–178. [78] W. Sun, V. Davidson, Dynamic fuzzy-reasoning-based function estimator, Fuzzy Sets and Systems 79 (1996) 357–366. [79] T. Sundic, S. Marco, J. Samitier, P. Wide, Electronic tongue and electronic nose data fusion in classification with neural networks and fuzzy logic based models, Proceedings of the 17th IEEE Instrum. Measurement Tech. Conf. 3 (2000) 1474–1479. [80] M. Syfert, J. Nowak, Diagnostics of evaporator in a sugar factory based on fuzzy models, fault detection, supervision and safety for technical processes, Proceedings volume from the 4th IFAC Symp., vol. 1, 2001, pp. 337–342. [81] T. Takagi, M. Sugeno, Fuzzy identification of systems and its application to modelling and control, IEEE Trans. Systems Man Cybernetics 15 (1985) 116–132. [82] J. Tan, X. Gao, D.E. Gerrard, Application of fuzzy sets and neural networks in sensory analysis, J. Sensory Stud. 14 (1999) 119–138. [83] F. Tian-Jin, D. Xiang-Qian, T. Rui, F. Zuen-Cheng, A hybrid expert system for formulated product designs, Proceedings of the IASTED International Conference, Artificial Intell. Appl. (2001) 336–341. [84] O. Tominaga, F. Ito, T. Hanai, H. Honda, T. Kobayashi, Modeling of consumer’s preferences for regular coffee samples and its application to product design, Food Sci. Tech. Res. 8 (2002) 281–285. [85] G. Trystram, F. Courtois, Food processing control: reality and problem, Food Res. Internat. 27 (1994) 173–185. [86] I.B. Turksen, I.A. Willson, A fuzzy set model for market share and preference prediction, European J. Oper. Res. 82 (1995) 39–52. [87] L. Valet, G. Mauris, P. Bolon, A Statistical Overview of Recent Literature in Information Fusion, ISIF, Paris, France, July 2000. [88] G.K. Venayagamoorthy, D. Naidoo, P. Govender, An industrial food processing plant automation using a hybrid of PI and fuzzy logic control, IEEE Internat. Conf. Fuzzy Systems 2 (2003) 1059–1062. [89] H. Voos, L. Litz, H. Konig, Fuzzy control of a drying process in sugar industry, 6th European Congress Intell. Tech. Soft Comput. EUFIT ’98, vol. 3, 1998, pp. 1476–1480. [90] G. Xie, R. Xiong, I. Church, Comparison of kinetics, neural network and fuzzy logic in modelling texture changes of dry peas in long time cooking, Lebensmittel Wissenschaft Technologie 31 (1998) 639–647. [91] B.Yea, R. Konishi, T. Osaki, K. Sugahara, The discrimination of many kinds of odor species using fuzzy reasoning and neural networks, Sensors Actuators 45 (1994) 159–165. [92] L. Zadeh, Fuzzy sets, Inform. Control 8 (1965) 338–353. [93] Q. Zhang, J. Litchfield, Applying fuzzy mathematics to product development and comparison, Food Tech. 45 (1991) 108–115. [94] Q. Zhang, J. Litchfield, Fuzzy logic control for a continuous crossflow grain dryer, J. Food Process Engrg. 16 (1993) 59–77.
Fuzzy concepts applied to food product quality control: A review N. Perrota,∗ , I. Ioannoub , I. Allaisc , C. Curtc , J. Hossenloppc , G. Trystramc a UMR GMPA, INRA, France b INPL Nancy, France c REQUALA, UMR Génie Industriel Alimentaire—Automatic applied to food processes, France
Available online 20 January 2006
Abstract Fuzzy logic is now a wide field of study and different tools have been developed over the last 10 years. Its implementation in food quality control for the food industry has been highlighted by several authors that have focused on different applications designed specifically for this task. This is especially true in the case of taking into account the reasoning process, expressed in linguistic terms, of operators and experts. Nevertheless, applications are still limited and few reviews on this topic are available. Consequently, the aim of this paper is to provide an overview of the application of fuzzy concepts to the control of the product quality in the food industry over the past 10 years. Future interesting developments and trends in this area are also emphasized. © 2006 Elsevier B.V. All rights reserved. Keywords: Food processes; Fuzzy logic; Control; Quality; Review
1. Introduction In the food industry, end-products must achieve a compromise between several properties, including sensory, sanitary and technological properties. Among the latter, sensory and sanitary properties are essential because they influence consumer choice and preference. Nevertheless, managing these properties right from the fabrication stage with the aim of controlling them is no easy task for several reasons: • The food industry works with many parameters that must be taken into account in parallel. A single sensory property like colour or texture can be linked individually to several dimensions registered by the human brain. • The food industry works with non-uniform, variable raw materials that, when processed, should lead to a product that satisfies a fixed standard. • The phenomena involved in the processing are highly non-linear and variables are coupled. • The food industry operates with very diverse processes and products and has requirements in terms of the portability and adaptability of the systems developed. • Little data are available in traditional manufacturing plants that produce, for example, sausage or cheese and this situation is general throughout the food industry. Furthermore, even when databases do exist, it is not always possible to use them for controlling food product quality. ∗ Corresponding author. INRA, UMR GMPA, 78850 Thiverval-Grignon, France. Tel.: +33 1 30 81 53 79.
E-mail addresses: [email protected] (N. Perrot), [email protected] (I. Ioannou), [email protected] (G. Trystram). 0165-0114/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.fss.2005.12.013
1146
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
number of articles in the field food and fuzzy
16 14 12 10 8 6 4 2 0 1991 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 year Fig. 1. Number of articles published in the field crossing the fuzzy logic and the control of the food processes.
In this context, despite the fact that the design of standards and reliable procedures for controlling the quality of products is a major objective for the food industry, automation is limited: (i) Few sensors are available to carry out such measurements. Although new sensors have been developed such as artificial noses, the road is difficult and long and inaccessible for SMEs. (ii) For several processes, it is difficult to established models sufficiently representative of the phenomenon involved, even for control purposes. (iii) Classical automated approaches are limited for the reasons mentioned below. At present many production processes rely to a great extent on the skill and experience of the operator, something that no system will be capable of replacing in the foreseeable future. Consequently, in practice, operators often play an important role and cooperate with automation so as to (1) make on-line evaluations of the sensory properties of the product and/or (2) adjust the on-line process. Moreover, experienced operators make macroscopic interpretations of the physicochemical phenomena that appear during processing, which can act in synergy with classical engineering knowledge on the process. Integrating operator and expert skill in a control framework is a relevant direction, especially for traditional processes. Nevertheless, it leads to designing mathematical tools that have to integrate (i) reasoning based on the use of linguistic symbols such as “over-coated”, “good colour”, etc., expressed not on a numerical scale but on a discontinuous graduated scale and referring to an evaluation of a deviation in comparison to a set point; (ii) an uncertainty on these symbols that is translated after fusion in a specific action; and (iii) an action that is the result of an implicit or explicit interpolation between two specific state recorded by the operator over time. Fuzzy sets and possibility theories were introduced by Zadeh in 1965 [92] as an extension of the set theory by the replacement of the characteristic function of a set by a membership function whose values range from 0 to 1. It is now a wide field of study that has seen the development of different tools over the last 10 years. Applied to the control of product quality in the food industry, it has been considered as pertinent by several authors for different applications and especially for taking into account the reasoning process, expressed in linguistic terms, of operators and experts [18,24,48,66,77,94]. Nevertheless, applications are still limited and few reviews on this topic are available. In this framework, the aim of this paper is to provide an overview of the application of fuzzy concepts for controlling product quality in the food industry over the last 10 years. The first papers on this topic appeared 15 years ago although the volume of literature really began to increase from 1996 (Fig. 1). All in all, 78 applications have been dedicated to this topic over the last 12 years. This topic involves different subjects: (1) representation of the descriptive sensory evaluation performed by a quality team, an operator, or a consumer; (2) indirect measurement of the properties of a food product; (3) diagnosis, supervision, and control of food quality. The proportion of papers dedicated to each of these research fields is illustrated in Fig. 2, which shows more than 80% of papers being dedicated to fields (2) and (3), thus they are well represented in comparison
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
1147
contribution to sensory analysis 14%
contribution to food process control 51%
contribution to indirect measurement of the food quality 35%
Percentage of assessors
Fig. 2. Proportion of the four typical fields encountered in the topic fuzzy and food processes in the whole papers.
55% 25% 15% 5%
0
a
b
c
d
e
Colour Fig. 3. Distribution of probability obtained from the assessors on a given descriptor of a chocolate: the colour upon [63].
to the others. Nevertheless, even though fewer papers are currently dedicated to the field of sensory analysis and process modelling, the emergence of fuzzy concepts is apparent in 75% of the papers dealing with this field published over the last 6 years.
2. Representation of descriptive sensory evaluation performed by a quality team, an operator and a consumer The application of fuzzy concepts is quite recent in this field. Works are more especially focused on the power of fuzzy logic to represent the semantics of human assessment in the field of sensory analysis. For example, Davidson et al. [20] suggests a linguistic format for the sensory assessment of foods as well as processing methods to analyse taste panel opinions within the framework of the fuzzy set theory. Tan et al. [82] describes an initial attempt to examine fuzzy formalism for sensory analysis and demonstrates how fuzzy sets may lead to a natural way of interpreting sensory data. They both establish that analogies exist between fuzzy entities and sensory entities, the same hierarchy being present in sensory analysis as in fuzzy logic, i.e. sensory scales as fuzzy sets, sensory attributes as fuzzy variables, and sensory answers as membership grades. Omri et al. [62] uses fuzzy models and software that replaces numeric scales with semantic ones. For example, the colour (Fig. 3) and the corresponding distribution of possibility are built on the basis of a distribution of probability obtained from assessors for a given descriptor of a chocolate. Fuzzy mathematical t-norm and t-conorm are thus handled in the place of the statistical tools used classically by the sensory analysis community.
1148
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
3. Indirect measurement or estimation of foodstuff quality Indirect measurement can be used alone for inspection, as in Chao et al. [12] for poultry postmortem classification of colour images of viscera to detect airsacculitis livers from normal livers or as in Nielsen and Paul [59] for tomato quality grading. It is also used coupled with a controller or embedded in a decision help system to control the quality of the food product as in Davidson et al. [21] to control peanut roasting (Fig. 4) or as in Perrot et al. [66] to control the quality of biscuits in an industrial tunnel oven. Requirements in terms of measurement precision and reliability are therefore different. In both cases, it is most often a “feature level fusion” as defined in Valet et al. [87]. The fusion is performed on specific features extracted from raw information generally provided by sensors such as cameras, electronic noses, NMR sensors, etc. For example, Yea et al. [91] used three commercial gas sensors to detect four kinds of fragrant smells with a discrimination rate of 99.2%. In the first step, fuzzy reasoning detects the fragrant smells followed by odour discrimination done by a neural network. De Silva et al. [23] used features, extracted by PCA, from acoustic video images to determine the firmness of herring roe. A fuzzy decision-making system test on 160 samples gave results with about 84% good discrimination. Sundic et al. [79] uses features extracted from an electronic tongue coupled with an electronic nose to mimic the human sensory perception of potato chips and cream. In some applications, when no sensors are available for technical or economic reasons, the only inputs of the indirect measurement module are human operators. In this case, the symbolic space handled by the operators to evaluate the product is used directly by the system as in Ioannou et al. [38]. In this paper, the operator sensory perception of the product is evaluated through three variables expressed on a semantic scale directly computed by the module of indirect measurement to evaluate the degree of browning. Results show about 92.5% good discrimination. This application is representative of situations often encountered in the food industry, where it is important to propose mathematical approaches capable of coping directly with symbolic data handled by the operators. The fuzzy mathematical tool used can be a fuzzy classifier like the fuzzy C means, as in Loufti et al. [50], and the fuzzy k-nearest neighbour algorithm as in Raptis et al. [69]. In this application, for example, it is used to compute the aroma and taste of wine to discriminate its age. The training set is constituted by about 140 data. Globally, the size of the training set required in the papers concerned with these types of approach is from 50 to 150 data. The fuzzy mathematical tool used can also be a fuzzy rule basis. The classical Mamdani [51] format and Takagi–Sugeno [81] formats are often handled, as in [20,23,31,76]. For example, in Harris [31], different types of milk are classified on the basis of two biochemical measurements and two microbiological measurements. For eight out of 13 papers, the parameters of the fuzzy rule basis are calibrated using a data driven approach, as in Shahin et al. [76] for sorting apples and Perrot et al. [67] for biscuit quality evaluation. For the other applications (five papers out of 13), the fuzzy rule bases are built from expert knowledge as in Nielsen and Paul [59] for tomato grading, Mauris et al. [55] for sausage crusting
Colour Defects
Nut Size (fuzzy) Fuzzy Logic Controller
Crisp Temperature
Gp/Ga (fuzzy)
Set Point Crisp Residence Time
PID Heater Control
Feedback
Feedforward
Colour Setpoint (fuzzy)
Roasting Process
Fig. 4. Indirect measurement of the residence time to control the peanut roasting upon [22].
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
Color measurement by the operator
1149
Classes of color Diagnostic and action – Fuzzy symbolic approach
Thickness measurement by a sensor
Fuzzy classification
Moisture content measurement by a sensor
Fuzzy classification
Temperature set points for the continuous oven
INDUSTRIAL TUNNEL OVEN Biscuits Fig. 5. Estimation of the state of the biscuits using both the operators and the sensors measurements in [68].
evaluation, and Da-Ven-Sun and Brosnan [17] for pizza quality evaluation. In certain examples, fuzzy rule bases are sometimes combined with other tools, such as neural networks. For example, Yea et al. [91] uses this configuration to establish discrimination of gas odours. Nevertheless, few papers deal with the problem of fusing data of different natures (numeric and linguistic), although this is a key problem in the food industry. In these papers, the ability of fuzzy logic theory to cope with such problems seems highly significant and interesting for several reasons: (i) It is a simple way of unifying different scales of measurement (linguistic and numeric scales); (ii) The mathematical formalism is easy to understand by the experts; (iii) It allows capturing different types of imprecision from sensors and human perceptions. For example, Perrot et al. [66] proposes estimating the state of a biscuit in an industrial tunnel oven by using human and sensor information (Fig. 5). Biscuit colour is evaluated by the expert, whereas information on moisture and thickness comes from sensors.
4. Diagnosis, supervision, and control of food quality This subject represents a fairly large number of published papers in the field “fuzzy and food” (51% of the total number). Most of them are classical applications of the Takagi–Sugeno controller, such as Linko et al. [47] for extrusion cooking, Zhang and Litchfield [94] for drying, Norback [60] for cheese-making, Alvarez et al. [4] for controlling isomerised hop pellet production, Honda et al. [32] for controlling the sake brewing process, and O’connor et al. [61] for controlling the brewing process. Two specific approaches are used by the authors to develop the diagnosis, control, and supervision modules: (i) data-driven approaches and (ii) “expert knowledge”-driven approaches. The first case includes 11 out of 40 papers written on this topic. We have chosen to include the fuzzy PID implementations in this category. Identification of the system is done automatically using a data basis and an optimisation algorithm. For example, Honda et al. [32] developed a fuzzy neural network to control the temperature of the Ginjo sake mashing process, O’connor et al. [61] developed a fuzzy PID to control the brewing process, Guillaume et al. [29] optimised a fuzzy rule basis using a genetic algorithm to establish a decision-aid system for the cheese-making process. In this case, the risk can be to develop models composed with a large number of rules (from 50 to 100 rules or more) and as a consequence a large number of parameters to identify. For example in Honda et al. [32], 283 data points are used to identify the model. This can lead to difficulties in handling such quantities of data to set all the parameters and is a major limitation for applications used in food processes. Moreover, the tools built can only be used to interpolate
1150
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
and not extrapolate new cases. Nevertheless, the added value brought about by using fuzzy logic in this case relies on (i) a friendly operator interface that handles data in a space close to the symbolic space handled by the operators; (ii) the possibility of automatically adapting the module to new processing conditions (such as a new formula for a biscuit) if a model or a set with enough data to characterize the process are available. In the “expert knowledge”-driven approaches, fuzzy logic is used as a way of providing a mathematical formalization of expert knowledge and embedding it in decision-aid help algorithms and controllers. For example, Davidson et al. [21] develops a fuzzy control system for continuous crossflow peanut roasting, Perrot et al. [66] proposes a fuzzy logic approach to control the quality of the biscuits in an industrial tunnel oven, Voos et al. [89] develops a fuzzy control of a drying process in the sugar industry based on operator experience, and Curt et al. [16] develops five Takagi–Sugeno modules to control the quality of the sausage during ripening. It is also used to perform supervisory tasks such as Acosta-Lazo et al. [1] for the supervision of a sugar factory. In all, 19 out of 40 papers have been written on this topic. In this case, modules are used directly without adaptation using a database while the data sets are only used to validate the approaches. The rules bases are generally compact with a limited number of rules (around 20 rules on average). For example, 11 rules are implemented to control the quality of the biscuits in Perrot et al. [66]. According to the authors Davidson [18] or O’connor et al. [61], fuzzy logic formalism is particularly well adapted for capturing how operators think and for capitalising human knowledge. Moreover, the robustness of certain specific models are underlined in some studies (for example Ioannou et al. [37]). Nevertheless, these properties depend on each specific model built and no theoretical demonstration on the robustness generally provided by fuzzy formalism is available. Adaptation of such models is generally achieved by hand although in some studies (seven papers out of 40), an initial structure of the fuzzy model is defined with the experts and an optimisation tool is used to optimise it or identify the fuzzy model’s parameters as in Perrot et al. [64] on a filtration process or Davidson et al. [21] on a peanut roasting process. The difficulty in this case is to keep the fundamental knowledge brought by the experts. Last but not least, we have chosen to highlight the papers focusing on the semantic scope of fuzzy logic and the ability of such methodology to capture expert reasoning and knowledge, and its uncertainty. Thus, Davidson et al. [21] used a fuzzy arithmetic that estimates peanut eating time and browning to control peanut roasting, Perrot et al. [63] developed a decision help system to control the cheese ripening process, integrating the uncertainty of human measurements. Petermeier et al. [68] used a hybrid approach to develop a model of the fouling behaviour of an arbitrary heat treatment device for milk. This is developed by combining deterministic differential equations with cognitive elements for the unknown parts of the knowledge model. These authors emphasize the relevance of this open field of research in the context of food processes and the interest of fuzzy symbolic representation of expert reasoning. Nevertheless, they call into question the optimality of the approaches developed on the basis of imperfect and incomplete expert knowledge. 5. Conclusion As we have seen, fuzzy logic is used in food applications to (i) capture and formalise the descriptive sensory evaluation performed by a quality team, an operator, or a consumer, (ii) develop an indirect measurement of the properties of a food product, and (iii) control food processes. Fig. 6 presents a classification of the different papers written in this area on the different research topics. If we focus on the type of approaches developed (Fig. 7), on the one hand, 33 papers deal exclusively with data-driven approaches including fuzzy PID. A total of 66% of these papers are above all dedicated to indirect measurement tasks, while 34% are dedicated to the control and modelling of processes. This category represents only a small percentage out of the total, which can be explained by a key difficulty in the food industry, especially in traditional manufacturing plants: that of constituting databases that can be easily used for control purposes. On the other hand, “expert knowledge”-driven approaches are dealt with in 35 papers. Mixed approaches are encountered in seven papers. If we focus on the type of fuzzy concepts applied (Fig. 8), 74% of the applications dealt with stem from the fuzzy set theory and most of them implement classical fuzzy logical functions (Mamdani type). On the contrary, the theory of possibility is hardly used, being dedicated to a restricted task in only three of the total of 78 papers [39,37,23]. Nevertheless, the authors underline a really interesting and open field of research. For example, in De Silva et al. [23], a firmness sensor for an automated herring roe grader is developed using fuzzy concepts. The theory of possibility is used in this case to estimate the fuzzy membership functions of the fuzzy decision-making system. The results show that this approach is more efficient than the use of classical trapezoidal membership functions.
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
1151
Food and fuzzy 1993-2004 78 papers 11 papers
40 papers
27 papers
Sensory analysis
Indirect measurement
13 papers
Supervision Decision help system Control
Process knowledge improvement
14 papers
Fuzzy rule basis
4 papers
Classification
Knowledge driven approaches
Data driven approaches 8 papers
5 papers
Data driven approaches
Knowledge driven approaches
Mixed or unknown approach
11 papers
15 papers
10 papers
Fig. 6. Classification of the papers written in the topic “fuzzy logic and the quality control of the food products”.
Mixed approaches 8%
Data driven approaches 42 %
Knowledge driven approaches 45 % Fig. 7. Proportion of the three different types of identification of the fuzzy models in the whole papers.
fuzzy adaptation of classification tools 18% Fuzzy-neural tools 8%
53% 21% Takagi-Sugeno
Mamdani or logical operators associated MaxMin
Fig. 8. Proportion of the different fuzzy concepts manipulated in the whole papers.
1152
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
List of the applications on the four topics Representation of the descriptive sensory evaluation [20,22,28,31,39,57,62,82–84,86] Indirect measurement Fuzzy rule basis Data-driven approaches: [12,13,17,23,49,76,79,91] Knowledge driven approaches: [11,36,55,59,93] Classification [2,5–7,14,33,41,46,50,52,67,69,73,75] Diagnosis, supervision or control Data-driven approaches [4,25,29,32,43–45,70,72,88,90] Knowledge-driven approaches [1,8,16,26,35,37,38,42,47,53,54,65,66,71,94] Mixed or unknown approaches Mixed: [10,21,60,61,64,80,89] Unknown: [3,40,74] Process knowledge improvement [21,30,63,68]
References [1] G.G. Acosta-Lazo, C.J. Alonso-Gonzales, B. Pulido-Junquera, Knowledge based diagnosis of a sugar process with teknolid, International Sugar J. 103 (2001) 44–51. [2] G. Adorni, D. Bianchi, S. Cagnoni, Ham quality control by means of fuzzy decision trees: a case study, IEEE International Conference on Fuzzy Systems, Proc. IEEE World Congress on Computational Intelligence, vol. 2, 1998, pp. 1583–1588. [3] L. Aldini, Utilization of fuzzy logic in diffusers, Ind. Sacc. Italiana 91 (1998) 83–89. [4] E. Alvarez, M.A. Cancela, J.M. Correa, J.M. Navaza, C. Riverol, Fuzzy logic control for the isomerized hop pellets production, J. Food Engrg. 39 (1999) 145–150. [5] C. Amza, Bones detection from chicken breast meat using a competitive Hopfield neural network and fuzzy reasoning, in: Scientific Bulletin Series C: Electrical Engineering, University Politehnica of Bucharest, 2001, pp. 53–64. [6] G. Andreotti, R. Lamanna, E. Trivellone, A. Motta, 13C NMR spectra: an easy way to distinguish milks from different animal species, J. Amer. Oil Chemist’s Soc. 79 (2002). [7] L. Ballerini, A. Hoegberg, K. Lundstroem, G. Borgefors, Color image analysis technique for measuring of fat in meat: an application for the meat industry, Proc. SPIE—Internat. Soc. Optical Engrg. 4301 (2001) 113–124. [8] R.B. Brown, T.M. Rothwell, V.J. Davidson, A fuzzy controller for infrared roasting of cereal grain, Canadian Biosyst. Engrg. 43 (2001) 3.9–3.15. [9] P. Buche, C. Dervin, A. Brouillaud-Delattre, N. Gnanou-Besse, Combining fuzzy querying of imprecise data and predictive microbiology using category-based reasoning for prediction of the possible microbial spoilage in foods: application to Listeria Monocytogenes, Internat. J. Food Microbiol. 73 (2002) 171–185. [10] D.A. Campbell, M.A. Pecar, M.J. Lees, Intelligently controlled beer filtration, Joint Conf. on Intelligent Systems, vol. 1, 1998, pp. 313–316. [11] O. Castillo, P. Melin, Intelligent quality control for manufacturing in the food industry using a new fuzzy-fractal approach, Proc. 2nd Internat. Conf. on Intelligent Processing and Manufacturing of Materials, vol. 1, 1999, pp. 151–156. [12] K. Chao, Y.R. Chen, H. Early, B. Park, Color image classification systems for poultry viscera inspection, Appl. Eng. Agric. 15 (1999) 363–369. [13] W. Chong-Yaw, R. Paramesran, F. Takeda, T. Tsuzuki, H. Kadota, S. Shimanouchi, Classification of rice grains using fuzzy artmap neural network, Proc. APCCAS. Asia-Pacific Conf. Circuits Syst. 2 (2002) 223–226. [14] Y. Chtioui, S. Panigrahi, L.F. Backer, Self-organizing map combined with a fuzzy clustering for color image segmentation of edible beans, Trans. ASAE 46 (2003) 831–838. [15] D. Corney, Food bytes: intelligent systems in the food industry, British Food J. 104 (2002) 787–805. [16] C. Curt, J. Hossenlopp, N. Perrot, G. Trystram, Dry sausage ripening control. Integration of sensory related properties, Food Control 13 (2002) 151–159. [17] Da-Ven-Sun, T. Brosnan, Pizza quality evaluation using computer vision—Part 1: pizza base and sauce spread, J. Food Engrg. 57 (2003) 81–89. [18] V. Davidson, Fuzzy control of food processes, in: G. Mittal (Ed.), Computerized Control Systems in the Food Industry, New York, Basel, Hong Kong, 1996, pp. 179–205.
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
1153
[19] V. Davidson, J. Ryks, Comparison of Monte Carlo and fuzzy math simulation methods for quantitative risk assessment, J. Food Protection 10 (2003) 1900–1910. [20] V. Davidson, J. Ryks, T. Chu, Fuzzy models to predict consumer ratings for biscuits based on digital image features, IEEE Trans. Fuzzy Systems 9 (2001). [21] V.J. Davidson, R.B. Brown, J.J. Landman, Fuzzy control system for peanut roasting, J. Food Engrg. 41 (1999) 141–146. [22] V.J. Davidson, W. Sun, A linguistic method for sensory assessment, J. Sensory Stud. (1998) 315–330. [23] C.W. De Silva, L.B. Gamage, R.G. Gosine, An intelligent firmness sensor for an automated herring roe grader, Intell. Automat. Soft Comput. 1 (1996) 99–114. [24] D. Dubois, L. Foulloy, S. Galichet, H. Prade, Performing approximate reasoning with words, in: J. Kacprzyk, L. Zadeh (Eds.), Computing with Words in Information Intelligent Systems 1, Physica-Verlag, 1999, pp. 24–29. [25] M. Estaben, M. Polit, J.P. Steyer, Fuzzy control for an anaerobic digester, Control Engrg. Practice 5 (1997) 1303–1310. [26] E.R. Fowler, New techniques for commercial bread dough mixing, IEEE Instrum. Measurement Magazine 3 (2000) 21–25. [27] F. Goyache, A. Bahamonde, J. Alonso, S. Lopez, J.J. Del Coz, J.R. Quevedo, J. Ranilla, O. Luaces, I. Alvarez, L.J. Royo, J. Diez, The usefulness of artificial intelligence techniques to assess subjective quality of products in the food industry, Food Sci. Tech. 12 (2001) 370–381. [28] A.S. Guillard, F. Blon, J.L. Vendeuvre, Influence de la matière première et du procédé, VPC 20 (1999). [29] S. Guillaume, B. Charnomordic, Knowledge discovery for control purposes in food industry databases, Fuzzy Sets and Systems 122 (2001) 487–497. [30] T. Hanai, N. Ueda, H. Honda, H. Tohyama, T. Kobayashi, Deciding the temperature course during sake mashing using a GA-FNN for quality control of sake, Seibutsu Kogaku Kaishi 76 (1998) 331–337. [31] J. Harris, Raw milk grading using fuzzy logic, Soc. Dairy Tech. 51 (1998) 52–56. [32] H. Honda, T. Hanai, A. Katayama, T.H.T. Kobayashi, Temperature control of Ginjo sake mashing process by automatic fuzzy modeling using fuzzy neural networks, J. Fermentation Bioengrg. 85 (1998) 107–112. [33] B.-G. Hu, R.G. Gosine, L.X. Cao, C.W. de Silva, Application of a fuzzy classification technique in computer grading of fish products, IEEE Trans. Fuzzy Systems 6 (1998) 144–152. [34] S.V. Iiyukhin, T.A. Haley, R.K. Singh, A survey of automation practices in the food industry, Food Control 12 (2001) 285–296. [35] I. Ioannou, N. Perrot, C. Curt, G. Mauris, G. Trystram, Development of a control system using the fuzzy set theory applied to a browning process—A fuzzy symbolic approach for the measurement of product browning: development of a diagnosis model—Part I, J. Food Engrg. 64 (2004) 497–506. [36] I. Ioannou, N. Perrot, J. Hossenlopp, G. Mauris, G. Trystram, The fuzzy set theory: A helpful tool for the estimation of sensory properties of crusting sausage appearance by a single expert, Food Quality Preference 13 (2002) 589–595. [37] I. Ioannou, N. Perrot, G. Mauris, G. Trystram, Experimental analysis of sensory measurement imperfections impact for a cheese ripening fuzzy model, IFSA Fuzzy logic, soft computing and computational intelligence theories and applications congress, Turquie (2003) 595–602. [38] I. Ioannou, N. Perrot, G. Mauris, G. Trystram, Development of a control system using the fuzzy set theory applied to a browning process—Towards a control system of the browning process combining a diagnosis model and a decision model—Part II, J Food Engrg. 64 (2004) 507–514. [39] S. Jaya, H. Das, Sensory evaluation of mango drinks using fuzzy logic, J. Sensory Stud. 18 (2003) 163–176. [40] B. Kett, Optimising recipes for price and performance, Innovations Food Tech. 18 (2003) 59–61. [41] L. Khodja, L. Foulloy, E. Benoit, T. Talou, Fuzzy techniques for coffee flavour classification, Inform. Process. Management Uncertainty Knowledge-Based Systems 1 (1996) 709–719. [42] S. Kim, S.I. Cho, Neural network modeling and fuzzy control simulation for bread-baking process, Trans. ASAE 40 (1997) 671–676. [43] A.B. Koc, P.H. Heinemann, G.R. Ziegler, W.B. Roush, Fuzzy logic control of whole milk powder processing, Trans. ASAE 45 (2002) 153–163. [44] T. Kurz, M. Fellner, T. Becker, A. Delgado, Observation and control of the beer fermentation using cognitive methods, J. Inst. Brewing 107 (2001) 241–252. [45] Lahtinen, Identification of fuzzy controller for use with a falling film evaporator, Food Control 12 (2001) 175–180. [46] M. Leunissen, V. Davidson, Y. Kakuda, Analysis of volatile flavour components roasted peanuts using supercritical fluid extraction and gaz chromatography mass spectrometry, J. Agric. Food Chem. 44 (1996) 2694–2699. [47] P. Linko, K. Uemura, Y. Zhu, T. Eerikainen, Application of neural modeling in fuzzy extrusion control, Trans. I. Chem. E 70 (1992) 131–137. [48] S. Linko, Expert systems—What can they do for the food industry, Trends Food Sci. Tech. 9 (1998) 3–12. [49] E. Llobet, E.L. Hines, J.W. Gardner, S. Franco, Non-destructive banana ripeness determination using a neural network-based electronic nose, Measurement Sci. Tech. 10 (1999) 538–548. [50] A. Loufti, P. Wide, Symbolic estimation of food odors using fuzzy techniques, IPMU Information Processing and management of uncertainty in knowledge based systems, Annecy (France) 2 (2002) 919–925. [51] E. Mamdani, Applications of fuzzy algorithms for control of simple dynamic plant, IEE 121 (12) (1974) 1585–1588. [52] F. Marcelloni, Recognition of olfactory signals based on supervised fuzzy C-means and k-NN algorithms, Pattern Recognition Lett. 22 (2001) 1007–1019. [53] G. Martinez, A. Lopez, A. Esnoz, P. Virseda, J. Ibarrola, A new fuzzy control system for white wine fermentation, Food Control 10 (1999) 175–180. [54] G. Mauris, N. Perrot, L. Berrah, S. Hamlaoui, Determining a confidence index of cheese ripening prediction by fuzzy trend sensory indicators, IEEE Instrum. Measurement Tech. Conf. 2 (2003) 1005–1008. [55] G. Mauris, N. Perrot, P. Lambert, J. Philippe, Fuzzy techniques evaluate sausage quality, IEEE Instrum. Measurement 3 (2000) 14–17. [56] M.J. McGrath, J.F. O’Connor, S. Cummins, Implementing a process control strategy for the food processing industry, J. Food Engrg. 35 (1998) 313–321. [57] T. Naes, E. Kubberod, H. Sivertsen, Identifying and interpreting market segments using conjoint analysis, Food Quality Preference 12 (2001) 133–143.
1154
N. Perrot et al. / Fuzzy Sets and Systems 157 (2006) 1145 – 1154
[58] C. Nicolas, P. Carel, J. Hossenlopp, G. Trystram, D.N. Rutledge, C. Emonet, Correlation between sensory data, instrumental data (gas sensors, physico-chemical analysis) and palatability measurements of twelve moist foods for cats, Sci. Aliments 19 (1999) 135–141. [59] H.M. Nielsen, W. Paul, Modelling image processing parameters and consumer aspects for tomato quality grading, IFAC Mathematical and Control Appl. Agric. Horticulture (1997) 141–146. [60] J.P. Norback, Natural language computer control of crucial steps in cheese making, Artificial Neural Networks Dairy Industry 49 (2) (1994) 119–122. [61] B. O’Connor, C. Riverol, P. Kelleher, N. Plant, R. Bevan, E. Hinchy, J. D’Arcy, Integration of fuzzy logic based control procedures in brewing, Food Control 13 (2002) 23–31. [62] M.N. Omri, I. Urdapilleta, J. Barthelemy, B. Bouchon-Meunier, C.A. Tijus, Semantic scales and fuzzy processing for sensorial evaluation studies, IPMU Information Processing and Management of Uncertainty in Knowledge-Based Systems, Spain II (1996) 715–719. [63] N. Perrot, L. Agioux, I. Ioannou, G. Mauris, G. Corrieu, G. Trystram, Decision support system design using the operator skill to control cheese ripening—Application of the fuzzy symbolic approach, J. Food Engrg. 64 (2004) 321–333. [64] N. Perrot, L. Mé, G. Trystram, J. Trichard, M. Decloux, Optimal control of the microfiltration of sugar product using a controller combining fuzzy and genetic approaches, Fuzzy Sets and Systems 94 (1998) 309–322. [65] N. Perrot, G. Trystram, F. Guely, Food product quality feed-back control and fuzzy sets—Application to an industrial baking process, in: Second European Congress of Chemical Engineering, 7ème congrès Français de Génie des Procédés-ECCE2, Montpellier, France, 1999. [66] N. Perrot, G. Trystram, F. Guely, F. Chevrie, N. Schoesetters, E. Dugre, Feed-back quality control in the baking industry using fuzzy sets, J. Food Process Engrg. 23 (2000) 249–279. [67] N. Perrot, G. Trystram, D. Le Gennec, F. Guely, Sensor fusion for real time quality evaluation of biscuit during baking. Comparison between Bayesian and Fuzzy approaches, J. Food Engrg. 29 (1996) 301–315. [68] H. Petermeier, R. Benning, A. Delgado, U. Kulozik, J. Hinrichs, T. Becker, Hybrid model of the fouling process in tubular heat exchangers for the dairy industry, J. Food Engrg. 55 (2002) 9–17. [69] C.G. Raptis, C.I. Siettos, C.T. Kiranoudis, G.V. Bafas, Classification of aged wine distillates using fuzzy and neural network systems, J. Food Engrg. (2000) 267–275. [70] Riza, D. Priyasta, F.R. Triputra, I. Ismail, PLC-based fuzzy controller for sterilizing process in crude palm oil mill, intelligent control for agricultural applications 2001, Proceedings volume from the 2nd IFAC/CIGR Workshop, 2002, pp. 247–252. [71] J. Ruger, B. Langhans, S. Alender, Fuzzy control in pulp drying, Zuckerindustrie 120 (1995) 387–398. [72] R. Rywotycki, Food frying process control system, J. Food Engrg. 4 (2003) 339–342. [73] C. Sarbu, H.W. Zwanziger, Fuzzy classification and comparison of some Romanian and German mineral waters, Analytical Lett. 34 (2001) 1541–1552. [74] Seung-Ju-Lee, Chang-Gi-Hong, Tack-Su-Han, Jun-Young-Kang, Young-An-Kwon, Application of fuzzy control to start-up of twin screw extruder, Food Control 13 (2002) 301–306. [75] M.A. Shahin, E.W. Tollner, R.W. Mcclendon, Artificial intelligence classifiers for sorting apples based on watercore, J. Agric. Engrg. Res. 73 (2001) 265–274. [76] M.A. Shahin, B.P. Verma, E.W. Tollner, Fuzzy logic model for predicting peanut maturity, Trans. ASAE 43 (2000) 483–490. [77] R. Singh, F. Ou-Yang, Neuro-fuzzy technology for computerized automation, in: G. Mittal (Ed.), Computerized Control Systems in the Food Industry, New York, Basel, Hong Kong, 1996, pp. 119–178. [78] W. Sun, V. Davidson, Dynamic fuzzy-reasoning-based function estimator, Fuzzy Sets and Systems 79 (1996) 357–366. [79] T. Sundic, S. Marco, J. Samitier, P. Wide, Electronic tongue and electronic nose data fusion in classification with neural networks and fuzzy logic based models, Proceedings of the 17th IEEE Instrum. Measurement Tech. Conf. 3 (2000) 1474–1479. [80] M. Syfert, J. Nowak, Diagnostics of evaporator in a sugar factory based on fuzzy models, fault detection, supervision and safety for technical processes, Proceedings volume from the 4th IFAC Symp., vol. 1, 2001, pp. 337–342. [81] T. Takagi, M. Sugeno, Fuzzy identification of systems and its application to modelling and control, IEEE Trans. Systems Man Cybernetics 15 (1985) 116–132. [82] J. Tan, X. Gao, D.E. Gerrard, Application of fuzzy sets and neural networks in sensory analysis, J. Sensory Stud. 14 (1999) 119–138. [83] F. Tian-Jin, D. Xiang-Qian, T. Rui, F. Zuen-Cheng, A hybrid expert system for formulated product designs, Proceedings of the IASTED International Conference, Artificial Intell. Appl. (2001) 336–341. [84] O. Tominaga, F. Ito, T. Hanai, H. Honda, T. Kobayashi, Modeling of consumer’s preferences for regular coffee samples and its application to product design, Food Sci. Tech. Res. 8 (2002) 281–285. [85] G. Trystram, F. Courtois, Food processing control: reality and problem, Food Res. Internat. 27 (1994) 173–185. [86] I.B. Turksen, I.A. Willson, A fuzzy set model for market share and preference prediction, European J. Oper. Res. 82 (1995) 39–52. [87] L. Valet, G. Mauris, P. Bolon, A Statistical Overview of Recent Literature in Information Fusion, ISIF, Paris, France, July 2000. [88] G.K. Venayagamoorthy, D. Naidoo, P. Govender, An industrial food processing plant automation using a hybrid of PI and fuzzy logic control, IEEE Internat. Conf. Fuzzy Systems 2 (2003) 1059–1062. [89] H. Voos, L. Litz, H. Konig, Fuzzy control of a drying process in sugar industry, 6th European Congress Intell. Tech. Soft Comput. EUFIT ’98, vol. 3, 1998, pp. 1476–1480. [90] G. Xie, R. Xiong, I. Church, Comparison of kinetics, neural network and fuzzy logic in modelling texture changes of dry peas in long time cooking, Lebensmittel Wissenschaft Technologie 31 (1998) 639–647. [91] B.Yea, R. Konishi, T. Osaki, K. Sugahara, The discrimination of many kinds of odor species using fuzzy reasoning and neural networks, Sensors Actuators 45 (1994) 159–165. [92] L. Zadeh, Fuzzy sets, Inform. Control 8 (1965) 338–353. [93] Q. Zhang, J. Litchfield, Applying fuzzy mathematics to product development and comparison, Food Tech. 45 (1991) 108–115. [94] Q. Zhang, J. Litchfield, Fuzzy logic control for a continuous crossflow grain dryer, J. Food Process Engrg. 16 (1993) 59–77.