- Email: [email protected]
Identification of waste packaging profiles using fuzzy logic
Resources, Conservation and Recycling 52 (2008) 1022–1030
Contents lists available at ScienceDirect
Resources, Conservation and Recycling journal homepage: www.elsevier.com/locate/resconrec
Identification of waste packaging profiles using fuzzy logic ´ Castro-Rodr´ıguez a , Gabriela Lozano-Olvera a,∗ , Sara Ojeda-Ben´ıtez b , Juan Ramon Miguel Bravo-Zanoguera b , Antonio Rodr´ıguez-Diaz c a PhD Program in Sciences and Engineering, Engineering Institute UABC, Boulevard Benito Ju´ arez y Calle de la Normal S/N, Col. Insurgentes Este, C.P. 21280, Mexicali Baja California, Mexico b Researcher of the Engineering Institute UABC, Boulevard Benito Ju´ arez y Calle de la Normal S/N, Col. Insurgentes Este, C.P. 21280, Mexicali Baja California, Mexico c Researcher of the Faculty of Chemical Sciences and Engineering – UABC, Calzada Universitaria S/N Delegaci´ on Mesa de Otay, Tijuana B.C. M´exico, C.P. 22390, Mexico
a r t i c l e
i n f o
Article history: Received 8 December 2007 Received in revised form 20 March 2008 Accepted 26 March 2008 Available online 21 May 2008 Keywords: Profiles of generation Composition of waste packaging Fuzzy logic
a b s t r a c t An important factor in the environmental crisis faced by modern civilization is the waste produced by a great diversity of commercially available basic household products. An important step in measuring the impact of these organic and inorganic wastes in a community is to characterize the products consumed by families of different socioeconomic levels. In this work we present a method to identify consumption and waste packaging profiles based on household demographic data and type-specific consumption products. Our approach is based on fuzzy logic techniques to deal with quantitative and qualitative information for assisting in the decision making process when dealing with solid waste. This approach was implemented to process the information from a database containing domestic waste produced by families of different socioeconomic levels in a Mexican city. The data were taken from products consumed in 123 dwellings. Different types of packaging materials were classified as metals, cardboard, plastics, and glass. Household demographic variables were included such as level of schooling, number of inhabitants per dwelling, family income, and amount of waste generated. The results achieved show that when correctly combining the variables of a family, a more accurate approximation to the real packaging waste generation profiles may be obtained. The techniques applied in this research were successfully adapted to face problems that until now were difficult to treat, allowing to process information in which inexact values or subjective terms are handled. © 2008 Elsevier B.V. All rights reserved.
1. Introduction The 20th century society requires a great deal of satisfaction provided by means of industrialization and marketing processes. These processes represent the exploitation of natural resources and the generation of large amounts of waste with different ´ compositions. Grodzinska-Jurczak et al. (2004) point in that this can be attributed to a boom in product packaging (mainly plastic). Therefore, the consumption practice of our society in this century causes problems of contamination by the amount of waste that it generates without recovering the value which they have as residue. On this matter Tsiliyannis (2005) indicates that the packaging waste constitutes a significant component of the municipal residues based on its volume. Thus, the generation of packaging waste is related to consumerism, closely related to the system of basic needs in which
∗ Corresponding author. Tel.: +52 686 5664150. E-mail address: [email protected] (G. Lozano-Olvera). 0921-3449/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.resconrec.2008.03.008
modern society exerts influence to drag the consumer into the disposable culture (Ojeda, 2005). The wastes produced by human activity, lose their value when discarded. McDougall et al. (2004) indicate that the lack of value is related to the often unknown mixed composition of the waste. Therefore, waste segregation generally increases its value when there are possible uses for the recovered materials. At present, consumption is one of the largest distractions of society because each family tries to satisfy its needs in the most practical and possible way (Ojeda, 2006). Likewise, waste generation continues to increase because the market economy offers a larger quantity of disposable products, which in addition are offered in a wide variety of designs and of packaging, generating packaging wastes and additional packing such as bags, cans, cardboard, glass, plastic, wood, paper, aluminum and tetra-pack among others, generally thrown away in the waste. Calver (2004) indicates that the transformation packaging is due to the changing lifestyle of consumers. Packaging is essential in modern society as it allows an enormous variety of products to reach the consumer intact, in proper hygienic conditions and general with pertinent information regarding the brand, the product and its usage.
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Packaging waste represents around 17% of urban solid waste by weight and 3% of the total waste flow. In recent years, this percentage has undergone an important increase as a result of the changes produced in the habits of consumers and the types of commercial distribution, which seem to favor more and more the packaging of products and in smaller quantities (individual portions, product units, small packs, etc.). Packaging is a material meant to provide specific services, to contain, protect, transport and preserve a product. Therefore, its function will depend on the way the different needs for which they have been created will be satisfied. In this sense, Vidales (2003) indicates that since its first commercial appearance to the present time, packing has become an element very useful within the complex network that has been woven to satisfy many needs of human beings. For that reason it is important to analyze the generation of packaging wastes taking as a reference the characteristics of generators, which in this case is the family. To investigate the generation and composition of packaging that are commonly found in domestic waste, represents a challenge for the waste management research. As the composition of packaging materials, the function of the packaging and the volume generated, among others, helps to propose strategies for waste management practices. In this sense, it is important to find ways of analyzing the problem in order to offer alternative solutions. Therefore, in this paper, fuzzy logic is applied, being a tool which has been tested in expert systems (Jian-Da et al., 2007) and in other applications (Chang, 1997; ´ 1999; Yukun and Yufeng, 2006; Ami et al., Mohamed and Cˆote, 2008; Pei-Chann and Chen-Hao, 2008; Shih-Ming and Shyi-Ming, 2008). Fuzzy inference systems have also been applied in the field of solid waste. Chen and Ni-Bin (2000) developed a dynamic model by applying fuzzy theory to predict waste generation in an urban area taking a limited number of samples. Al-Jarrah and Abu-Qdais (2006) used an intelligent system based in the fuzzy inference system to approach the problem of selecting a site for a sanitary landfill. In another study, Nie et al. (2007) developed and applied a model for the planning of a solid waste management system under uncertainty, using Interval-Parameter Fuzzy-Robust programming (IFRP). In this study, fuzzy logic is applied because it happens to work with mixed data type (quantitative and qualitative) of two different database structures. Fuzzy logic works with rules to generate knowledge from the databases, and the rules are recognized as an effective and natural means to transmit knowledge between humans, to make and to justify decisions (Moreno et al., 2006). One of the problems which the experts in solid waste face, is to know the waste generation rates and the composition of the waste, for that reason it is important to work with tools that advance knowledge about the household generation behaviors extracting information from existing databases which have been developed. One of the types of wastes that are generated is packaging; therefore fuzzy logic is applied to investigate the profiles of generation by family, considering that it involves qualitative and quantitative variables. This tool is adapted to produce generation profiles of the amount and weight of packaging as a function of family characteristics.
The sampling work for this research was done in three stages. In the first stage, a survey was used in order to learn about the family’s basic characteristics, such as number of members, level of education, income, etc. The second stage consisted in the collection of samples during 8 consecutive days from the households that had accepted to participate; and, in the third stage, another instrument was applied, providing information related to the consumption habits characterizing each inhabitant by house, as well as the environmental knowledge that they have. The information obtained from the three stages studied was stored in two especially designed databases to allow us make a descriptive analysis of the consumption habits, family structure and the generation of waste, among others. Nevertheless, the databases have not been data mining in their totality; it was necessary to apply tools to obtain information of the analyzed databases. To advance in the data analysis validation of the information required a normalization processes applied to the original database in order to eliminate, capture errors or redundancy in the information, preventing wrong interpretation of the information. Once validation was achieved, basic statistical analyses (average, variances, correlation and factorial analysis) were performed to identify diverse panoramas of data behavior and to find existing dependent and independent variables, to obtain the proposed objective of finding profiles of generation of packaging by family. After identifying the dependent and independent variables (see Table 1), a new data base matrix was structured containing the information necessary to obtain the first profiles of the families. In order to generate the input data (x) as the demographic data of the house, defined by the number of inhabitants by house, level of education, family type, number of bags analyzed and income by house, and the output data (y), defined to know the generation of packaging waste by house by weight and by the number of packaging materials. The output variables are: the amount and weight of packaging discard by house and by type of material (plastic, cardboard, metal and glass) (Table 2). The data matrix presented in Table 2 groups the characteristics of the households, including data related to the generation of waste packaging per household in order to define rules that may reflect the quantity and weight of the waste packaging that the households with those attributes tend to discard. To do so, matrix data are inferred. This matrix consists of 123 tuples with 13 attributes Table 1 Definition for the matrix variables Variables
Definition
Inhabitants Level of education
The total members of a single house. Last scholastic degree of the family parents.
Income
Income reported by house. With this variable the socioeconomic levels are inferred, the criterion to determine it is based on the minimum wage (MW).a Low—income of 1–2 minimum wages. Medium—income of 3–5 minimum wages. High—superior to 5 minimum wages.
Family type
Nuclear—the typical family formed by father, mother and son. One-parent—only a father or mother and children. Extended—family who shares home with one or more relatives (grandparents, uncles or nephews).
Number of bags Quantity of waste packaging
The number of bags analyzed by house. Amount of articles by type of packing (plastic, cardboard, metal, glass).
2. Methodology In order to complete this study, the results of a research carried out by Ojeda (2007) were analyzed. This research was conducted in three neighborhoods of the city of Mexicali, which were selected for their socioeconomic stratum (high, medium and low). The research was done in these three neighborhoods, with the waste generated by the selected families during 8 consecutive days.
1023
a The minimum wage is the salary unit in Mexico to pay a worker, and is equivalent to 50.57 Mexican pesos.
1024
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Table 2 Matrix of input and output data Input variables (household demographic data) No of bags
X1 9 7 8 8 6 9 8 6 8 7 9 6 5 5 6 7 8 7 8 9 7 7 a
Family typology
Inhabitants
X2 1 3 3 1 3 1 3 3 3 1 3 1 2 1 3 3 3 3 3 3 3 3
X3 3 6 5 3 4 4 4 9 5 4 6 6 6 4 6 6 5 6 3 6 8 6
Output variables (generation of waste packaging per household in weight and quantity)
Level of education
Income (pesos/week)a
Plastic Weight
Quantity
Cardboard Weight
Quantity
Metal Weight
Quantity
Glass Weight
Quantity
X4 3 2 14 8 4 1.5 2 9 3 6 3 5.5 2 2.5 2 12 3 3.5 1 1.5 5.5 4
X5 3779.61 1691.44 1222.20 2706.30 916.65 611.10 400.00 916.65 2444.40 611.10 1130.14 1527.75 1931.51 611.10 1527.75 1416.65 916.65 1527.75 1833.30 1130.14 2444.40 1014.86
Y1 3779.61 1691.44 1222.20 2706.30 916.65 611.10 400.00 916.65 2444.40 611.10 1130.14 1527.75 1931.51 611.10 1527.75 1416.65 916.65 1527.75 1833.30 1130.14 2444.40 1014.86
Y2 45 32 34 18 31 96 27 61 40 28 21 57 59 11 13 44 35 34 47 66 33 36
Y3 1946.33 488.39 1263.22 836.03 1005.64 2964.13 612.92 1163.61 1115.47 540.05 498.13 815.06 1086.03 348.35 115.58 1158.48 797.4 853.6 923.96 1498.28 744.74 641.9
Y4 15 5 11 7 7 8 2 18 10 9 21 5 20 4 4 7 4 8 11 14 5 8
Y5 218.13 304.47 303.73 171.21 240.5 220.98 15 407.06 234.27 186.6 273.2 36.16 577.8 72.31 70.4 171.52 112.5 401.3 535.94 358.42 196.9 560.11
Y6 22 2 22 0 1 17 2 5 2 0 15 5 11 0 2 20 2 2 0 15 3 1
Y7 926.35 424.6 3131.93 0 14.2 2324.63 348.1 275.4 170 0 1540.9 638.87 1366.33 0 145 529.24 482.9 153.3 0 3237.66 535.01 14.9
Y8 4 1 3 0 0 8 2 2 0 0 7 4 9 0 1 1 2 1 0 13 2 0
11.50 Mexican pesos is equivalent to $1.
of which five describe household characteristics (input variables), the following eight variables describe the quantity and weight of the four categories of waste packaging analyzed in this paper. Once the information analyzed, the computing tools were applied; since they have the capacity of automatic and intelligent analysis of great volumes of data to find knowledge. The first step was to try and obtain knowledge from the generated databases; the proposal by Moreno et al. (2006) was used, and a search process was carried out including all data available. In this process, the grouping technique (C-means, C-fuzzy-means, subtractive) was used in order to find the most significant and relevant patterns and a system of fuzzy logic for the extraction of production rules (IF–THEN). The grouping technique was used to obtain the background of the rules. From the results of these groupings, inferences are established, using the Takagi-Sugeno-Kang (TSK) method, which are fuzzy expressions, and the outputs are deterministic functions (Pasinno and Yurkovich, 1997). This method has been successfully applied in a large number of practical problems. Its main advantage is that it shows a compact/constraint equation of the system thanks to which it is possible to calculate the parameters (Pi ) with classical methods, making its design easier. The database extraction process is based on the following steps: 1. Input data was grouped by means of grouping techniques (Ca means, C-fuzzy-means, subtractive) to calculate the centroid ci,j a and the standard deviation i,j given by Eq. (1) which will be the parameters of the membership functions ground given by Eq. (2). a i,j =
r × (max(X) − min(X)) a √ 8
ij (xj )
of the back-
(1)
where X is a matrix. In this case, it is the matrix shown in Table 2, representing the system’s input and output variables; i are rules from the membership function and j the variables of matrix X necessary to evaluate the rule; a is a parameter defining the
association rules, and ra , a positive constant, is the rule a output representing a neighborhood radius. 2. The following step was to calculate the consequents by means of the TSK fuzzy reasoning method. This model consists in analyzing the rules, each rule has a numerical output, the total output is obtained by means of “weighted average”, actually the operator of weighted average sometimes is replaced by the operator of “weighted Sum” to reduce the times of analysis, especially in the training of a system of diffuse inference.
∧
ij xj
=e
2
−(1/2)((|| xp,j −c a ||/( a ) )) i,j
ij
(2)
3. The rules based on the parameters of the membership functions of antecedents were extracted, ij (xj ), as well as those of consequents, fi (x), from the rules of the TSK fuzzy inference system shown in Eq. (3). Ri = IF x1 is i1 and . . . and xj is ij and . . . and xn is in THEN y is f1 = ai1 x1 + · · · + aij xj + · · · + ain xn + ai
(3)
4. The TSK fuzzy inference system was assessed with the input data ∧
(Xi ) shown in Table 2 to obtain the output (y) and calculate the mean squared error (Eq. (4)). RMSE =
MSE(e)
(4)
where RMSE is the comparison of the real input used by the system to carry out the assessment. MSE is the mean of squared errors. If RMSE was zero, real data would equal the assessed data. This indicates that the fuzzy comparison equals real data, then the error range tends to zero. The Fuzzy Logic Toolbox (FLT) was used to develop and apply the steps to the problem we are analyzing in this paper. This is a tool for the development of user-friendly fuzzy programs in the Matlab environment; the graphic called Fuzzy Interference System (FIS)
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
1025
In this paper, the socioeconomic stratum and the level of education are the fuzzy variables. The stratum is the economic level with which a family manages to satisfy its needs and reflects its consumption practices. It is an uncertain variable because it may take a range of values for each family as the data used are not an exact number. In connection with the level of education, it was assigned according to the level of education of the inhabitants, which could be complete or incomplete. For example, at the secondary school level of education, the inhabitants could have the secondary school level complete or incomplete; in this range the attribute level of education may lose defining features. 3. Results
Fig. 1. Function of FIS editor.
was used. It allows to build, edit and observe fuzzy interferences within an FLT. The operation of the FIS editor is shown in Fig. 1; it can be observed the sequence of data processing with the fuzzy logic tool, the matrix of data pre-processed is the input to the fuzzy system under a clustering process, which generates new information that allows to define the rules. For the problem that we are analyzing is important to generate rules, because these give a special vision about the knowledge and forms in which this can be represented. The usage of rules implies particular ways of understanding problems and of solving them. When applying rules to the problem of waste, it helps to assess the behavior or reasoning the system will have when assessing the input, qualitative and quantitative parameters, depending on the behavior of the different types of waste packaging and of the family generating them, so that an output consistent with the basic statistical data obtained may be achieved. After obtaining the rules, the fuzzy logic method was used for the assessment of the system based in rules; and, therefore, an output was obtained. This tool allows to assess variables considered inaccurate or uncertain, as it is the case analyzed in this paper, where the determination of generation profiles for a family implies to work with variables associated to the family’s behavior and to the particular characteristics of each of them, which become uncertain when combining them. In addition, fuzzy logic facilitates working with these types of variables since it makes possible the establishment of sets with restrictions or limits that stay effective in a given universe. It is for that reason that was applied to this work, since the family has a structure and composition difficult to define, these vary constantly, and it is reflected in the members of each family.
When applying fuzzy logic tools to restructure the collected information in the two databases, a matrix with input and output data was obtained. The presented results belong to the matrix in Table 2, which was created to analyze data from 123 households and the generation of packaging of each household per type of material. Data related to packaging generation from 850 bags collected for the sampling were analyzed. Of the total number of households analyzed, 26% belong to the low stratum, 48% to the medium and 25% to the high. 3.1. Composition of packaging The total weight of the analyzed packaging equaled 501.9 kg, which is equivalent to a total of 8062 packages discarded. In Table 3, packaging generation is shown by composition, where the type of material used in the packaging is shown by weight and quantity. It was observed that plastic packaging is the ones most commonly discarded, followed by cardboard and metal packaging. The results related to the generation of packaging by usage category are shown in Fig. 2. Among them, food, medicines and beverages can be mentioned. It can be observed that most generation is concentrated in the category of food, being the low stratum the one which generates the most. This behavior explains consumption practices per socioeconomic stratum. The households belonging to the low stratum consume products in small-sized packaging because their economic capacity is low and they discard a larger quantity of packaging in shorter periods, whereas, in the high stratum, consumers buy products in larger-sized containers or packaging, for example soft drinks of 3 l instead as opposed to the 600 ml bought by the ones in the low stratum. This implies that the high stratum generates lesser packaging, while the low one discards a larger quantity of packaging. 3.2. Family structure In order to analyze the family profiles, the type of family has been considered: nuclear, one-parent and extended. It was found that 49% of all households belong to nuclear families, 12% to one-parent families and 39% to extended families.
Table 3 Average quantity of packaging waste sampled by socioeconomic stratum Composition
Plastic Cardboard Metal Glass
Low (packaging (family/week))
Medium (packaging (family/week))
High (packaging (family/week))
Quantity
Weight (g)
Quantity
Weight (g)
Quantity
Weight (g)
38.9 9.2 7.9 3.4
1031.5 251.2 841.5 659.1
39.2 11.2 7.4 3.3
1560.6 550.8 770.7 687.8
288.8 382.5 442.7 495.1
542.3 4071.6 6077.6 10091.9
1026
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Fig. 2. Packaging generation by socioeconomic stratum by usage categories.
Another variable included was the level of education of the parents. This variable was obtained by means of the average education achieved by both parents, which was analyzed as a fuzzy group. An analysis was carried out of each of the levels of education achieved, as well as the levels reported as incomplete by the members of the family. A membership function was created for the analysis of the education variable in order to evaluate each of the fuzzy situations that may appear in the analyzed sampling. This function of membership is generated after obtaining the assessed clustering.The membership functions of the education variable is represented by a Gaussian function, allowing us to assess each of the levels of education completed, as well as the incomplete levels of education achieved (see Eq. (5)). 2
A (x) = e−k(x−m)
(5)
In the membership function, the value of X corresponds to all the levels of education assessed and m represents the mean values. Through this analysis, each of the situations presented in the sampling analyzed was assessed in detail. In this case, it was related
to all the combinations concerning the levels of education of the members of each family included in this research. Concerning the income variable of each household included in the analysis, this variable was associated to the socioeconomic stratum and the packaging generation, as well as the classification of the type of material of the packaging generated per household. Fig. 3 shows graphically the system proposed, based on the modeling of fuzzy rules and fuzzy control. 3.3. Profiles of packaging generation Once the variables “packaging generation” and “household general characteristics” were analyzed, we proceeded with the creation of a matrix (Table 2). In this table, five input variables were defined (X1 , X2 , X3 , X4 , X5 ), the output variables corresponding to Y1 , . . ., Yn , representing the data expected to be obtained when assessing the characteristics of each family, as well as the composition and quantification of the packaging generated. Using this matrix created from the two databases employed in this research, the four input variables were evaluated, as well as the packaging generation per type of material (see Fig. 4).
Fig. 3. Basic structure of the system based on fuzzy rules applying TSK to determine profiles of packaging generation.
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Fig. 4. Application of the system to determine profiles of packaging generation in the Matlab Fuzzy Logic Toolbox.
The input variables (X1 , X2 , X3 , X4 , X5 ) of the system are presented in Fig. 4. These inputs are processed by the TSK method, allowing the identification of the output. One of the disadvantages of this process is that it only manages one output; therefore, the process is repeated for each of the outputs that the user may require. In this case, eight outputs are required (Y1 , . . ., Y8 ). This repetitive process is not performed by the user, because there is a platform designed for the user, which allows him/her to execute it only once and to obtain the information required.
3.3.1. Profiles of packaging generation per type of material This section describes the results of the profiles of packaging generation, both per quantity–frequency of packaging, and by weight. The results in evaluating packaging generation by quantity are shown in Figs. 5 and 6; generation by weight is shown in Fig. 8.
1027
Fig. 5 shows the behavior of packaging generation in packaging made out of plastic, cardboard, glass and metal in households, taking as the variable the level of education of the heads of household. This diagram shows that plastic, metal and glass packaging are generated more frequently in households with a higher level of education, from medium to high level; while cardboard, according to the results obtained, is a packaging more widely used by inhabitants with a lower level of education. When adding another variable to the analysis, changes in the generation of packaging are observed. Fig. 6 shows the behavior of container generation including the input variable. When developing this assessment, a similarity regarding the behavior with plastic and cardboard is observed, as well as with glass and metal, though there is no such similarity concerning the quantity of packaging generated. This shows that when combining the variables of a family correctly, a more accurate approximation of packaging generation by family is obtained. The results of assessing the weight of plastic packaging and the income variable show that the behavior is different to the one observed in the analysis carried out with the frequencies of packaging. As in the assessment of weight, it is necessary to consider other factors or variables in the families, such as income, shopping frequency, presentations of the products, among others. When analyzing packaging generation in the higher stratum, it was generally found that families in this socioeconomic stratum do their shopping by acquiring products in large-sized presentations (2 l, 5 kl, etc.). Therefore, generation is higher, though the frequency with which they discard these packages is within longer periods of time. However, in the lower stratum, the opposite happens: products are consumed in packaging of smaller-sized presentations. Therefore, they generate a larger amount of these items, and lesser weight was found in this generation (Fig. 7). Fig. 8 shows the assessment results concerning the generation of packaging made out of plastic, cardboard, glass, metals with two variables of the family structure: number of dwellers and income.
Fig. 5. Assessment of packaging generation per level of education and type of material. (a) Plastic, (b) cardboard, (c) glass and (d) metals.
1028
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Fig. 6. Packaging generation by type of material vs. education and income. (a) Plastic, (b) cardboard, (c) glass and (d) metals.
In subsection (a) of Fig. 8, the results of plastic packaging generation are shown, taking as input variables the number of inhabitants per household and the income. The results show that the families with a higher income and fewer members generate larger amounts of plastic packaging, while families made up of four, five or six members and the same income, keep a more homogenous generation profile. In subsection (b) of Fig. 8, the diagram shows the generation of packaging made out of cardboard is higher in low-income families
with three to seven inhabitants per family. In subsections (c) and (d), the results of the analysis related to the generation of packaging made of glass and metal is shown. The diagrams show that, in terms of weight, the generation of these packaging materials is very similar. When evaluating these variables an acceptable RMSE was obtained in each of the different outputs assessed. Table 4 shows the values obtained with the assessment of the system, as well as the mean squared errors. When observing these values, one
Fig. 7. Generation average regarding the weight of packages.
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
1029
Fig. 8. Packaging generation by type of material vs. number of inhabitants and income. (a) Plastic, (b) cardboard, (c) glass and (d) metals.
Table 4 Assessment of the errors of the system Variable
Plastic
Cardboard
Glass
Metal
RMSE MSE
2.43 5.93
1.35 1.82
1.17 1.37
1.80 3.23
may notice that they are acceptable and very similar to the real ones. 4. Discussion and conclusion One of the main problems encountered when studying household solid waste is to obtain all household data and manage the components of the variables constantly varying, which makes it uncertain and complex. In this application, the variable family structure is very complex as time produces a constant variation; for example, the members of a family may increase their number when a new member is born, gets married or it may decrease if a member dies. The other variable included is income, which has a similar behavior, as the income may grow when the salary of one of the members is increased, or one of them begins to work. The same happens with the level of education, which varies with time. Fuzzy logic provides us with elements to generate predictions concerning changes in the family structure. One of the goals of the tools related to the search of knowledge in databases is to identify the patterns of behavior of data, and to infer knowledge and predict situations (Pei-Chann and Chen-Hao, 2008). In this research, the aim was to predict the profiles of pack-
aging generation per family, combining qualitative and quantitative variables. Throughout the processing of the information, it was possible to prove that the database contained a large number of redundancies among its variables, but also a lack of information in some of its fields. Under this situation, and using traditional statistical, the generated information has a high degree of uncertainty since data did not exist or it was scarce. But as indicated in Chen and Ni-Bin (2000) work, where with few simples they were able to forecast waste generation, fuzzy logic represents a powerful tool to efficiently handle these problems. That is why fuzzy logic can infer the information using available data, meanwhile analysis with conventional statistics does not include all available input when data set is not complete. With the tools applied in this research to these situations of uncertainty due to the lack of information or the combination of qualitative and quantitative variables, knowledge from the databases used was inferred. The generation and composition of household solid waste has a strong relationship with the patterns and levels of consumption of a given population. The results found in this paper show that there is a direct correlation between the composition of packaging generated from household waste flow and the features identifying a family. In this sense, Schiffman and Kanuk (2005) indicate that the family may be analyzed as a variable. Also, as the absence or presence of a new member produces an impact on the behavior of the family, another factor determining changes in the family structure is the age of its members, as according to their age basic and non-basic needs are modified over time. The profiles of packaging generation of the studied families reflect consumerism practices because, as indicated by Ojeda
1030
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
(2005), the system of basic needs is related to this consumerism practices. It is important to know the composition and quantity of packaging generated by a family, in order to give the discarded packaging the real value it has and to establish policies for the management and better usage of this waste in Mexican municipalities, because as it was pointed out by McDougall et al. (2004) the unawareness of the type of packaging generated reduces the value of waste. Fuzzy logic was applied, as the family dynamic is complex and it impacts on packaging generation, producing variations in weight or frequency and depending on how the family is structured. This is the reason why it is important to have a tool that may allow the assessment of any situation that may occur at the core of a family, in case there is any kind of change, both physical and intellectual. Or even that it may predict any change which might occur within the family structure in the future. The fact of applying fuzzy logic to determine the profiles of generation has allowed us to set up a system that simultaneously processes quantitative data, as packaging generation, and qualitative data, as the composition profile per structure and type of family, removing the inaccuracy obtained with the basic statistics. References Al-Jarrah O, Abu-Qdais H. Municipal solid waste landfill siting using intelligent system. Waste Management 2006;26(3):299–306. Ami HIL, Wen-Chin Ch, Ching-Jan Ch. A Fuzzy AHP and BSC approach for evaluating performance of IT department in the manufacturing industry in Taiwan. Expert Systems with Applications 2008;34:96–107, On line. Chang PT. Fuzzy seasonality forecasting. Fuzzy Sets Systems 1997;90:1–10. Chen HW, Ni-Bin Ch. Prediction analysis of solid waste generation based on grey fuzzy dynamic modeling. Resource, Conservation and Recycling 2000;20(1–2):1–18. ´ Calver G. What is packing? Mexico: Ediciones Pili; 2004.
´ Grodzinska-Jurczak M, Hanna Z, Read A. Management of packaging waste in Polan—development agenda and accession to the EU. Waste Management & Research 2004;22:212–23. Jian-Da W, Yu-Hsuan W, Mingsian RB. Development of an expert system for fault diagnosis in scooter engine platform using fuzzy-logic inference. Expert Systems with Applications 2007;33:1063–75. McDougall F, White M, Franke M, Hindle M. Integral management of solid residues: inventory of the life cycle. Venezuela: Procter & Gamble Industrial; 2004. Mohamed AMO, Cˆote´ K. Decision analysis of polluted sites—a fuzzy set approach. Waste Management 1999;19(7–8):519–33. Moreno JE, Castillo O, Castro JR. 2006 Data Mining for the extraction of if-then rules using fuzzy logic based on the FIS of Mamdani and Takagi-Sugeno Kang. In: IEEE-Computational Intelligence Society Mexico Chapter. Proceedings of international seminar on computational intelligence ISCI 2006; 2006. p. 62–8. Nie XH, Li YP, Liu L. IFRP: A hybrid interval-parameter fuzzy robust programming approach for waste management planning under uncertainty. Journal Environmental Management 2007;84(1):1–11. Ojeda BS. Flow of packages and packing present in the domiciliary solid remainders. An analysis of the life cycle. Unpublished report. UABC—Instituto de Ingenier´ıa; 2005. Ojeda BS. The consumption source for waste generation and contamination. In: ´ y Medio ambiente. Serie editorial CinMargarito QN, editor. Contaminacion ´ ´ ISBN cuentenario de la UABC; 2006, p. 227–49. Editorial Miguel Angel Porrua, 970-701-859-3. Ojeda BS. Dimensions of the consumer behavior and his relation with the generated solid residues: strategies for its recovery in a sustainable house. (2005–2007) Financed by CONACYT. Project in process; 2007. ˜ Pasinno K, Yurkovich S. Fuzzy control. 1st ed. Addison Wesley Publishing Company; 1997. Pei-Chann Ch, Chen-Hao L. A TSK type fuzzy rule based system for stock price prediction. Expert Systems with Applications 2008;34:135–44. ´ ´ 2005. Schiffman L, Kanuk L. Consumer behavior. Mexico: Press Pearson Educacion; Shih-Ming B, Shyi-Ming Ch. Evaluation students’ learning achievement using fuzzy membership functions and fuzzy rules. Expert Systems with Applications 2008;34:399–410. Tsiliyannis C. A new rate index for environmental monitoring of combined reuse/recycle packaging systems. Waste Management Research 2005;23:304–13. Vidales GMD. World of packing, a manual for the design and production of packages and packing. Mexico: Press G. Gili S.A. de CV; 2003. Yukun C, Yufeng L. An intelligent fuzzy based recommendation system for consumer electronic products. Expert Systems with Applications 2006;33(1):230–40.
Contents lists available at ScienceDirect
Resources, Conservation and Recycling journal homepage: www.elsevier.com/locate/resconrec
Identification of waste packaging profiles using fuzzy logic ´ Castro-Rodr´ıguez a , Gabriela Lozano-Olvera a,∗ , Sara Ojeda-Ben´ıtez b , Juan Ramon Miguel Bravo-Zanoguera b , Antonio Rodr´ıguez-Diaz c a PhD Program in Sciences and Engineering, Engineering Institute UABC, Boulevard Benito Ju´ arez y Calle de la Normal S/N, Col. Insurgentes Este, C.P. 21280, Mexicali Baja California, Mexico b Researcher of the Engineering Institute UABC, Boulevard Benito Ju´ arez y Calle de la Normal S/N, Col. Insurgentes Este, C.P. 21280, Mexicali Baja California, Mexico c Researcher of the Faculty of Chemical Sciences and Engineering – UABC, Calzada Universitaria S/N Delegaci´ on Mesa de Otay, Tijuana B.C. M´exico, C.P. 22390, Mexico
a r t i c l e
i n f o
Article history: Received 8 December 2007 Received in revised form 20 March 2008 Accepted 26 March 2008 Available online 21 May 2008 Keywords: Profiles of generation Composition of waste packaging Fuzzy logic
a b s t r a c t An important factor in the environmental crisis faced by modern civilization is the waste produced by a great diversity of commercially available basic household products. An important step in measuring the impact of these organic and inorganic wastes in a community is to characterize the products consumed by families of different socioeconomic levels. In this work we present a method to identify consumption and waste packaging profiles based on household demographic data and type-specific consumption products. Our approach is based on fuzzy logic techniques to deal with quantitative and qualitative information for assisting in the decision making process when dealing with solid waste. This approach was implemented to process the information from a database containing domestic waste produced by families of different socioeconomic levels in a Mexican city. The data were taken from products consumed in 123 dwellings. Different types of packaging materials were classified as metals, cardboard, plastics, and glass. Household demographic variables were included such as level of schooling, number of inhabitants per dwelling, family income, and amount of waste generated. The results achieved show that when correctly combining the variables of a family, a more accurate approximation to the real packaging waste generation profiles may be obtained. The techniques applied in this research were successfully adapted to face problems that until now were difficult to treat, allowing to process information in which inexact values or subjective terms are handled. © 2008 Elsevier B.V. All rights reserved.
1. Introduction The 20th century society requires a great deal of satisfaction provided by means of industrialization and marketing processes. These processes represent the exploitation of natural resources and the generation of large amounts of waste with different ´ compositions. Grodzinska-Jurczak et al. (2004) point in that this can be attributed to a boom in product packaging (mainly plastic). Therefore, the consumption practice of our society in this century causes problems of contamination by the amount of waste that it generates without recovering the value which they have as residue. On this matter Tsiliyannis (2005) indicates that the packaging waste constitutes a significant component of the municipal residues based on its volume. Thus, the generation of packaging waste is related to consumerism, closely related to the system of basic needs in which
∗ Corresponding author. Tel.: +52 686 5664150. E-mail address: [email protected] (G. Lozano-Olvera). 0921-3449/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.resconrec.2008.03.008
modern society exerts influence to drag the consumer into the disposable culture (Ojeda, 2005). The wastes produced by human activity, lose their value when discarded. McDougall et al. (2004) indicate that the lack of value is related to the often unknown mixed composition of the waste. Therefore, waste segregation generally increases its value when there are possible uses for the recovered materials. At present, consumption is one of the largest distractions of society because each family tries to satisfy its needs in the most practical and possible way (Ojeda, 2006). Likewise, waste generation continues to increase because the market economy offers a larger quantity of disposable products, which in addition are offered in a wide variety of designs and of packaging, generating packaging wastes and additional packing such as bags, cans, cardboard, glass, plastic, wood, paper, aluminum and tetra-pack among others, generally thrown away in the waste. Calver (2004) indicates that the transformation packaging is due to the changing lifestyle of consumers. Packaging is essential in modern society as it allows an enormous variety of products to reach the consumer intact, in proper hygienic conditions and general with pertinent information regarding the brand, the product and its usage.
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Packaging waste represents around 17% of urban solid waste by weight and 3% of the total waste flow. In recent years, this percentage has undergone an important increase as a result of the changes produced in the habits of consumers and the types of commercial distribution, which seem to favor more and more the packaging of products and in smaller quantities (individual portions, product units, small packs, etc.). Packaging is a material meant to provide specific services, to contain, protect, transport and preserve a product. Therefore, its function will depend on the way the different needs for which they have been created will be satisfied. In this sense, Vidales (2003) indicates that since its first commercial appearance to the present time, packing has become an element very useful within the complex network that has been woven to satisfy many needs of human beings. For that reason it is important to analyze the generation of packaging wastes taking as a reference the characteristics of generators, which in this case is the family. To investigate the generation and composition of packaging that are commonly found in domestic waste, represents a challenge for the waste management research. As the composition of packaging materials, the function of the packaging and the volume generated, among others, helps to propose strategies for waste management practices. In this sense, it is important to find ways of analyzing the problem in order to offer alternative solutions. Therefore, in this paper, fuzzy logic is applied, being a tool which has been tested in expert systems (Jian-Da et al., 2007) and in other applications (Chang, 1997; ´ 1999; Yukun and Yufeng, 2006; Ami et al., Mohamed and Cˆote, 2008; Pei-Chann and Chen-Hao, 2008; Shih-Ming and Shyi-Ming, 2008). Fuzzy inference systems have also been applied in the field of solid waste. Chen and Ni-Bin (2000) developed a dynamic model by applying fuzzy theory to predict waste generation in an urban area taking a limited number of samples. Al-Jarrah and Abu-Qdais (2006) used an intelligent system based in the fuzzy inference system to approach the problem of selecting a site for a sanitary landfill. In another study, Nie et al. (2007) developed and applied a model for the planning of a solid waste management system under uncertainty, using Interval-Parameter Fuzzy-Robust programming (IFRP). In this study, fuzzy logic is applied because it happens to work with mixed data type (quantitative and qualitative) of two different database structures. Fuzzy logic works with rules to generate knowledge from the databases, and the rules are recognized as an effective and natural means to transmit knowledge between humans, to make and to justify decisions (Moreno et al., 2006). One of the problems which the experts in solid waste face, is to know the waste generation rates and the composition of the waste, for that reason it is important to work with tools that advance knowledge about the household generation behaviors extracting information from existing databases which have been developed. One of the types of wastes that are generated is packaging; therefore fuzzy logic is applied to investigate the profiles of generation by family, considering that it involves qualitative and quantitative variables. This tool is adapted to produce generation profiles of the amount and weight of packaging as a function of family characteristics.
The sampling work for this research was done in three stages. In the first stage, a survey was used in order to learn about the family’s basic characteristics, such as number of members, level of education, income, etc. The second stage consisted in the collection of samples during 8 consecutive days from the households that had accepted to participate; and, in the third stage, another instrument was applied, providing information related to the consumption habits characterizing each inhabitant by house, as well as the environmental knowledge that they have. The information obtained from the three stages studied was stored in two especially designed databases to allow us make a descriptive analysis of the consumption habits, family structure and the generation of waste, among others. Nevertheless, the databases have not been data mining in their totality; it was necessary to apply tools to obtain information of the analyzed databases. To advance in the data analysis validation of the information required a normalization processes applied to the original database in order to eliminate, capture errors or redundancy in the information, preventing wrong interpretation of the information. Once validation was achieved, basic statistical analyses (average, variances, correlation and factorial analysis) were performed to identify diverse panoramas of data behavior and to find existing dependent and independent variables, to obtain the proposed objective of finding profiles of generation of packaging by family. After identifying the dependent and independent variables (see Table 1), a new data base matrix was structured containing the information necessary to obtain the first profiles of the families. In order to generate the input data (x) as the demographic data of the house, defined by the number of inhabitants by house, level of education, family type, number of bags analyzed and income by house, and the output data (y), defined to know the generation of packaging waste by house by weight and by the number of packaging materials. The output variables are: the amount and weight of packaging discard by house and by type of material (plastic, cardboard, metal and glass) (Table 2). The data matrix presented in Table 2 groups the characteristics of the households, including data related to the generation of waste packaging per household in order to define rules that may reflect the quantity and weight of the waste packaging that the households with those attributes tend to discard. To do so, matrix data are inferred. This matrix consists of 123 tuples with 13 attributes Table 1 Definition for the matrix variables Variables
Definition
Inhabitants Level of education
The total members of a single house. Last scholastic degree of the family parents.
Income
Income reported by house. With this variable the socioeconomic levels are inferred, the criterion to determine it is based on the minimum wage (MW).a Low—income of 1–2 minimum wages. Medium—income of 3–5 minimum wages. High—superior to 5 minimum wages.
Family type
Nuclear—the typical family formed by father, mother and son. One-parent—only a father or mother and children. Extended—family who shares home with one or more relatives (grandparents, uncles or nephews).
Number of bags Quantity of waste packaging
The number of bags analyzed by house. Amount of articles by type of packing (plastic, cardboard, metal, glass).
2. Methodology In order to complete this study, the results of a research carried out by Ojeda (2007) were analyzed. This research was conducted in three neighborhoods of the city of Mexicali, which were selected for their socioeconomic stratum (high, medium and low). The research was done in these three neighborhoods, with the waste generated by the selected families during 8 consecutive days.
1023
a The minimum wage is the salary unit in Mexico to pay a worker, and is equivalent to 50.57 Mexican pesos.
1024
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Table 2 Matrix of input and output data Input variables (household demographic data) No of bags
X1 9 7 8 8 6 9 8 6 8 7 9 6 5 5 6 7 8 7 8 9 7 7 a
Family typology
Inhabitants
X2 1 3 3 1 3 1 3 3 3 1 3 1 2 1 3 3 3 3 3 3 3 3
X3 3 6 5 3 4 4 4 9 5 4 6 6 6 4 6 6 5 6 3 6 8 6
Output variables (generation of waste packaging per household in weight and quantity)
Level of education
Income (pesos/week)a
Plastic Weight
Quantity
Cardboard Weight
Quantity
Metal Weight
Quantity
Glass Weight
Quantity
X4 3 2 14 8 4 1.5 2 9 3 6 3 5.5 2 2.5 2 12 3 3.5 1 1.5 5.5 4
X5 3779.61 1691.44 1222.20 2706.30 916.65 611.10 400.00 916.65 2444.40 611.10 1130.14 1527.75 1931.51 611.10 1527.75 1416.65 916.65 1527.75 1833.30 1130.14 2444.40 1014.86
Y1 3779.61 1691.44 1222.20 2706.30 916.65 611.10 400.00 916.65 2444.40 611.10 1130.14 1527.75 1931.51 611.10 1527.75 1416.65 916.65 1527.75 1833.30 1130.14 2444.40 1014.86
Y2 45 32 34 18 31 96 27 61 40 28 21 57 59 11 13 44 35 34 47 66 33 36
Y3 1946.33 488.39 1263.22 836.03 1005.64 2964.13 612.92 1163.61 1115.47 540.05 498.13 815.06 1086.03 348.35 115.58 1158.48 797.4 853.6 923.96 1498.28 744.74 641.9
Y4 15 5 11 7 7 8 2 18 10 9 21 5 20 4 4 7 4 8 11 14 5 8
Y5 218.13 304.47 303.73 171.21 240.5 220.98 15 407.06 234.27 186.6 273.2 36.16 577.8 72.31 70.4 171.52 112.5 401.3 535.94 358.42 196.9 560.11
Y6 22 2 22 0 1 17 2 5 2 0 15 5 11 0 2 20 2 2 0 15 3 1
Y7 926.35 424.6 3131.93 0 14.2 2324.63 348.1 275.4 170 0 1540.9 638.87 1366.33 0 145 529.24 482.9 153.3 0 3237.66 535.01 14.9
Y8 4 1 3 0 0 8 2 2 0 0 7 4 9 0 1 1 2 1 0 13 2 0
11.50 Mexican pesos is equivalent to $1.
of which five describe household characteristics (input variables), the following eight variables describe the quantity and weight of the four categories of waste packaging analyzed in this paper. Once the information analyzed, the computing tools were applied; since they have the capacity of automatic and intelligent analysis of great volumes of data to find knowledge. The first step was to try and obtain knowledge from the generated databases; the proposal by Moreno et al. (2006) was used, and a search process was carried out including all data available. In this process, the grouping technique (C-means, C-fuzzy-means, subtractive) was used in order to find the most significant and relevant patterns and a system of fuzzy logic for the extraction of production rules (IF–THEN). The grouping technique was used to obtain the background of the rules. From the results of these groupings, inferences are established, using the Takagi-Sugeno-Kang (TSK) method, which are fuzzy expressions, and the outputs are deterministic functions (Pasinno and Yurkovich, 1997). This method has been successfully applied in a large number of practical problems. Its main advantage is that it shows a compact/constraint equation of the system thanks to which it is possible to calculate the parameters (Pi ) with classical methods, making its design easier. The database extraction process is based on the following steps: 1. Input data was grouped by means of grouping techniques (Ca means, C-fuzzy-means, subtractive) to calculate the centroid ci,j a and the standard deviation i,j given by Eq. (1) which will be the parameters of the membership functions ground given by Eq. (2). a i,j =
r × (max(X) − min(X)) a √ 8
ij (xj )
of the back-
(1)
where X is a matrix. In this case, it is the matrix shown in Table 2, representing the system’s input and output variables; i are rules from the membership function and j the variables of matrix X necessary to evaluate the rule; a is a parameter defining the
association rules, and ra , a positive constant, is the rule a output representing a neighborhood radius. 2. The following step was to calculate the consequents by means of the TSK fuzzy reasoning method. This model consists in analyzing the rules, each rule has a numerical output, the total output is obtained by means of “weighted average”, actually the operator of weighted average sometimes is replaced by the operator of “weighted Sum” to reduce the times of analysis, especially in the training of a system of diffuse inference.
∧
ij xj
=e
2
−(1/2)((|| xp,j −c a ||/( a ) )) i,j
ij
(2)
3. The rules based on the parameters of the membership functions of antecedents were extracted, ij (xj ), as well as those of consequents, fi (x), from the rules of the TSK fuzzy inference system shown in Eq. (3). Ri = IF x1 is i1 and . . . and xj is ij and . . . and xn is in THEN y is f1 = ai1 x1 + · · · + aij xj + · · · + ain xn + ai
(3)
4. The TSK fuzzy inference system was assessed with the input data ∧
(Xi ) shown in Table 2 to obtain the output (y) and calculate the mean squared error (Eq. (4)). RMSE =
MSE(e)
(4)
where RMSE is the comparison of the real input used by the system to carry out the assessment. MSE is the mean of squared errors. If RMSE was zero, real data would equal the assessed data. This indicates that the fuzzy comparison equals real data, then the error range tends to zero. The Fuzzy Logic Toolbox (FLT) was used to develop and apply the steps to the problem we are analyzing in this paper. This is a tool for the development of user-friendly fuzzy programs in the Matlab environment; the graphic called Fuzzy Interference System (FIS)
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
1025
In this paper, the socioeconomic stratum and the level of education are the fuzzy variables. The stratum is the economic level with which a family manages to satisfy its needs and reflects its consumption practices. It is an uncertain variable because it may take a range of values for each family as the data used are not an exact number. In connection with the level of education, it was assigned according to the level of education of the inhabitants, which could be complete or incomplete. For example, at the secondary school level of education, the inhabitants could have the secondary school level complete or incomplete; in this range the attribute level of education may lose defining features. 3. Results
Fig. 1. Function of FIS editor.
was used. It allows to build, edit and observe fuzzy interferences within an FLT. The operation of the FIS editor is shown in Fig. 1; it can be observed the sequence of data processing with the fuzzy logic tool, the matrix of data pre-processed is the input to the fuzzy system under a clustering process, which generates new information that allows to define the rules. For the problem that we are analyzing is important to generate rules, because these give a special vision about the knowledge and forms in which this can be represented. The usage of rules implies particular ways of understanding problems and of solving them. When applying rules to the problem of waste, it helps to assess the behavior or reasoning the system will have when assessing the input, qualitative and quantitative parameters, depending on the behavior of the different types of waste packaging and of the family generating them, so that an output consistent with the basic statistical data obtained may be achieved. After obtaining the rules, the fuzzy logic method was used for the assessment of the system based in rules; and, therefore, an output was obtained. This tool allows to assess variables considered inaccurate or uncertain, as it is the case analyzed in this paper, where the determination of generation profiles for a family implies to work with variables associated to the family’s behavior and to the particular characteristics of each of them, which become uncertain when combining them. In addition, fuzzy logic facilitates working with these types of variables since it makes possible the establishment of sets with restrictions or limits that stay effective in a given universe. It is for that reason that was applied to this work, since the family has a structure and composition difficult to define, these vary constantly, and it is reflected in the members of each family.
When applying fuzzy logic tools to restructure the collected information in the two databases, a matrix with input and output data was obtained. The presented results belong to the matrix in Table 2, which was created to analyze data from 123 households and the generation of packaging of each household per type of material. Data related to packaging generation from 850 bags collected for the sampling were analyzed. Of the total number of households analyzed, 26% belong to the low stratum, 48% to the medium and 25% to the high. 3.1. Composition of packaging The total weight of the analyzed packaging equaled 501.9 kg, which is equivalent to a total of 8062 packages discarded. In Table 3, packaging generation is shown by composition, where the type of material used in the packaging is shown by weight and quantity. It was observed that plastic packaging is the ones most commonly discarded, followed by cardboard and metal packaging. The results related to the generation of packaging by usage category are shown in Fig. 2. Among them, food, medicines and beverages can be mentioned. It can be observed that most generation is concentrated in the category of food, being the low stratum the one which generates the most. This behavior explains consumption practices per socioeconomic stratum. The households belonging to the low stratum consume products in small-sized packaging because their economic capacity is low and they discard a larger quantity of packaging in shorter periods, whereas, in the high stratum, consumers buy products in larger-sized containers or packaging, for example soft drinks of 3 l instead as opposed to the 600 ml bought by the ones in the low stratum. This implies that the high stratum generates lesser packaging, while the low one discards a larger quantity of packaging. 3.2. Family structure In order to analyze the family profiles, the type of family has been considered: nuclear, one-parent and extended. It was found that 49% of all households belong to nuclear families, 12% to one-parent families and 39% to extended families.
Table 3 Average quantity of packaging waste sampled by socioeconomic stratum Composition
Plastic Cardboard Metal Glass
Low (packaging (family/week))
Medium (packaging (family/week))
High (packaging (family/week))
Quantity
Weight (g)
Quantity
Weight (g)
Quantity
Weight (g)
38.9 9.2 7.9 3.4
1031.5 251.2 841.5 659.1
39.2 11.2 7.4 3.3
1560.6 550.8 770.7 687.8
288.8 382.5 442.7 495.1
542.3 4071.6 6077.6 10091.9
1026
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Fig. 2. Packaging generation by socioeconomic stratum by usage categories.
Another variable included was the level of education of the parents. This variable was obtained by means of the average education achieved by both parents, which was analyzed as a fuzzy group. An analysis was carried out of each of the levels of education achieved, as well as the levels reported as incomplete by the members of the family. A membership function was created for the analysis of the education variable in order to evaluate each of the fuzzy situations that may appear in the analyzed sampling. This function of membership is generated after obtaining the assessed clustering.The membership functions of the education variable is represented by a Gaussian function, allowing us to assess each of the levels of education completed, as well as the incomplete levels of education achieved (see Eq. (5)). 2
A (x) = e−k(x−m)
(5)
In the membership function, the value of X corresponds to all the levels of education assessed and m represents the mean values. Through this analysis, each of the situations presented in the sampling analyzed was assessed in detail. In this case, it was related
to all the combinations concerning the levels of education of the members of each family included in this research. Concerning the income variable of each household included in the analysis, this variable was associated to the socioeconomic stratum and the packaging generation, as well as the classification of the type of material of the packaging generated per household. Fig. 3 shows graphically the system proposed, based on the modeling of fuzzy rules and fuzzy control. 3.3. Profiles of packaging generation Once the variables “packaging generation” and “household general characteristics” were analyzed, we proceeded with the creation of a matrix (Table 2). In this table, five input variables were defined (X1 , X2 , X3 , X4 , X5 ), the output variables corresponding to Y1 , . . ., Yn , representing the data expected to be obtained when assessing the characteristics of each family, as well as the composition and quantification of the packaging generated. Using this matrix created from the two databases employed in this research, the four input variables were evaluated, as well as the packaging generation per type of material (see Fig. 4).
Fig. 3. Basic structure of the system based on fuzzy rules applying TSK to determine profiles of packaging generation.
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Fig. 4. Application of the system to determine profiles of packaging generation in the Matlab Fuzzy Logic Toolbox.
The input variables (X1 , X2 , X3 , X4 , X5 ) of the system are presented in Fig. 4. These inputs are processed by the TSK method, allowing the identification of the output. One of the disadvantages of this process is that it only manages one output; therefore, the process is repeated for each of the outputs that the user may require. In this case, eight outputs are required (Y1 , . . ., Y8 ). This repetitive process is not performed by the user, because there is a platform designed for the user, which allows him/her to execute it only once and to obtain the information required.
3.3.1. Profiles of packaging generation per type of material This section describes the results of the profiles of packaging generation, both per quantity–frequency of packaging, and by weight. The results in evaluating packaging generation by quantity are shown in Figs. 5 and 6; generation by weight is shown in Fig. 8.
1027
Fig. 5 shows the behavior of packaging generation in packaging made out of plastic, cardboard, glass and metal in households, taking as the variable the level of education of the heads of household. This diagram shows that plastic, metal and glass packaging are generated more frequently in households with a higher level of education, from medium to high level; while cardboard, according to the results obtained, is a packaging more widely used by inhabitants with a lower level of education. When adding another variable to the analysis, changes in the generation of packaging are observed. Fig. 6 shows the behavior of container generation including the input variable. When developing this assessment, a similarity regarding the behavior with plastic and cardboard is observed, as well as with glass and metal, though there is no such similarity concerning the quantity of packaging generated. This shows that when combining the variables of a family correctly, a more accurate approximation of packaging generation by family is obtained. The results of assessing the weight of plastic packaging and the income variable show that the behavior is different to the one observed in the analysis carried out with the frequencies of packaging. As in the assessment of weight, it is necessary to consider other factors or variables in the families, such as income, shopping frequency, presentations of the products, among others. When analyzing packaging generation in the higher stratum, it was generally found that families in this socioeconomic stratum do their shopping by acquiring products in large-sized presentations (2 l, 5 kl, etc.). Therefore, generation is higher, though the frequency with which they discard these packages is within longer periods of time. However, in the lower stratum, the opposite happens: products are consumed in packaging of smaller-sized presentations. Therefore, they generate a larger amount of these items, and lesser weight was found in this generation (Fig. 7). Fig. 8 shows the assessment results concerning the generation of packaging made out of plastic, cardboard, glass, metals with two variables of the family structure: number of dwellers and income.
Fig. 5. Assessment of packaging generation per level of education and type of material. (a) Plastic, (b) cardboard, (c) glass and (d) metals.
1028
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
Fig. 6. Packaging generation by type of material vs. education and income. (a) Plastic, (b) cardboard, (c) glass and (d) metals.
In subsection (a) of Fig. 8, the results of plastic packaging generation are shown, taking as input variables the number of inhabitants per household and the income. The results show that the families with a higher income and fewer members generate larger amounts of plastic packaging, while families made up of four, five or six members and the same income, keep a more homogenous generation profile. In subsection (b) of Fig. 8, the diagram shows the generation of packaging made out of cardboard is higher in low-income families
with three to seven inhabitants per family. In subsections (c) and (d), the results of the analysis related to the generation of packaging made of glass and metal is shown. The diagrams show that, in terms of weight, the generation of these packaging materials is very similar. When evaluating these variables an acceptable RMSE was obtained in each of the different outputs assessed. Table 4 shows the values obtained with the assessment of the system, as well as the mean squared errors. When observing these values, one
Fig. 7. Generation average regarding the weight of packages.
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
1029
Fig. 8. Packaging generation by type of material vs. number of inhabitants and income. (a) Plastic, (b) cardboard, (c) glass and (d) metals.
Table 4 Assessment of the errors of the system Variable
Plastic
Cardboard
Glass
Metal
RMSE MSE
2.43 5.93
1.35 1.82
1.17 1.37
1.80 3.23
may notice that they are acceptable and very similar to the real ones. 4. Discussion and conclusion One of the main problems encountered when studying household solid waste is to obtain all household data and manage the components of the variables constantly varying, which makes it uncertain and complex. In this application, the variable family structure is very complex as time produces a constant variation; for example, the members of a family may increase their number when a new member is born, gets married or it may decrease if a member dies. The other variable included is income, which has a similar behavior, as the income may grow when the salary of one of the members is increased, or one of them begins to work. The same happens with the level of education, which varies with time. Fuzzy logic provides us with elements to generate predictions concerning changes in the family structure. One of the goals of the tools related to the search of knowledge in databases is to identify the patterns of behavior of data, and to infer knowledge and predict situations (Pei-Chann and Chen-Hao, 2008). In this research, the aim was to predict the profiles of pack-
aging generation per family, combining qualitative and quantitative variables. Throughout the processing of the information, it was possible to prove that the database contained a large number of redundancies among its variables, but also a lack of information in some of its fields. Under this situation, and using traditional statistical, the generated information has a high degree of uncertainty since data did not exist or it was scarce. But as indicated in Chen and Ni-Bin (2000) work, where with few simples they were able to forecast waste generation, fuzzy logic represents a powerful tool to efficiently handle these problems. That is why fuzzy logic can infer the information using available data, meanwhile analysis with conventional statistics does not include all available input when data set is not complete. With the tools applied in this research to these situations of uncertainty due to the lack of information or the combination of qualitative and quantitative variables, knowledge from the databases used was inferred. The generation and composition of household solid waste has a strong relationship with the patterns and levels of consumption of a given population. The results found in this paper show that there is a direct correlation between the composition of packaging generated from household waste flow and the features identifying a family. In this sense, Schiffman and Kanuk (2005) indicate that the family may be analyzed as a variable. Also, as the absence or presence of a new member produces an impact on the behavior of the family, another factor determining changes in the family structure is the age of its members, as according to their age basic and non-basic needs are modified over time. The profiles of packaging generation of the studied families reflect consumerism practices because, as indicated by Ojeda
1030
G. Lozano-Olvera et al. / Resources, Conservation and Recycling 52 (2008) 1022–1030
(2005), the system of basic needs is related to this consumerism practices. It is important to know the composition and quantity of packaging generated by a family, in order to give the discarded packaging the real value it has and to establish policies for the management and better usage of this waste in Mexican municipalities, because as it was pointed out by McDougall et al. (2004) the unawareness of the type of packaging generated reduces the value of waste. Fuzzy logic was applied, as the family dynamic is complex and it impacts on packaging generation, producing variations in weight or frequency and depending on how the family is structured. This is the reason why it is important to have a tool that may allow the assessment of any situation that may occur at the core of a family, in case there is any kind of change, both physical and intellectual. Or even that it may predict any change which might occur within the family structure in the future. The fact of applying fuzzy logic to determine the profiles of generation has allowed us to set up a system that simultaneously processes quantitative data, as packaging generation, and qualitative data, as the composition profile per structure and type of family, removing the inaccuracy obtained with the basic statistics. References Al-Jarrah O, Abu-Qdais H. Municipal solid waste landfill siting using intelligent system. Waste Management 2006;26(3):299–306. Ami HIL, Wen-Chin Ch, Ching-Jan Ch. A Fuzzy AHP and BSC approach for evaluating performance of IT department in the manufacturing industry in Taiwan. Expert Systems with Applications 2008;34:96–107, On line. Chang PT. Fuzzy seasonality forecasting. Fuzzy Sets Systems 1997;90:1–10. Chen HW, Ni-Bin Ch. Prediction analysis of solid waste generation based on grey fuzzy dynamic modeling. Resource, Conservation and Recycling 2000;20(1–2):1–18. ´ Calver G. What is packing? Mexico: Ediciones Pili; 2004.
´ Grodzinska-Jurczak M, Hanna Z, Read A. Management of packaging waste in Polan—development agenda and accession to the EU. Waste Management & Research 2004;22:212–23. Jian-Da W, Yu-Hsuan W, Mingsian RB. Development of an expert system for fault diagnosis in scooter engine platform using fuzzy-logic inference. Expert Systems with Applications 2007;33:1063–75. McDougall F, White M, Franke M, Hindle M. Integral management of solid residues: inventory of the life cycle. Venezuela: Procter & Gamble Industrial; 2004. Mohamed AMO, Cˆote´ K. Decision analysis of polluted sites—a fuzzy set approach. Waste Management 1999;19(7–8):519–33. Moreno JE, Castillo O, Castro JR. 2006 Data Mining for the extraction of if-then rules using fuzzy logic based on the FIS of Mamdani and Takagi-Sugeno Kang. In: IEEE-Computational Intelligence Society Mexico Chapter. Proceedings of international seminar on computational intelligence ISCI 2006; 2006. p. 62–8. Nie XH, Li YP, Liu L. IFRP: A hybrid interval-parameter fuzzy robust programming approach for waste management planning under uncertainty. Journal Environmental Management 2007;84(1):1–11. Ojeda BS. Flow of packages and packing present in the domiciliary solid remainders. An analysis of the life cycle. Unpublished report. UABC—Instituto de Ingenier´ıa; 2005. Ojeda BS. The consumption source for waste generation and contamination. In: ´ y Medio ambiente. Serie editorial CinMargarito QN, editor. Contaminacion ´ ´ ISBN cuentenario de la UABC; 2006, p. 227–49. Editorial Miguel Angel Porrua, 970-701-859-3. Ojeda BS. Dimensions of the consumer behavior and his relation with the generated solid residues: strategies for its recovery in a sustainable house. (2005–2007) Financed by CONACYT. Project in process; 2007. ˜ Pasinno K, Yurkovich S. Fuzzy control. 1st ed. Addison Wesley Publishing Company; 1997. Pei-Chann Ch, Chen-Hao L. A TSK type fuzzy rule based system for stock price prediction. Expert Systems with Applications 2008;34:135–44. ´ ´ 2005. Schiffman L, Kanuk L. Consumer behavior. Mexico: Press Pearson Educacion; Shih-Ming B, Shyi-Ming Ch. Evaluation students’ learning achievement using fuzzy membership functions and fuzzy rules. Expert Systems with Applications 2008;34:399–410. Tsiliyannis C. A new rate index for environmental monitoring of combined reuse/recycle packaging systems. Waste Management Research 2005;23:304–13. Vidales GMD. World of packing, a manual for the design and production of packages and packing. Mexico: Press G. Gili S.A. de CV; 2003. Yukun C, Yufeng L. An intelligent fuzzy based recommendation system for consumer electronic products. Expert Systems with Applications 2006;33(1):230–40.