- Email: [email protected]
Hybrid modeling and simulation for complementing Lifecycle Assessment
Computers & Industrial Engineering 69 (2014) 77–88
Contents lists available at ScienceDirect
Computers & Industrial Engineering journal homepage: www.elsevier.com/locate/caie
Hybrid modeling and simulation for complementing Lifecycle Assessment Bochao Wang a, Séverin Brême b,⇑, Young B. Moon a a b
Department of Mechanical and Aerospace Engineering, Syracuse University, Syracuse, NY 13244, USA Institut Supérieur de l’Aéronautique et de l’Espace (ISAE), Toulouse, France
a r t i c l e
i n f o
Article history: Received 22 July 2013 Received in revised form 27 December 2013 Accepted 30 December 2013 Available online 7 January 2014 Keywords: Sustainability Lifecycle assessment (LCA) Agent-based modeling System dynamics Discrete-event simulation
a b s t r a c t This paper presents a new complementary lifecycle assessment (LCA) approach to address several limitations of the standard LCA methodology. An integrated approach of agent-based modeling, system dynamics and discrete event simulation was adopted to complement the standard LCA methodology. A hybrid simulation model was developed as a proof-of-concept system, then it was validated using a case study of bottled water and alternative drink products. The model was based on the assumption that parameters and relationships were constant regardless of local uniqueness. The research demonstrates that the hybrid modeling and simulation method can address several limitations of the standard LCA. Also, it is also proven that the method has a potential to address social and economic aspects. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction As energy consumption and pollution become critical issues worldwide, people are more and more concerned with sustainability issues through various attempts ranging from creating environmental-friendly products to changing habits to reduce waste. Governments and corporations also make numerous efforts such as managing energy usage, waste, emission, etc. While such initiatives deserve commendations, there is a danger of focusing on local improvement only thus unintentionally worsening the situation unless a holistic systems thinking guides those executions. One of the useful tools in evaluating overall environmental impacts throughout a product’s entire life is lifecycle assessment (LCA). LCA is the ‘‘compilation and evaluation of the inputs, outputs and the potential environmental impacts of a product system throughout its life cycle’’ (ISO 14040, 1997). In a LCA study, many aspects throughout a product’s lifecycle can be considered including production, transportation, distribution, usage and end-of-life activities. LCA has been used to identify potential opportunities for improvement such as in better design, better manufacturing processes and better management in order to minimize negative sustainability impacts (Brezet & Hemel, 1997). LCA is a unique tool that comprehensively examines the environmental impacts of a product or service throughout its life ⇑ Corresponding author. E-mail addresses: [email protected] (B. Wang), [email protected] (S. Brême), [email protected] (Y.B. Moon). 0360-8352/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.cie.2013.12.016
cycle. It is an ISO standardized method and proved to be a useful aid in decision-making for designers, managers, government and consumers (Brezet, Stevels, & Rombouts, 1999; Rahimifard & Clegg, 2007). However, the standard LCA method has a number of limitations and some of these problems can be critical (Reap et al., 2008). One of them is that the LCA method takes a static viewpoint that its parameters and internal relations among entities remain constant. Also, the social and economic impact, local environmental uniqueness, effects of dynamic environment, and temporal perspectives cannot be easily considered in the LCA. In other words, the standard LCA is useful as a high level tool, but not necessarily for dealing with dynamics and uncertainties. This research aims to demonstrate that a hybrid simulation model can address these limitations of the standard LCA approach. All the steps in LCA – goal and scope definition, inventory analysis, and impact assessment, including interpretation – are considered in our research. However, the focus is placed on the third step of the LCA, that is, impact assessment. A hybrid simulation model combining agent-based modeling, system dynamics and discreteevent simulation methods was developed as a proof-of-concept system. The validity of the developed approach was done on comparing bottled water alternatives such as tap water and vitamin water along with different bottle options. The paper is organized as follows. First, limitations of traditional LCA approach will be addressed, followed by explanation of an integrated hybrid modeling and simulation approach. Drinking water and beverages are chosen to illustrate how the framework is developed and modeled. Impact analysis based on the simulation
78
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
results is explained to show the value and potential of the new complementary LCA approach. Conclusion and future work is provided at the end. 2. Limitations of standard Lifecycle Assessment (LCA) The standard LCA method has been adopted to address the needs of research and practical analyses for a long time (Azapagic, 1999; Guinée et al., 1993; Van den Heede & De Belie, 2012). It is useful because it constructs a grounded structure and procedure to analyze an overall environmental impact that a product can make during its whole life cycle. However, its advantages through adopting a systematic viewpoint (Forrester, 1968) became its limitations because many systems of interest are never static, but evolving. Stasinopoulos et al. (2012) points out one kernel limitation of standard LCA method is that its life cycle inventory studies assume that parameters used are always constant or fixed functions of time which prevents taking temporal effects and dynamic response of the system into consideration. The report that the simplified assumption made in estimating energy benefits is not reasonable later. In comprehensive discussion of unresolved problems in LCA, Reap et al. (2008) explains the reason and progress made to overcome numerous disadvantages, and classifies those critical problems and difficulties into several categories. Categories are also rated for severity and adequacy of traditional LCA solutions. Powell criticizes the laggard traditional inventory analysis and local environmental uniqueness with impacts that could potentially lead to a faulty decision-making process (Powell, 2000). Mayyas et al. (2012) points out the disadvantages of static point of view, not only on the system’s boundary selection and inventory, but also on the data and time horizons. He further uses an example on how to make a sustainable lightweight body-in-white design with improved quick solutions to the standard LCA approach. The uncertainties such as in data and function are also overlooked in the traditional LCA study. However, these uncertainties can play important roles in affecting outputs and estimated environmental impact severity and thus impact decision making for enterprises and governments. While these are all valid critiques and points for improvement for the standard LCA method to be addressed eventually, we address only a few of these problems in our research, that is, social and economic impacts, alternative scenario considerations, local environmental uniqueness and dynamics and time horizons. Table 1 highlights what this research addresses in the list of problems that Reap and his colleagues pointed out (Reap et al., 2008). 3. Integrated hybrid modeling and simulation method Three commonly used systems modeling and simulation methods are Discrete Event Modeling and Simulation (DEMS), System Dynamics Modeling and Simulation (SDMS) and Agent-based Modeling and Simulation (ABMS). These are integrated to simulate the life cycle process and study short-term and long-term performance under various scenarios. The new approach has potential for building a unique model that can combine each model’s uniqueness and advantages into one model and take their differences into consideration. Discrete Event Modeling and Simulation (DEMS) can simulate multiple events in a time sequence (Zeigler, Kim, & Praehofer, 2000). They are built in the form of entities, flowcharts and resources. DEMS is a natural choice when linear processes in a complex environment is modeled and an entity’s action is triggered by other entities or a certain time. Many service facilities, production systems, maintenance and recycling facilities, and transportation
and material handling systems are best described and simulated via DEMS. System Dynamics Modeling and Simulation (SDMS) is a methodology used to model and simulate a system from a higher system-level viewpoint (Doebelin, 1998; Sterman, 2001). It comprises stocks, flows and unique feedback loops. The state of the whole system could be observed from various stocks at a given time. Aggregates are linked through aggregated mechanisms implemented as flows in SDMS. Stock and flow are natural choice for modeling beverage inventory and production flow. Feedback loops link each module of the system with defined relations and influence. Agent-based modeling and simulation (ABMS) is a methodology used to model and simulate individual actions and interactions of agents in a complex adaptive system, focusing on their effects on the system as a whole (North & Macal, 2007). ABMS are also called individual based models, due to their bottom-up individual-level modeling approach. They are constructed in the form of active objects, individual behavior rules, and direct or indirect interaction within a dynamic environment. All of these elements can be used to represent agents and their interactive operations in an environment such as a competitive beverage sale market. Therefore, it can be used to bridge marketing and engineering activities. It also suits complex and flexible situations that need to be modeled, especially when taking customers, retailers and competitors into consideration. These three methodologies can be combined into an integrated hybrid modeling and simulation method to complement each methodology’s respective strengths. The most promising part of the integrated modeling approach is its flexibility that can handle dynamic and evolving requirements of a system. SDMS can deal with aggregates at the highest abstraction level while DEMS can be used at middle levels of abstraction and possibly at lower levels as well. ABMS can be used across all levels of abstraction. In our research, we utilized the flexibility of the integrated hybrid modeling and simulation and developed a proof-of-concept system that can complement the standard LCA method.
4. Lifecycle Assessment of Drinking Water and Beverages Tap water is still one of the major drinking sources in daily life. However, the bottled water market in developed countries such as the United States and Japan has grown rapidly. A report from the Beverage Marketing Corporation (THE 2006 STATS) reveals that there was a 50% growth in America’s bottled water consumption between 2002 and 2007. For health, quality and convenience, bottled water has become a popular choice in the drinking water and beverage market. The rise in popularity of bottled water has created a burden on its sustainability (Gleick, 2010), however. For example, bottled water produces waste by product in production, transportation, distribution, refrigeration and recycling. Also, many other drinking alternatives (for example, vitamin water) have become available in the market recently. So the basic questions are: (i) Which one of the available options (e.g. tap water, bottled water, and other drinking alternatives) is the most sustainable? (ii) Will consumers’ choice make a difference? From the related important aspect of packaging, Lee and Xu (2005) addressed sustainable packaging issues in general. Büsser and Jungbluth (2009) studied the role of flexible packaging in the lifecycle of coffee. Vellini and Savioli (2009) focused particularly on glass containers from production to recycling. All of these reveal that the packaging issue is a significant one in evaluating
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
79
Table 1 LCA Problems (Reap et al., 2008) addressed by this paper and potentially by the hybrid modeling approach. Traditional LCA problems Functional unit definition Boundary selection Social and economic impacts Alternative scenario considerations Allocation Negligible contribution criteria Local technical uniqueness Impact category selection Spatial variation Local environmental uniqueness Dynamics of the environment Time horizons Weighting and valuation Uncertainty in the decision process Data availability and quality
Problems addressed in this paper
Potential solvable problems p
p p p p p p p p p p p
Fig. 1. The boundary of the system and the stages of sport drink systems.
environmental impacts. We incorporate all the important elements and steps in assessing drinking water alternatives with the idea that multiple impacts are interrelated (Humbert et al., 2009). LCA consists of three distinct steps with associated interpretation of results. The three steps are: (i) goal and scope definition, (ii) inventory analysis, and (iii) assessment of impacts associated with these inputs and outputs.
4.1. Goal and scope The goal of our study is to measure the environmental impacts of beverage consumption habits under different scenarios. In this study, the critical environmental issues and responsibilities were identified along the entire lifecycle chain of five specific drinking alternatives: 1. 2. 3. 4. 5.
Tap water in a glass bottle, Tap water in a reusable aluminum bottle, Ecoshape bottled water, Sport drink, and Vitamin water.
Consumer behavior is also taken into consideration. Consumers are allowed to choose freely from drinking alternatives and switch among them over time to reflect the trend of consumers. Two particular impacts, energy consumption and global warming potential, are assessed to reveal the practicality of the new methodology.
4.2.2. Proper functional unit and emission data A reasonable water/beverage consumption amount is chosen as 3 liters (L) per day per person. This is equivalent to 6 bottled water volumes or vitamin water bottles. Data for material/energy consumption, water usage, waste generation, greenhouse gases emissions, distribution selection and quantitative relations between entities and parameters are collected from a company’s study report (Nestlé Waters North America, 2010), GaBi databases (LBP, 2006) and published LCA papers (Azoulay, Garzon, & Eisenberg, 2001; Keoleian, Bulkley, & Dettore, 2009). 4.3. Assumptions The whole life cycle stages from beverage and packaging production to end-of-life are considered and summarized in Table 2. In order to effectively reflect and compare the results of each product in different scenarios, the two most commonly used and important factors are chosen: (i) Energy is the amount of energy used during each phase of the lifecycle (Pasqualino, Meneses, & Castells, 2011). (ii) Global-warming potential (GWP) is a relative measure of how much heat a greenhouse gas traps in the atmosphere. It compares the amount of heat trapped by a certain mass of the gas in question to the amount of heat trapped by a similar mass of carbon dioxide. GWP is expressed as a factor of carbon dioxide (whose GWP is standardized to 1). Additional assumptions are made to simplify the model constraints and set a reference standard to start with:
4.2. Inventory analysis 4.2.1. Product system boundary The whole lifecycle of drinking alternatives will be covered. For the sport drink as an example, its lifecycle ranges from beverage production, package production, transportation, distribution, refrigeration to recycling or landfilling disposal. The boundary of the system and stages are shown in Fig. 1. There are five stages in the sub-model of sport drink.
(i) It is assumed that each consumer consumes 6 servings a day. Each serving is in a 500 mL bottle or glass, which is the functional unit used. (ii) For containers purchases, it is assumed that, on average, (a) for tap water in glass bottles, one glass is bought every 4 months; (b) for tap water in reusable aluminum bottles, one reusable bottle is bought every 1000 days;
80
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Table 2 Life cycle stages and descriptions. Stage
Description
Beverage production
This phase includes the pumping and the treatment of tap water at a municipal water plant. It also includes processes, extraction and transportation of all materials that are needed to transform water into the final beverage This phase includes the extraction and transportation of raw materials and the production of the container (glass, aluminum or plastic bottle)
Container and packaging production Distribution and marketplace
Transportation
Transportation of the containers and bottled beverages from the production plant to a supermarket, with a truck
Refrigeration Marketing Storage
Refrigeration of the bottled beverages at a supermarket, in a refrigerator Activities related to marketing the containers and the bottled beverages Storage of the containers and the bottled beverages at a supermarket
Consumer use
Transportation Refrigeration Dishwashing
Transportation of the containers and bottled beverages from the supermarket to the user’s home, in the user’s car Refrigeration of the beverages at the user’s home in a refrigerator Dishwashing of the reusable container (glass or aluminum bottle) at the user’s home in a dishwasher. Frequency of dishwashing is different for glasses and reusable bottles
End-of-life
Hauling Recycling Landfilling Waste-toenergy
Hauling of the discarded empty containers to a landfill or to a recycling facility Recycling of the empty containers at a recycling facility Landfilling of the empty containers Conversion of the empty containers into energy
Table 3 The beverage selection percentage by customers according to the six scenarios.
Table 4 Value comparison of the simulation result and a company’s study report.
Scenarios
Tap water Ecoshape bottle Reusable aluminum bottle Sport drink Vitamin drink Total (%)
1 and 2
3
4
5 and 6
40 36 20 2 2 100
0 96 0 2 2 100
57 0 35 4 4 100
40 36 20 2 2 100
(c) for the other drinks, no additional container is purchased, that is, the beverage is drunk directly from the bottle. (iii) It is assumed that all beverages except tap water are refrigerated. (iv) For washing, it is assumed that (a) for tap water in glass bottles, the consumer washes his/her glass bottle 3 times a day (for 6 servings a day) (b) for tap water in reusable aluminum bottles, consumer washes his/her bottle every 4 days (or 0.25 times a day) and washing takes place in a dishwasher. (v) For energy consumption calculation, it is assumed that an average US dishwasher consumes 2 kW h per load (Energy Star, 2013), so (a) a glass bottle is assumed to represent 1/44 of a load (2/44 kW h= 0.045 kW h = 162.551 kJ) (b) an aluminum bottle is assumed to represent 1/22 of a load (2/22 kW h= 0.09 kW h = 325.102 kJ).
Model results
A company’s study report results
ENERGY (MJ)
TapWater EcoShape Reusable Sport Drink VitaminWater
0.089 2.854 0.049 4.972 5.049
0.14 3 Not available 5 5
GWP (kg CO2 eq)
TapWater
0.014
0.01
EcoShape Reusable SportDrink VitaminWater
0.147 0.007 0.266 0.288
0.15 0.06 0.2475 0.2565
The energy used for the refrigeration of one serving is 0.00907 kW h = 32.763 kJ. 5. Modeling and simulation System dynamics is used to model the workflow and calculate the energy consumption and GWP for each step continuously. Since the overall process is a continuous procedure in straightforward workflow which involves plenty of feedbacks such as material flow and energy usage, system dynamics is used because it is best suited to analyze a system with dynamic stocks, flows and feedbacks. Beyond that, the detailed work processing procedure and the connection linking consumer behaviors with actions are described via discrete-event and agent-based methods (see Tables 3 and 4). 5.1. Scenarios
The reference standard is set using the data from a company’s study report: Volume: 615 L. Energy consumption of the refrigerator: 422 kW h/year = 1.156164385 kW h/day for 615 L. Room occupied: 2 L of the refrigerator for a 0.5 L serving. Energy consumption for the refrigeration of one serving: 0.00375988 kW h per day. Refrigeration duration: 2.4 days.
Scenario 1 is the ‘‘reference scenario’’. This scenario represents the base pattern of beverage consumption in New York State assuming that only the five kinds of beverages are available. All beverages are refrigerated for 2.4 days on average, except tap water, which is not refrigerated. The glasses used for tap water and the reusable bottles are washed in a dishwasher. Scenario 2 is similar to the reference scenario, but during the winter it is assumed that no beverage is refrigerated by the consumer.
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Scenario 3: In this scenario, tap water is somehow unavailable due to pollution. ‘‘Tap water in glass bottle’’ and ‘‘tap water in reusable aluminum bottle’’ are then replaced by ‘‘ecoshape bottled water’’. All drinks are refrigerated. Scenario 4: Under this scenario, bottled water is banned. Scenario 5: Same as the reference scenario, but refrigeration conditions are different. It is assumed that refrigeration takes place for 7.2 days instead of 2.4, in a 20-year-old refrigerator that consumes three times more energy. Furthermore, it is assumed that the beverage occupies about 1/10 of its refrigerator content. Scenario 6: Same as the reference scenario, but glasses and bottles are washed by hand with cold water.
81
Fig. 3. Bottled water production feedbacks in system dynamics.
5.2. Framework and modeling 5.2.1. Overall hybrid modeling framework The model is developed in two main parts: (i) modeling the lifecycle of each beverage and (ii) modeling the behavior of each consumer of the population. The lifecycles of the five beverages are first modeled in SDMS, followed by the consumer behavior modeled in ABMS and DEMS. The integrated model is then established in order to compare the environmental impacts of beverage consumption under the different scenarios. The structure of the hybrid model is presented in Fig. 2. 5.2.2. SDMS part Bottles, glasses and cardboard used to produce bottled water can be recycled into new containers as a feedback system illustrates in Fig. 3. This increases the production rates of beverages and containers. Four kinds of raw materials are needed to manufacture bottles and their packaging: packaging cardboard, PPlid, PET Resin, and Pallet. The production rate of each beverage and each container is based on average consumer consumption (6 servings a day for one consumer) and on averaged losses (some bottles and glasses are stolen or damaged in the supermarkets or during manufacturing and transportation). (See Figs. 4–7). After the production stage, the beverages are transported to the marketplace, which is a supermarket. There, bottled beverages are refrigerated before being sold to the consumers. After people drink the beverages, there are three ways to dispose of all materials: landfilling, waste-to-energy or recycling. The recycle rates depend on the container used. For tap water, the SDMS diagram is composed of two independent SDMS diagrams: one for tap water and the other for the containers. Tap water is processed and generated at a municipal water plant, then distributed through pipes to the consumers. The tap water consumption is determined by adding: (i) the tap water that is drunk by the consumer, (ii) the tap water that is used for dishwashing and, (iii) the losses during the production and
Fig. 4. Distribution and market stages of the bottled water system.
distribution processes. The production rate of tap water is determined to meet customer’s demand. On the other hand, glass bottles are produced from raw materials and reusable aluminum bottles are from recycled materials. Their production rates are chosen with respect to consumer demand. They are then distributed to the marketplace, bought by consumers, used and then discarded. For the bottled beverages (ecoshape bottle, sport drink and vitamin water), the SDMS diagram is similar to the one of tap water.
5.2.3. ABMS and DEMS part Starting with potential users, when they go to the supermarket to shop for their favorite products, every purchase transaction is based on the availability of these products. Customers will recycle used products and buy new products after a certain period, such as the product’s lifecycle length or any replenishing time. DEMS is embedded in the statechart of agent behavior, where the state of agent will change to another state when time elapses or by certain
Fig. 2. Hybrid model structure.
82
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Fig. 5. Energy flows in a hybrid model.
Fig. 6. State chart of customer behaviors on two competing alternative products.
Fig. 7. State chart of customer behaviors with five alternative selections.
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
83
rates depending on various variables. The state such as purchase or discard may happen by triggering certain requirements, such as the word-of-mouth effect or the availability of the products. The sequence of discrete events is usually one-way and follows a time sequence, while two-way conversion is comprised of two one-way discrete events. Concerning the agent-based model, a state chart of consumers is used to describe personal behaviors. This model is based on the following assumptions: – All the beverages are always available to the consumers and both marketing and word-of-mouth from other customers are not taken into account. – Each customer is a potential buyer of the five beverages. – At the beginning of each day, each consumer buys one of the five beverages for one day. It means that he will consume six servings of this beverage, on average. – At the end of the day, the consumer becomes a potential buyer again and he will buy a new beverage at the beginning of the next day. – As all the beverages are always available, the consumer sticks to his initial choice everyday. 5.3. Results and validation 5.3.1. Observations The three most important factors affecting the simulation outcomes are the supply chain capability, the customer involvement impact and the market specialty. In the model, environment-sensitive behaviors are considered. There are a certain adjustable percentage of customers who prefer energy-efficient products over alternative choices. The energy efficiency calculation is based on the ratio of each product’s energy consumption over total consumption amount, which leads some people to make a conversion when a more energy-efficient product becomes available. First, the characteristic of the supply chain capability is studied. Fig. 8 shows the market share of each product in different colors. It compares three scenarios, the replenishing period time changes from the longest in scenario (a) to the shortest in scenario (c) while keeping all other parameters constant. The X-axis represents the timeline of simulation. The Y-axis represents the number of customers. The total number of customers reaches 300,000, which is the asymptotic value of the total number of customers who would make the purchase. Customers waiting replenishment are represented in yellow color. They are potential customers in the market but without a decision to purchase any product. The yellow zone shrinks when the replenishing time decreases, indicating that the supply chain is capable of meeting the customer demand in longer replenishment time scenarios. The supply chain, especially the manufacturer, is a bottleneck in terms of production and transportation when it needs to meet customers’ demands faster. Secondly, the impact of customer involvement is evaluated. Each loyal customer is assumed to bring five new customers to a certain kind of product or brand on average, which is called the word-of-mouth effect. When the word-of-mouth effect is not taken into consideration, the amount of potential customers is even larger than customers who made the purchase in a short period of time. There is a big gap between supply and demand (the yellow zone) at a very early stage in scenario (a), when the replenishing time is long. However, the result is quite different when the word-of-mouth effect is introduced in order to make it an active competitive market, even with a long replenishing time (scenario (d)). As the reason for the gap (yellow zone) is mentioned above, the capability of the supply chain limits the ability to satisfy the customers’ needs. This study shows that the word-of-mouth effect contributes to 20–25% of sales. The comparison between scenarios
Fig. 8. Comparison of market share in three different replenishment time scenarios: (a) the longest time, (b) the average time, and (c) the shortest time.
(a) and (d) (see scenario (d) in Fig. 9) indicates that the favorite choice for daily environment-sensitive customers is tap water, given its round the clock availability and energy efficiency. Finally, another important factor, market specialty, is analyzed. When tap water is no longer an option, such as when a natural disaster or water pollution occurs, bottled water, vitamin drink
84
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Fig. 11. Energy consumption per serving changes along as time elapses in scenario 1.
Fig. 9. Scenario (d) : market share comparison as a result of word-of-mouth.
and sport drink become the only choices (scenario (e), with a short replenishing time). The results of the preference for favorite choices are different (see Fig. 10). Regarding tap water, bottled water gets the largest market share proving that bottled water is a safe choice. It becomes the second overall favorite choice considering energy efficiency and waste. Although vitamin drink (in pink) still falls behind bottled water (in red) in market share, it runs ahead of sport drink (in cyan). 5.3.2. Simulation scenario comparison In scenario 1, it is observed that as time elapses, the energy consumption per serving reduces, as shown in Fig. 11. A very slight reduction is seen in the first 4 to 5 years due to reusable glass and aluminum bottles usage, however the reduction speed slows down due to the cumulative amount of other beverage containers production and recycling. A similar shape, with a very slight reduction, could be found in other scenarios (see Figs. 12 and 13). It is found that the energy consumption is high in the fifth scenario due to poor refrigeration efficiency both in energy consumption and GWP. The result of scenario 6 is very close to scenarios 1
Fig. 10. Market share comparison when tap water is unavailable in scenario (e).
and 2. Dishwashing is a component of 60% of all consumed bottled beverages, but these are tap water cases, where dishwashing has a small impact compared to the others. That can be a reason why the difference is very small. 5.3.3. Verification and validation The results obtained from our hybrid model for the three bottled beverages (ecoshape, sport and vitamin) are reasonably close to those reported in a company’s study report (Nestlé Waters North America, 2010), for both Energy and GWP. Results from using the new LCA approach were compared with a company’s study report and the findings show that there exists some value differences on the impact assessments of the tap water and reusable Aluminum bottle cases. The difference of energy consumption and GWP is 32 times and 10 times larger, respectively, between drinking tap water with a glass and Ecoshape bottled water. These values are significant and make a severe impact on the environment. We believe that the difference could be due to uncertainty and to the dynamic environment, which our model could capture but not the standard LCA method. Table 5 lists the LCA problems addressed in this paper and the corresponding verification and validation methods adopted to caliber the hybrid model and make sure it reflects the characteristics of the competitive bottled water market. A Design-Of-Experiment analysis was completed to find out the significant correlations and dependencies. For example, in Fig. 14(a), it was found that the refrigeration energy consumption efficiency plays the most important role in affecting total energy consumption, followed by the frequency of purchase and the recycling rate. The GWP assessment in Fig. 14(b) is almost the same as the energy analysis except that the driving factor changes to refrigeration GWP efficiency. The amount of beverage and corresponding container consumption in Fig. 14(c) also rises along with the frequency of purchase, followed by the loss percentage during transportation and storage and the recycling rate, making the final energy consumption and GWP per serving interesting in Fig. 14(d) and (e) respectively. Refrigeration and loss percentage are two significant factors that need attention in order to bring down energy consumption and GWP (see Fig. 15). The optimization test was also carried out to uncover the best recycling rate, refrigeration energy consumption and emission, loss percentage and frequency of purchase. Using larger containers or reusable containers with a longer lifespan could reduce the GWP of the whole system. The former three factors are linearly correlated with GWP output, while the latter two factors contribute to GWP reduction nonlinearly and with less influence.
85
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Fig. 12. Comparison of simulation results in six different scenarios.
GWP (kg CO2 equivalent)
ENERGY (MJ) Model Value
Model Value
Nestle Report Value
VitaminWa…
Reusable
SportDrink
TapWater
EcoShape
Nestle Report Value
Fig. 13. Comparison of Energy and GWP from the model and from a company’s study report.
Table 5 Traditional LCA problems addressed in the paper and their modeling methods with potential result validation. Traditional LCA problems addressed in the paper
Modeling method
Result verification and validation
Social and economic impacts Alternative scenario considerations Local environmental uniqueness Dynamics of the environment Time horizons Uncertainty in the decision process
ABMS DEMS, SDMS and ABMS DEMS and ABMS SDMS and ABMS DEMS, SDMS and ABMS DEMS, SDMS and ABMS
Customer behaviors, regulation or incentives and competitive market Different local uniqueness, season and preference, customer behaviors Different inputs represent various local environments Customer-driven market and agents make decisions on feedback The trend alters according to different time spans Most parameters and relations have a certain uncertainty range with robust analysis
It was found that the variation of some factors led to a wide range in the environmental impact of water consumption. Communication with customers is very effective toward educating a much more environment-friendly market. Findings match the published LCA report, that is, tap water is the best long-term energy-efficient and environment-friendly choice among all products. Most of the energy consumption for tap water is in dishwashing and glass production. Tap water is not the best choice in short-term studies when the replenishing time is very short. However in reality, after a certain amount of
time, tap water becomes the best choice for people who care about the environment and the energy efficiency. 5.3.4. Summary Bottled water is regarded as a major pollution source given its energy consumption and after-use disposal method such as landfill. Contrary to what one might think, it was found that bottled water, which is a very common choice for users in the US, is the third best choice after tap water with a glass and reusable aluminum bottles, or the second best choice as a beverage source for
86
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Fig. 14. Design-Of-Experiment impact significance output.
the environment compared with vitamin water, juice, sports drinks or similar soft drink products. Water production takes a large portion of the energy consumed for the production of bottled water (larger than the distribution and transportation consumption), while the production of other soft drinks takes extra steps to
produce beverage or packaging and to recycle, which makes them less energy-efficient or environment-friendly, if the recycling rate is the same for all the products except tap water. The word-of-mouth effect is non-negligible and the initial customer inclination is not sensitive according to the market shares
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
87
Fig. 15. Correlation between input factors and GWP per serving.
and the energy use, since the conversion from one product to another along with the customer’s influence is complementarily strong. It was also found that having a regional supply chain helps national or international manufacturers to bring down costs and environmental impacts and also attract more loyal customers.
7. Conclusion and future work Usually, bottled water is not thought of as an environmentalfriendly choice. This hybrid modeling study confirms that the energy consumption and GWP of ecoshape bottled water can be quite larger (32 times and 10 times, respectively) than tap water from a glass bottle. It is also indicated that bottled water is a good thirdbest choice although it is far behind the two best choices of drinking tap water. This is far better than sports drink and vitamin water both in energy efficiency and GWP emission.
Energy usage and greenhouse gas emission are found to be positively correlated, so energy use can be used as the gauge to measure and evaluate efficiency and environmental friendliness. The results of the hybrid DEMS-SDMS-ABMS model are good and promising that not only the results match published reports but also the model can extend further to incorporate customer’s behaviors and market responses. For future work, other factors such as customer behavior comparison, reusable product introduction and model validation can be incorporated into the model to provide more comprehensive results. Washing glasses and bottles plays an important role in the lifecycles, as water needs to be heated regardless whether dishwashing takes place in a dishwasher or in a sink. It would be interesting to compare the reference scenario with one in which the dishwashing water is not heated. Customer behavior comparison is another potential subject of further study. Some customers may choose the least expensive products, while others prefer to
88
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
follow the currently fashionable products. Such customer behaviors would be interesting to model and compare. When additional data becomes available, more quantitative analysis beyond Design-Of-Experiment can be implemented. Products such as sports drinks, coffee, and tea can be added to investigate a more comprehensive beverage market. Cost analysis can also be done if pricing data is available. Parameters such as the transportation distance variance are another consideration for future studies. There are critical values of some parameters that exist which optimize the performance. To minimize the impact on the environment, public advertisement can be effective through the word-of-mouth effect as it is seen in the study. Government’s environmental policies can also be incorporated into future models as they might impact environmental issues by restraining certain types of products and promoting others through various incentives if they are more environmental-friendly and energy efficient. References Azapagic, A. (1999). Life cycle assessment and its application to process selection, design and optimization. Chemical Engineering Journal, 73(1), 1–21. Azoulay, A., Garzon, P., & Eisenberg, M. J. (2001). Comparison of the mineral content of tap water and bottled waters. Journal of General Internal Medicine, 16(3), 168–175. Brezet, H., & Hemel, C. V. (1997). Ecodesign – A promising approach to sustainable production and consumption. United Nations Environment Programme, Paris: Industry and Environment, Cleaner Production. Brezet, H., Stevels, A., & Rombouts, J. (1999). LCA for EcoDesign: The Dutch experience, EcoDesign’ 99: First international symposium on environmentally conscious design and inverse manufacturing. IEEE Computer Society, 36–40. Büsser, S., & Jungbluth, N. (2009). The role of flexible packaging in the life cycle of coffee and butter. International Journal of Life Cycle Assessment, 14(1), 80–91. Doebelin, E. (1998). System Dynamics: Modeling, Analysis, Simulation. Design: CRC Press. Energy Star, 2013. Dishwashers Key Product Criteria, (Retrieved 27.12.13). Forrester, J. W. (1968). Principles of systems. MIT, Cambridge, Mass: Wright-Allen Press. GaBi databases, 2006. PE International GmbH; LBP-GaBi, University of Stuttgart: GaBi Software System. Software or database. Gleick, P. H. (2010). Bottled and sold: The story behind our obsession with bottled water. Island Press.
Guinée, J. B. et al. (1993). Quantitative life cycle assessment of products: 2 Classification, valuation and improvement analysis. Journal of Cleaner Production, 1(2), 81–91. Humbert, S., et al. 2009. A case study on multi-impact life cycle assessment of coffee alternatives. Quantis study paper. ISO 14040, 1997. American National Standard. Environmental management, Life cycle assessment, Principles and framework. Keoleian, G., Bulkley, J., & Dettore, C., 2009. Comparative life-cycle assessment of bottled versus tap water systems. Deep Blue Library-University of Michigan at Ann Arbor. Lee, S. G., & Xu, X. (2005). Design for the environment: Life cycle assessment and sustainable packaging issues. International Journal of Environmental Technology and Management, 5(1), 14–41. Mayyas, A. T. et al. (2012). Life cycle assessment-based selection for a sustainable lightweight body-in-white design. Energy, 39(1), 412–425. Nestlé Waters North America, 2010. Environmental Life Cycle Assessment of Drinking Water Alternatives & Consumer Beverage Consumption in North America. Research performed for NWNA. (Accessed 20.02.12). North, M. J., & Macal, C. M. (2007). Managing business complexity: Discovering strategic solutions with agent-based modeling and simulation (1st ed.). USA: Oxford University Press. Pasqualino, J., Meneses, M., & Castells, F. (2011). The carbon footprint and energy consumption of beverage packaging selection and disposal. Journal of Food Engineering, 103, 357–365. Powell, J. (2000). The potential for using life cycle inventory analysis in local authority waste management decision making. Journal of Environmental Planning and Management, 43(3), 351–367. Rahimifard, S., & Clegg, A. J. (2007). Aspects of sustainable design and manufacture. International Journal of Production Research, 45(18), 4013–4019. Reap, J. et al. (2008). A survey of unresolved problems in life cycle assessment. The International Journal of Life Cycle Assessment, 13, 374–388. Stasinopoulos, P. et al. (2012). A system dynamics approach in LCA to account for temporal effects-a consequential energy LCI of car body-in-whites. The International Journal of Life Cycle Assessment, 17, 199–207. Sterman, J. D. (2001). System dynamics modeling: Tools for learning in a complex world. California Management Review, 43(4), 8–25. THE 2006 STATS, Source. Commentary and content provided by Beverage Marketing Corporation. Van den Heede, P., & De Belie, N. (2012). Environmental impact and life cycle assessment (LCA) of traditional and ‘green’ concretes: Literature review and theoretical calculations. Cement and Concrete Composites, 34(4), 431–442. Vellini, M., & Savioli, M. (2009). Energy and environmental analysis of glass container production and recycling. Energy, 34(12), 2137–2143. Zeigler, B. P., Kim, T. G., & Praehofer, H. (2000). Theory of modeling and simulation: integrating discrete event and continuous complex dynamic systems (2nd ed.). Academic Press.
Contents lists available at ScienceDirect
Computers & Industrial Engineering journal homepage: www.elsevier.com/locate/caie
Hybrid modeling and simulation for complementing Lifecycle Assessment Bochao Wang a, Séverin Brême b,⇑, Young B. Moon a a b
Department of Mechanical and Aerospace Engineering, Syracuse University, Syracuse, NY 13244, USA Institut Supérieur de l’Aéronautique et de l’Espace (ISAE), Toulouse, France
a r t i c l e
i n f o
Article history: Received 22 July 2013 Received in revised form 27 December 2013 Accepted 30 December 2013 Available online 7 January 2014 Keywords: Sustainability Lifecycle assessment (LCA) Agent-based modeling System dynamics Discrete-event simulation
a b s t r a c t This paper presents a new complementary lifecycle assessment (LCA) approach to address several limitations of the standard LCA methodology. An integrated approach of agent-based modeling, system dynamics and discrete event simulation was adopted to complement the standard LCA methodology. A hybrid simulation model was developed as a proof-of-concept system, then it was validated using a case study of bottled water and alternative drink products. The model was based on the assumption that parameters and relationships were constant regardless of local uniqueness. The research demonstrates that the hybrid modeling and simulation method can address several limitations of the standard LCA. Also, it is also proven that the method has a potential to address social and economic aspects. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction As energy consumption and pollution become critical issues worldwide, people are more and more concerned with sustainability issues through various attempts ranging from creating environmental-friendly products to changing habits to reduce waste. Governments and corporations also make numerous efforts such as managing energy usage, waste, emission, etc. While such initiatives deserve commendations, there is a danger of focusing on local improvement only thus unintentionally worsening the situation unless a holistic systems thinking guides those executions. One of the useful tools in evaluating overall environmental impacts throughout a product’s entire life is lifecycle assessment (LCA). LCA is the ‘‘compilation and evaluation of the inputs, outputs and the potential environmental impacts of a product system throughout its life cycle’’ (ISO 14040, 1997). In a LCA study, many aspects throughout a product’s lifecycle can be considered including production, transportation, distribution, usage and end-of-life activities. LCA has been used to identify potential opportunities for improvement such as in better design, better manufacturing processes and better management in order to minimize negative sustainability impacts (Brezet & Hemel, 1997). LCA is a unique tool that comprehensively examines the environmental impacts of a product or service throughout its life ⇑ Corresponding author. E-mail addresses: [email protected] (B. Wang), [email protected] (S. Brême), [email protected] (Y.B. Moon). 0360-8352/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.cie.2013.12.016
cycle. It is an ISO standardized method and proved to be a useful aid in decision-making for designers, managers, government and consumers (Brezet, Stevels, & Rombouts, 1999; Rahimifard & Clegg, 2007). However, the standard LCA method has a number of limitations and some of these problems can be critical (Reap et al., 2008). One of them is that the LCA method takes a static viewpoint that its parameters and internal relations among entities remain constant. Also, the social and economic impact, local environmental uniqueness, effects of dynamic environment, and temporal perspectives cannot be easily considered in the LCA. In other words, the standard LCA is useful as a high level tool, but not necessarily for dealing with dynamics and uncertainties. This research aims to demonstrate that a hybrid simulation model can address these limitations of the standard LCA approach. All the steps in LCA – goal and scope definition, inventory analysis, and impact assessment, including interpretation – are considered in our research. However, the focus is placed on the third step of the LCA, that is, impact assessment. A hybrid simulation model combining agent-based modeling, system dynamics and discreteevent simulation methods was developed as a proof-of-concept system. The validity of the developed approach was done on comparing bottled water alternatives such as tap water and vitamin water along with different bottle options. The paper is organized as follows. First, limitations of traditional LCA approach will be addressed, followed by explanation of an integrated hybrid modeling and simulation approach. Drinking water and beverages are chosen to illustrate how the framework is developed and modeled. Impact analysis based on the simulation
78
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
results is explained to show the value and potential of the new complementary LCA approach. Conclusion and future work is provided at the end. 2. Limitations of standard Lifecycle Assessment (LCA) The standard LCA method has been adopted to address the needs of research and practical analyses for a long time (Azapagic, 1999; Guinée et al., 1993; Van den Heede & De Belie, 2012). It is useful because it constructs a grounded structure and procedure to analyze an overall environmental impact that a product can make during its whole life cycle. However, its advantages through adopting a systematic viewpoint (Forrester, 1968) became its limitations because many systems of interest are never static, but evolving. Stasinopoulos et al. (2012) points out one kernel limitation of standard LCA method is that its life cycle inventory studies assume that parameters used are always constant or fixed functions of time which prevents taking temporal effects and dynamic response of the system into consideration. The report that the simplified assumption made in estimating energy benefits is not reasonable later. In comprehensive discussion of unresolved problems in LCA, Reap et al. (2008) explains the reason and progress made to overcome numerous disadvantages, and classifies those critical problems and difficulties into several categories. Categories are also rated for severity and adequacy of traditional LCA solutions. Powell criticizes the laggard traditional inventory analysis and local environmental uniqueness with impacts that could potentially lead to a faulty decision-making process (Powell, 2000). Mayyas et al. (2012) points out the disadvantages of static point of view, not only on the system’s boundary selection and inventory, but also on the data and time horizons. He further uses an example on how to make a sustainable lightweight body-in-white design with improved quick solutions to the standard LCA approach. The uncertainties such as in data and function are also overlooked in the traditional LCA study. However, these uncertainties can play important roles in affecting outputs and estimated environmental impact severity and thus impact decision making for enterprises and governments. While these are all valid critiques and points for improvement for the standard LCA method to be addressed eventually, we address only a few of these problems in our research, that is, social and economic impacts, alternative scenario considerations, local environmental uniqueness and dynamics and time horizons. Table 1 highlights what this research addresses in the list of problems that Reap and his colleagues pointed out (Reap et al., 2008). 3. Integrated hybrid modeling and simulation method Three commonly used systems modeling and simulation methods are Discrete Event Modeling and Simulation (DEMS), System Dynamics Modeling and Simulation (SDMS) and Agent-based Modeling and Simulation (ABMS). These are integrated to simulate the life cycle process and study short-term and long-term performance under various scenarios. The new approach has potential for building a unique model that can combine each model’s uniqueness and advantages into one model and take their differences into consideration. Discrete Event Modeling and Simulation (DEMS) can simulate multiple events in a time sequence (Zeigler, Kim, & Praehofer, 2000). They are built in the form of entities, flowcharts and resources. DEMS is a natural choice when linear processes in a complex environment is modeled and an entity’s action is triggered by other entities or a certain time. Many service facilities, production systems, maintenance and recycling facilities, and transportation
and material handling systems are best described and simulated via DEMS. System Dynamics Modeling and Simulation (SDMS) is a methodology used to model and simulate a system from a higher system-level viewpoint (Doebelin, 1998; Sterman, 2001). It comprises stocks, flows and unique feedback loops. The state of the whole system could be observed from various stocks at a given time. Aggregates are linked through aggregated mechanisms implemented as flows in SDMS. Stock and flow are natural choice for modeling beverage inventory and production flow. Feedback loops link each module of the system with defined relations and influence. Agent-based modeling and simulation (ABMS) is a methodology used to model and simulate individual actions and interactions of agents in a complex adaptive system, focusing on their effects on the system as a whole (North & Macal, 2007). ABMS are also called individual based models, due to their bottom-up individual-level modeling approach. They are constructed in the form of active objects, individual behavior rules, and direct or indirect interaction within a dynamic environment. All of these elements can be used to represent agents and their interactive operations in an environment such as a competitive beverage sale market. Therefore, it can be used to bridge marketing and engineering activities. It also suits complex and flexible situations that need to be modeled, especially when taking customers, retailers and competitors into consideration. These three methodologies can be combined into an integrated hybrid modeling and simulation method to complement each methodology’s respective strengths. The most promising part of the integrated modeling approach is its flexibility that can handle dynamic and evolving requirements of a system. SDMS can deal with aggregates at the highest abstraction level while DEMS can be used at middle levels of abstraction and possibly at lower levels as well. ABMS can be used across all levels of abstraction. In our research, we utilized the flexibility of the integrated hybrid modeling and simulation and developed a proof-of-concept system that can complement the standard LCA method.
4. Lifecycle Assessment of Drinking Water and Beverages Tap water is still one of the major drinking sources in daily life. However, the bottled water market in developed countries such as the United States and Japan has grown rapidly. A report from the Beverage Marketing Corporation (THE 2006 STATS) reveals that there was a 50% growth in America’s bottled water consumption between 2002 and 2007. For health, quality and convenience, bottled water has become a popular choice in the drinking water and beverage market. The rise in popularity of bottled water has created a burden on its sustainability (Gleick, 2010), however. For example, bottled water produces waste by product in production, transportation, distribution, refrigeration and recycling. Also, many other drinking alternatives (for example, vitamin water) have become available in the market recently. So the basic questions are: (i) Which one of the available options (e.g. tap water, bottled water, and other drinking alternatives) is the most sustainable? (ii) Will consumers’ choice make a difference? From the related important aspect of packaging, Lee and Xu (2005) addressed sustainable packaging issues in general. Büsser and Jungbluth (2009) studied the role of flexible packaging in the lifecycle of coffee. Vellini and Savioli (2009) focused particularly on glass containers from production to recycling. All of these reveal that the packaging issue is a significant one in evaluating
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
79
Table 1 LCA Problems (Reap et al., 2008) addressed by this paper and potentially by the hybrid modeling approach. Traditional LCA problems Functional unit definition Boundary selection Social and economic impacts Alternative scenario considerations Allocation Negligible contribution criteria Local technical uniqueness Impact category selection Spatial variation Local environmental uniqueness Dynamics of the environment Time horizons Weighting and valuation Uncertainty in the decision process Data availability and quality
Problems addressed in this paper
Potential solvable problems p
p p p p p p p p p p p
Fig. 1. The boundary of the system and the stages of sport drink systems.
environmental impacts. We incorporate all the important elements and steps in assessing drinking water alternatives with the idea that multiple impacts are interrelated (Humbert et al., 2009). LCA consists of three distinct steps with associated interpretation of results. The three steps are: (i) goal and scope definition, (ii) inventory analysis, and (iii) assessment of impacts associated with these inputs and outputs.
4.1. Goal and scope The goal of our study is to measure the environmental impacts of beverage consumption habits under different scenarios. In this study, the critical environmental issues and responsibilities were identified along the entire lifecycle chain of five specific drinking alternatives: 1. 2. 3. 4. 5.
Tap water in a glass bottle, Tap water in a reusable aluminum bottle, Ecoshape bottled water, Sport drink, and Vitamin water.
Consumer behavior is also taken into consideration. Consumers are allowed to choose freely from drinking alternatives and switch among them over time to reflect the trend of consumers. Two particular impacts, energy consumption and global warming potential, are assessed to reveal the practicality of the new methodology.
4.2.2. Proper functional unit and emission data A reasonable water/beverage consumption amount is chosen as 3 liters (L) per day per person. This is equivalent to 6 bottled water volumes or vitamin water bottles. Data for material/energy consumption, water usage, waste generation, greenhouse gases emissions, distribution selection and quantitative relations between entities and parameters are collected from a company’s study report (Nestlé Waters North America, 2010), GaBi databases (LBP, 2006) and published LCA papers (Azoulay, Garzon, & Eisenberg, 2001; Keoleian, Bulkley, & Dettore, 2009). 4.3. Assumptions The whole life cycle stages from beverage and packaging production to end-of-life are considered and summarized in Table 2. In order to effectively reflect and compare the results of each product in different scenarios, the two most commonly used and important factors are chosen: (i) Energy is the amount of energy used during each phase of the lifecycle (Pasqualino, Meneses, & Castells, 2011). (ii) Global-warming potential (GWP) is a relative measure of how much heat a greenhouse gas traps in the atmosphere. It compares the amount of heat trapped by a certain mass of the gas in question to the amount of heat trapped by a similar mass of carbon dioxide. GWP is expressed as a factor of carbon dioxide (whose GWP is standardized to 1). Additional assumptions are made to simplify the model constraints and set a reference standard to start with:
4.2. Inventory analysis 4.2.1. Product system boundary The whole lifecycle of drinking alternatives will be covered. For the sport drink as an example, its lifecycle ranges from beverage production, package production, transportation, distribution, refrigeration to recycling or landfilling disposal. The boundary of the system and stages are shown in Fig. 1. There are five stages in the sub-model of sport drink.
(i) It is assumed that each consumer consumes 6 servings a day. Each serving is in a 500 mL bottle or glass, which is the functional unit used. (ii) For containers purchases, it is assumed that, on average, (a) for tap water in glass bottles, one glass is bought every 4 months; (b) for tap water in reusable aluminum bottles, one reusable bottle is bought every 1000 days;
80
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Table 2 Life cycle stages and descriptions. Stage
Description
Beverage production
This phase includes the pumping and the treatment of tap water at a municipal water plant. It also includes processes, extraction and transportation of all materials that are needed to transform water into the final beverage This phase includes the extraction and transportation of raw materials and the production of the container (glass, aluminum or plastic bottle)
Container and packaging production Distribution and marketplace
Transportation
Transportation of the containers and bottled beverages from the production plant to a supermarket, with a truck
Refrigeration Marketing Storage
Refrigeration of the bottled beverages at a supermarket, in a refrigerator Activities related to marketing the containers and the bottled beverages Storage of the containers and the bottled beverages at a supermarket
Consumer use
Transportation Refrigeration Dishwashing
Transportation of the containers and bottled beverages from the supermarket to the user’s home, in the user’s car Refrigeration of the beverages at the user’s home in a refrigerator Dishwashing of the reusable container (glass or aluminum bottle) at the user’s home in a dishwasher. Frequency of dishwashing is different for glasses and reusable bottles
End-of-life
Hauling Recycling Landfilling Waste-toenergy
Hauling of the discarded empty containers to a landfill or to a recycling facility Recycling of the empty containers at a recycling facility Landfilling of the empty containers Conversion of the empty containers into energy
Table 3 The beverage selection percentage by customers according to the six scenarios.
Table 4 Value comparison of the simulation result and a company’s study report.
Scenarios
Tap water Ecoshape bottle Reusable aluminum bottle Sport drink Vitamin drink Total (%)
1 and 2
3
4
5 and 6
40 36 20 2 2 100
0 96 0 2 2 100
57 0 35 4 4 100
40 36 20 2 2 100
(c) for the other drinks, no additional container is purchased, that is, the beverage is drunk directly from the bottle. (iii) It is assumed that all beverages except tap water are refrigerated. (iv) For washing, it is assumed that (a) for tap water in glass bottles, the consumer washes his/her glass bottle 3 times a day (for 6 servings a day) (b) for tap water in reusable aluminum bottles, consumer washes his/her bottle every 4 days (or 0.25 times a day) and washing takes place in a dishwasher. (v) For energy consumption calculation, it is assumed that an average US dishwasher consumes 2 kW h per load (Energy Star, 2013), so (a) a glass bottle is assumed to represent 1/44 of a load (2/44 kW h= 0.045 kW h = 162.551 kJ) (b) an aluminum bottle is assumed to represent 1/22 of a load (2/22 kW h= 0.09 kW h = 325.102 kJ).
Model results
A company’s study report results
ENERGY (MJ)
TapWater EcoShape Reusable Sport Drink VitaminWater
0.089 2.854 0.049 4.972 5.049
0.14 3 Not available 5 5
GWP (kg CO2 eq)
TapWater
0.014
0.01
EcoShape Reusable SportDrink VitaminWater
0.147 0.007 0.266 0.288
0.15 0.06 0.2475 0.2565
The energy used for the refrigeration of one serving is 0.00907 kW h = 32.763 kJ. 5. Modeling and simulation System dynamics is used to model the workflow and calculate the energy consumption and GWP for each step continuously. Since the overall process is a continuous procedure in straightforward workflow which involves plenty of feedbacks such as material flow and energy usage, system dynamics is used because it is best suited to analyze a system with dynamic stocks, flows and feedbacks. Beyond that, the detailed work processing procedure and the connection linking consumer behaviors with actions are described via discrete-event and agent-based methods (see Tables 3 and 4). 5.1. Scenarios
The reference standard is set using the data from a company’s study report: Volume: 615 L. Energy consumption of the refrigerator: 422 kW h/year = 1.156164385 kW h/day for 615 L. Room occupied: 2 L of the refrigerator for a 0.5 L serving. Energy consumption for the refrigeration of one serving: 0.00375988 kW h per day. Refrigeration duration: 2.4 days.
Scenario 1 is the ‘‘reference scenario’’. This scenario represents the base pattern of beverage consumption in New York State assuming that only the five kinds of beverages are available. All beverages are refrigerated for 2.4 days on average, except tap water, which is not refrigerated. The glasses used for tap water and the reusable bottles are washed in a dishwasher. Scenario 2 is similar to the reference scenario, but during the winter it is assumed that no beverage is refrigerated by the consumer.
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Scenario 3: In this scenario, tap water is somehow unavailable due to pollution. ‘‘Tap water in glass bottle’’ and ‘‘tap water in reusable aluminum bottle’’ are then replaced by ‘‘ecoshape bottled water’’. All drinks are refrigerated. Scenario 4: Under this scenario, bottled water is banned. Scenario 5: Same as the reference scenario, but refrigeration conditions are different. It is assumed that refrigeration takes place for 7.2 days instead of 2.4, in a 20-year-old refrigerator that consumes three times more energy. Furthermore, it is assumed that the beverage occupies about 1/10 of its refrigerator content. Scenario 6: Same as the reference scenario, but glasses and bottles are washed by hand with cold water.
81
Fig. 3. Bottled water production feedbacks in system dynamics.
5.2. Framework and modeling 5.2.1. Overall hybrid modeling framework The model is developed in two main parts: (i) modeling the lifecycle of each beverage and (ii) modeling the behavior of each consumer of the population. The lifecycles of the five beverages are first modeled in SDMS, followed by the consumer behavior modeled in ABMS and DEMS. The integrated model is then established in order to compare the environmental impacts of beverage consumption under the different scenarios. The structure of the hybrid model is presented in Fig. 2. 5.2.2. SDMS part Bottles, glasses and cardboard used to produce bottled water can be recycled into new containers as a feedback system illustrates in Fig. 3. This increases the production rates of beverages and containers. Four kinds of raw materials are needed to manufacture bottles and their packaging: packaging cardboard, PPlid, PET Resin, and Pallet. The production rate of each beverage and each container is based on average consumer consumption (6 servings a day for one consumer) and on averaged losses (some bottles and glasses are stolen or damaged in the supermarkets or during manufacturing and transportation). (See Figs. 4–7). After the production stage, the beverages are transported to the marketplace, which is a supermarket. There, bottled beverages are refrigerated before being sold to the consumers. After people drink the beverages, there are three ways to dispose of all materials: landfilling, waste-to-energy or recycling. The recycle rates depend on the container used. For tap water, the SDMS diagram is composed of two independent SDMS diagrams: one for tap water and the other for the containers. Tap water is processed and generated at a municipal water plant, then distributed through pipes to the consumers. The tap water consumption is determined by adding: (i) the tap water that is drunk by the consumer, (ii) the tap water that is used for dishwashing and, (iii) the losses during the production and
Fig. 4. Distribution and market stages of the bottled water system.
distribution processes. The production rate of tap water is determined to meet customer’s demand. On the other hand, glass bottles are produced from raw materials and reusable aluminum bottles are from recycled materials. Their production rates are chosen with respect to consumer demand. They are then distributed to the marketplace, bought by consumers, used and then discarded. For the bottled beverages (ecoshape bottle, sport drink and vitamin water), the SDMS diagram is similar to the one of tap water.
5.2.3. ABMS and DEMS part Starting with potential users, when they go to the supermarket to shop for their favorite products, every purchase transaction is based on the availability of these products. Customers will recycle used products and buy new products after a certain period, such as the product’s lifecycle length or any replenishing time. DEMS is embedded in the statechart of agent behavior, where the state of agent will change to another state when time elapses or by certain
Fig. 2. Hybrid model structure.
82
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Fig. 5. Energy flows in a hybrid model.
Fig. 6. State chart of customer behaviors on two competing alternative products.
Fig. 7. State chart of customer behaviors with five alternative selections.
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
83
rates depending on various variables. The state such as purchase or discard may happen by triggering certain requirements, such as the word-of-mouth effect or the availability of the products. The sequence of discrete events is usually one-way and follows a time sequence, while two-way conversion is comprised of two one-way discrete events. Concerning the agent-based model, a state chart of consumers is used to describe personal behaviors. This model is based on the following assumptions: – All the beverages are always available to the consumers and both marketing and word-of-mouth from other customers are not taken into account. – Each customer is a potential buyer of the five beverages. – At the beginning of each day, each consumer buys one of the five beverages for one day. It means that he will consume six servings of this beverage, on average. – At the end of the day, the consumer becomes a potential buyer again and he will buy a new beverage at the beginning of the next day. – As all the beverages are always available, the consumer sticks to his initial choice everyday. 5.3. Results and validation 5.3.1. Observations The three most important factors affecting the simulation outcomes are the supply chain capability, the customer involvement impact and the market specialty. In the model, environment-sensitive behaviors are considered. There are a certain adjustable percentage of customers who prefer energy-efficient products over alternative choices. The energy efficiency calculation is based on the ratio of each product’s energy consumption over total consumption amount, which leads some people to make a conversion when a more energy-efficient product becomes available. First, the characteristic of the supply chain capability is studied. Fig. 8 shows the market share of each product in different colors. It compares three scenarios, the replenishing period time changes from the longest in scenario (a) to the shortest in scenario (c) while keeping all other parameters constant. The X-axis represents the timeline of simulation. The Y-axis represents the number of customers. The total number of customers reaches 300,000, which is the asymptotic value of the total number of customers who would make the purchase. Customers waiting replenishment are represented in yellow color. They are potential customers in the market but without a decision to purchase any product. The yellow zone shrinks when the replenishing time decreases, indicating that the supply chain is capable of meeting the customer demand in longer replenishment time scenarios. The supply chain, especially the manufacturer, is a bottleneck in terms of production and transportation when it needs to meet customers’ demands faster. Secondly, the impact of customer involvement is evaluated. Each loyal customer is assumed to bring five new customers to a certain kind of product or brand on average, which is called the word-of-mouth effect. When the word-of-mouth effect is not taken into consideration, the amount of potential customers is even larger than customers who made the purchase in a short period of time. There is a big gap between supply and demand (the yellow zone) at a very early stage in scenario (a), when the replenishing time is long. However, the result is quite different when the word-of-mouth effect is introduced in order to make it an active competitive market, even with a long replenishing time (scenario (d)). As the reason for the gap (yellow zone) is mentioned above, the capability of the supply chain limits the ability to satisfy the customers’ needs. This study shows that the word-of-mouth effect contributes to 20–25% of sales. The comparison between scenarios
Fig. 8. Comparison of market share in three different replenishment time scenarios: (a) the longest time, (b) the average time, and (c) the shortest time.
(a) and (d) (see scenario (d) in Fig. 9) indicates that the favorite choice for daily environment-sensitive customers is tap water, given its round the clock availability and energy efficiency. Finally, another important factor, market specialty, is analyzed. When tap water is no longer an option, such as when a natural disaster or water pollution occurs, bottled water, vitamin drink
84
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Fig. 11. Energy consumption per serving changes along as time elapses in scenario 1.
Fig. 9. Scenario (d) : market share comparison as a result of word-of-mouth.
and sport drink become the only choices (scenario (e), with a short replenishing time). The results of the preference for favorite choices are different (see Fig. 10). Regarding tap water, bottled water gets the largest market share proving that bottled water is a safe choice. It becomes the second overall favorite choice considering energy efficiency and waste. Although vitamin drink (in pink) still falls behind bottled water (in red) in market share, it runs ahead of sport drink (in cyan). 5.3.2. Simulation scenario comparison In scenario 1, it is observed that as time elapses, the energy consumption per serving reduces, as shown in Fig. 11. A very slight reduction is seen in the first 4 to 5 years due to reusable glass and aluminum bottles usage, however the reduction speed slows down due to the cumulative amount of other beverage containers production and recycling. A similar shape, with a very slight reduction, could be found in other scenarios (see Figs. 12 and 13). It is found that the energy consumption is high in the fifth scenario due to poor refrigeration efficiency both in energy consumption and GWP. The result of scenario 6 is very close to scenarios 1
Fig. 10. Market share comparison when tap water is unavailable in scenario (e).
and 2. Dishwashing is a component of 60% of all consumed bottled beverages, but these are tap water cases, where dishwashing has a small impact compared to the others. That can be a reason why the difference is very small. 5.3.3. Verification and validation The results obtained from our hybrid model for the three bottled beverages (ecoshape, sport and vitamin) are reasonably close to those reported in a company’s study report (Nestlé Waters North America, 2010), for both Energy and GWP. Results from using the new LCA approach were compared with a company’s study report and the findings show that there exists some value differences on the impact assessments of the tap water and reusable Aluminum bottle cases. The difference of energy consumption and GWP is 32 times and 10 times larger, respectively, between drinking tap water with a glass and Ecoshape bottled water. These values are significant and make a severe impact on the environment. We believe that the difference could be due to uncertainty and to the dynamic environment, which our model could capture but not the standard LCA method. Table 5 lists the LCA problems addressed in this paper and the corresponding verification and validation methods adopted to caliber the hybrid model and make sure it reflects the characteristics of the competitive bottled water market. A Design-Of-Experiment analysis was completed to find out the significant correlations and dependencies. For example, in Fig. 14(a), it was found that the refrigeration energy consumption efficiency plays the most important role in affecting total energy consumption, followed by the frequency of purchase and the recycling rate. The GWP assessment in Fig. 14(b) is almost the same as the energy analysis except that the driving factor changes to refrigeration GWP efficiency. The amount of beverage and corresponding container consumption in Fig. 14(c) also rises along with the frequency of purchase, followed by the loss percentage during transportation and storage and the recycling rate, making the final energy consumption and GWP per serving interesting in Fig. 14(d) and (e) respectively. Refrigeration and loss percentage are two significant factors that need attention in order to bring down energy consumption and GWP (see Fig. 15). The optimization test was also carried out to uncover the best recycling rate, refrigeration energy consumption and emission, loss percentage and frequency of purchase. Using larger containers or reusable containers with a longer lifespan could reduce the GWP of the whole system. The former three factors are linearly correlated with GWP output, while the latter two factors contribute to GWP reduction nonlinearly and with less influence.
85
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Fig. 12. Comparison of simulation results in six different scenarios.
GWP (kg CO2 equivalent)
ENERGY (MJ) Model Value
Model Value
Nestle Report Value
VitaminWa…
Reusable
SportDrink
TapWater
EcoShape
Nestle Report Value
Fig. 13. Comparison of Energy and GWP from the model and from a company’s study report.
Table 5 Traditional LCA problems addressed in the paper and their modeling methods with potential result validation. Traditional LCA problems addressed in the paper
Modeling method
Result verification and validation
Social and economic impacts Alternative scenario considerations Local environmental uniqueness Dynamics of the environment Time horizons Uncertainty in the decision process
ABMS DEMS, SDMS and ABMS DEMS and ABMS SDMS and ABMS DEMS, SDMS and ABMS DEMS, SDMS and ABMS
Customer behaviors, regulation or incentives and competitive market Different local uniqueness, season and preference, customer behaviors Different inputs represent various local environments Customer-driven market and agents make decisions on feedback The trend alters according to different time spans Most parameters and relations have a certain uncertainty range with robust analysis
It was found that the variation of some factors led to a wide range in the environmental impact of water consumption. Communication with customers is very effective toward educating a much more environment-friendly market. Findings match the published LCA report, that is, tap water is the best long-term energy-efficient and environment-friendly choice among all products. Most of the energy consumption for tap water is in dishwashing and glass production. Tap water is not the best choice in short-term studies when the replenishing time is very short. However in reality, after a certain amount of
time, tap water becomes the best choice for people who care about the environment and the energy efficiency. 5.3.4. Summary Bottled water is regarded as a major pollution source given its energy consumption and after-use disposal method such as landfill. Contrary to what one might think, it was found that bottled water, which is a very common choice for users in the US, is the third best choice after tap water with a glass and reusable aluminum bottles, or the second best choice as a beverage source for
86
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
Fig. 14. Design-Of-Experiment impact significance output.
the environment compared with vitamin water, juice, sports drinks or similar soft drink products. Water production takes a large portion of the energy consumed for the production of bottled water (larger than the distribution and transportation consumption), while the production of other soft drinks takes extra steps to
produce beverage or packaging and to recycle, which makes them less energy-efficient or environment-friendly, if the recycling rate is the same for all the products except tap water. The word-of-mouth effect is non-negligible and the initial customer inclination is not sensitive according to the market shares
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
87
Fig. 15. Correlation between input factors and GWP per serving.
and the energy use, since the conversion from one product to another along with the customer’s influence is complementarily strong. It was also found that having a regional supply chain helps national or international manufacturers to bring down costs and environmental impacts and also attract more loyal customers.
7. Conclusion and future work Usually, bottled water is not thought of as an environmentalfriendly choice. This hybrid modeling study confirms that the energy consumption and GWP of ecoshape bottled water can be quite larger (32 times and 10 times, respectively) than tap water from a glass bottle. It is also indicated that bottled water is a good thirdbest choice although it is far behind the two best choices of drinking tap water. This is far better than sports drink and vitamin water both in energy efficiency and GWP emission.
Energy usage and greenhouse gas emission are found to be positively correlated, so energy use can be used as the gauge to measure and evaluate efficiency and environmental friendliness. The results of the hybrid DEMS-SDMS-ABMS model are good and promising that not only the results match published reports but also the model can extend further to incorporate customer’s behaviors and market responses. For future work, other factors such as customer behavior comparison, reusable product introduction and model validation can be incorporated into the model to provide more comprehensive results. Washing glasses and bottles plays an important role in the lifecycles, as water needs to be heated regardless whether dishwashing takes place in a dishwasher or in a sink. It would be interesting to compare the reference scenario with one in which the dishwashing water is not heated. Customer behavior comparison is another potential subject of further study. Some customers may choose the least expensive products, while others prefer to
88
B. Wang et al. / Computers & Industrial Engineering 69 (2014) 77–88
follow the currently fashionable products. Such customer behaviors would be interesting to model and compare. When additional data becomes available, more quantitative analysis beyond Design-Of-Experiment can be implemented. Products such as sports drinks, coffee, and tea can be added to investigate a more comprehensive beverage market. Cost analysis can also be done if pricing data is available. Parameters such as the transportation distance variance are another consideration for future studies. There are critical values of some parameters that exist which optimize the performance. To minimize the impact on the environment, public advertisement can be effective through the word-of-mouth effect as it is seen in the study. Government’s environmental policies can also be incorporated into future models as they might impact environmental issues by restraining certain types of products and promoting others through various incentives if they are more environmental-friendly and energy efficient. References Azapagic, A. (1999). Life cycle assessment and its application to process selection, design and optimization. Chemical Engineering Journal, 73(1), 1–21. Azoulay, A., Garzon, P., & Eisenberg, M. J. (2001). Comparison of the mineral content of tap water and bottled waters. Journal of General Internal Medicine, 16(3), 168–175. Brezet, H., & Hemel, C. V. (1997). Ecodesign – A promising approach to sustainable production and consumption. United Nations Environment Programme, Paris: Industry and Environment, Cleaner Production. Brezet, H., Stevels, A., & Rombouts, J. (1999). LCA for EcoDesign: The Dutch experience, EcoDesign’ 99: First international symposium on environmentally conscious design and inverse manufacturing. IEEE Computer Society, 36–40. Büsser, S., & Jungbluth, N. (2009). The role of flexible packaging in the life cycle of coffee and butter. International Journal of Life Cycle Assessment, 14(1), 80–91. Doebelin, E. (1998). System Dynamics: Modeling, Analysis, Simulation. Design: CRC Press. Energy Star, 2013. Dishwashers Key Product Criteria,
Guinée, J. B. et al. (1993). Quantitative life cycle assessment of products: 2 Classification, valuation and improvement analysis. Journal of Cleaner Production, 1(2), 81–91. Humbert, S., et al. 2009. A case study on multi-impact life cycle assessment of coffee alternatives. Quantis study paper. ISO 14040, 1997. American National Standard. Environmental management, Life cycle assessment, Principles and framework. Keoleian, G., Bulkley, J., & Dettore, C., 2009. Comparative life-cycle assessment of bottled versus tap water systems. Deep Blue Library-University of Michigan at Ann Arbor. Lee, S. G., & Xu, X. (2005). Design for the environment: Life cycle assessment and sustainable packaging issues. International Journal of Environmental Technology and Management, 5(1), 14–41. Mayyas, A. T. et al. (2012). Life cycle assessment-based selection for a sustainable lightweight body-in-white design. Energy, 39(1), 412–425. Nestlé Waters North America, 2010. Environmental Life Cycle Assessment of Drinking Water Alternatives & Consumer Beverage Consumption in North America. Research performed for NWNA.