Elsevier

Waste Management

Volume 18, Issue 2, April 1998, Pages 87-97
Waste Management

A prototype knowledge-based decision support system for industrial waste management: part I. The decision support system1

https://doi.org/10.1016/S0956-053X(97)10025-3Get rights and content

Abstract

Although there are a number of expert systems available which are designed to assist in resolving environmental problems, there is still a need for a system which would assist managers in determining waste management options for all types of wastes from one or more industrial plants, giving priority to sustainable use of resources, reuse and recycling. A prototype model was developed to determine the potentials for reuse and recycling of waste materials, to select the treatments needed to recycle waste materials or for treatment before disposal, and to determine potentials for co-treatment of wastes. A knowledge-based decision support system was then designed using this model. This paper describes the prototype model, the developed knowledge-based decision support system, the input and storage of data within the system and the inference engine developed for the system to determine the treatment options for the wastes. Options for sorting and selecting treatment trains are described, along with a discussion of the limitations of the approach and future developments needed for the system.

Introduction

Industrial ecology is a recent concept in engineering and management and is an attempt to manage an industrial estate as an ecosystem, with feedback loops and minimal use of resources and production of wastes[1]. Life cycle assessment, design for the environment, cleaner production technologies and sustainable waste management practices are all important aspects of industrial ecology. The concept of industrial ecology will develop industries which minimize use of resources and production of wastes, from resource extraction to manufacturing and to final consumer use and recycling of the product.

One of the more intricate implementations of industrial ecology lies in the establishment of industrial ecosystems, where an industrial estate operates as an ecosystem, with residual energy and materials being traded or sold between plants or co-treated for ultimate disposal. Under this system, industrial plants individually implement waste minimization programs. For the wastes which cannot be economically managed by individual plants, the estate as a whole looks for the potential for on-site reuse, recycling, recovery or co-treatment for disposal. A few such industrial ecosystems have been initiated, one in Kalundborg, Denmark[2], and one near Halifax, Nova Scotia, Canada[3]. Three Eco-Industrial Parks are being planned in the US and a similar initiative is being established near Rotterdam, Holland[4]. However, there are a number of obstacles to initiating and managing an industrial ecosystem which include company concerns with regard to proprietary or confidential information, negotiating balance of payment, supervision and operation of co-treatment facilities and the complexity of managing the wastes produced by the plants.

The latter issue is one of great importance since the number of wastes produced by a single plant may at first appear to be only four or five, consisting of aqueous effluents, air emissions and a few solid wastes but such outputs are often not segregated and each may contain 10 or more wastes from different processes. Consequently, a plant may be actually producing up to 100 wastes, combined to produce the four or five output waste streams. In looking at methods to recycle or reuse wastes it is important to segregate waste streams and reduce contamination as much as possible, therefore reducing the need to treat the waste. To then examine the potentials for reuse, recycling, recovery or co-treatment for disposal requires the assessment of all segregated wastes from all participating plants which, for a large industrial estate, may well be over 100 facilities.

In determining reuse, recycling, recovery or co-treatment potentials for wastes, the characteristics of each segregated waste must be determined as well as the required characteristics of input materials for each plant. Wastes which totally match input requirements can then be reused directly as inputs while wastes with matching characteristics can be combined for co-treatment before recycling, recovery or disposal. The potential of treatments to change the characteristics of the waste materials to match input material, recovery or disposal requirements is then assessed. The production of secondary wastes from each treatment must also be considered and the cost associated with each treatment path is determined. Finally, the optimum treatment path for each waste is then determined. This process may require examination of thousands of possible waste treatment paths, far beyond the capability of any industrial manager.

Consequently, to implement an industrial ecosystem successfully, industrial managers require a decision support system that would determine waste management options for all types of wastes, giving priority to recycling and reuse and minimizing cost and environmental impacts. Ideally, the system would require a minimum of analytical information and input of information would be straightforward. The output would identify the optimum treatments necessary to treat all the wastes produced, ideally by recycling or reusing those by-products or, if necessary, disposing of them. Data on costs or savings for each treatment and for the treatment system overall would also be provided.

Such a system would require much more information than is currently readily available and more than could be obtained for the research summarized within this paper. There are still discrepancies regarding input criteria for many treatments, including the most common treatment systems such as aerobic wastewater treatment5, 6. Cost data are not readily available for many treatments and treatment outputs and efficiencies often depend upon a variety of factors, including operator experience, temperature, type and age of equipment and process control facilities. A complete model for such a decision support system was considered beyond the scope of this research. However, a model for determining waste treatment options for a variety of waste types can be developed, although there must be some recognition of the limitations of the input criteria and the output formulae and efficiencies. Such a system does not optimize the treatments; it simply lists the options available for treating the wastes and indicates if there are any wastes that can not be treated by the treatments in the system. A manager can then select a treatment option that is common to most wastes and examine other options for reducing or eliminating non-treatable wastes.

By providing a decision support system that can act as a tool to educate and assist managers in making waste management decisions, it may be possible to encourage industries to recognize the potential markets for their waste materials. Moreover, industrial estate managers can assess the suitability of the current waste treatment system to treat wastes from a new industry or determine if additional treatments are required.

This paper describes a prototype system which is the basic building block of a larger knowledge based decision support system (KBDSS) to determine options for integrated management of wastes from a variety of sources and industries. The KBDSS could provide decision support to government or industry representatives for the development of an integrated waste management plan for a number of plants within a defined geographical area such as an industrial estate.

A typical plant operation involves inputting a raw material or input into a process which then produces outputs which may be either products or waste materials (Fig. 1). Processes can include many of the operations at a plant e.g. the manufacturing process (which produces the product), the cooling process or the maintenance process, while inputs can include raw materials such as iron ore or water as well as manufactured items such as filters. Inputs usually have to meet certain specifications for purity or content of specific materials; e.g. boiler water must contain less than 50 ppm particulates or iron ore must contain 90% iron. These inputs are then changed into products and/or into waste materials, through the processes. Waste materials may be either reused, if they can meet the specifications for an input material, treated and then recycled once they meet these criteria or treated until they meet the standards for disposal.

When assessing a material for reuse as an input, a number of criteria have to be met. First, the input must not be a manufactured item such as a filter. Used manufactured items usually require specific treatments or inspection and replacement of worn parts for recycling. Second, the waste material must contain the major input components at the percent required in the input. Third, the parameters (COD, dissolved solids, etc.) of the waste must lie within the acceptable range required for the input. Once these criteria have been met, the waste material can be considered for reuse.

To recycle a waste material as an input, treatments are usually required. Again, the input must not be a manufactured item since the treatment and refitting of new parts is specific to an item and would require an expert system in itself. The waste material must contain the major components of the input. However, treatments can be used to change the percentage of components in a waste material so they match the levels needed by the input. Similarly, the parameters of the waste can be changed by treatments so they, too, do not need to match for recycling.

Treatments usually have a primary effect on one or two parameters, but that change may affect the characteristics of other parameters. For example, removing oil from a waste will then change the percentage of water in the waste material. Moreover, many treatments are most efficient in treating wastes whose parameters lie within a specified range, e.g. solids that are less than 1000 ppm. Consequently, to effectively treat a waste, the parameters of that waste should lie within the range specified for a treatment.

The output of the treatment is also defined by its parameters and these can be determined by calculating the changes that have occurred. Treatments usually produce more than one output and an output is selected if it matches the required parameters. The other outputs are termed secondary wastes.

Once the inputs and outputs of a treatment have been defined, it is then possible to determine the treatments required to recycle a waste material or to meet disposal regulations. The parameters of the waste material are compared to the input requirements; treatments which primarily affect those parameters which do not fit the input requirements but will be able to treat the material are then selected. The output of the treatment is calculated and compared to the input requirements; if they do not match, another treatment is then selected. The sequence of wastetreatment 1treatment 2treatment 3→… treatment ninput is termed a treatment train.

Ideally, costs of materials and treatments would have to be included to ensure an optimal treatment policy. However, this was considered to be beyond the scope of this work. By allowing the user to select shorter treatment trains, or trains which produce low volumes of secondary wastes from treatments, costs may be reduced. By identifying all the possible means of treating wastes from a group of plants for reuse, recycling or disposal, possibilities for treating wastes using the same treatment option can be determined which can also significantly reduce costs. Using one or more of these criteria, possible options for treatment of all the wastes can be determined.

The objective of this work was to develop a KBDSS prototype that would assist a user in the development of an integrated waste management plan, specifically looking at waste reuse, recycling, co-treatment and disposal. The system had to meet a number of requirements which included:

  • requiring a minimum amount of analytical data for inputs, products and wastes;

  • allowing the user to input, change or specifically select industry data for assessment;

  • incorporating a number of treatment options, sufficient to treat a variety of waste types such as gases, liquids, sludges and solids;

  • including a framework to incorporate and calculate cost data;

  • enabling the user to include regulations for disposal of wastes which would be incorporated into the system;

  • enabling the user to change or add to the list of treatment options;

  • determining all potentials for reuse or recycling wastes as inputs;

  • determining all potentials for treating wastes for disposal;

  • assessing the treatment options for potentials for co-treatment;

  • identifying the mass of secondary wastes produced from treatment of each waste;

  • allowing the user to determine criteria by which the final treatment options are selected; and

  • listing those wastes which cannot be treated or disposed of with the considered treatments and disposal options.

The software selected for this system was Paradox, a database program compatible with Windows or DOS. This inexpensive program (∼US$299 for Paradox 7, 1996) uses an object-oriented programming system, and interactive forms can be designed to accept data directly into tables within the program. Moreover, checks on the data to ensure that they are in the correct format or within acceptable levels can be included with each object for verification purposes. The software is flexible, capable of handling significant volumes of data and easy to learn and program, making it ideal for this purpose.

Paradox is not set up as a rule-based shell. Rather than using IF...THEN rules, for the purposes of this application it was considered to be more effective to use tables to store and compare parameters. Since Paradox is a database, this capability was readily incorporated into the program. Paradox is not a spreadsheet so calculation formulae cannot be readily stored in tables except as strings, although it can be linked to a spreadsheet or to a language compiler (e.g. C). Paradox will also incorporate or translate DBase files.

Paradox does not contain an inference engine. Therefore, a major task of this study was the development of an inference engine that accepts the data input by the user, stores it, compares it with the knowledge base, calculates an output and determines the suitability of these outputs.

A knowledge-based system (KBS) adds a knowledge base to a computer-based information system[7], representing and manipulating knowledge about non-structured problems to find a solution[8]. The developed KBDSS incorporated three basic components: the knowledge base, the data base and the inference engine. The knowledge base defines the input requirements (e.g. pH, oil, water or metal content) for a number of treatments and the parameters of the output from each treatment. It also contains the discharge standard requirements for disposal to air, water or landfill. The data base is input by the system user and specifies the parameters of inputs to the processes and the outputs from the process. The inference engine compares the input requirements with the output requirements and, using the knowledge base, determines which treatments can be used to treat the outputs to match the process inputs or the discharge standard requirements for disposal.

The treatment and discharge standard knowledge base contains information from the literature, experts, manufacturers and other sources and is the main knowledge base used to determine the treatment options available for treating a waste. A total of 25 treatments are currently incorporated into the knowledge base as listed in Table 1. The system allows a user to change or input new data into this knowledge base. If a specific treatment exists on-site, this flexibility allows the user to input the known input limits and output formulae for available treatments, thus ensuring a more accurate determination of acceptance of the material and of the parameters of the output material.

The knowledge is held in two main tables, each table containing values or formulae for a number of parameters which define the treatment input requirements and treatment outputs. The parameters were selected for three main reasons—(a) their discharge into the environment is regulated; (b) they specify limits to treatments or (c) they specify recoverable materials. Other factors such as human health and/or environmental impact and possible effects on material used for equipment construction were also considered. Since many regulations limit parameters at the ppm level, this was considered to be the most effective means for recording the parameters. The user must translate any percentages into ppm values.

The first knowledge base table contains the minimum and maximum input values for each treatment. These values are the maximum and minimum quantities that the treatment can effectively handle. The fields in this table include all the parameters for all states of a material. The user first defines the input state (gas, liquid, sludge or solid) that a treatment will treat, then inputs data through forms specific to that input state. The specific parameter information depends upon the state of the input (Table 2) and is measured in ppm, unless such a measurement does not apply (e.g. pH). A field can be left blank—the program will fill it with 0 (for the minimum input value) or 1,000,000 (for the maximum input value) (except for pH, maximum input value=14 and COD, maximum input value=3,000,000).

The second table defines the outputs for a treatment and contains formulae for all parameters. The user is first asked how many outputs a treatment has and what state those outputs are; the appropriate form is then provided for the user to complete. The user inputs the formulae as a string using defined codes for parameters—e.g. “particulates” is coded as Part, “volatiles” is coded as Vol. The user must recognize certain concerns—mass must include a conversion from ppm, while the ppm of a substance in a material will change as the mass changes so that all treatment parameters will require some formulae. The output can be either a formula, a percentage or a specified level. A subroutine is used to calculate the results of the formulae string.

The user interface operates through a series of menus, from which the user makes choices, and a series of forms which the user completes. As data are entered into forms, they are placed into tables in the database. The user is initially requested to input general data about the plant (company, plant name, contact, telephone number etc.). The user then selects a specific plant process (e.g. manufacturing, maintenance, cooling etc.) and the composition and characteristics of the process inputs are obtained. If the selected process produces a product, a basic mass balance must be completed for the process by the user. This acts as a verification for the inputs and outputs.

The following data is required for each input:

  • the state of the input (gas, liquid, sludge or solid);

  • if the input is simple (a raw material such as iron ore or a non-complex material such as white spirit which contains four or less major components), or complex (a material such as a cartridge filter);

  • the mass of the input per time period (this is translated into mass per annum);

  • the four main components of the input and the average, minimum and maximum percent of each component and;

  • the cost of each input.

The system does not allow a minimum greater than the average or a maximum less than the average to be entered and the average total must be less than or equal to 100%.

If the input is simple, then information about the input parameters is then obtained; otherwise the system excludes that input and moves to the next input. The parameters for inputs and wastes are the same as those in the knowledge base. Once data for all inputs have been obtained, then, if the process produces a product, the user inputs data on the product (including product state, mass produced and the four main components in the product). Otherwise for each waste produced by the process, the mass produced, the four main components of each waste, the average percentage of each component and the parameters of each waste are obtained.

An attempt was made to lump parameters (e.g. all heavy metals) together to reduce the number of inputs the user would have to make. This may lead to some inadequacies in the system since, for example, different heavy metals have different properties and may not react to treatments in the same manner. However, it was felt that the increase in convenience for the user would outweigh the inadequacies; otherwise the user may need to input over 50 parameters for each material which would make the system appear to be too complicated to use effectively. The treatment results were selected to be conservative to ensure that most metals would be treated adequately. Future refinements may incorporate parameter subsets to obtain more accuracy.

For processes which produce products, once all inputs, wastes and products of a process have been entered, an approximate mass balance is calculated to determine if inputs and outputs balance using the following equationMi=∑Mp+∑Mw

where

The user is then provided with a breakdown of the number and quantity of inputs and outputs that have been entered. It must be recognized that this is only a simple mass balance and, when using data estimated for a year, there can be major discrepancies in the balance. However, these discrepancies serve to inform the user that more accurate information should be obtained or that there are inputs or wastes that are not being included, which is particularly important in the case of fugitive emissions.

Once all the information has been entered for one process, another process is selected and data obtained for inputs, products and wastes for the second process. If all processes are complete for a plant, another set of plant information is entered. When data for all the plants have been entered, the user can then move to the inference engine to determine possibilities for reuse, recycling, co-treatment and disposal.

Many KBDSSs use IF…THEN rules to control the flow through the decision-making process. These rules are effective when dealing with a small knowledge base or a system where the parameters considered at each branch are different. Such a system could potentially have been used for this program. However, the developed system was designed to take specific information about a set number of parameters (Table 2) and use that information to define which treatments could be appropriate. Treatments were defined by the input parameter requirements and the parameters of the outputs produced. These input and output parameters constituted the knowledge base of the system. Consequently, rather than a series of IF…THEN rules, data tables were used to store the knowledge base and the inference engine was designed to compare the input parameters for each treatment in the knowledge base with the material input and waste parameters stored in the database by the user for each process. The inference engine is depicted in a flow chart in Fig. 2.

Exhaustive enumeration was used to generate the treatment options, and multiple sort mechanisms are utilized for defining final options and to propose a model for the calculation of costs for the final options. The program first compares the major components and characteristics of a selected waste with previously assessed wastes. If they are the same, the wastes can be treated using the same treatment options, so these options are copied into the new waste file. Masses are then calculated for the outputs from the treatment of the new waste using a mass ratio from the previously assessed waste.

A waste which does not match previously assessed wastes is compared with each “simple” input. If the main components of the input are found in the waste, then the parameters are compared; otherwise another input is considered. If the waste parameters lie within the minimum and maximum of the input parameters, then the waste is considered to be directly reuseable as the input. If not, then some treatment is necessary.

The needed treatments are defined by those parameters which did not match the input parameters. Therefore, treatments which change those parameters and treat materials in the state of the waste (i.e. gas, liquid, sludge or solid) are then selected and stored in a temporary table. Once all the treatments are placed in this temporary table, the system moves to the first uncompleted treatment train. The parameters of the output from that treatment are then calculated and compared with the input parameters (Fig. 3). If the parameters match then no further treatments are required. Otherwise, treatments which change non-matching parameters are selected. All treatments whose input limits accept the output from the first treatment are placed as possible second treatments in the temporary table (Fig. 4). The output parameters from these second treatments are then compared with those of the input parameter and the cycle is continued until:

  • no treatments are found to fit the parameters to be changed;

  • 10 treatments have been assessed or;

  • the treatment output matches the input parameters (refer to Fig. 5).

Once a treatment train has been completed, the system moves down a record in the temporary table and follows this train to the end in the same manner, as shown in Fig. 6. The process continues until all treatment trains have been found that treat the waste to the input requirement and another input can then be considered.

To determine which output from a treatment is selected for recycling, the calculated outputs are compared with the selected input. The output with parameters which match the input parameters most closely is then selected. The mass of the second output, which is termed a “secondary waste”, is then added to other secondary waste masses for that treatment train. The completed record would include the total mass of secondary wastes produced by all treatments in that treatment train. As each treatment output is calculated, the output mass is also calculated and stored separately in the table, as is the material state of the output.

Once all inputs have been evaluated, selected standards are then evaluated. These standards dictate the parameter limits that must be met to release gaseous emissions or liquid effluents, to dispose of material in a landfill or to apply material to land for land treatment. The waste parameters can then be compared to these parameters and the required treatments to match parameters be determined.

Upon evaluation of all standards, the total treatment train options for the specific waste have been determined. If no trains have been found for a waste, that waste is placed in a no-treatment table. The next waste is then assessed.

Once all wastes have been evaluated, the final treatment train table should contain all possible treatment trains for all wastes. These then have to be sorted according to the criteria selected by the user in order to select the final options.

The user is given a choice of the following criteria for sorting the treatment trains:

  • a maximum train length equal to the shortest, second shortest or third shortest train for each waste;

  • a maximum mass of secondary wastes equal to the lowest, second lowest or third lowest masses of secondary wastes per treatment train for each waste;

  • the maximum number of matching treatments for all wastes.

The user can select more than one criterion, although the third criterion can only be selected as the final criterion as it produces the final lists of treatment trains. The first option determines the shortest, second shortest or third shortest (as selected by the user) treatment train for each waste and selects all trains of that length or shorter for the specified waste. This criterion minimizes the treatments used in the final options. The second option sorts the treatment trains for each waste according to the total mass of secondary waste and selects all trains with secondary wastes equal to or lower than the lowest, second lowest or third lowest (as determined by the user) secondary waste. The amount of secondary waste that has to be treated or disposed of is thus minimized.

For matching treatments, trains which reuse wastes were automatically selected. This allowed wastes to be co-treated, reducing the costs of treatment. Treatments were matched by first examining all trains which recycle wastes. Those which had the greatest number of matching last treatments and final outputs were then selected; these were subsequently selected for the maximum of matching second last treatments, then the matching third last treatments etc. until all the trains had been matched and sorted. Once the treatments have been sorted, all of the trains for the selected wastes are deleted from the main list and the remaining treatment trains are then matched.

For the final result, the program provides three tables which are the three most optimal overall waste management options for all the wastes, as selected using the criteria set by the user. Each table includes treatment trains for all treatable wastes which, together, are the most effective means of treating the wastes using the specified criteria. Wastes which could not be treated are also identified. The total mass of secondary wastes produced per treatment train are also included in the results.

Both the system and the data require some form of external validation. The system was first examined on a stepwise basis, examining the data and results at each step to ensure that it was functioning correctly. A case study was also used (see accompanying Part II) to further verify that the program functioned as intended and represented the case study situation in a reasonable manner.

Selection of the parameters plays a critical role in the functioning of the system. Three errors can occur with regard to selection of the parameters:

  • a treatment is accepted although it will not actually treat a waste;

  • a treatment is not accepted although it will actually treat a waste;

  • the output from a treatment is not properly calculated since a parameter datum is not available or the treatment is less or more efficient than expected.

These errors could occur if acceptance or rejection by a treatment is dependent upon a parameter that is not included in the knowledge base. The more parameters available in the system, the less the chance that the system will cause these three errors. In discussion with Woods[10] the other major parameter that could be included for gases and liquids is temperature, necessary for some treatments (baghouse filtration) and for energy recovery. Since few treatments would require this parameter and energy recovery has not been included in this system, the temperature parameter was not included.

The trade-off with selection of more parameters is the cost requirement for the analysis and the time required for data input by the user. It must also be recognized that most materials must be tested to determine how effectively the treatment will treat them. Moreover, this program is designed to provide treatment suggestions, not definitive solutions, to users. Consequently, the parameters selected were considered to be an acceptable trade-off between data input requirements and potential errors in the suggested solution sets.

The parameters should be able to incorporate most types of waste materials. However, it may be necessary to change the parameters to accommodate different wastes. Further testing will be required to determine how the wastes from a range of different industries are handled by the developed system.

The system has a number of limitations and further development of the system could address some of the following issues.

A limit of 10 treatments was allowed for the treatment trains which may mean that some materials may not be treated within that limit. However, it was considered that most treatment trains would require fewer than 10 treatments and this level was a reasonable upper limit. With more than 10 treatments, the number of permutations would be extremely high and require additional computer time to determine acceptable trains. Further trains could be allowed if necessary in the future.

When selecting the parameters, it was assumed that any change in major components would be reflected in the parameters and vice versa. Therefore, if the material parameters met those of the input and the main components of the waste match the input, then the final treated output should match the desired input. The system does not consider the complex chemical reactions that may occur, possibly yielding a result that does not match the desired input or producing some hazardous by-products. To include such a database within this system would require significant additional work.

The program is a compromise between a very specific, large and complex expert system and a general system which gives decision support to a user. In the former case, each different type of treatment could be included. For example, instead of precipitation being classed as a treatment, it could be separated into the different types of precipitation: ferrous, alumina, etc. The same could be done to determine possible membrane treatments. Each individual treatment could concurrently require an expert system module to determine if the material is acceptable for that treatment and if the treatment will perform the function described by the system. Consequently, it must be recognized that with this system, although the parameters may match, the components of the resulting material may not match the components of the input.

In addition, the efficiency of the treatments may not be as high or as low as the output formulae dictates. Very few materials will act exactly as those studied in research for a variety of reasons:

  • they may be different than the studied material;

  • it is not uncommon for waste parameters to vary widely;

  • the operator may not follow the same procedure in handling and treating the material;

  • environmental conditions and equipment materials may differ;

  • larger vessels may create different treatment conditions; and

  • supplementary materials may not be of the same quality (e.g. distilled water versus untreated water).

It is important that the user recognize that the suggested treatment trains must be tested to determine the efficiency of the treatments and the quality of the results. Fine tuning will be required to ensure that a standard product can be produced for reuse or recycling. Industries must also recognize that materials that are to be reused or recycled have to meet quality standards. Improving the consistency and reducing impurities in these former waste materials may reduce the need for treatment for recycling.

Another factor that must be considered is that the parameters included in this system are only a few of the parameters that characterize wastes or inputs. Although all of the parameters included in the system may match, other parameters, important to a specific treatment or for reuse or recycling to a particular input, may not lie within the treatment or input specification and the waste may require further treatment. Future developments in the system may include subsets of some parameters such as heavy metals, to provide greater accuracy for the system and improve the matching capabilities of the system.

Section snippets

Concluding remarks

The developed KBDSS prototype provides a tool for determining reuse, recycling, co-treatment and disposal potentials for a multitude of wastes from a variety of industries. The program allows for simplified input of data and for the user to select the criteria which determine the final treatment trains selected. Treatment input and output parameters may be defined by the user to incorporate existing facilities and their known operating parameters. The final results do not yet provide economic

Acknowledgements

The authors would like to thank Canada’s Tri-Council Secretariat for their provision of an Eco-Research Fellowship towards this research, and McMaster University for its funding contribution.

References (12)

  • Frosch RA, Gallopoulos NE. Strategies for Manufacturing. In Managing Planet Earth, special ed. Scientific American,...
  • Co ̂té R, Plunkett T. Industrial Ecology: Efficient and Excellent Production. School for Resource and Environmental...
  • Co ̂té R. et al. Designing and Operating Industrial Parks as Ecosystems. School for Resource and Environmental Studies,...
  • Co ̂té RP, Kelley T, Reid H, Smolenaars T. The industrial park as an ecosystem: cross-sectoral case studies. School for...
  • Grady CPL, Lim HC. Biological Wastewater Treatment. Marcel Dekker Inc., New York...
  • Metcalf and Eddy, Inc. Wastewater Engineering Treatment, Disposal, Reuse. McGraw-Hill Book Company, New York...
There are more references available in the full text version of this article.

Cited by (16)

  • Investigation of public's perception towards rural sustainable development based on a two-level expert system

    2009, Expert Systems with Applications
    Citation Excerpt :

    Comparatively, expert system is considered as one of the effective tools in supporting decision making related to sustainable development at multiple scales. It can link a combination of decision-support tools (e.g., optimization modeling, and cost-benefit analysis) and various information components into a general decision-making process (Athanasiadis & Mitkas, 2007; Boyleand & Baetz, 1998; Carswell, Gardiner, Bertolotto, Rizzini, & Mandrak, 2008; Huang, Batez, & Patry, 1992, 1995a, 1995b; Lewis & Bardon, 1998; Tremblay, Hester, Mcleod, & Huot, 2004; Zeng, Zhang, Liu, & Yang, 2007; Zeng & Trauth, 2005; Yao, Liu, Fan, Yao, & Yang, 2006). With such an effective tool, both the knowledge of agricultural development and eco-environmental preservation, as well as public’s perception towards sustainable development can be interactively and dynamically acquired.

  • An expert system for decision support of municipal water pollution control

    2003, Engineering Applications of Artificial Intelligence
  • Industrial symbiosis waste exchange identification and optimization

    2021, Proceedings of the Annual Hawaii International Conference on System Sciences
View all citing articles on Scopus
1

Part II of this article has already been published in this journal: Boyle CA, Baetz BW. A prototype knowledge-based decision support system for industrial waste management; part II. Application to a Trinidadian industrial estate case study. Waste Management 1997;17(7):411–428.

View full text