AgriBuilding 2001 September 3–7, 2001 – Campinas, SP, Brazil
Intelligent decision support system applied to environmental pollution caused by swine manure Carlos Renato V. de Oliveira1 Paulo Armando V. de Oliveira3 1 Veterinarian,
Jorge Muniz Barreto2 Flávio Bello Fialho3
Prof. EAFC, Concórdia, SC, Brazil; 2 Electronic Eng., Ph.D., UFSC; 3 Researcher, Ph.D., Embrapa Swine and Poutry
[email protected] [email protected]
1
Abstract
23
In the past years, swine production in Brazil has been growing rapidly, competing with cattle and poultry in meat production. The West of the State of Santa Catarina (SC), a region with 30 000 km2 , has the most technically advanced swine production in Brazil. Within this region live 80% of the state’s swine producers, of which only 60% have some form of manure storage system. In brazilian literature, papers on the use of artificial intelligence and expert systems in swine production are rare. No specific paper on waste handling was found. The success of any program to reduce pollution caused by swine waste depends on its capacity to evaluate the current situation, and to anticipate pollution risks, in order to choose which technology to use. This paper describes the development of a decision support system (DSS), using as reference the data base of swine producers of Western Santa Catarina and their characteristics. Individual swine production systems will be classified with respect to meeting State and National legal requirements for waste management. The decision support system will allow the identification of high pollution risk producers and pollution sources. This information may be used to restrict the issue of new swine production licences in high risk areas, as well as to identify current producers which must improve their waste management system, thus reducing pollution. The system may also be used to simulate the consequences of changes in animal density and management techniques on the severity of pollution ("what-if" analysis). The system will give support to technicians, researchers and environmental organizations for making decisions on pollution issues involving swine production.
24
Keywords: artificial intelligence, expert system, swine, waste management,
25
manure, pollution
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
26
27 28 29 30 31
1
Introduction
In the past years, swine production in Brazil has been growing rapidly, competing with cattle and poultry in meat production. The West of the State of Santa Catarina (SC) has the most technically advanced swine production in Brazil. Within this 30 000 km2 region live 80% of the state’s swine producers, of which only 60% have some form of manure storage system. 1
AgriBuilding 2001 September 3–7, 2001 – Campinas, SP, Brazil
46
The environmental impact of this intensive production is enormous. Inadequate handling of liquid swine manure has caused great damage to the environment, and is the main cause of pollution in the region. The greatest challenge in intensive swine production regions is to find an alternative system of production which reduces smell, the emission of hazardous gases, the risk of groundwater pollution by nitrates and the emission of ammonia to the atmosphere. In brazilian literature, works about the use of artificial intelligence and expert systems in swine production are rare. This is specially true for works dealing with swine manure management. The success of any pollution control program which deals with swine manure depends greatly on its capacity to appraise the situation, antecipate the occurence of problems and to make decisions about the adoption of one or another technology or management system. The purpose of this work is to develop a Decision Support System (DSS) using data about caracteristics of swine production units, swine growers and current legislation on waste handling and management.
47
2
32 33 34 35 36 37 38 39 40 41 42 43 44 45
48 49 50 51 52
Material and methods
The decision making process may be divided into four main phases: information gathering, problem structuring, choice of the best course of action and implemetation of the chosen alternative (TURBAN, 1988, cited by MOURA, 1995). The methodology used to develop and build the Decision Support System for swine waste pollution control and the corresponding Expert System is shown on Figure 1. Analysis and planning
Design and project
Construction and tests
Implementation
Operation and maintenance
Figure 1 — Life cycle of the development of the decision support system and the expert system (Adapted from Turban (1988), cited by Filho (1997)). 2
AgriBuilding 2001 September 3–7, 2001 – Campinas, SP, Brazil
82
In the first phase of the development cycle (Analysis and Planning), a general understanding of the problem is developed. This phase is being carried out in the National Research Center for Swine and Poultry (Embrapa Swine and Poultry), where information is being gathered with the objective of constructing a data model. Embrapa Swine and Poultry has been developing a data base called Project Fragosos. This data base contains information about the polluting potential of swine producing facilities located within the watershed of Lageado Fragosos, in the county of Concórdia, Santa Catarina. A survey (called “Survey for registration of swine producing properties in the Fragosos watershed”) with 75 questions was applied to the producers, and the results were stored in a spreadsheet. The information considered useful for processing by the expert system was extracted from this spreadsheet. The data is used to classify producers into one of three categories: polluters, potential polluters and non-polluters. The resulting data is used to feed the system’s data base. Another database being used is that of FATMA (Foundation for Support of Technology and Environment of the State of Santa Catarina), which contains data about swine producers in the whole region. In the Construction and Tests phase, the Expert System Shell chosen was “Sinta”, develobed by the Federal University of Ceará. This shell was chosen because it is free software, easy to use and without limits on the number of rules in the knowledge base. Being a user-friendly environment, the shell may be used by technicians who deal with swine production, who typically do not have much background in working with computers. During the Implementation phase, if-then type rules were defined. In the creation of rules, care must be taken with the use of imprecise knowledge. Generally, imprecise knowledge is associated with any decision making system, due to the process of gathering information from many sources, sometimes with inconsistencies. In swine production, there are many concepts which are not computationally adequate. These were transformed into more precise concepts, and those in which this was not possible were discarded. The system is currently only being used experimentally.
83
3
53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81
84 85 86 87 88 89 90 91 92 93 94 95 96
Results and discusion
Is was possible to verify, up to the present moment, that treatment of imprecise information is a very important factor in the develpopment of an intelligent system in agriculture. Thus future versions of the system should use fuzzy logic to treat imprecise variables. The system is being tested for robustness and for how well it takes the same decisions a human specialist would. The system is being validated by comparing the classification given to producers (polluters, potential polluters or non-polluters) to the situation observed in the field. The degree of accuracy of the system is yet to be determined, but initial studies indicate a good degree of agreement of the predictions with reality. This system should be tested and compared with other sectors of agriculture. The use of computer technology is not so common among swine producers, compared to other agricultural activities. Specifically in the West of Santa Catarina, the financial 3
AgriBuilding 2001 September 3–7, 2001 – Campinas, SP, Brazil
98
condition of swine producers is a barrier to investments in technology, and the number of computer professionals in swine production is small.
99
4
97
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
Bibliography
ACCS. Relatório Anual 2000 da Associação Catarinense de Criadores de Suínos, ACCS, Concórdia SC, Dezembro 2000. Cox, E. The “Fuzzy” Systems Handbook: a practitioner’s guide to building, using, and maintaining “fuzzy” systems. Academic Press Professional Inc. 1994 Filho, M. A. R.; Braga, J. L.; Fontes, C. A. A. Um sistema de apoio à decisão para o gerenciamento de confinamentos de bovinos de corte. Revista de Economia e Sociologia Rural, v. 36, n. 2, 1997. Miranda, C. R. de; Santos Filho, J. I. dos; A Situação dos Dejetos Suínos na Região da AMAUC SC, X Congresso Brasileiro de Veterinários Especialistas em Suínos, 26 a 29 de outubro de 1999, Belo Horizonte, MG. Anais. . . 1999. Moura A. D. de; Sistema inteligente de apoio à decisão aplicado ao gerenciamento da produção de frangos de corte. Dissertação de Mestrado, Curso de Economia Rural, Universidade Federal de Viçosa. 1995. Oliveira, M. A., Gottgtroy, M. P. B.; A Importância do Tratamento do Conhecimento Impreciso no Desenvolvimento de Sistemas para a Agricultura. Agrosoft’95. Juiz de Fora MG. 1995. Oliveira, P. A. V. de (coord.). Manual de Manejo e Utilização dos Dejetos de Suínos. Concórdia SC:, EMBRAPA-CNPSA, 1983. 188 p. (EMBRAPA-CNPSA, Documentos n. 27).
4