【作者】 王靖飞；

【导师】 王洪斌； 熊范纶；

【作者基本信息】 东北农业大学 ， 临床兽医学， 2002， 博士

【摘要】 专家系统是一个智能计算机程序，它利用知识和经验，通过推理来解决领域中只有人类专家才能解决的难题。专家系统的研究和应用已经成为全世界人工智能研究的热点和焦点，它们在农业领域的应用为农业生产带来的很大的经济效益。我国政府提出的发展农业智能系统的战略为我们创造了广阔的发展空间，同时也为智能农业系统的发展带来了无限的发展契机。动物疾病诊断专家系统的研究是在结合国内外研究现状，针对我国畜牧业的发展的实际需求和以解决生产实践中的难题为目的的基础上提出的，具有很重要的现实意义。 针对目前动物疾病诊断专家知识的自动获取还无法实现的状况，本研究对各种获取源讨论之后，确定了以下主要的知识获取源：书籍、专家、杂志和网络，并且对这些知识源的重要性进行了分析，其重要性从大到小排列顺序为：书籍＞专家＞杂志＞网络。根据知识源的特点，确定了由研究生负责知识的主要获取任务，学科专家进行知识加工和专家评判小组进行知识评判的知识获取策略。这种方法可以有效地利用现有的人力和物质资源，提高知识获取的效率和准确性。 寻求适合农业知识特点的表示方法是农业专家系统研究的热点问题之一。本研究针对动物疾病诊断知识的特点，提出了“IF＜症状i＞THEN＜疾病a＞CFai”知识表示方法，其中“症状i”为“疾病a”所表现的症状集合中的一个元素；“CFai”为当“症状i”存在是动物患疾病“疾病a”的概率；sum from i=1 to n(cfai)=100。这种知识表示方法可以很好的表示动物疾病诊断专家知识。 针对动物疾病诊断的特点，本研究设计了基于数据库的规则库和其它信息库，在规则库的建设中，用一维数据库来模仿了脑信息的存贮与激活方式，以这种方式建立的规则库适合动物疾病诊断知识的特点，具有实用性强、操作和维护简单的优点。 在传统的专家系统精确推理和模糊推理的理论指导下，通过对领域专家在诊断动物疾病过程中思维方式的研究，提出了适用于动物疾病诊断过程的混合推理模式，即先进行正向推理，由正向推理得出一组假设结论，然后进行反向推理，去验证假设结论集合中的元素是否可以成为结论。在整个推理过程中，本研究对症状和疾病之间的关系进行了量化，由一定的数字来表示这两者之间关系的强弱，实现了动物疾病的数值诊断。由于疾病与症状之间是一一对应的关系，并且又有数值作为它们之间联系的纽带，从而实现了系统与用户的动态交互，从一定程度上模拟了专家在解决问题中的思维过程和提问方法，使系统更加人性化、效率更高、准确性更好、实用性也更强。 界面是用户同系统交流的媒介，界面的友好程度直接决定着用户是否乐意使用该软件。为此，本研究制定的界面设计目标是：简洁、美观，符合人性化；易于使用掌握；适应性强；和具有帮助能力。 根据动物疾病诊断的特点以及生产实践的需求，本研究提出了软件复用技术在动物疾病诊断专家系统研制中的应用策略，对一些关键技术通过逐步完善，可以逐渐形成动 王靖飞 动物疾病诊断专家系统的研究与应用2002年5月一物疾病诊断专家系统开发平台，将会大大提高动物疾病诊断专家系统的开发效率。 最后，利用以上对动物疾病诊断专家系统的研究结果，初步完成了奶牛疾病诊断专家系统的研制。在奶牛诊断专家系统的研制中，根据奶牛疾病诊断知识的特点和不同用户、不同情况的需求，本研究在奶牛疾病诊断系统中首次实现了两套推理策略、两个不同特点的知识库的集成应用。从系统的运行情况和运行结果来看，本文中所提出的开发动物疾病诊断专家系统的各种方法均具有实际应用价值，可以作为一般动物疾病诊断专家系统开发的基本方法。 本研究从解决生产实践中的难题出发，以实际应用为目的，对动物疾病诊断专家系统的各项技术做了全面的论述，提出了一些适合动物疾病诊断特点合要求的研制动物疾病诊断专家系统的方法，为专业人员开发其他动物疾病诊断专家系统积累了重要的基础资料，也为专家系统在农业领域的应用做了一些基础性研究，对专家系统的应用与研究具有重要的实践意义。更多还原

【Abstract】 Expert System is an intelligent computer program, which can use expertise and reasoning to solve the problems that can only be solved by domain experts. It has been the researching focus of the artificial intelligence of the whole world.Great benefits have been brought by its application in the fields of agriculture. When an animal shows disease symptoms, it is important to make an accurate diagnosis to support control strategies. Diagnosing disease in animals requires considerable expertise. Only a few experts have the ability to do this, and each expert has his own- specific domain. To retain expertise and to make it more generally accessible, expert system for animal disease diagnosis should be developed.Knowledge acquisition is the "bottle neck" problem in the development of artificial intelligence. It is also a problem in developing diagnostic expert system for animal disease. To acquire accurate knowledge, this paper presents four knowledge acquisition resources for developing diagnostic expert system for animal diseases. They are books, experts, journals and internet/intranets. Their importance are listed as follows: books> experts > journals > internet/intranets. Many domain experts were consulted and their expertise were acquired.In this paper, a knowledge representation method is presented according to the characters of diagnostic knowledge of animal diseases, which represented as IF <signs> THEN <disease> CF. Signs is a symptom set, and it has all symptoms of a disease. Disease represents the name of one diseases, and every disease has its own number of symptoms. CF is the probability of the disease to be the result of inference when the symptoms is presented. The sum of CF to one disease is 100.Database-based rulebase and other information-bases are designed in this paper. The advantages of the rule-base with such structure is that it can be used and maintained easily, and operated simply. The rule-base with such structure is the result of simulating the information storage and activation of brain.A new way of inference with both forward and backward chains is presented in the paper, it is the result of simulating experts’ thinking during the processes of animal diseases diagnosis. This inference contains two parts. On the first stage, forward chain will be used. A set of hypothesis will be got from the information provided by user through forward chain of this stage. Then the inference mechanism will turn to the second stage to test the hypothesisbe presented to user. Otherwise, the inference will repeat this process.Interface,which is important in determining whether a user prefer to use the system.is a communication intermedium between user and system. A designing object of interface is presented in this paper, which is concise,simple and easy to learn.Reuse is the bases of all scientific research. Expert systems of diagnosis for animal diseases share common parts, therefore, if one of such system has been developed, its thought of construction, controls and components can be used in other expert system’s construction. In this paper, some reusing technologies have been discussed, which can be used in the development of diagnostic expert system for any diseases of all animal.Finally, all the methods described before have been used in the development of expert system diagnosis for cow diseases. This expert system contains disease descriptions and offers the ability to diagnose diseases in cows. Digitalized photographic pictures and video materials can be showed by system to support an interaction session. The paper describes the development of the system, the structure of the knowledge-base and the inference mechanism. A prototype has been built for diagnosing adult cow diseases. It is proved that those methods are correct and usable.更多还原