分层的基于模型诊断及其于假设部分序的优先诊断
|
摘要
众所周知，由传统专家系统方法建立的诊断系统存在诸多缺陷，如所收集知识的不完备性、系统对知识的依赖性、获得专家知识不一致性等。基于模型诊断推理正是克服传统专家系统的这些严重缺陷而提出的一种智能诊断推理技术。基于模型诊断方法使用描述系统部件的内部结构和外部行为的模型，该模型具有很强的设备独立性，同时使用费用（cost）也比较低。如在最初产品设计和生产时就能给出模型；修改此产品模型就可应用到彼产品上去，缩短产品设计和开发周期。正是由于基于模型诊断有如此多的优点及实用价值，八十年代中期吸引了越来越多的研究者，到九十年代已经成为人工智能研究领域的一个十分活跃的分支。
从理论的角度可把基于模型诊断划分为两大类：一是基于一致性诊断；另一个是溯因诊断。基于一致性诊断是处理诊断系统的正常行为模型；溯因诊断是处理诊断系统的故障行为模型。但无论是哪类诊断，在处理复杂设备时，从缩小诊断搜索空间的角度出发建立高效的模型描述是提高诊断效率的关键。分层基于模型诊断正是基于上述两种诊断方法而提出的，其基本思想是应用抽象将诊断系统进行分层描述，经过抽象处理后得到的新的抽象层(系统)只包含有限个抽象层部件；而细化到具体层时，只生成与抽象层相关的候选，这样就可大大减小诊断系统搜索空间，提高系统诊断效率。
我们对Mozetic的工作进行了研究，他主要定义了三个典型的行为抽象算子，即常量抽象、变量抽象和模型抽象，将待诊断系统的具体层模型描述映射到抽象层(同时故障被隔离到抽象层)。诊断在抽象层开始，此时系统描述比较简单，只包含少数几个部件或抽象的定性行为，得到的少数几个候选很容易被验证，再用抽象层的诊断结果指导具体层诊断。同时引入一致性条件：一是不完备约束条件，即假设具体层映射m，则有抽象状态的x不能映射到无抽象状态的y；二是保护映射条件，即如果在抽象层不是诊断的候选，在具体层一定不是诊断。通过一致性条件在诊断过程中首先过滤掉引入的不完备抽象和不可能诊断，在分层描述的基础上进一步缩小诊断搜索空间，提高诊断效率。另外，模型描述是待诊断系统任意状态到输入、输出观测值的映射，这与以往基于一致性诊断和溯因诊断明显不同。之后我们使用Borland JBuilder实现了该算法的基于模型诊断系统，并在或门电路上进行了测试。
Mozetic分层诊断是应用抽象对诊断系统进行分层描述，减少抽象层部件的数量；并进一步通过一致性条件消除那些不可能的诊断来减少诊
断代价。但这只是一种静态的、单一的、固定的分层描述技术，没有充分利用当前各层有用的观测值，致使分层诊断效率不够理想。例如当最抽象层不是最顶层，而是中间的某一层或第0层，后者是Mozetic算法失效的情况(与一般诊断效率相同)。Mozetic 改进算法提出了一项动态确定分层描述的技术，即以有用的观测值为基础，通过使用相应结构树重新安排各层给定的分层描述来创建一个新的分层描述。我们把任意分层描述H（由结构抽象建立的）与其结构树相关联，结构树中每个节点代表H的部件，它的子孙代表H的子部件。结构树ST(H)中每个节点c是第i层的一个部件，把部件的描述和观测值（当它们有用时）与节点相结合，因此这个节点就与分层描述H中的一个诊断问题的子集相关联。这样我们就把系统诊断问题看作是对树的遍历问题。需要强调的是，它不是修剪Mozetic算法中的分层描述，而是以观测值为指导重新动态地安排最初的分层描述技术。使用改进算法，Mozetic算法不理想的情况得到很好的解决。接着我们又进一步提出一个改进算法，在过程TOP-DOWN-DIAGNOSIS开始前，通过有用观测值排除与观测值相关联的节点的不可能候选来增强各层的诊断能力。一般地，任何部件都可能有多个行为模式，由其所聚合的部件也一定有多个行为模式。在给定观测值的情况下，这些行为模式中某些是不可能的候选，应该首先被排除。我们的做法是:先将结构树中与观测值相关联的节点分离出来，若其某一行为模式与观测值不一致，则排除此行为模式，当进行诊断时就不用处理被排除的行为模式了。此外，由于从下向上处理结构树时只考虑被观测的节点，当到达描述部件c的节点时，其所有被观测的子节点已经处理完毕。由于部件c的每个行为模式都是其所聚合的子部件行为模式的抽象，因此，对超部件c的某个行为模式mi，如果mi所聚合的所有子部件的行为模式已经被排除，就可排除c的行为模式mi。注意：进一步改进算法虽然排除了大量的不可能诊断，进一步缩小搜索空间，但是不能保证排除所有不可能候选。
最后，我们介绍用假设的优先权控制诊断，这些优先权是关于假设的部分序，并按顺序扩展到关于假设集的优先关联，通过选择极好假设集来减少可能诊断数量。通过搜索和剪枝Reiter碰树策略来计算优先解决方案，进而计算优先诊断。另外单一优先解决方案也可通过快速hill-climbing策略计算。
As we know, diagnostic systems constructed using traditional expert system approaches have many shortcomings, such as knowledge collected in completeness, system dependent on knowledge,and the difficulties of the acquisition of necessary diagnostic knowledge from experts etc. Model-based diagnosis is a new reasoning technique of intelligent diagnosis to overcome the shortcomings of traditional diagnostic methods model-based diagnostic method employ the model of the internal structure and behavior model of system, this method is strongly device independent,at the same time,this approach can be less costly to use.For example,the mode is given when product is designed and manufacturing at first;this mended model of product is use to that model of product. At the same time, the designing and exploitation cycle of product are shorten. Model-based diagnosis is attracted by many researchers in 1980s because of advantages and actual value. It has been an active branch in the area of artificial intelligence until 1990s.
Two representative logical methods of model-based diagnosis are consistency-based diagnosis and abductive diagnosis. Consistency- based diagnosis deal with normal behavior models of system, and abductive diagnosis deal with faulty behavior models of system. No matter what it is , model representation of high efficiency is important to improve efficiency of diagnosis when we deal with complicated device. Hierarchical model-based diagnosis is put forward based on above two diagnosis methods. It’s basic idea is hierarchical representation to diagnose by abstraction. The new system of abstraction have been obtained only contains finite components of abstract level. But when the components of abstraction level is elaborated to the detail levels, Candidates are made in relation to abstract levels. Such can reduce the search space of diagnosis system in order to improve diagnosis efficiency.
We studied the work of “Mozetic”, in which he defined three typical behavior abstraction operators mainly, including refinement/ collapse of values, introduction/deletion variables, eleboration/ simplification of mapping, and defined to model
representation of detail levels in the diagnoses system which is mapped to abstraction level. The faults are isolated one level at a time by hierarohical diagnoses. At the same time, consistency condition is introduced . The first is restriction of incompleteness given a detailed level mapping m the condition, it prohibits cases where an x with an abstraction is mapped to a y without an abstraction; the second is preservation for mapping, which basically says that diagnosis which are impossible at the abstract level(where the search space is smaller)are impossible at the detailed level as well. Restriction of imcompleteness and impossible diagnoses are filtered by consistency condition in the process of diagnosis. Based on hierarohyical representation, restriction of imcompleteness and impossible diagnoses are filtered by consistency condition in the process of diagnosis, and searching space is reduced.
Moreover, model representation is defined that any state of diagnoses system is mapped to observation of input and output system, which is different from consistency-based diagnoses and abductive diagnoses. We implement the algorithm using Borland Jbuider and tested it on or gate of circuit.
Diagnoses system is represented hierarchically by abstraction in order to reduce the number of component of abstracted level and eliminate impossible diagnoses in detailed level by consistency condition to reduce diagnostic costs. But it is a kind of static 、 single、and fixed technics of hierarchical representation, which make no use of the set of currently available observations so that the efficiency of hierarchical diagnoses is not high. Such as,when the most abstraction level is not the top level but certain middle level and level 0.The level 0 is the worst condition (which is the same as general diagnoses efficiency.) We improve Mozetic’s algorithm by structure abstractio
从理论的角度可把基于模型诊断划分为两大类：一是基于一致性诊断；另一个是溯因诊断。基于一致性诊断是处理诊断系统的正常行为模型；溯因诊断是处理诊断系统的故障行为模型。但无论是哪类诊断，在处理复杂设备时，从缩小诊断搜索空间的角度出发建立高效的模型描述是提高诊断效率的关键。分层基于模型诊断正是基于上述两种诊断方法而提出的，其基本思想是应用抽象将诊断系统进行分层描述，经过抽象处理后得到的新的抽象层(系统)只包含有限个抽象层部件；而细化到具体层时，只生成与抽象层相关的候选，这样就可大大减小诊断系统搜索空间，提高系统诊断效率。
我们对Mozetic的工作进行了研究，他主要定义了三个典型的行为抽象算子，即常量抽象、变量抽象和模型抽象，将待诊断系统的具体层模型描述映射到抽象层(同时故障被隔离到抽象层)。诊断在抽象层开始，此时系统描述比较简单，只包含少数几个部件或抽象的定性行为，得到的少数几个候选很容易被验证，再用抽象层的诊断结果指导具体层诊断。同时引入一致性条件：一是不完备约束条件，即假设具体层映射m，则有抽象状态的x不能映射到无抽象状态的y；二是保护映射条件，即如果在抽象层不是诊断的候选，在具体层一定不是诊断。通过一致性条件在诊断过程中首先过滤掉引入的不完备抽象和不可能诊断，在分层描述的基础上进一步缩小诊断搜索空间，提高诊断效率。另外，模型描述是待诊断系统任意状态到输入、输出观测值的映射，这与以往基于一致性诊断和溯因诊断明显不同。之后我们使用Borland JBuilder实现了该算法的基于模型诊断系统，并在或门电路上进行了测试。
Mozetic分层诊断是应用抽象对诊断系统进行分层描述，减少抽象层部件的数量；并进一步通过一致性条件消除那些不可能的诊断来减少诊
断代价。但这只是一种静态的、单一的、固定的分层描述技术，没有充分利用当前各层有用的观测值，致使分层诊断效率不够理想。例如当最抽象层不是最顶层，而是中间的某一层或第0层，后者是Mozetic算法失效的情况(与一般诊断效率相同)。Mozetic 改进算法提出了一项动态确定分层描述的技术，即以有用的观测值为基础，通过使用相应结构树重新安排各层给定的分层描述来创建一个新的分层描述。我们把任意分层描述H（由结构抽象建立的）与其结构树相关联，结构树中每个节点代表H的部件，它的子孙代表H的子部件。结构树ST(H)中每个节点c是第i层的一个部件，把部件的描述和观测值（当它们有用时）与节点相结合，因此这个节点就与分层描述H中的一个诊断问题的子集相关联。这样我们就把系统诊断问题看作是对树的遍历问题。需要强调的是，它不是修剪Mozetic算法中的分层描述，而是以观测值为指导重新动态地安排最初的分层描述技术。使用改进算法，Mozetic算法不理想的情况得到很好的解决。接着我们又进一步提出一个改进算法，在过程TOP-DOWN-DIAGNOSIS开始前，通过有用观测值排除与观测值相关联的节点的不可能候选来增强各层的诊断能力。一般地，任何部件都可能有多个行为模式，由其所聚合的部件也一定有多个行为模式。在给定观测值的情况下，这些行为模式中某些是不可能的候选，应该首先被排除。我们的做法是:先将结构树中与观测值相关联的节点分离出来，若其某一行为模式与观测值不一致，则排除此行为模式，当进行诊断时就不用处理被排除的行为模式了。此外，由于从下向上处理结构树时只考虑被观测的节点，当到达描述部件c的节点时，其所有被观测的子节点已经处理完毕。由于部件c的每个行为模式都是其所聚合的子部件行为模式的抽象，因此，对超部件c的某个行为模式mi，如果mi所聚合的所有子部件的行为模式已经被排除，就可排除c的行为模式mi。注意：进一步改进算法虽然排除了大量的不可能诊断，进一步缩小搜索空间，但是不能保证排除所有不可能候选。
最后，我们介绍用假设的优先权控制诊断，这些优先权是关于假设的部分序，并按顺序扩展到关于假设集的优先关联，通过选择极好假设集来减少可能诊断数量。通过搜索和剪枝Reiter碰树策略来计算优先解决方案，进而计算优先诊断。另外单一优先解决方案也可通过快速hill-climbing策略计算。
As we know, diagnostic systems constructed using traditional expert system approaches have many shortcomings, such as knowledge collected in completeness, system dependent on knowledge,and the difficulties of the acquisition of necessary diagnostic knowledge from experts etc. Model-based diagnosis is a new reasoning technique of intelligent diagnosis to overcome the shortcomings of traditional diagnostic methods model-based diagnostic method employ the model of the internal structure and behavior model of system, this method is strongly device independent,at the same time,this approach can be less costly to use.For example,the mode is given when product is designed and manufacturing at first;this mended model of product is use to that model of product. At the same time, the designing and exploitation cycle of product are shorten. Model-based diagnosis is attracted by many researchers in 1980s because of advantages and actual value. It has been an active branch in the area of artificial intelligence until 1990s.
Two representative logical methods of model-based diagnosis are consistency-based diagnosis and abductive diagnosis. Consistency- based diagnosis deal with normal behavior models of system, and abductive diagnosis deal with faulty behavior models of system. No matter what it is , model representation of high efficiency is important to improve efficiency of diagnosis when we deal with complicated device. Hierarchical model-based diagnosis is put forward based on above two diagnosis methods. It’s basic idea is hierarchical representation to diagnose by abstraction. The new system of abstraction have been obtained only contains finite components of abstract level. But when the components of abstraction level is elaborated to the detail levels, Candidates are made in relation to abstract levels. Such can reduce the search space of diagnosis system in order to improve diagnosis efficiency.
We studied the work of “Mozetic”, in which he defined three typical behavior abstraction operators mainly, including refinement/ collapse of values, introduction/deletion variables, eleboration/ simplification of mapping, and defined to model
representation of detail levels in the diagnoses system which is mapped to abstraction level. The faults are isolated one level at a time by hierarohical diagnoses. At the same time, consistency condition is introduced . The first is restriction of incompleteness given a detailed level mapping m the condition, it prohibits cases where an x with an abstraction is mapped to a y without an abstraction; the second is preservation for mapping, which basically says that diagnosis which are impossible at the abstract level(where the search space is smaller)are impossible at the detailed level as well. Restriction of imcompleteness and impossible diagnoses are filtered by consistency condition in the process of diagnosis. Based on hierarohyical representation, restriction of imcompleteness and impossible diagnoses are filtered by consistency condition in the process of diagnosis, and searching space is reduced.
Moreover, model representation is defined that any state of diagnoses system is mapped to observation of input and output system, which is different from consistency-based diagnoses and abductive diagnoses. We implement the algorithm using Borland Jbuider and tested it on or gate of circuit.
Diagnoses system is represented hierarchically by abstraction in order to reduce the number of component of abstracted level and eliminate impossible diagnoses in detailed level by consistency condition to reduce diagnostic costs. But it is a kind of static 、 single、and fixed technics of hierarchical representation, which make no use of the set of currently available observations so that the efficiency of hierarchical diagnoses is not high. Such as,when the most abstraction level is not the top level but certain middle level and level 0.The level 0 is the worst condition (which is the same as general diagnoses efficiency.) We improve Mozetic’s algorithm by structure abstractio
引文
[1] Hamscher W,Console L,de Kleer J eds. Readings in Model-based Diagnosis. San Mateo:Morgan Kaufmann Publishers,1992.
[2] Bublin S,Kashyap R L. Diagnosing electronics system by integrating models of normal and abnormal functioning. Proceedings of IEA/AIE-89,1989,pages 79-84.
[3] Console L, Friedrich G eds. Model-based Diagnosis. Basel-Switzerland, Science Publishers, 1994.
[4] 李占山，姜云飞. 基于模型诊断推理的回顾与展望. 计算机科学, 1988,6(25):54-57.
[5] de Kleer J. Local methods for localizing faults in electronic circuits. MIT Aritificial Intelligence Loboratory,1976.
[6] Patil R, Szolovits P, Schwartz W. Causal understanding of patient illness in medical diagnosis. Proceedings of IJCAI-81,BC 1981,pages 893-899.
[7] Davis R,Shrobe H,et al. Diagnosis based on structure and function. Proceedings of AAAI-82,1982,pages 137-142.
[8] Reiter R. A theory of diagnosis from first principles. ArtificialIntelligence,1987, 32:57–95.
[9] Apt K R, eds. Meta-logic and logic programming.Morgan Kaufmann Publisher, 1995.
[10] Nejdl W, eds. A formal framework for representings diagnosis stratigies in model-based diagnostic system. IJCAI,1995.
[11] Sacerdoti Earl D. Planning in a hierarchy of abstraction spaces. Artificial Intelligence,1974,5:115-135.
[12] Giunchiglia F,Walsh T. Abstract theorem proving. Joint Conf on Artificial Intelligence, IJCAI-89,1989,11:372-377.
[13] Plaisted D A. Theorem proving with abstractions. Artificial Intelligence,1981, 16:47-108.
[14] Genesereth M R. The use of design descriptions in automated diagnosis. Artificial Intelligence,1984,24:411-436.
[15] Tenenberg J D. Abstraction in Planning. Ph.D. Thesis, Computer Science Department,
University of Rochester, 1988.
[16] Autio K. Abstraction of behavior and structure in model-based diagnosis. Proceedings of the Sixth International Workshop on Principles of Diagnosis (DX-95), 1995,pages 1-7.
[17] Davis R. Diagnostic reasoning based on structure and behavior. Artificial Intelligence,1984,24:347-410.
[18] De Kleer J, Williams B C. Diagnosing multiple faults. Artificial Intelligence,1987,
32:97–130.
[19] de Kleer J, Williams B C. Diagnosis with behavioral modes. Proceedings of IJCAI-89, 1989,pages 1324-1330.
[20] Struss P and Dressler O. Physical negation- Integrating fault models into the general diagnostic engine. Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI),1989, pages 1318–1323.
[21] 欧阳丹彤 姜云飞. 刻画基于模型的中心诊断. 软件学报, 1999,10(1):73-77.
[22] 欧阳丹彤 姜云飞. 基于扩展的因果理论的鉴别诊断. 软件学报, 1999,10(7):719-723.
[23] Hamscher W. Model-based troubleshooting of digital systems.MIT Technical Report AI-TR 1074. MTI Press,Cambridge,MA,1988.
[24] Kumar A N and Upadhyaya S J. Automating representation. Working Pagers of the Second AAAI Workshop on Model Based Reasoning,1990,pages 124-130.
[25] Lbath R, Giambiasi N and Delorme C. Framework for reasoning about hierarchical/ multi-vues models. In: System Fault Diagnostics, Reliability and Related Knowledge- Based Approaches, 1987,2: 29-41.
[26] Igor Mozetic. Reduction of diagnostic complexity through model abstraction. In Working Notes “First International Workshop on Principles of Diagnosis”,1990,pages 102-111.
[27] Igor Mozetic and Bernhard Pfahringer. Improving diagnostic efficiency in KARDIO: abstractions, constraint propagation,and model compilation. In Keravnou E.(ed.), Deep Models for Medical Knowledge Engineering,1992,pages 1-25.
[28] Heintze N, Michaylov S ,Stuckey P. CLP(R) and some electrical engineering problems. Proceedings of 4th International Conference on Logic Programming,1997,pages675-703.
[29] Igor Mozetic Hierarchical Model-Based Diagnosis.In W.Hamscher L.Console, de Kleer J, editors,Readings in Model-Based Diagnosis,1992,pages 354-372.
[30] Luca Chittaro, Roberto Ranon. Hierarchical diagnosis guided by observation. IJCAI, 2001.
[31] Hamscher W C.Modeling digital circuits for troubleshooting. Artificial Intelligence, 1991,51:223-271.
[32] Peter Struss. What's in sd? towards a theory of diagnosis. In Hamscher W, Console L, de Kleer J, editors, Readings in Model-based Diagnosis, 1992,pages 419-449.
[33] Autio K and Reiter R. Structural abstraction in model-based diagnosis. Proceedings of the Thirteenth European Conference on Artificial Intelligence,1998,pages 269–273.
[34] Poole D. Noramlity and faults in logic-based diagnosis, Proceedings of 11th International Joint Conference On Artificial Intelligence,1989, pages 1304-1310.
[35] Reggia J A,Nau D S and Wang P Y. Diagnostic expert system based on a set covering model, International Joint Man-Machine Stud,1983,19:437-460.
[36] de Kleer J, Mackworth A K, and Reiter R. Characterizing diagnoses and
systems.Artificial Intelligence, 1992,56:197–222.
[37] Console L and Torasso P. Integrating model of the correct behavior into abductive diagnosis. Proceedings of 9th European Conference On Artificial Intelligence,1990, pages 160-166.
[38] Junker U and Brewka G. Handling partially ordered defaults in TMS. in: Kruse R and Siegel P,eds.Symbolic and Quantitative Aspects for Uncertainty, Proceedings of European Conference ECSQAU, 1991,pages 211-218.
[39] Junker U. Relationships between assumptions.Dissertation,University of Kaiserslautern, Kaiserslautern 1992.
[40] Geffner H and Pearl J. Conditional entailment: Bridging two approaches to default reasoning. Artificial Intelligence,1992, 53(2-3): 209-244.
[41] de Kleer J. Local methods for localizing faults in electronic circuits. MIT AI Memo 394,Cambridge,MA,1976.
[2] Bublin S,Kashyap R L. Diagnosing electronics system by integrating models of normal and abnormal functioning. Proceedings of IEA/AIE-89,1989,pages 79-84.
[3] Console L, Friedrich G eds. Model-based Diagnosis. Basel-Switzerland, Science Publishers, 1994.
[4] 李占山，姜云飞. 基于模型诊断推理的回顾与展望. 计算机科学, 1988,6(25):54-57.
[5] de Kleer J. Local methods for localizing faults in electronic circuits. MIT Aritificial Intelligence Loboratory,1976.
[6] Patil R, Szolovits P, Schwartz W. Causal understanding of patient illness in medical diagnosis. Proceedings of IJCAI-81,BC 1981,pages 893-899.
[7] Davis R,Shrobe H,et al. Diagnosis based on structure and function. Proceedings of AAAI-82,1982,pages 137-142.
[8] Reiter R. A theory of diagnosis from first principles. ArtificialIntelligence,1987, 32:57–95.
[9] Apt K R, eds. Meta-logic and logic programming.Morgan Kaufmann Publisher, 1995.
[10] Nejdl W, eds. A formal framework for representings diagnosis stratigies in model-based diagnostic system. IJCAI,1995.
[11] Sacerdoti Earl D. Planning in a hierarchy of abstraction spaces. Artificial Intelligence,1974,5:115-135.
[12] Giunchiglia F,Walsh T. Abstract theorem proving. Joint Conf on Artificial Intelligence, IJCAI-89,1989,11:372-377.
[13] Plaisted D A. Theorem proving with abstractions. Artificial Intelligence,1981, 16:47-108.
[14] Genesereth M R. The use of design descriptions in automated diagnosis. Artificial Intelligence,1984,24:411-436.
[15] Tenenberg J D. Abstraction in Planning. Ph.D. Thesis, Computer Science Department,
University of Rochester, 1988.
[16] Autio K. Abstraction of behavior and structure in model-based diagnosis. Proceedings of the Sixth International Workshop on Principles of Diagnosis (DX-95), 1995,pages 1-7.
[17] Davis R. Diagnostic reasoning based on structure and behavior. Artificial Intelligence,1984,24:347-410.
[18] De Kleer J, Williams B C. Diagnosing multiple faults. Artificial Intelligence,1987,
32:97–130.
[19] de Kleer J, Williams B C. Diagnosis with behavioral modes. Proceedings of IJCAI-89, 1989,pages 1324-1330.
[20] Struss P and Dressler O. Physical negation- Integrating fault models into the general diagnostic engine. Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI),1989, pages 1318–1323.
[21] 欧阳丹彤 姜云飞. 刻画基于模型的中心诊断. 软件学报, 1999,10(1):73-77.
[22] 欧阳丹彤 姜云飞. 基于扩展的因果理论的鉴别诊断. 软件学报, 1999,10(7):719-723.
[23] Hamscher W. Model-based troubleshooting of digital systems.MIT Technical Report AI-TR 1074. MTI Press,Cambridge,MA,1988.
[24] Kumar A N and Upadhyaya S J. Automating representation. Working Pagers of the Second AAAI Workshop on Model Based Reasoning,1990,pages 124-130.
[25] Lbath R, Giambiasi N and Delorme C. Framework for reasoning about hierarchical/ multi-vues models. In: System Fault Diagnostics, Reliability and Related Knowledge- Based Approaches, 1987,2: 29-41.
[26] Igor Mozetic. Reduction of diagnostic complexity through model abstraction. In Working Notes “First International Workshop on Principles of Diagnosis”,1990,pages 102-111.
[27] Igor Mozetic and Bernhard Pfahringer. Improving diagnostic efficiency in KARDIO: abstractions, constraint propagation,and model compilation. In Keravnou E.(ed.), Deep Models for Medical Knowledge Engineering,1992,pages 1-25.
[28] Heintze N, Michaylov S ,Stuckey P. CLP(R) and some electrical engineering problems. Proceedings of 4th International Conference on Logic Programming,1997,pages675-703.
[29] Igor Mozetic Hierarchical Model-Based Diagnosis.In W.Hamscher L.Console, de Kleer J, editors,Readings in Model-Based Diagnosis,1992,pages 354-372.
[30] Luca Chittaro, Roberto Ranon. Hierarchical diagnosis guided by observation. IJCAI, 2001.
[31] Hamscher W C.Modeling digital circuits for troubleshooting. Artificial Intelligence, 1991,51:223-271.
[32] Peter Struss. What's in sd? towards a theory of diagnosis. In Hamscher W, Console L, de Kleer J, editors, Readings in Model-based Diagnosis, 1992,pages 419-449.
[33] Autio K and Reiter R. Structural abstraction in model-based diagnosis. Proceedings of the Thirteenth European Conference on Artificial Intelligence,1998,pages 269–273.
[34] Poole D. Noramlity and faults in logic-based diagnosis, Proceedings of 11th International Joint Conference On Artificial Intelligence,1989, pages 1304-1310.
[35] Reggia J A,Nau D S and Wang P Y. Diagnostic expert system based on a set covering model, International Joint Man-Machine Stud,1983,19:437-460.
[36] de Kleer J, Mackworth A K, and Reiter R. Characterizing diagnoses and
systems.Artificial Intelligence, 1992,56:197–222.
[37] Console L and Torasso P. Integrating model of the correct behavior into abductive diagnosis. Proceedings of 9th European Conference On Artificial Intelligence,1990, pages 160-166.
[38] Junker U and Brewka G. Handling partially ordered defaults in TMS. in: Kruse R and Siegel P,eds.Symbolic and Quantitative Aspects for Uncertainty, Proceedings of European Conference ECSQAU, 1991,pages 211-218.
[39] Junker U. Relationships between assumptions.Dissertation,University of Kaiserslautern, Kaiserslautern 1992.
[40] Geffner H and Pearl J. Conditional entailment: Bridging two approaches to default reasoning. Artificial Intelligence,1992, 53(2-3): 209-244.
[41] de Kleer J. Local methods for localizing faults in electronic circuits. MIT AI Memo 394,Cambridge,MA,1976.