J. d. Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....D is a diagnosis for S i (1) D [ Sd [ Ctx j V 2Obs abd , 2) D [ Sd [ Ctx [ Obscon 6j . The symbol j denotes the possibly limited reasoning capabilities of a diagnostic agent. I.e: f j j g f j j g. If Obsabd = and Obscon = Obs, then we have a pure consistency based diagnosis [4, 5], and if Obscon = and Obsabd = Obs, we have a pure abductive diagnosis [1] 3. MULTI AGENT DIAGNOSIS In a Multi Agent setting, knowledge of the system S is distributed over the agents, ether semantically and spatially. Combinations are, of course, also possible. The distribution of the ....

J. d. Kleer and B. C. Williams. Diagnosing multiple faults. Arti cial Intelligence, 32:97-130, 1987.

....functions. The key idea underlying model based diagnosis is that a combination of component modes is a possible description of the current state of the spacecraft only if the set of models associated with these modes is consistent with the observed sensor values. Following de Kleer and Williams [9], MI uses a conflict directed best first search to find the most likely combination of component modes consistent with the Helium Fuel Oxidizer inflow = outflow = 0 ###### ##################### #### ##### #### ##### ###### ############################################## ....

J. de Kleer and B. C. Williams, Diagnosing Multiple Faults," Artificial Intelligence, Vol 32, Number 1, 1987.

....Functionality of the ATMS. The ATMS universal propagation algorithm computes the minimal sets of assumptions necessary to derive a given formula. This feature is useful in device diagnosis where one wants to find the minimal number of possible faults that explains a given observed behavior [ de Kleer and Williams, 1987 ] de Kleer and Williams, 1989 ] In fault diagnosis, however, one is often interested in premise sets that contain only a single fault. 3 It should be noted that the ATMS described here is different from the clause management system described in [de Kleer and Reiter, 1987] The clause ....

J. de Kleer and B. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97-- 130, 1987.

....and on a utility function taking into account breakdown, observations, and repair costs. Upon executing an action and obtaining new observations, the diagnostic reasoner updates the candidate set. Friedrich and Nejdl 1992; Sun and Weld 1992) are representative of this approach and generalize (de Kleer and Williams 1987) dedicated to the choice of the best next measurement. An important limit of these works is their dependence on the assumption that every relevant observation can be made reliably when needed, and on very basic reliable repair actions. Furthermore, the examples used for their illustration are ....

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32(1):97--130, 1987.

....it to handle defaults. Examples are Reiter s default logic [Reiter, 1980] Moore s autoepistemic logic [Moore, 1985] McCarthy s circumscription [McCarthy, 1980] In the meantime, truth maintenance systems have been improved. De Kleer developed ATMS [de Kleer, 1986a] and applied it to diagnosis [De Kleer and Williams, 1987], and Goodwin presented an improved TMS [Goodwin, 1987] A first link to our work has been the ATMS extension of Oskar Dressler [Dressler, 1988] Using his nogood inference rule , it became possible to handle negative and disjunctive information by ATMS. However, these developments are not ....

....or cause this malfunctioning. The faults can be broken parts in technical systems or diseases. There are various methods how to find them: The so called consistency based approaches take the normal of the correct behaviour of a system, and use it to predict values for observable attributes (cf. [De Kleer and Williams, 1987], Reiter, 1987] If the predicted values are different to the observed values then contradictions are detected and some parts do not behave as their description. Other approaches use special knowledge about possible faults. In abductive diagnosis, faults are abduced from abnormal observations ....

J. De Kleer and B.C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

.... prime covers are a way for either exhaustively, or concisely, representing the causes of failure of a system [13, 9] In automated reasoning, a prime cover can be either a set of minimal demonstrations, or the normalization of a formula into conjunctive normal form for further computations [12, 15, 16, 23, 18]. The problem of computing (irredundant, minimal) prime covers of Boolean functions was first addressed by Quine in the 1950 s [21] Since that date, efforts have been done to elaborate efficient prime cover computation procedures [22, 17, 2, 24, 25, 14, 3] For some problems, for instance for ....

J. De Kleer, B. C. Williams, "Diagnosing Multiple Faults", in Artificial Intelligence, Vol 32. pp. 97-- 130, 1987.

....must be utilised. A lot of research has been done on planning techniques with some exciting results [1,3,4,6,13,14,16,17] In this paper, a heuristic Partial Order Planner with Universal quantification s and Conditional effects (UCPOP) is described. Model Based Reasoning (MBR) 2, 5, 7 12,15] has been widely accepted as the principal diagnostic technique in several domains. However, little emphasis has been put on its application to network and system management. In MBR, it is assumed that the structure of a system is known and we can use that knowledge to reason about its ....

J.de Kleer and B.C. Williams, Diagnosing multiple faults, Artificial Intelligence, Vol. 32, No. 1, 97-130, April 1987.

.... such systems as an analogue of Reiter s model based theory of diagnosis [13] We chose Reiter s theory as a point of departure because it provides a formal characterization of diagnosis shared to some extent by most of the model based systems described in the literature, including DART [8] GDE [4] and its descendents [5, 9] and the work of Davis [2] Our approach follows from two basic insights: first, the generality of Reiter s theory of diagnosis makes it applicable to other domains; second, a productive analogy exists between the problem of diagnosis and that of reconfiguration. ....

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32(1):97--130, April 1987.

....of computing the prime implicants of a monotone Boolean function in conjunctive normal form. This problem is also investigated in [9] which gives some new results. 6.5. Model based diagnosis. Basic techniques for model based diagnosis have been developed within AI by de Kleer and Williams [13], and by Reiter [45] To introduce the necessary concepts briefly, a system is a pair (SD; COMP ) where SD, the system description, is a set of usually first order sentences and COMP is a set of constants which model the MINIMAL TRANSVERSALS OF A HYPERGRAPH 1301 system components. This general ....

....set of diagnoses for (SD; COMP;OBS) Given C and D for input, deciding if there is an additional diagnosis not contained in D is p m equivalent to co SIMPLE H SAT. The determination of diagnoses from the given minimal conflict sets is essential in popular algorithms for model based diagnosis [45, 13]. Deciding if an already computed set of diagnoses is complete with respect to a given set of minimal conflict sets, i.e. consists of all diagnoses, is an important subproblem if diagnoses are computed incrementally. Therefore, the complexity of the additional diagnosis problem is of crucial ....

J. de Kleer and B. C. Williams, Diagnosing multiple faults, Artificial Intelligence, 32 (1987), pp. 97--130.

.... Ames Research Center (ARC) and the Jet Propulsion Lab (JPL) combines high level planning and scheduling, robust multi threaded execution, and modelbased fault detection isolation and recovery, into an integrated architecture that is able to robustly control a spacecraft over long periods of time [6, 12]. One of the primary components of the Remote Agent architecture is the Livingstone model based health management system. Livingstone is an advanced inference engine that uses a high level declarative model of a physical device to monitor the state of that device, detect off nominal behavior, ....

....off nominal behavior, isolate failures to individual components, and reason about alternative recovery actions. The key benefit provided by Livingstone is the use of a first principles model that describes the behavior of each component within the device and the interactions between the components [6,7,8]. By reasoning generatively about the behavior of the device using the model, Livingstone is able to detect failures whenever a discrepancy occurs between the observations and predictions. In addition, Livingstone is able to use the same model to generate the most likely hypothesis that is ....

J. de Kleer and B. C. Williams, Diagnosing Multiple Faults, Artificial Intelligence, Vol 32, Number 1, 1987.

....Furthermore, our monitoring method is directly applicable to fault diagnosis in dynamic systems. Fault hypotheses can be proposed for monitoring based on initial weak information such as the signs of discrepancies between observations and predictions, by using existing methods such as [ de Kleer and Williams, 1987; Ng, 1991 ] Automatic model building methods can select relevant model fragments from a background knowledge base to express initially weak knowledge about a fault as an SQDE [ Crawford et al. 1990; Rickel and Porter, 1994 ] The observation stream is then used to refine or refute each ....

J. de Kleer and B. C. Williams. Diagnosing Multiple Faults. Artificial Intelligence, 32:97--130, 1987.

....of the rover, MI infers the most likely current state (see Figure 3) MI also provides a layer of abstraction to the executive, allowing plans to be specified in terms of component modes, rather than in terms of low level sensor values. MIR uses algorithms adapted from model based diagnosis [ de Kleer and Williams, 1987; 1989 ] to provide the above functions. MIR extends the basic ideas of modelbased diagnosis by modeling each component as a finite state machine (see Figure 4) and modeling the whole rover as a set of concurrent, synchronous state machines. 4.3 Resource Management Resources on rovers are ....

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:100--117, 1987.

....hypotheses [8, 10] However, they must be given an a priori model of each possible state. One of the strengths of the approach presented in this paper is its ability to construct models from training data and then use them for state identi cation. Qualitative model based diagnosis techniques [2, 6] consider a snapshot of the system rather than its history. In addition, the system state is assumed to be consistent with a propositional description of one of a set of possible states. The presence of noisy data and temporal patterns negates these assumptions. Hidden Markov Models have been ....

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artif. Intelligence, 32:100-117, 1987.

....25000 30000 10 20 30 40 50 60 70 80 90 100 Number of system nodes in the ROBDD Number of system variables in the domain 2 faults 3 faults 4 faults 5 faults Figure 8: The number of nodes in the layers containing system nodes as a function of the number of system variables. works like [de Kleer and Williams, 1987] and [Williams and Nayak, 1996] For instance, in circuit diagnosis [de Kleer and Williams, 1987] uses a logical model of the system to be diagnosed and determines the next action based on expected Shannon entropy. To calculate the expected Shannon entropy they require the conditional probability ....

....of system variables in the domain 2 faults 3 faults 4 faults 5 faults Figure 8: The number of nodes in the layers containing system nodes as a function of the number of system variables. works like [de Kleer and Williams, 1987] and [Williams and Nayak, 1996] For instance, in circuit diagnosis [de Kleer and Williams, 1987] uses a logical model of the system to be diagnosed and determines the next action based on expected Shannon entropy. To calculate the expected Shannon entropy they require the conditional probability of a set of failed components (termed a candidate in [de Kleer and Williams, 1987] given some ....

[Article contains additional citation context not shown here]

de Kleer, J. and Williams, B. (1987). Diagnosing multiple faults. Articial Intelligence, 32(1):97-130.

....hypotheses [8, 10] However, they must be given an a priori model of each possible state. One of the strengths of the approach presented in this paper is its ability to construct models from training data and then use them for state identification. Qualitative model based diagnosis techniques [2, 6] consider a snapshot of the system rather than its history. In addition, the system state is assumed to be consistent with a propositional description of one of a set of possible states. The presence of noisy data and temporal patterns negates these assumptions. Hidden Markov Models have been ....

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artif. Intelligence, 32:100--117, 1987.

....In the AI literature on hypothetical reasoning there are relatively few results on the design of experiments for discriminating between competing hypotheses, or on the conclusions one may draw from the outcome of an experiment. There are exceptions, of course. Among these are de Kleer and Williams [4] who provide a probabilistic analysis to decide what measurement to take next. The DART system of Genesereth [5] was capable of proposing circuit inputs and observations to be made in order to confirm or refute a possible diagnosis. TraumAID (Webber et al. 18] is a system for treating trauma ....

....the system under analysis. For example, in the case of circuits, Sigma might describe the individual circuit components, their normal input output behaviour, their fault models, the topology of their interconnections, and the legal combinations of circuit inputs (e.g. deKleer and Williams [4], Reiter [14] We also assume a fixed set HY P of hypotheses. In the case where Sigma describes a circuit, HY P might be the set of diagnoses which we currently hold for this device. How we arrived at the set HY P will be largely irrelevant for our purposes. HY P could be a set of abductive ....

[Article contains additional citation context not shown here]

J. de Kleer and B.C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

....branches of work for state estimation and fault diagnosis are Kalman filters from control theory and qualitative model based diagnosis from artificial intelligence. State estimation for rovers exceeds the capabilities of the current approaches. Qualitative model based techniques for diagnosis [2, 6] rely on the system transitioning occasionally from one steady state to another. Rovers receive rapidlychanging streams of continuous valued sensor data. In addition, the model based techniques often rely on a snapshot of the system, disregarding history. But in fact the history may be critical to ....

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artif. Intelligence, 32:100--117, 1987.

....on hypothetical reasoning there are relatively few results on the design of tests for discriminating a space of hypotheses, or on the conclusions one may draw from the outcome of a test. There are exceptions of course, particularly in the area of diagnosis. Among these are de Kleer and Williams [ de Kleer and Williams, 1987 ] who provide a probabilistic analysis to decide what measurement to take next. The DART system of Genesereth [ Genesereth, 1984 ] was capable of proposing circuit inputs and observations to be made in order to confirm or refute a possible diagnosis. TraumAID [ Webber et al. 1990 ] is a system ....

....knowledge describing the system under analysis. For example, in the case of circuits, Sigma might describe the individual circuit components, their normal input output behaviour, their fault models, the topology of their interconnections, and the legal combinations of circuit inputs (e.g. de Kleer and Williams, 1987 ] Reiter, 1987 ] We also assume a fixed set HY P of hypotheses. In the case where Sigma describes a circuit, HY P might be the set of diagnoses which we currently hold for this device. How we arrived at the set HY P will be largely irrelevant for our purposes. HY P could be a set of ....

[Article contains additional citation context not shown here]

J. de Kleer and B.C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

....the batteries. Either way, my flashlight is now emitting a light, which was my ultimate goal. Traditionally, the AI research on diagnosis has focused on the problem of determining a set of candidate diagnoses, given a description of system behavior and an observation of aberrant behavior (e.g. (de Kleer and Williams, 1987), Reiter, 1987) Testing could subsequently be performed to acquire sufficient discriminatory observations in order to identify a unique diagnosis. Recently, some researchers have cast diagnostic problem solving in a more purposive role (e.g. Provan and Poole, 1991) Friedrich and Nejdl, ....

J. de Kleer and B. Williams (1987). Diagnosing multiple faults. In Artificial Intelligence Journal 32:97--130.

....with uncertainties. It is necessary to let the ATMS have the ability to cope with uncertainty problems. In order to overcome this problem, some research on the integration of symbolic reasoning with numerical inference has been carried out to associate numerical uncertainties with ATMS [3] 4] [7], 10] 13] 14] 15] 16] 20] 21] In [7] De Kleer and Williams use probability theory to deal with uncertainty associated with assumptions. In [10] 15] the authors use possibilistic logic to handle this problem. In [10] both assumptions and justifications are associated with ....

....have the ability to cope with uncertainty problems. In order to overcome this problem, some research on the integration of symbolic reasoning with numerical inference has been carried out to associate numerical uncertainties with ATMS [3] 4] 7] 10] 13] 14] 15] 16] 20] 21] In [7], De Kleer and Williams use probability theory to deal with uncertainty associated with assumptions. In [10] 15] the authors use possibilistic logic to handle this problem. In [10] both assumptions and justifications are associated with uncertainty measures. The uncertainty values associated ....

[Article contains additional citation context not shown here]

de Kleer,J. and B.C.Williams, Diagnosing multiple faults, Artificial Intelligence 32 (1987) 97-130.

....reason about their physical environment [33] to engineering artifacts that can reason about physics [49] Its main goals are the prediction of behavior, analysis, design, control, monitoring, and fault diagnosis. Many QR programs, particularly the ones that perform monitoring [21] or diagnosis [18], infer the behavior of a physical system from its structure or vice versa [15] What distinguishes QP from other formalisms that represent physics knowledge, such as differential equations, is the abstraction to a qualitative level. For example, so called landmarks divide the continuum of real ....

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32(1), 1987.

....problem domain, the performance of such algorithms can be bounded in advance. We present empirical results comparing SAB with two modelbased algorithms, MBD1 and MBD2, for the task of finding one or all minimal cardinality diagnoses. MBD1 uses the same computing strategy as algorithm GDE [ 9 ] . MBD2 adopts a breadth first search strategy similar to the algorithm DIAGNOSE [ 24 ] The main conclusion is that for nearly acyclic circuits, such as the N bit adder, the performance of SAB being linear provides definite advantages as the size of the circuit increases. 1 Introduction ....

....results in an algorithm called structure based abduction (SAB) 12 ] which will be empirically investigated here. The performance of SAB is compared with two model based diagnosis algorithms called here MBD1 and MBD2. MBD1 uses the same strategy for finding predictions and conflicts as GDE [ 9 ] . MBD2 is a focused version of MBD1 geared toward computing minimal cardinality diagnoses. MBD2 adopts a breadth first search strategy for searching the hypotheses space and then computes conflicts sequentially, as in DIAGNOSE by Reiter [ 24 ] The main contribution of this work is in ....

[Article contains additional citation context not shown here]

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

....to bridge faults in the topology itself. This is one of the landmark papers in the area of model based reasoning. Quite a bit of work has followed Davis examples and theories. De Kleer and Williams published a landmark paper on MBR for diagnosis describing GDE, the General Diagnostic Engine [de Kleer and Williams 87] GDE infers behaviour test device model observed behaviour behavioural discrepancies predicted behaviour correctness assumptions Generate Not to be Quoted without Author Permission 8 A Perspective on Explanation in Diagnosis from device structure and functionality. It is applied to ....

....its behaviour as explicitly as RATIONALE since causality is more difficult to establish in JETA. However, both approaches support the user with contextual help while a diagnosis is achieved. It is more difficult to visualize contextual help in model based diagnosis. If we follow the de Kleer [de Kleer and Williams 87] approach which represents a device with functionality as a set of components with behaviour. The device can be diagnosed by assuming a faulty component and enumerating the behavioural states that the fault propagates in the remainder of the device. This is compared to the behaviour that a ....

de Kleer, J., Williams, B.C., "Diagnosing Multiple Faults", Artificial Intelligence, vol. 32, (1987).

....categories of failure. He also advocates the use of the relaxation of assumptions to bridge faults in the topology itself. Quite a bit of work has followed Davis examples and theories. De Kleer and Williams published a key paper on MBR for diagnosis describing GDE, the General Diagnostic Engine [de Kleer and Williams 87] GDE infers behaviour from device structure and functionality. It is applied to digital circuits and makes use of an ATMS (Assumption Based Truth Maintenance System) This work forms the cornerstone of ATMS based model based reasoning systems. It was followed by many papers that criticized the ....

....Learning the device component model, its behaviour and functionality using the FBR knowledge provides the technician with a tool that can achieve model based diagnosis. For these reasons it was concluded that the Hakeem algorithm should be implemented. Background If we follow the de Kleer [de Kleer and Williams 87] approach which represents a device with functionality as a set of components with behaviour. The device can be diagnosed by assuming a faulty component and enumerating the behavioural states that the fault propagates in the remainder of the device. This is compared to the behaviour that a ....

de Kleer, J., Williams, B.C., "Diagnosing Multiple Faults", Artificial Intelligence, vol. 32, (1987).

....OBS OK(C i ) C COMPS . 1) The diagnosis procedure is then organized as a search for revised mode assignments to the components that eliminate inconsistency (see e.g. Reiter, 1987] SD OBS mode i (C i ) C COMPS . 2) Implementations such as GDE (modeling correct behavior only, [de Kleer and Williams, 1987]) and GDE (exploiting also fault models, Struss and Dressler, 1989] employ an Assumption based Truth Maintenance System (ATMS) in order to record inferential dependencies and to determine (minimal) sets of mode assumptions that conflict with the observations and compute diagnosis candidates as ....

de Kleer, J. and Williams, B. C., Diagnosing Multiple Faults. Artificial Intelligence, 32:97-130, 1987.

....unexpected behavior is identified (fault isolation) based 8 on analytical redundancy [16, 70, 99, 245] or symbolic techniques [34, 112, 182] and then the failure is accommodated based on a post fault system model. Related work on failure detection, identification and recovery may be found in [34, 113, 172, 183, 189, 244]. From the control system theory point of view, failure detection and isolation methods may be broadly classified as: i) methods which use a system model, and, ii) methods which do not use a system model. Gertler in [79] presents a thorough review of model based and model free approaches to FDI. ....

....G. and Pattipati, K. Resource Allocation and Performance Evaluation in Large Human Machine Organizations , IEEE Transactions on Systems, Man and Cybernetics, Vol. 21, No. 3, May June 1991. 112] Kleer, J. D. An Assumption Based FMS , Artificial Intelligence, Vol. 28 , 1986, pp. 127 162. [113] Kleer, J. D. and Williams, B. C. Diagnosing Multiple Faults , Elsevier Science Publishers, Vol. AI 32 , 1987. 114] Kludge, W. E. and Lautenbach, K. The Orderly Resolution of Memory Access Conflicts Among Conflicting Channel Processes , IEEE Transactions on Computers, Vol. C 31, No. 3, ....

Kleer, J. D., and Williams, B. C., "Diagnosing Multiple Faults", Elsevier Science Publishers, Vol. AI-32 , 1987.

....corresponding formal theory of diagnosis. DNSB diagnosis has been developed in the context of troubleshooting in electronic circuits [Davis Hamscher, 1988] A well known program that supports DNSB diagnosis, and includes various strategies to do so eciently, is the General Diagnostic Engine (GDE) De Kleer Williams, 1987; Forbus De Kleer, 1993] Above, we have reviewed the conceptual basis of diagnosis based on a model of normal 3 CONCEPTUAL BASIS OF DIAGNOSIS 8 real world observed ndings observation model of abnormal behaviour predicted ndings prediction match Figure 3: ....

....to provide a formal underpinning of diagnostic problem solving using knowledge of the normal structure and behaviour of technical devices, i.e. DNSB diagnosis. The theory of diagnosis may be viewed as the logical foundation of earlier work in DNSB diagnosis by J. de Kleer et al. De Kleer, 1976; De Kleer Williams, 1987] Brown and colleagues [Brown et al. 1982] R. Davis and H. Shrobe [Davis Shrobe, 1983; Davis, 1984] and M.R. Genesereth [Genesereth, 1984] The logical formalisation uses results from earlier work by R. Reiter, Reiter, 1980] and J. McCarthy, McCarthy, 1986] on nonmonotonic reasoning. We ....

J. de Kleer and B.C. Williams (1987). Diagnosing multiple faults. Articial Intelligence, 32, 97-130.

....tricks (like constraints) can be seen as the form of abduction that is carried out in the logic induced by the semantics of the arrow and negation as failure operators. Recently, much attention has been paid to abductive reasoning in different fields of Artificial Intelligence, such as diagnosis [4, 5, 13, 12], interpretation of natural language sentences [7, 15] plan recognition [1] scene interpretation [14] This work is a contribution to the characterization of abduction as a form of inference whose context is only the logic it depends upon. Abstraction from any particular problem helps to focus ....

J. de Kleer and C. B. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

.... the early 1980s, diagnosis and treatment have been central problems in theoretical and applied AI [2] Given that a device is not working properly or a patient has some complaint, the automated diagnostic system is charged with determining the set of faults or diseases that explain the symptoms [3, 4, 6, 1]. The diagnostician is able to ask questions about the behavior of the device, or test individual components in order to determine if they are working properly. As new information is gained, the procedure updates its current view of the world. Inference focuses on identifying the set of faults ....

....seeks to identify the most likely causes of a malfunction, but also generates a plan of action for repair. This plan consists of repairing or replacing individual components of a composite device or system, as well as making observations or tests. We and others call this process troubleshooting [3]. Optimal Troubleshooting and Decision Trees An optimal troubleshooting plan is a sequence of observations and repairs that minimizes expected costs. The classic way to compute the expected cost of a plan is to use a decision tree. 1 In this section, we show how this computation is done. In the ....

J. de Kleer and B. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

....methods (e.g. Dubuisson [2] These approaches are mainly concerned with fault detection. In total contrast, some diagnosis methods oriented towards fault isolation focus on a logical model of the system and perform abductive inference or consistency analysis (e.g. De Kleer and Williams [3], Poole [4] Struss and Dressler [5] Console Torasso [6] In a similar spirit, yet another class of approaches exploits causal information between failures and their symptoms; the causal knowledge is then represented as a Bayesian network (e.g. Pearl [7] or more simply in terms of causal ....

De Kleer J, Williams B., "Diagnosing Multiple Faults", Artificial Intelligence, 32, 1987, 97-130.

....clearly separates the model interpretation from the lattice search. The TP function does not require ATMS style dependency recording and can be easily realized by a logic programming system like Prolog. This also avoids incomplete constraint propagation which occurs in most ATMS based systems [6]. Furthermore, without any change to the diagnostic algorithm, TP can be realized by different instances of the Constraint Logic Programming (CLP) scheme [10] depending on the domain of application. In section 2 we give a new characterization of models, diagnoses and conflicts, and show how to ....

....4 the example is expanded and experimental results are compared to de Kleer s HTMS based system [4] The basic algorithm which computes all minimal diagnoses is described in section 3. In contrast to most consistency based approaches where minimal diagnoses are computed from conflicts (e.g. [18, 6]) our algorithm computes minimal diagnoses directly from diagnoses. Conflicts are computed as a side effect, and are used to prune the search space. In section 4 we show that the algorithm computes the k 1 st minimal diagnosis in time O(n 2k ) where n is the number of the model components. ....

[Article contains additional citation context not shown here]

de Kleer, J., Williams, B.C. Diagnosing multiple faults. Artificial Intelligence 32 (1): 97--130, 1987.

....is to assume each component to be faulty which does not produce the expected behavior. The second is to explicitly define faults by introducing fault models. Model based diagnosis currently concentrates on finding single or multiple device faults at hardware troubleshooting [Davis 1984, de Kleer and Williams 1987, Reiter 1987, de Kleer et al. 1990] A detailed description of the structure and behavior of devices enables to use more principled methods of reasoning. That s what [Davis 1983] called reasoning from first principles. This works fine for technical domains as the correct or faulty behavior of ....

Kleer J.de, Williams B.C. (1987): Diagnosing Multiple Faults, Artificial Intelligence, 32(1)97-130.

....formal methods is a way to overcome these problems, and formalization becomes essential when we have to ensure that system specifications are met. This work presents a conceptual model and a formal model of GDE( Struss Dressler, 1989] which is an extension of GDE (General Diagnostic Engine) [de Kleer Williams, 1987]. A proof is given which shows that the high level specification of GDE( is met by the conceptual and formal model presented. The work has been carried out in cooperation with Dr. Manfred Aben of the SWI department of the University of Amsterdam. Dr. Aben has developed a framework for formally ....

....in a uniform way. The analysis identifies the relevant goals (tasks) in diagnosis, and the way in which these goals can be achieved (by problem solving methods) We showed how the analysis can be used to generate conceptual models of diagnostic strategies [Benjamins Jansweijer, 1994] including GDE [de Kleer Williams, 1987] and GDE [Struss Dressler, 1989] However, a conceptual model is only a first step in the development process of a knowledge based system. Current software and knowledge engineering approaches acknowledge this gap between a conceptual model its implementation, and, to bridge it, they view KBS ....

de Kleer, J. & Williams, B. (1987). Diagnosing multiple faults. Artificial Intelligence, 32:97--130.

....knowledge about how components are structured and work normally. There is no knowledge as to how malfunctions occur and manifest themselves. Diagnosis consists of isolating deviations from normal behaviour. This has normally been the preserve of consistency based 1 approaches [ Genesereth, 1984, de Kleer, 1987 ] 2. We have just information on faults (diseases) and their symptoms, and want to account for abnormal observations. This has traditionally been 1 This term and abduction are used as technical terms defined in section 2. the preserve of abductive approaches [ Popl, 1973, Reggia, 1983, ....

....case. the) diagnoses. As one would expect the sort of knowledge that has to be specified for each is different. 2.1 What does normal mean There are two different meanings that have been associated with the notion of normality. 1. A component is normal if it works correctly all of the time [ de Kleer, 1987 ] We never conclude some component is normal (it may act abnormally in the next observation) Saying a component is abnormal gives no information. 2. We localise our discussion to a particular case. By normal we mean that the correct answer is being produced in this case. We should parameterise ....

[Article contains additional citation context not shown here]

J. de Kleer and B. C. Williams, "Diagnosing Multiple Faults", Artificial Intelligence, Vol. 32, No. 1, pp. 97-130.

....diagnoses are the only ones we have to know for determining the best point in which to make a further measurement in order to reduce the conflict sets number. 1 INTRODUCTION Recently, a new diagnostic paradigm, Model Based Diagnosis (see [1] for a survey) moved from its theoretical foundation [3][6] to some practical application [7] In almost every model based approach, diagnoses are found from discrepancies between observation and prediction. The intermediate step is the exhaustive generation of the conflict sets for the system (SD,COMPS,OBS) in which the System Description and the ....

de Kleer, B., C., Williams, Diagnosing Multiple Faults, Artificial Intelligence 32, pp. 97-130, 1987.

....not just flat sets of atomic pieces of information but alternative proof structures. A major conceptual improvement would be that of automatic extraction of advice by the system, regarding the future course of the inquiry. We could apply some techniques (like the minimum entropy one, see [35]) to drive the acquisition of new evidence regarding the case under consideration, in order to further differentiate among the goods. ....

J. de Kleer and B. C. Williams, Diagnosing Multiple Faults, Artificial Intelligence, 32(1), 1987.

....and many others aspects yet not deeply investigated, like program debugging. The first efforts to formalize explanation based reasoning come from the area of diagnostic problem solving, where two fundamental approaches have been devised: the consistency based approach pioneered by de Kleer [6] and formalized by Reiter [30] and the abductive approach introduced by Reggia[29] Cox and Pietrzykowsky [5] and Poole [28] Poole [26] and successively Console and Torasso [4] have shown under which restrictions the two approaches compute the same explanations, and Konolige [18] generalized ....

J. de Kleer and C. B. Williams. Diagnosing multiple faults. Artificial Intelligence 32:97--130, 1987.

....is oil below the car. The authors study two abduction problems corresponding to this DP : 1. Psi = foil below car(present)g and Psi Gamma = fg (Poole s view of a diagnostic problem [22] with minimal solution W 1 = fholed(oil cup)g. 2. Psi = Psi Gamma = fg (De Kleer s DP view [4]) with minimal solution W 2 = fg. To solve abduction problem 1 it is necessary to add the following rules: not oil below car(present) correct(oil cup) not ab(oil cup) holed(oil cup) ab(oil cup) fault mode( oil cup, holed ) The above program has only one minimal revision fab(oil cup) ....

J. de Kleer and B.C. Williams. Diagnosing multiple faults. AI, 32:97--130, 1987.

....there is a general diagnosing algorithm to compute diagnosis candidates, which actually serves as a logical foundation of general diagnosis Published in: John Stewman (ed. Proc. 8th Florida Artificial Intelligence Research Symposium, Melbourne (FL, US) 1995, pp123 127 systems such as GDE [3]. As Selman and Levesque showed in [15] it is NP hard to decide if there is an explanation containing a given hypothesis. In order to generate diagnosis candidates more practically and quickly, some efforts have been reported in, e.g [4, 12, 5] In Reiter [14] it was assumed that the description ....

....description may vary. Multiple system models seem to play a central role in diagnosis as a process. Under different diagnostic assumptions or taking different diagnostic strategies, we may have different system models [16, 17, 7] After a certain system model is picked up, we can follow, e.g. [14, 3] to compute the diagnosis candidates. This paper discusses the use and importance of category theory in system descriptions for modelbased diagnostic reasoning. We argue that category theory provides a precise and convenient conceptual language and tool to model the process of diagnosing complex ....

de Kleer, J. and Williams, B.C., Diagnosing multiple faults, Artificial Intelligence, 32, 1987, 97130

....mechanism which systematically generates subsets of Comps, with minimal cardinality first, is too inefficient for systems with large numbers of components. Instead, Reiter [1987] proposes a diagnostic method based on the concept of a conflict set, originally due to de Kleer [1976] De Kleer and Williams [1987] have independently implemented General Diagnostic Engine (GDE) which effectively realizes the above ideas. In the abductive approach [Poole, 1989] SD contains just different modes of behavior and does not distinguish between the normal and abnormal behavior. An abductive diagnosis is then a ....

....structure and behavior A model interpreter is domain independent, but depends on the representation formalism chosen for modeling. Reasoning is typically based on theorem proving if a model is represented by first order logic [Reiter, 1987] or on constraint propagation coupled with an ATMS [de Kleer and Williams, 1987]. Alternatively, one can represent and interpret models by logic programs [Lloyd, 1987] or by constraint logic programs [Jaffar et al. 1986, Cohen, 1990] The origin of this representation paradigm goes back to the KARDIO model [Bratko et al. 1989] similar representation was proposed by ....

[Article contains additional citation context not shown here]

de Kleer, J., Williams, B.C. Diagnosing multiple faults. Artificial Intelligence 32, pp. 97-130, 1987.

....between the cause and the contextual condition) to produce an effect and that the start of the effect is delayed. In a similar way behavior formulae can be used to express the component oriented behavioral models typical of the model based tradition (since dart, 16] ht [9] gde, and [11]) In this case some selected atoms in the behavior formulae denote a mode of behavior of the components being modeled and each formula describes a rule of behavior (usually input output behavior) of a component. In this case the temporal constraints express delays in such a behavior (e.g. delays ....

J. de Kleer and B.C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32(1):97--130, 1987.

....provide a brief historical background to abduction. Then, we define our model of abduction problems and show how it applies to other theories of abduction. Next, we describe our complexity results, proofs 1 The latter constraint is not the same as eliminating candidates in de Kleer Williams [6] or inconsistency in Reiter [26] If a hypothesis is insufficient to explain all the observations, the hypothesis is not ruled out because it can still be in composite hypotheses. of which are given in the appendix. Finally, we consider the relationship of these results to one abduction ....

....which this task remains intractable. In contrast to maintaining explicit links between hypotheses and data, Davis Hamscher s model based diagnosis [5] determines at run time what data need to be explained and what hypotheses can explain the data. Much of this work, such as de Kleer Williams [6] and Reiter [26] place an emphasis on generating all minimal composite hypotheses that explain all the data. However, there can be an exponential number of such hypotheses. Current research is investigating how to focus the reasoning on the most relevant composite hypotheses [7, 8, 30] ....

[Article contains additional citation context not shown here]

J. de Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32(1):97--130, 1987.

....these knowledge types in diagnosis should be investigated. In other words the FAN diagnostic method might be improved. ffl Its formalisation offers the possibility to compare the FAN reasoner to other diagnostic reasoners described in the literature and expressed in the framework (e.g. 2, 7] [3, 8, 4]) To enable further comparisons alternative representations of the FAN notions can be analysed. A link between phenomena A and B can be expressed propositionally as A B when it is pathognomonic, as A B when it is obligatory and as A ff B when it is facultative. The obligatory and the ....

J. H. de Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

....rather than in terms of low level sensor values. MR supports the run time generation of novel reconfiguration actions to return components to the desired mode or to re enable high level capabilities such as able to produce thrust . Livingstone uses algorithms adapted from model based diagnosis [11, 12] to provide the above functions. The key idea underlying model based diagnosis is that a combination of component modes is a possible description of the current state of the spacecraft only if the set of models associated with these modes is consistent with the observed sensor values. Following de ....

J. de Kleer and B. C. Williams, Diagnosing Multiple Faults, Artificial Intelligence, Vol 32, Number 1, 1987.

....of Karlsruhe, D 76128 Karlsruhe hnamei ira.uka.de http: i12www.ira.uka.de hnamei In [11] Reiter laid the foundations of the formal theory of an approach to diagnosis known as diagnosis from first principles. His work was based on the work of many researchers, notably that of de Kleer [4] and Genesereth [5] In diagnosis from first principles, we have a logic based description of some system (e.g. a circuit) and an observation of the system s behavior. We then try to find a set of components in the system which, when assumed to be abnormal, explains the discrepancy between the ....

J. de Kleer and B. Williams. Diagnosing multiple faults, Artificial Intelligence, 56:197--222, 1987.

No context found.

J. d. Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

No context found.

J. Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32(1):97--130, 1987.

No context found.

J. d. Kleer and B. C. Williams. Diagnosing multiple faults. Artificial Intelligence, 32:97--130, 1987.

No context found.

J. de Kleer, B. Williams, Diagnosing Multiple Faults, Artificial Intelligence, 1987

No context found.

Intelligence 115:145--214. de Kleer, J., and Williams, B. 1987. Diagnosing multiple faults. Artificial Intelligence 32:97--130.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC