Data processing: measuring – calibrating – or testing – Measurement system – Performance or efficiency evaluation
Reexamination Certificate
1999-11-16
2002-01-29
Hoff, Marc S. (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system
Performance or efficiency evaluation
C709S241000
Reexamination Certificate
active
06343261
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to a device for automatic diagnosis of a technical system.
The state of the technology is described below.
The object of the technical diagnosis is to localize the component errors that occur in a specific technical system (tS). For this purpose, values of parameters must be measured at tS.
The state of the technology encompasses different methods and devices for diagnosing a technical system tS through the evaluation of a model of tS. It is known that it is often costly to construct a model that is adequately realistic. To reduce the expenses when a plurality of similar components is installed in the technical system, or when a plurality of similar technical systems is to be modeled and subsequently diagnosed, approaches have been developed for describing the object to be diagnosed with a component-oriented model. The model reflects the structure of the technical system in components. The generic scope of the invention falls into this category. In contrast, numerous modeling approaches from the regulating technique, such as differential equations, are not component-oriented.
Prior to the diagnosis, a human expert must input once the information about the structure and function of the technical system tS, and the possible errors and the determination of these errors, that the diagnosis device requires for searching the errors. This process, which is often referred to as knowledge acquisition, is time consuming and error-prone. Therefore, a principal challenge is to design the diagnosis device to operate with as little input information as possible and automatically generate as much of the required information as possible from the input information.
The attached drawings illustrate the function of a model-based diagnosis device by way of an example.
The foundation of the present claim to patentability is described in the overview article by Randall Davis and Walter Hamscher: “Model-Based Reasoning; Troubleshooting” in: Walter Hamscher and Luca Console and Randall Davis: “Readings in Model-Based Diagnosis,” Morgan Kaufmann Publ. (1992), pp. 1-24. Each component of tS possesses different behavioral modes; in the simplest case, the modes are “intact” and “defective.” A diagnosis allocates each component a behavioral mode, thereby specifying which components are defective.
The term “diagnosis” has two meanings: First, it refers to the process of finding the errors in the technical system tS and, second, it describes the result of this search.
An important class of model-based diagnosis methods is presented in the above-cited article by Randall Davis and Walter Hamscher, namely the General Diagnostic Engine (GDE). The aforementioned figures illustrate the function of such a GDE by way of a simple example.
In
FIG. 1
, C
1
and C
2
double their respective input value. C
3
forms the difference between its two input values, and C
4
increases its input value by 5. The values
2
and
5
were supplied to this circuit, and
8
and
20
were measured as output values. Under the assumption that all of the components operate error-free,
6
or
15
would have to be the attained result. Consequently, a component is defective.
If a measurement has revealed that the value x is present at the input of the component C
2
, and if f
2
and f
4
, the behavioral models for the components C
2
and C
4
, respectively, are in the “intact” behavioral mode, the model predicts the value z=f
4
(f
2
(x)) for the output of C
4
if it is assumed that C
2
and C
4
are intact. This is because y=f
2
(x) in this example is simultaneously the output of C
2
and the input of C
4
.
For each model value (every value that is predicted by the model),
FIG. 2
notes which assumptions must be met so that the model yields this value. For example, the value
15
is obtained under the assumption that both C
2
and C
4
are intact. Two desired/actual differences were determined; therefore, there are two conflict sets. On the one hand, C
1
, C
2
and C
3
cannot be intact simultaneously because of the deviation at Output_
1
; on the other hand, C
2
and C
4
cannot simultaneously be intact because of Input_
2
.
The candidate set prior to the beginning of the diagnosis includes only one candidate, namely the one that allocates all components the behavioral mode “intact.” The candidate set is expanded with the aid of the conflict set to subsequently reduce the candidate set to its minimum through assumptions and measures.
In the first step, the conflict set is evaluated as follows:
{(C
1
, intact), (C
2
, intact, (C
4
, intact)},
and in the second step, the conflict set is evaluated as
{(C
2
, intact), (C
4
, intact)}.
The candidate set after the first step with the first conflict set includes the candidates [(C
1
, defective), (C
2
, intact), (C
3
, intact), (C
4
, intact)].
This candidate set is refuted by the second conflict set. The corrected, direct successors are [(C
1
, defective), (C
2
, defective), (C
3
, intact), (C
4
, intact)] and [(C
1
, defective), (C
2
, intact), (C
3
, intact), (C
4
, defective)]. The two successors are consistent with the two observations, and are therefore incorporated into the candidate set after Step
2
.
The following options are available for continuing the diagnosis in this situation and reducing the candidate set:
The first option determines further measured values. In the event that the desired value
4
is present at the output of C
1
, GDE supplies {(C
1
, defective)} as the third conflict set, and excludes all candidates that allocate C
1
the defective mode.
A further option involves a “single fault assumption,” i.e., it is assumed that, at the most, one component is defective at any one time. The diagnosis of [(C
1
, intact), (C
2
, defective), (C
3
, intact), (C
4
, intact)] follows immediately afterward.
The physical behavior of each component that appears at least once in tS is described one time in the definition of the input and output parameters and the internal parameters of the component, and in the description of the connections between the parameters as relations (constraints). A special type of component parameter is the behavioral modes of the component. The constraints of a component can be universal, and be allocated to certain behavioral modes of the component, which means that they are only applicable when the component has assumed the respective behavioral mode.
Model-based diagnosis presupposed that the behavioral of the components can be described locally, that is, the allocation of values to each component's parameters is solely dependent on the allocation of values to other parameters of the same component. Then, it is only necessary to describe each type of component once; the diagnosis device can re-use this description, which significantly reduces the amount of required input information.
The technique of model-based diagnosis is especially advantageous in comparison to other methods if the components to be described are of a simple nature (e.g., electrical or hydraulic components) and appear numerous times in tS. The diagnosis device thus evaluates a library with the description of all component types, as well as a construction model of the technical system. This construction model describes how the components are connected to one another. A component Comp_
1
is connected to a component Comp_
2
through the identification of the parameter Para_
1
at Comp_
1
with a parameter Para_
2
at Comp_
2
. The construction model further describes the type of each component. The diagnosis device automatically constructs the model of the technical system tS from the descriptions of the component types and the construction model.
Most of the known diagnosis methods, and the diagnosis device of the invention, presuppose that the technical description of the system tS to be diagnosed does not change during a diagnosis, in other words, the library with models of component types and the construction model are valid for the entire d
Iwanowski Sebastian
John Ute
May Volker
Tatar Mugur
Daimler-Chrysler AG
Duong Khoi
Hoff Marc S.
Kunitz Norman N.
Venable
LandOfFree
Apparatus and method for automatically diagnosing a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Apparatus and method for automatically diagnosing a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Apparatus and method for automatically diagnosing a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2828338