Diagnosis of networks of components, with stripwise modeling

Details Diagnosis of networks of components, with stripwise modeling Diagnosis of networks of components, with stripwise modeling

1998-01-14

2001-01-09

Olms, Douglas W. (Department: 2732)

Multiplex communications

Diagnostic testing

Determination of communication parameters

C370S242000, 36, 36, 37, 37, 37

Reexamination Certificate

active

06172966

ABSTRACT:

BACKGROUND OF THE INVENTION
The invention relates to the automatic diagnosis of networks of components.
A diagnosis device capable of operating on analogue signals is described in an European Application (EP-A-408 425). It works on the basis of functional models. These comprise component-expressions pertaining to physical quantities relating to a component, such as Ohm's law (V=I.R); they also comprise law-expressions representing general relations between physical quantities, such as the Kirchhoff's first law; the algebraic sum of the currents at a node is zero.
Diagnosis begins from acquired physical quantities. These are stored in memory in the form of samples, matched up with a corresponding specification of the relevant nodes and components of the network.
If the device works “blind” (without knowing a priori the layout of the network), the specification of the nodes and components must be sufficient to allow it to learn the layout of the network along with its exploration of the latter. If, on the contrary, the layout of the network is known in advance, it is sufficient to designate each measurement point in this known layout.
For each acquisition, the samples are referred to a chosen working time interval. The device searches therein for anomalies with respect to the functional models (law-expressions and component-expressions). This is done for each sampling instant in the working time interval.
Generally, the samples of physical quantities acquired do not lend themselves directly to the detection of anomalies. First, it is necessary to compute estimated physical quantities, as many as necessary so as no longer to have any unknowns. A violated expression then makes it possible in principle, bearing in mind the uncertainties, to locate the fault.
The aim is to locate a defective component in the relevant network and to do so in minimum time.
However, it has transpired that difficulties persisted for certain applications.
First, the prior device functions by “propagation”: for each sampling instant it is necessary, starting from the measured samples, to compute as many estimated quantities as necessary, and then to work iteratively on the models until the faulty component is found. It is clear that the number of operations, and consequently the time elapsed, grow very quickly, as the complexity of the circuit to be analysed increases.
Moreover, to produce the functional models it is necessary to “bracket” certain variables, that is to say fix a minimum and a maximum in respect thereof. Now, in certain situations, analogue signals (voltage or current) vary very rapidly. It transpired that bracketing their derivatives then becomes particularly tricky, especially when uncertainty is taken into account.
These difficulties are made worse by the fact that, in most cases, there is furthermore reason to consider uncertainty ranges rather than raw values.
They are made worse also each time that it is necessary to bring into the models not only the quantities themselves, but also their first derivatives (or higher-order derivatives), as is often the case. It is often difficult to bracket such derivatives by uncertainty ranges, in the presence of fast variations in the basic quantities themselves.
SUMMARY OF THE INVENTION
The purpose of the present invention is especially to afford a solution to these problems.
The device proposed is of the known type including:
an input memory for receiving blocks of measurements (in the prior art, packets of digital samples), each of which represents measurements of a physical quantity made at a known location of the network,
a base memory for storing the definition of functional models relating to the network, and
processing means capable of determining whether a functional model is or is not satisfied by one or more blocks of measurements, so as to identify a possible fault in one at least of the components of the network.
The invention makes use first of a specific representation of the measurement blocks, in the form of n-tuples. An n-tuple represents the time evolution of a variable, on the basis of a string of numbers, which defines a time interval, and an envelope of the variable over this time interval. The device furthermore includes means of computing on n-tuples, which means may be regarded as subsumed within the processing means, or as tool separate from the latter.
The device of the invention also includes test means able to receive one or more n-tuples (Nf, Ng), and to yield an output relating to the satisfying of a standard expression by these n-tuples. It will be seen that these test means can operate without exhaustively examining the contents of the envelopes which this or these n-tuples (Nf, Nf) represent.
In the input memory, each block of measurements is stored in the form of at least one initial n-tuple.
In the base memory, the definition of a functional model comprises a list of components (L
v
), and the designation (Df, Dg) of at least one particular function of the measurable quantities (L
p
). Associated with the standard expression of the test means, this particular function defines the equation of the model.
Finally, the processing means comprise:
management means capable, repetitively,
of cooperating with the base memory, so as to match up the available initial n-tuples, and at least one model applicable to the measurable quantities which these n-tuples represent,
of deploying the computations on n-tuples, with the function designation (Df, Dg) of the model, and these n-tuples, to obtain at least one result n-tuple (Nf, Ng),
of applying the result n-tuple (Nf, Ng) to the test means, and
of associating the output from the test means with the list of components (L
v
) of the model,
as well as decision logic, capable of analysing various outputs from the test means, with a view to determining one or more defective components.
Very advantageously, the models pertain solely to directly measurable quantities; similarly, each measurable quantity enters at most once into each function designation (Df, Dg) of a model.
Subsequently, for the initial n-tuples, the envelope is delimited as a trapezium, with vertical bases in an amplitude/time graph. It will be seen that this corresponds to an n-tuple of order
1
.
The device can also include means for receiving or constructing a map of a network to be diagnosed, and a generator of model-relations from this map and from physical laws regarding the components of the network.
In the currently preferred embodiment, the test means and the function designations together define a standard structure of differential equations, such as f( )+g′( )=0. Consequences to which we shall return later result from this.
Other characteristics and advantages of the invention will emerge on examining the detailed description below, and the drawings.


REFERENCES:
patent: 5046034 (1991-09-01), Stark et al.
patent: 5214653 (1993-05-01), Elliott, Jr. et al.
patent: 5327437 (1994-07-01), Balzer
patent: 5412802 (1995-05-01), Fujinami et al.
patent: 5566092 (1996-10-01), Wang et al.
patent: 5602761 (1997-02-01), Spoerre et al.
patent: 0408425 (1991-01-01), None

Affiliated with

Also associated with

Rating

Diagnosis of networks of components, with stripwise modeling does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Diagnosis of networks of components, with stripwise modeling, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Diagnosis of networks of components, with stripwise modeling will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2479462