Sensor arrangement including a neural network and detection meth

Electricity: measuring and testing – Magnetic – With means to create magnetic field to test material

Patent

Rate now

  [ 0.00 ] – not rated yet Voters 0   Comments 0

Details

324202, 32420716, 324225, 324230, 324234, 324607, 702 38, 702 85, 702170, 702189, G01B 706, G01N 2790, G01R 3312, G06F 1900

Patent

active

058983040

DESCRIPTION:

BRIEF SUMMARY
FIELD OF THE INVENTION

The invention relates to a sensor arrangement comprising at least one measuring coil, at least one source of voltage for the measuring coil, and an evaluation unit with means for detecting, processing, and evaluating measured signals.


BACKGROUND OF THE INVENTION

Such sensor arrangements, for example with eddy-current sensors having a measuring coil, have been used for years for a large variety of measurements, such as, for example, for measuring the distance to an object, for measuring the thickness of a coating layer on an object, for measuring the conductivity and magnetic. permeability of a target, or for examining the homogeneity, and for detecting damage in the structure of the target surface. Normally, measurements with the known sensor arrangements require an extensive knowledge of physical relations between quantity being determined, the measured value, and possible disturbance variables. This knowledge must often be converted to measuring and evaluation electronics that are specially adapted to the measuring problem. Some measurements cannot be carried out at all, since different influence variables superpose, so that no clear statement of the measurements is obtained. When measuring the thickness of a conductive layer on a likewise conductive carrier with a known sensor arrangement, measuring errors will occur, for example, despite an adaptation to the particular target material. This applies, for example, to the existence of local inhomogeneities, magnetization, conductivity, effective permeability, temperature gradients, etc. When using known sensor arrangements of the state of the art, it will be necessary to find again for each measuring problem the mathematical relations between influence variables and measured values. The mathematical relations are in part complex and extremely nonlinear, so that this procedure, if possible at all, will take an enormous amount of time.
The complexity of the mathematical relations between influence variables and measured values is to be demonstrated by the example of the noncontacting distance measurement by the eddy-current principle. The impedance of a coil (real part and imaginary part) varies upon approaching an electrically and/or magnetically conductive object. Thus, by measuring the impedance of a measuring coil, it is possible to determine the distance between the coil and the object being measured. In the object being measured, a current is induced, which counteracts the excitation of the coil. This reaction is again dependent on the electric conductivity and on the magnetic behavior of the measuring object, namely the material parameters. Same are again temperature-dependent. Furthermore, the impedance values are frequency-dependent, and nonlinearly related to the measuring distance. Reliable results of the measurement may be obtained only when all these influence variables are considered.


SUMMARY OF THE INVENTION

It is now the object of the invention to describe a sensor arrangement of the kind under discussion, which permits the user to perform largely material-independent measurements, without knowledge of the mathematical background.
The sensor arrangement of the present invention accomplishes the foregoing object by providing a characteristic features of claim 1. Accordingly, providing a sensor arrangement which is configured such that the evaluation unit for evaluating the measured signals comprises a neural network of neurons or nodes arranged into layers including an input layer, at least one hidden layer, and an output layer, and connection weights for the nodes of the individual layers, and that the connection weights are determined and stored during a learning phase by measuring the outputs of the network produced when a plurality of different suitable learning objects with known actual values are input to the network, and using an algorithm to arrive at connection weights which produce the known actual values as outputs of the network.
In accordance with the invention, it has been recognized that a detailed

REFERENCES:
patent: 4084136 (1978-04-01), Libby et al.
patent: 4263551 (1981-04-01), Gregory et al.
patent: 5055784 (1991-10-01), Jaeger et al.
Patent Abstracts of Japan, vol. 13, No. 210 (P-872), May 17, 1989 for "Shape Detector".
Patent Abstracts of Japan, vol. 15, No. 79 (P-1170), Feb. 25, 1991 for "AC-Loss Measuring Apparatus".
Patent Abstracts of Japan, vol. 12, No. 266 (P-735), Jul. 26, 1988 for "Nondestructive Measuring Method for Tube Thickness".
Patent Abstracts of Japan, vol. 5, No. 32 (P-050), Feb. 27, 1981 for "Device for Measuring Magnetic Permeability".
Materials Evaluation, "Eddy Current Defect Characterization Using Neural Networks", vol. 48, Mar. 1990, pp. 342-347.
SICE '93, "Neural Network Based Sensor Integration for On-Line Weld Quality Control"; pp. 1483-1486.

LandOfFree

Say what you really think

Search LandOfFree.com for the USA inventors and patents. Rate them and share your experience with other people.

Rating

Sensor arrangement including a neural network and detection meth does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Sensor arrangement including a neural network and detection meth, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Sensor arrangement including a neural network and detection meth will most certainly appreciate the feedback.

Rate now

     

Profile ID: LFUS-PAI-O-687712

  Search
All data on this website is collected from public sources. Our data reflects the most accurate information available at the time of publication.