Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Mechanical measurement system
Reexamination Certificate
2001-07-17
2003-04-22
Barlow, John (Department: 2863)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Mechanical measurement system
C702S081000
Reexamination Certificate
active
06553324
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to a method and a device for implementation in a measuring system containing a sensor system and an arithmetic unit for detection of a defect in the measuring system, wherein at least one state variable of the system is derived from at least one sensor value in the measuring system, and wherein the defect may be an error of measurement, function, algorithm, software, hardware, calibration or modeling.
In systems whose state variables can be sensed by means of at least one sensor of a sensor system with an associated arithmetic unit the most accurate possible determination thereof from at least one measured variable per measurement or iteration step is of great importance for the functionality of the system. This applies in particular for systems which are critical to safety, such as air data systems of an aircraft since, if the factor of safety for accuracy of the measured variables is not sufficient, further measures such as further sensors are necessary in order to achieve the required factor of safety of the overall system. Corruption of an accuracy proposition in a system which is critical to safety, such as an air data system or the system for determination of chemical or nuclear states in reactions, can lead to a catastrophe such as an aircraft crash or an explosion. Furthermore, to assure the required factor of safety, sensor defects must be detected and eliminated with high reliability.
According to the general prior art, modern sensor systems that must meet safety requirements in corresponding applications are redundantly designed, and for this purpose both the sensors and the data-processing system for calculating the state variables of the system may be provided redundantly. In special applications, such as in controlled aircraft, particularly stringent safety requirements the arithmetic units such as flight computers associated with the sensors are designed with dissimilar hardware and also software. Thus the expense for meeting stringent safety requirements is very high.
In systems with at least one sensor system which is provided with at least one sensor and an arithmetic unit associated therewith for determination of at least one state variable of the system, there are known, for example, methods which correlate a particular state variable with a measured value of the sensor, possibly as a function of further parameters of the system, via calibrations of the sensor. In particular, a method for an air data system is known in which a determination of the instantaneous n-dimensional state variables x of the system from at least one sensor signal is undertaken by means of a cost function &khgr;
2
(x, y, u), which comprises calibration curves or surfaces, in order to optimize the accuracy and reliability of the states x to be determined. This method is published in, for example, Friehmelt, H. and Jost, M., Flush Air Data System—An Advanced Air Data System for Aerospace, 1999 Annual Meeting of the German Aerospace Association, Berlin, Proceedings, 99-180, page 5. In this method there are used calibration curves or surfaces of pressure measurements, representing the measured variable y of the sensor, which is generally a vector, and which depends on the air data as the sought state variables x and on known configurations or controller inputs u. The calibration curves or surfaces are then used in the form of an already known mathematical relationship y=y(x, u).
This cost function is recreated for each measurement y, since it depends on the measured value y which is relevant for the respective measurement and possibly on the system parameter u which is relevant for the particular situation. For each cost function of a measurement, the minimum of the cost function is calculated in order to determine the instantaneous state variables x. As an example, this is accomplished by beginning from one or more start vectors x
o
of the cost function by means of prior art methods, for example by means of a gradient method. In the process, the cost function &khgr;
2
(x
o
, y, u), beginning from a randomly selected first initialization x
o
of the state, is decreased by recursive variation of x along the gradient of &khgr;
2
until a local minimum of &khgr;
2
relative to x is reached.
These prior art methods indeed lead to an accuracy prediction or covariance P for the states x to be determined. This accuracy prediction is correct, however, only if all measured signals and calibration surfaces exhibit normal accuracies. If, for example, a measured signal becomes inaccurate due to a sensor defect or the system becomes decalibrated, the accuracy propositions P are no longer true, since the covariance is modeled only for the state of the intact and not of the defective sensor or calibration.
SUMMARY OF THE INVENTION
An object of the invention is to provide a method as well as a device which permits detection of errors of the measured variables y and assessment of the quality of the calculated accuracy propositions P of the state x even in defective sensor systems.
According to the invention a method for implementation in a measuring system containing a sensor system and an arithmetic unit for detection of a defect in the measuring system, wherein at least one state variable of the system is derived from at least one sensor value in the measuring system, the improvement comprising calculating a cost function value based on deviation of the measured value y from the calibration as a function of a previously calculated state x for the respective measurement and comprising the calculated cost function value with a corresponding threshold value to identify an error if said threshold value is exceeded.
REFERENCES:
patent: 5235527 (1993-08-01), Ogawa et al.
patent: 5532950 (1996-07-01), Moses et al.
patent: 5680409 (1997-10-01), Qin et al.
patent: 5870486 (1999-02-01), Choate et al.
patent: 5991525 (1999-11-01), Shah et al.
patent: 6016465 (2000-01-01), Kelly
patent: 6424985 (2002-07-01), Lee
Xinmin, Z; Yongsheng, N; “Sensor Failure Detection Based On ON-Line Learning Neural Predictor”; Conference Proceedings of 2nd InternationalSymposium on Test and Measurement; Jun. 1997; pp. 298-301.
Langmeier Andreas
Lohmiller Winfried
Barlow John
Eads Deutschland GmbH
Ladas & Parry
Washburn Douglas N
LandOfFree
Method and device for detection of a defect in a sensor system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and device for detection of a defect in a sensor system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and device for detection of a defect in a sensor system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3072294