Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Mechanical measurement system
Reexamination Certificate
2000-07-13
2003-06-10
Barlow, John (Department: 2863)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Mechanical measurement system
C706S015000, C706S023000
Reexamination Certificate
active
06577960
ABSTRACT:
BACKGROUND OF THE INVENTION
A liquid gauging system using a static neural network has been disclosed in the above referenced co-pending U.S. patent application Ser. No. 08/996,858 (hereinafter the “858 Application”). In the 858 Application, a system was disclosed for estimating a quantity of liquid in a container, which could not be measured directly, using a neural network with inputs based on the parameter measurements of a plurality of sensors at a single instant in time, i.e. a single snapshot of sensor output values. The performance of such static neural network liquid gauging systems have been considered adequate, but inherently limited due to the lack of accounting for transient conditions. For those liquid container systems affected by motion, the snapshot of sensor measurement data does not uniquely determine the state of the system during transients, like attitude changes due to acceleration, for example, which may lead to a degradation in the estimation of liquid quantity by the neural network during such conditions.
Aircraft fuel tanks, for example, undergo attitude changes frequently during flight which render the fuel surface in motion causing such conditions as fuel slosh, liquid surface vibration and wave motion just to name a few. Under these conditions, the sensor fuel measurements may lag the attitude or acceleration measurement rendering an input pattern different from that for which the neural network was trained. Such surface effects are not taken into account during training thereof. Consequently, the static neural network may estimate a different quantity of liquid than what actually exists in the tank.
In addition, the sensors of a liquid gauging system generally include built in damping that may cause the liquid parameter measurements thereof to lag behind any rapid changes in liquid state. Accordingly, even such transient conditions as a rapid change in liquid quantity due to a transfer of liquid between containers, for example, may lead to inaccurate liquid quantity estimates from the static neural network. Not only are the errors associated with such dynamic effects not accounted for during training, but any additional information available from such effects which may make the quantity determination of the static neural network more reliable from a training perspective are effectively filtered out as a result of such built in sensor damping.
Accordingly, what is most desirable is a liquid gauging system that can take into account not only the current sensor parameter measurements, but past sensor parameter measurements as well. Including past histories of the parameter measurements will result in an improved liquid quantity estimate even if the most recent sensor measurements alone do not provide for an actual current state of the liquid due to transient conditions such as those exemplified above. The present invention provides for such an improvement in performance.
SUMMARY OF THE INVENTION
In accordance with one aspect of the present invention, liquid gauging apparatus using a time delay neural network for determining a quantity of liquid in a container that is not directly measurable by sensors comprises: a plurality of sensors; each of said sensors for measuring a respective parameter of said liquid and for producing a time varying sensor output signal representative of the respective parameter measured thereby; and processing means for processing said sensor output signals by a time delay neural network algorithm to determine a current quantity of the liquid in the container based on current and past parameter measurements of said sensor output signals.
In accordance with another aspect of the present invention, a method of training a time delay neural network algorithm for computing a quantity of liquid in a container from current and past liquid parameter sensor measurements comprises the steps of: establishing a dynamic model of liquid behavior in the container and parameter measurements of said liquid behavior sensed by a plurality of sensors; deriving from said dynamic model training data sets for a plurality of liquid quantity values, each said data set comprising current and past liquid parameter sensor measurement values corresponding to a liquid quantity value of the plurality, and said corresponding liquid quantity value; and training said time delay neural network algorithm with said derived training data sets.
REFERENCES:
patent: 4426616 (1984-01-01), Maier
patent: 4965731 (1990-10-01), Weitz, Jr.
patent: 4968946 (1990-11-01), Maier
patent: 5832468 (1998-11-01), Miller et al.
patent: 6157894 (2000-12-01), Hess et al.
patent: 683208 (1992-03-01), None
patent: 4223346 (1992-07-01), None
patent: WO99/32857 (1999-07-01), None
Elements of Artificial Neural Networks (Kishan Mehrotra, Chilukuri K. Mohan, Sanjay Ranka) pp. 139-141 (MIT Press).
Fundamentals of Artificial Neural Networks (Mohamad H. Hassoun) pp. 254-259 (MIT Press).
Rabelo Luis C.
Walker Mark
Zakrewski Radoslaw R.
Barlow John
Calfee Halter & Griswold LLP
Dougherty Anthony
Percio David R.
Simmonds Precision Products Inc.
LandOfFree
Liquid gauging apparatus using a time delay neural network does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Liquid gauging apparatus using a time delay neural network, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Liquid gauging apparatus using a time delay neural network will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3129372