Data processing: measuring – calibrating – or testing – Calibration or correction system – Circuit tuning
Reexamination Certificate
2002-06-18
2004-02-03
Shah, Kamini (Department: 2863)
Data processing: measuring, calibrating, or testing
Calibration or correction system
Circuit tuning
C706S026000, C342S041000, C701S003000
Reexamination Certificate
active
06687637
ABSTRACT:
FIELD OF THE INVENTION
The invention relates generally to data sensor validation and more particularly to a method of data sensor validation for use in environmental applications.
BACKGROUND OF THE INVENTION
A neural network is an artificial neural circuit network that, either in circuitry or in software, performs correlation processing. In a typical neural network, there is one or more intermediate layers between a data input layer and a data output layer, each of these layers being made up of a plurality of units, network-like connections being made between the input/output sections and the intermediate layers by means of the input/output systems. Because this neural network has non-linear components, it is capable of performing extremely complex correlations with respect to a variety of data types. These correlations are then useful in determining approximations, projections, and so forth. Because of this, neural networks are currently used in many industries, including manufacturing and service industries.
In these industries, a neural network is selected for a particular process and is then trained using known input data and known output responses. For example, in a process control circuit, a neural network is trained to provide a desired process control signal in response to a plurality of sensor data received at an input port thereof. Through training, weights within the neural network are modified to ensure that each sensor input value is appropriately accounted for in the control signal provided at the output of the neural network. Of course, some neural networks are manufactured with their weights integrated therein when their use is known and fixed.
Conventional neural network training and testing methods require complete patterns such that they are required to discard patterns with missing or bad data. Experimental results have shown that neural network testing performance generally increases with more training data when trained.
Most methods of training and using neural networks do not account for the relationship between measurements by one sensor relative to another sensor measurement in unrelated systems. Often, in conjunction with increasing the reliability of measurement data, fault detection techniques such as sensor redundancy are used to increase a control system's ability to recognize that measurement data is unreliable. If measurement data from a sensor in a group of redundant sensors is inconsistent with measurement data from other sensors in the group, the inconsistent data is considered as unreliable and are ignore.
Data sensor validation is an important part of feedback based control systems and of large scale monitoring systems. In data sensor validation, data from each of a plurality of sensors is validated to avoid decisions or monitoring being dependent upon erroneous sensor data. Effective detection of erroneous measurements and recovery of missing data are important as erroneous or missing data may disrupt operations and may cause severely abnormal operating conditions and result in incorrect safety, control, and economic decisions.
When a neural network is trained, weighting coefficients and biases are randomly applied with respect to the input data for each of units that accepts data. As data is input under these conditions, judgments are made with regard to the correctness of the output resulting from calculation according to these weighting coefficients. Whether or not the output results are correct is fed back using a learning method such as back-propagation, the originally set weighting coefficients and biases being corrected, and data being re-input. By repeating this process of input and correction of weighting coefficients and biases a large number of times, the weighting coefficients and biases that will obtain an appropriate output for a prescribed data input are established.
By installing a trained neural network into character recognition, image processing or other system that is implemented by a computer, the neural network can be put into practical use in many industries, including manufacturing and service industries. These neural networks are used in closed environments wherein the sensors sense known parameters as for example the amount of carbon monoxide or other gases along a manufacturing process. Such a restricted environment facilitates the identification and the replacement or repair/adjustment/calibration of a faulty sensor when erroneous data are sensed.
Conversely, when considering large-scale neural network, i.e. open field control system, it is important to precisely point out which sensor is deficient when erroneous data are received at a control operating system. Since in an open field neural network the sensors are remotely located, sending a technician to an isolated remote location for replacing a faulty sensor is an expensive process that most organizations tend to avoid if the faulty sensor is not precisely identified.
Furthermore, a major problem with existing validation system using neural network is when a sensor data is close to an extreme value—lower or upper—within or outside a pre-determined range of values, the sensor data is attributed a value corresponding to an extreme value, without consideration of the real value of the sensor data validation of the sensor data. Therefore, the attributed value is not representative of an event occurring at the sensor, there is no indication to which extreme the sensed value is close to.
It would be advantageous to provide a method of validating data that is improved over the data limit approach but not as costly to implement as the duplicate sensor approach.
Furthermore, it would be advantageous to provide a method for suggesting a value for replacing a sensed data, which shows a shift from a predictable sensed data.
OBJECT OF THE INVENTION
It is another object of the present invention to provide a method for verifying the validity of sensor data for use in environmental type applications.
SUMMARY OF THE INVENTION
In accordance with the invention there is provided a method of data sensor validation comprising the steps of: pre-processing data sensor from each sensor from a plurality of sensors for at least segmenting the data sensors into a plurality of groups, each group for grouping sensors for sensing highly relevant data one to another; providing the pre-processed data sensor to a correlation processor, the correlation processor for determining from pre-processed data sensor, pre-processed data that is other than correlated, the determination made in dependence upon redundant pre-processed data other than pre-processed data from two sensors for sensing an identical parameter; and, when pre-processed data that is other than correlated is detected, providing an indication to an operator that the sensor data is other than correlated.
In accordance with the invention there is provided a method of data sensor validation comprising the steps of: pre-processing data sensor from each sensor from a plurality of sensors; providing the pre-processed data sensor to a correlation processor, the correlation processor for determining from pre-processed data sensor, pre-processed data that is other than correlated, the determination made in dependence upon redundant pre-processed data other than pre-processed data from two sensors for sensing an identical parameter; and, when pre-processed data that is other than correlated is detected, providing an indication to an operator that the sensor data is other than correlated.
In accordance with the invention there is provided a sensor for use in geographically remote sensor applications comprising a sensing circuitry for sensing data; a transmitter for transmitting sensed data to a correlation processor, the correlation processor for determining from pre-processed sensed data, pre-processed data that is other than correlated, the determination made in dependence upon redundant pre-processed data other than pre-processed data from two sensors for sensing an identical parameter at an approximately same geographic location; a
Freedman & Associates
GlobVision Inc.
Shah Kamini
LandOfFree
Data sensor validation system and method does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Data sensor validation system and method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Data sensor validation system and method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3347533