Discriminate reduction data processor

• US 5,619,432 A
• Filed: 04/05/1995
• Issued: 04/08/1997
• Est. Priority Date: 04/05/1995
• Status: Expired due to Term

• Current Assignee: United States Department of the Navy
• Original Assignee: the united states of america as represented by the secretary of the navy
• Inventors: Chandler, Larry S.
• Primary Examiner(s): Voeltz, Emanuel T.
• Assistant Examiner(s): MILLER, CRAIG S
• Application Number:
  US08/417,340
• Time in Patent Office: 734 Days
• Field of Search: 364/570, 364/572-577, 364/581, 364/582, 364/556, 364/550, 364/602, 364/604, 364/605, 364/723, 364/732, 364/733, 364/735, 341/155, 341/156
• US Class Current: 702/189
• CPC Class Codes: G06F 17/17 Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method G06F17/18 takes precedence ; interpolation for numerical control G05B19/18

Abstract

Data processing systems are disclosed which use arithmetic processors with final stage discriminate rectifiers and discriminate data sampling to evaluate parameters which effectively minimize parametric expressions representing sums of normalized datum variances to provide corresponding data representations which account for such items as nonlinearities, nonuniform error distributions, and variable precision uncertainties. Conformal analysis is provided as an optional replacement for regression analysis to account for significant errors in more than a single variable parameter. The applications include general forms of curve fitting, transformations, function approximations and corresponding forms of real time data processing. The systems provide for both linear and nonlinear forms of data reduction.

Claims

1. A data processing system, comprising:

a memory; and

means for accessing and processing information whereby a data representation of variation in characteristic measurement is generated;

said data representation being stored in said memory;

said data representation being generated by the method comprising;

accessing at least some said information which characterizes said variation;

effectuating at least one form of discriminate reduction data processing, said effectuating including establishing correspondence between at least some said information and at least one approximating parameter which will substantially minimize a parametric expression; and

evaluating said at least one approximating parameter which substantially minimizes said parametric expression in correspondence with said information.

2. A data processing system as in claim 1, wherein said means for accessing and processing information includes:

means for executing an application program; and

a database used by said application program;

said database including intermittent measure and related sample time generated by the method comprising;

performing analog signal preprocessing including criteria-related signal conditioning for triggering analog-to-digital conversion of incoming data at data-related time intervals; and

accessing and selectively sampling real time data by analog-to-digital conversion and simultaneously acquiring representation of time.

3. A data processing system as in claim 1, wherein said establishing correspondence includes at least one form of conformal analysis comprising a method of relating said at least one approximating parameter to said information by rendering and including analytical expression in form to effectuate said evaluating;

said analytical expression reflecting said method of relating;

said method of relating including steps of;

taking partial derivatives of a representative parametric sum with respect to at least two approximative parameters which represent at least one resolute parameter and at least one unknown parameter;

establishing a system of equations by using said information to provide expressions for at least two said partial derivatives in terms of at least two approximating parameters, and by equating said expressions for at least two said partial derivatives to zero;

consolidating equations by assuming a relative value for said resolute parameter and imposing interrelated resolute parameter restraints upon at least one said unknown parameter, said imposing being accomplished by combining at least one equation containing said resolute parameter with at least one equation containing an unknown said non-resolute parameter;

reducing the total number of equations in said system of equations by dropping said at least one equation containing said resolute parameter from said system of equations; and

combining the remaining equations of said system of equations to establish said correspondence.

4. A data processing system as in claim 3, wherein said conformal analysis is type 1 conformal analysis.

5. A data processing system as in claim 3, wherein said conformal analysis is type 2 conformal analysis.

6. A data processing system as in claim 1, wherein said parametric expression represents the sum of a plurality of squared coordinate normalizing approximation deviations;

and wherein said establishing correspondence includes;

accessing at least some said information which provides representative measure for at least one derivative; and

subjecting said representative measure to operations including absolute value rectification, whereby numerical correspondence to said derivative is effectively converted to include representation of at least one transformation weight factor coordinate normalizing proportion in said evaluating said at least one approximating parameter, said transformation weight factor coordinate normalizing proportion, as considered within said processing accuracy, substantially corresponding in magnitude to the absolute value of the inverse of said derivative being raised to a power other than two;

including representation of said at least one transformation weight factor coordinate normalizing proportion in said establishing correspondence; and

evaluating at least one set including at least one said approximating parameter, said set rendering said plurality of squared coordinate normalizing approximation deviations as representing measure of a corresponding plurality of coordinate-normalized datum variances.

7. A data processing system as in claim 6, wherein:

said parametric expression includes at least one nested parameter;

at least one represented coordinate normalizing approximation deviation in said parametric expression is represented as a parametric first order Taylor series approximation; and

said evaluating includes evaluation of at least one estimate of a first order correction for a nested parameter.

8. A data processing system as in claim 6, wherein said establishing correspondence includes at least one form of conformal analysis comprising a method of relating said at least one approximating parameter to said information by rendering and including analytical expression in form to effectuate said evaluating;

said analytical expression reflecting said method of relating;

said method of relating including steps of;

taking partial derivatives of a representative parametric sum with respect to at least two approximative parameters which represent at least one resolute parameter and at least one unknown parameter;

establishing a system of equations by using said information to provide expressions for at least two said partial derivatives in terms of at least two approximating parameters, and by equating said expressions for at least two said partial derivatives to zero;

consolidating equations by assuming a relative value for said resolute parameter and imposing interrelated resolute parameter restraints upon at least one said unknown parameter, said imposing being accomplished by combining at least one equation containing said resolute parameter with at least one equation containing an unknown said non-resolute parameter;

reducing the total number of equations in said system of equations by dropping said at least one equation containing said resolute parameter from said system of equations; and

combining the remaining equations of said system of equations to establish said correspondence.

9. A data processing system as in claim 6, wherein said power represents a value of two divided by the count of independent variables which are represented by data that is included in said evaluating.

10. A data processing system as in claim 1, wherein at least some said information is analog data being subjected to analog preprocessing whereby an analog representation of integrated sums is generated;

representation of said integrated sums being included in said establishing correspondence;

said analog representation of integrated sums being generated by the method comprising;

accessing at least one real time analog signal;

subjecting said real time analog signal to operations including differentiation and absolute value rectification, thereby providing a rectified analog signal with amplitude proportionately corresponding over at least one interval of time to the absolute value of the inverse of the derivative of said real time analog data being raised to a power other than two as considered within the accuracy of said analog preprocessing;

formulating at least one analog signal, which is the product of at least two analog signals, one of said at least two analog signals being said rectified analog signal; and

integrating at least one said rectified analog signal and at least one said product of at least two analog signals, over said at least one interval of time.

11. A final stage discriminate rectifier comprising means for accessing and processing information whereby a data representation of at least one transformation weight factor coordinate normalizing proportion is generated;

said data representation being generated by the method comprising;

accessing at least some said information which establishes representative measure for at least one derivative; and

subjecting said representative measure to operations including absolute value rectification, whereby represented numerical correspondence to said derivative is effectively converted to represent at least one positive value;

said at least one positive value, within said processing accuracy, substantially corresponding in proportion to the absolute value of the inverse of said derivative being raised to a power other than two.

12. A final stage discriminate rectifier as in claim 11, comprising means for generating and transferring proportionality of said at least one transformation weight factor coordinate normalizing proportion, wherein:

said proportionality of at least one transformation weight factor coordinate normalizing proportion is provided in digital form;

said means for accessing and processing information includes a computer chip; and

control signaling from a logic control unit activates logic gate circuitry to effectuate at least some said operations.

13. A final stage discriminate rectifier as in claim 11, wherein said means for accessing and processing information includes analog circuitry whereby:

said data representation is provided as an analog signal;

said represented measure is real time analog data; and

said at least one positive value is included in a continuum of positive values which are represented by a rectified analog signal, said rectified analog signal being continuous over at least one time interval wherein the amplitude of said rectified analog signal proportionately represents the absolute value of the inverse of the said derivative being raised to a power other than two.

14. A logic control unit comprising means for providing control signals to effectuate a least one form of data processing whereby a data representation of variation in characteristic measurement is generated, said data representation being generated by:

accessing at least some information which characterizes said variation;

effectuating at least one form of discriminate reduction data processing, said effectuating including establishing correspondence between at least some said information and at least one approximating parameter which will substantially minimize a parametric expression; and

evaluating said at least one approximating parameter which substantially minimizes said parametric expression in correspondence with said information.

15. A logic control unit as in claim 14, wherein said parametric expression represents the sum of a plurality of squared coordinate normalizing approximation deviations, and wherein said establishing correspondence includes:

accessing at least some said information which provides representative measure;

subjecting said representative measure to operations, whereby numerical correspondence is effectively converted to include representation of at least one coordinate normalizing proportion in said evaluating said at least one approximating parameter, said coordinate normalizing proportion, as considered within said processing accuracy, substantially corresponding in magnitude to the absolute value of the inverse of said numerical correspondence being raised to a power other than two;

including representation of said at least one coordinate normalizing proportion in said establishing correspondence; and

evaluating at least one set including at least one said approximating parameter, said set rendering said plurality of squared coordinate normalizing approximation deviations as representing measure of a corresponding plurality of coordinate-normalized datum variances.

16. A logic control unit as in claim 15, wherein said power represents a value of two divided by the count of independent variables that are represented by data that is included in said evaluating.

17. A logic control unit as in claim 15, wherein said means for providing control signals to effectuate said at least one form of discriminate reduction data processing includes a computer chip.

18. A logic control unit as in claim 15, wherein:

said parametric expression includes at least one nested parameter;

at least one represented coordinate normalizing approximation deviation in said parametric expression is represented as a parametric first order Taylor series approximation; and

said evaluating includes evaluation of at least one estimate of a first order correction for a nested parameter.

19. A logic control unit as in claim 15, wherein said establishing correspondence includes at least one form of conformal analysis comprising a method of relating said at least one approximating parameter to said information by rendering and including analytical expression in form to effectuate said evaluating;

said analytical expression reflecting said method of relating;

said method of relating including steps of;

taking partial derivatives of a representative parametric sum with respect to at least two approximative parameters which represent at least one resolute parameter and at least one unknown parameter;

establishing a system of equations by using said information to provide expressions for at least two said partial derivatives in terms of at least two approximating parameters, and by equating said expressions for at least two said partial derivatives to zero;

consolidating equations by assuming a relative value for said resolute parameter and imposing interrelated resolute parameter restraints upon at least one said unknown parameter, said imposing being accomplished by combining at least one equation containing said resolute parameter with at least one equation containing an unknown said non-resolute parameter;

reducing the total number of equations in said system of equations by dropping said at least one equation containing said resolute parameter from said system of equations; and

combining the remaining equations of said system of equations to establish said correspondence.

20. A logic control unit as in claim 14, wherein:

said means for providing control signals to effectuate said at least one form of discriminate reduction data processing includes a computer chip; and

said establishing correspondence includes at least one form of conformal analysis comprising a method of relating said at least one approximating parameter to said information by rendering and including analytical expression in form to effectuate said evaluating;

said analytical expression reflecting said method of relating;

said method of relating including steps of;

taking partial derivatives of a representative parametric sum with respect to at least two parameters which include at least one resolute parameter and at least one unknown parameter;

establishing a system of equations by using said information to provide expressions for at least two said partial derivatives in terms of at least two approximating parameters, and by equating said expressions for at least two said partial derivatives to zero;

consolidating equations by assuming a relative value for said resolute parameter and imposing interrelated resolute parameter restraints upon at least one said unknown parameter, said imposing being accomplished by combining at least one equation containing said resolute parameter with at least one equation containing an unknown said non-resolute parameter;

reducing the total number of equations in said system of equations by dropping said at least one equation containing said resolute parameter from said system of equations; and

combining the remaining equations of said system of equations to establish said correspondence.

21. A logic control unit as in claim 14, wherein said establishing correspondence includes at least one form of conformal analysis comprising a method of relating said at least one approximating parameter to said information by rendering and including analytical expression in form to effectuate said evaluating;

said at least one form of conformal analysis being type 1 conformal analysis;

said analytical expression reflecting said method of relating;

said method of relating including steps of;

taking partial derivatives of a representative parametric sum with respect to at least two approximative parameters which represent at least one resolute parameter and at least one unknown parameter;

establishing a system of equations by using said information to provide expressions for at least two said partial derivatives in terms of at least two approximating parameters, and by equating said expressions for at least two said partial derivatives to zero;

consolidating equations by assuming a relative value for said resolute parameter and imposing interrelated resolute parameter restraints upon at least one said unknown parameter, said imposing being accomplished by combining at least one equation containing said resolute parameter with at least one equation containing an unknown said non-resolute parameter;

reducing the total number of equations in said system of equations by dropping said at least one equation containing said resolute parameter from said system of equations; and

combining the remaining equations of said system of equations to establish said correspondence.

22. A logic control unit as in claim 14, wherein said establishing correspondence includes at least one form of conformal analysis comprising a method of relating said at least one approximating parameter to said information by rendering and including analytical expression in form to effectuate said evaluating;

said at least one form of conformal analysis being type 2 conformal analysis;

said analytical expression reflecting said method of relating;

said method of relating including steps of;

taking partial derivatives of a representative parametric sum with respect to at least two approximative parameters which represent at least one resolute parameter and at least one unknown parameter;

establishing a system of equations by using said information to provide expressions for at least two said partial derivatives in terms of at least two approximating parameters, and by equating said expressions for at least two said partial derivatives to zero;

consolidating equations by assuming a relative value for said resolute parameter and imposing interrelated resolute parameter restraints upon at least one said unknown parameter, said imposing being accomplished by combining at least one equation containing said resolute parameter with at least one equation containing an unknown said non-resolute parameter;

reducing the total number of equations in said system of equations by dropping said at least one equation containing said resolute parameter from said system of equations; and

combining the remaining equations of said system of equations to establish said correspondence.

23. A logic control unit as in claim 14, wherein said effectuating includes:

dividing some said information into at least two data subsets;

establishing correspondence between said at least two data subsets and at least four sets of at least two parameters which substantially minimize a parametric expression in respective correspondence with each of said at least two said data subsets and each of at least two different term coefficients being selected as the resolute;

evaluating said at least four sets of at least two parameters which substantially minimize said parametric expression;

determining points of commonality as represented by approximating equations respectively corresponding to each of said at least two data subsets; and

evaluating said approximating parameter in correspondence with said points of commonality.

24. A logic control unit as in claim 14, wherein said evaluating includes iteration whereby:

the number of iterations allowed between improved estimates is determined by a designated number;

the iteration is terminated when said number of iterations allowed is exceeded;

the extrapolation ratio is a function of count of iteration cycles;

at least one difference is established for each iteration, said difference being the difference between a successive approximation and a corresponding computed value; and

said iteration includes a decision-making for providing successive approximations, said decision-making including;

if means for comparing at least two successive differences is not available then new estimates are set equal to the most recently computed values, said means for comparing at least two successive differences is established, the iteration count is incremented, and the iteration is continued;

if the absolute value of any most recently established difference is smaller than the absolute value of the corresponding previous said difference, then new estimate for the next corresponding approximation is set equal to the corresponding most recently computed value, the iteration count is incremented, and the iteration is continued;

if the absolute value of each most recently established difference is larger than the absolute value of each corresponding previous said difference, and if the sign of any most recently established difference is the same as the sign of the corresponding previous said difference, then at least one new estimate for the next corresponding approximation is extrapolated back, the iteration is continued, said means for comparing at least two successive differences is rendered unavailable, and the iteration count is not incremented.