Latent effects decision analysis

• US 6,782,372 B1
• Filed: 09/28/2000
• Issued: 08/24/2004
• Est. Priority Date: 09/28/2000
• Status: Active Grant

• Current Assignee: National Technology & Engineering Solutions Of Sandia LLC (Honeywell International Inc.)
• Original Assignee: Sandia Corporation (Martin Marietta Corporation)
• Inventors: Werner, Paul W., Cooper, J. Arlin
• Primary Examiner(s): Khatri, Anil
• Assistant Examiner(s): Booker, Kelvin
• Application Number:
  US09/675,900
• Time in Patent Office: 1,426 Days
• Field of Search: 706/48, 706/45, 706/60, 706/50, 706/919, 706/911, 706/925, 706/2, 706/7, 706/6, 706/5, 706/23, 706/8, 706/52, 706/47, 705/7
• US Class Current: 706/8
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling

Abstract

Latent effects on a system are broken down into components ranging from those far removed in time from the system under study (latent) to those which closely effect changes in the system. Each component is provided with weighted inputs either by a user or from outputs of other components. A non-linear mathematical process known as ‘soft aggregation’ is performed on the inputs to each component to provide information relating to the component. This information is combined in decreasing order of latency to the system to provide a quantifiable measure of an attribute of a system (e.g., safety) or to test hypotheses (e.g., for forensic deduction or decisions about various system design options).

Claims

1. A computer implemented method of utilizing latent effects of prior events in making decisions about a system having an output indicative of the state of the system, said method comprising:

decomposing the system into a series of subsystems, each of said subsystems having at least one input and an output, the outputs of all but one subsystems forming inputs for subsequent subsystems, the output of the last subsystem in the series being the system output;

arranging said subsystems in time order according to the immediacy with which a change in an output of a subsystem causes a change in the system output, the subsystems earlier in time order providing information representative of latent effects;

in series order starting with the subsystem furthest in time from the system output, using a computer to apply a non-linear function that asymptotically approaches each of a first value as the value of positive inputs increase and a second lower value as the value of negative inputs increase to determine the output for that subsystem.

2. The method of claim 1 further comprising decomposing any subsystem into a series of additional subsystems.

3. The method of claim 2 wherein each of the inputs are weighted.

4. The method of claim 3 wherein the sum of the weights of the positive inputs and the sum of the weights of the negative inputs each equal one.

5. The method of claim 4 wherein each output $f=[1-a^{-f}$

(q)]·

b-f

(r),

where a and b are constants, ƒ

(q) is a function of one of the positive negative inputs, and ƒ

(r) is a function of the other of the positive or negative inputs.

6. The method of claim 5 wherein $f=[1-{}^{}$

-∑

i=1n

kg

wi

xi]



-∑

j=1m

kb

vj

yj

where w_i and v_j are “

weights”

on the significance of the n “

positive” and

m “

negative”

factors being;

x_iare the positive inputs and y_i are the negative inputs on a scale of 0 to 1; and

the constants k_g and k_b are variables that controls how fast positive and negative information accumulates.

7. The method of claim 6 wherein $\sum _{}^{}$

i=1n

wi=1

 



and

 



∑

j=1m

vj=1.

8. The method of claim 6 wherein each input covers a range extending from a highest value to a lowest value, and each output is displayed as a range having a low value determined from the lowest values of inputs and a highest value determined from the highest values of inputs.

9. The method of claim 8 wherein said inputs and outputs are represented on a computer display.

10. The method of claim 6 further comprising enabling a warning indicator when a determined output exceeds a predetermined value.

12. The method of claim 3 wherein the system output is representative of a quantifiable measure of an attribute of a system.

13. The method of claim 12 further comprising using the computer implemented method to model historical data for a system, and adjusting the weights to give an output corresponding to the historical data.

11. The method of clam 10 wherein said warning indicator provides a first indication when said predetermined output is a predetermined amount below said predetermined value, a second indication when said predetermined output is approximately at said predetermined value, and a third indication when said predetermined output is a predetermined amount above said predetermined value.

14. A computer implemented method of utilizing latent effects of prior events in making decisions about a system having an output indicative of the state of the system, said method comprising:

decomposing the system into a series of subsystems, each of said subsystems having at least one input and an output, the outputs of all but one subsystems forming inputs for subsequent subsystems, the output of the last subsystem in the series being the system output;

further decomposing any subsystem into a series of subsystems;

arranging said subsystems in time order according to the immediacy with which a change in an output of a subsystem causes a change in the system output, the subsystems earlier in time order providing information representative of latent effects; and

in series order stating with the subsystem furthest in time from the system output, using a computer to apply a weighted non-linear function that asymptotically approaches each of a first value as the value of positive inputs increase and a second lower value as the value of negative inputs increase to determine the output for that subsystem, where the sum of the weights of the positive inputs and the sum of the weights of negative inputs each equal one and each output $f=\left[1-a^{-f<br><br>\left(q\right)}\right]·<br><br>b^{-f<br><br>\left(r\right)},$

where a and b am constants, ƒ

(q) is a function of one of the positive negative inputs, and ƒ

(r) is a function of the other of the positive or negative inputs.