User directed heuristic design optimization search

• US 6,086,617 A
• Filed: 07/18/1997
• Issued: 07/11/2000
• Est. Priority Date: 07/18/1997
• Status: Expired due to Fees

• Current Assignee: Engineous Software, Inc. (Dassault Systemes SA)
• Original Assignee: Engineous Software, Inc. (Dassault Systemes SA)
• Inventors: Waldon, Scott, Tong, Siu, Powell, Dave
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): Broda, Samuel
• Application Number:
  US08/898,041
• Time in Patent Office: 1,089 Days
• Field of Search: 364/578, 364/489, 364/488, 364/164, 395/500, 395/650, 395/22, 395/500.23, 371/27, 381/43, 706/46, 706/60, 706/902, 706/920
• US Class Current: 703/2
• CPC Class Codes: G05B 17/02 electric
• CPC Class Codes: Y10S 706/902 Application using ai with detail of the ai system
• CPC Class Codes: Y10S 706/92 Simulation

Abstract

A directed heuristic search (DHS) has been developed for an optimization design system. The DHS an optimization simulation system has been developed. The DHS directs a design optimization process that implements a user'"'"'s selections and directions, the order and directions in which the search for an optimal design is to be conducted, and how the search is to sequence through potential design solutions. In addition, DHS is tractable, in that it records a log of the sequence of design solutions that were evaluated during a particular design optimization search. The user may review this search log, to understand how and why the DHS made design optimization decisions. The user may change search parameter settings in a user defined dependency to change future search patterns for optimal designs.

Claims

1. A user directed search method for optimizing a model having input and output parameters, where the steps of the method are performed using a computer having a memory device and the computer executes at least one executable program that simulates the model and the input and output parameters are inputs and outputs to the at least one executable program, and the steps comprise the following:

a. Creating a dependency table stored in the memory device, where the dependency table correlates the input parameters to the output parameters of the model and the dependency table indicates how input parameter values should be permutated and the table is at least partially created by a user based on empirical data available to the user and based on user-defined goals;

b. For a selected output parameter, specifying a group of input parameters, including some but not all of the inputs to the model, that influence the selected output parameter;

c. Repeating step (b) for a plurality of the output parameters;

d. For each output parameter of the plurality of output parameters, permutating values for the input parameters in the group of input parameters that influence the selected output parameter;

e. Solving the model using the permutated values as the input parameters to the model by executing on the computer the at least one executable program using the permutated values as input values to the program;

f. Optimizing the model by repeating steps (d) and (e) iteratively in accordance with one or more of the goals;

g. Ending the optimization step (f) after an incremental gain in advancing the solution toward the goal is less than a predetermined minimal gain, and

h. Outputting an optimized solution to the model after step (g).

10. A user directed search method for optimizing a model as in claim 1 wherein the empirical data recited in step (a) includes data learned by the user from observation of a system corresponding to the model.

2. A user directed search method as set forth in method 1, further comprising the steps of:

f. Prior to step (d), ranking the output parameters, and performing step (d) with respect to each output parameter in an order dependent on the ranking of output parameters.

3. A user directed search method as set forth in method 1, further comprising the steps of:

i. Prior to step (d), ranking the input parameter in each of one or more of the groups of input parameters, and performing step (d) with respect to each group of input parameters in an order dependent on the ranking of input parameters.

4. A user directed search method as set forth in method 1, further comprising the steps of:

i. Prior to step (d), specifying in the dependency table a user selected direction for change of each input parameter in one or more of the groups of input parameters, and in step (d) permutating each input parameter in the direction specified in the dependency table.

5. A user directed search method as set forth in method 4, wherein in step (i) an amount of change to be made to each input parameter is selectable by the user.

6. A user directed search method as set forth in method 5, wherein in step (i) the amount of change to be made to each input parameter is defined as a percentage of a current value of the input parameter.

7. A user directed search method as set forth in method 4, wherein in step (i) the amount of change to be made to each input parameter increases if a current solution to the model is an improvement to the output parameter.

8. A user directed search method for optimizing a model embodied in at least one program executed on a computer having a memory, where the model has user-defined hard constraints on one or more output parameters of the model, user-defined soft constraints on one or more output parameters of the model, and user-defined goals for one or more output parameters of the model, and where a dependency table stored in the memory correlates input parameters to the model to the output parameters, where the input and output parameters and goals are stored in the memory and the steps of the method comprise the following:

a. Solving the model by permutating the input parameters corresponding to the output parameters subject to the hard constraints, until the hard constraints are satisfied;

b. After step (a), solving the model by permutating the input parameters corresponding to the output parameters subject to the soft constraints, to satisfy the soft constraints without violating any hard constraint;

c. After step (b), setting values corresponding to the soft constraints as hard constraints;

d. After step (c), optimizing a solution to the model in accordance with the goals of the model and without violating the constraints, by permutating the input parameters corresponding to the output parameters;

e. optimizing the solution in step (d) until an incremental gain in advancing the solution to the goal is less than an incremental gain specified for the goal, and

f. outputting an optimized solution after step (e).

9. A user directed search method as set forth in claim 8 further comprising the following steps:

g. Assigning a user selected give back value or values to one or more of the goals, and, in step (d), and permitting solutions to vary from goals by an amount equal to the give back value or values.

11. A user directed search method for optimizing a model as in claim 8 further comprising the step of (f) of creating the dependency table by determining relationships between the output parameters and input parameters based on empirical data known to a user.