Computer method and user interface for decision analysis and for global system optimization

• US 6,735,596 B2
• Filed: 06/07/2001
• Issued: 05/11/2004
• Est. Priority Date: 06/07/2001
• Status: Expired due to Fees

• Current Assignee: Guy Charles Corynen
• Original Assignee: Guy Charles Corynen
• Inventors: Corynen, Guy Charles
• Primary Examiner(s): Rones, Charles
• Assistant Examiner(s): Mofiz, Apu M
• Application Number:
  US09/875,119
• Publication Number:
  US 20030167265A1
• Time in Patent Office: 1,069 Days
• Field of Search: 707/1, 707/2, 707/3, 707/4, 707/5, 707/101, 707/102, 600/300, 703/2, 705/8, 705/35, 706/11, 706/15, 706/45, 771/129
• US Class Current: 1/1
• CPC Class Codes: G06Q 10/04 Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" market predictions or forecasting for commercial activities G06Q30/0202
• CPC Class Codes: Y10S 707/99942 Manipulating data structure, e.g. compression, compaction, compilation
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

A stand-alone computer application for solving decision problems, conducting decision and risk analyses, and for globally optimizing the operation of systems and processes is disclosed. Running on a personal computer, the application embodies three major methods. The first method is project oriented and includes highly modularized procedures for solving and analyzing sequential, probabilistic and multi-objective decision problems. It enables a user to specify and to seamlessly integrate prioritized objectives, performance measures, utility functions, decisions, choices, outcomes, probabilities, multi-dimensional costs and external parameters into an objectives-based decision model. It further includes procedures to assist a user in the construction of objectives hierarchies, utility functions, cost functions and probability expressions.

The second method generalizes the first method into a system controller for globally optimizing a process or system whose operation involves sequential probabilistic and multi-objective decisions that cannot be addressed using conventional linear, non-linear and stochastic control methods. It provides a framework for modeling hybrid discrete-continuous control systems where discrete events and processes are modeled at upper levels of a control hierarchy and continuous processes are modeled at lower levels.

The third method provides a multiple-window graphical user interface to assist a user in applying the first two methods, and includes procedures for manipulating and editing textual and graphical outputs.

Claims

1. A method for modeling, solving and analyzing sequential, probabilistic and multi-objective decision problems, said method stored and executed as a stand-alone application on a computer having a processor, a color display screen, a pointing and selection device, a keyboard, and means for storing data and information, the method comprising the steps of:

(a) receiving, as a first specification for a decision problem, one or more prioritized objectives for said decision problem, including a parent-child structure relating any and every parent objective to its offspring,

(b) organizing said objectives and parent-child structure received in Step (a) as a multi-level objectives hierarchy in accordance with said first specification,

(c) receiving, as a second specification, a collection of one or more performance measures for the decision problem together with a pairing of each performance measure with one and only one of said objectives,

(d) for each performance measure in said collection of performance measures, receiving, as a third specification, a not necessarily monotonic, linear or exponential mathematical function, said function assigning a real number to each of a plurality of values of said performance measure, or receiving a collection of parameters and parameter values enabling the construction of said mathematical function,

(e) receiving, as a fourth specification, a plurality of nodes, each node assigned a name and further specified as either a decision node, a chance node, a terminal node or a continuation node, together with a parent-child structure relating every parent node to its offspring, further receiving a selection of the principal node for the decision problem and, for every parent node, a specification of a contribution to each performance measure in said collection of performance measures, said contribution accrued during a transition from the parent node to each of its offspring and, for every chance node, a specification of a probability of transition from said chance node to each of its offspring and, for every continuation node, the name of the node to which it is continued,

(f) creating a decision model for said decision problem by processing said second through fourth specifications and organizing said specifications into a sequential decision network,

(g) solving the decision problem by processing said sequential decision network with the solver engine of the method, the resulting solution comprising an optimal decision policy for the problem and an optimal principal choice for a root node selected by a user,

whereby an optimal solution is obtained for any sequential, probabilistic and multi-objective decision problem specified in accordance with Steps (a) through (d), said solution comprising an optimal decision policy and an optimal choice for the principal decision node selected by a user.

2. The method of claim 1, further comprising procedures for performing one-way and two-way sensitivity analyses, said analyses providing the sensitivities of a decision problem solution to variations in parameters associated with said sequential decision network.

3. The method of claim 1, further comprising procedures for performing a Tornado Analysis, said analysis providing a combination of one-way sensitivities of a decision problem solution to a limited number of variations in a plurality of parameters associated with said sequential decision network.

4. The method of claim 1, wherein Step (a) further includes receiving a prioritization of said objectives and processing said prioritization with the method, comprising the steps of:

(a) eliciting a relative ranking of the objectives,

(b) further refining said relative ranking by eliciting an absolute ranking consistent with the relative ranking,

(c) numerically normalizing said absolute ranking, thereby obtaining an assignment of an absolute numerical priority to each objective, said assignment including the property that, for every parent objective, the sum of the normalized priorities assigned to its offspring equals unity.

5. The method of claim 1, further including procedures for performing a plurality of editing operations on the objectives hierarchy.

6. The method of claim 1, wherein said collection of performance measures further includes quantitative, discrete, continuous, qualitative or linguistic measures, and wherein Step (c) further comprises the steps of receiving for each in the collection of performance measures a specification of:

(a) upper bounds, lower bounds, and constraint values,

(b) a scale granularity and step size,

whereby an economical representation of said performance measures is obtained, said representation enabling efficient optimizations and a thorough management of constraint violations during said optimizations.

7. The method of claim 1, wherein Step (d) further includes procedures for aiding a user in the construction of utility functions, said functions not necessarily monotonic, linear or exponential, and said procedures including the application of the Mid-Value Splitting Method in accordance with the principles and practices of modern decision theory.

8. The method of claim 1, wherein Step (d) further includes procedures for aiding a user in the construction of utility functions, said functions not necessarily monotonic, linear or exponential, and said procedures including a point-by-point construction of piece-wise linear utility functions.

9. The method of claim 1, wherein step (d) further includes procedures for aiding a user employing a pointing and selection device and a keyboard to specify explicit mathematical expressions for probability functions, cost and benefit functions and utility functions, said probability, cost, benefit and utility functions not necessarily monotonic, linear or exponential, and said expressions comprised in a list of expressions included in the invention and selected from said list by activating a command button from the second plurality of command buttons, said command button selected in accordance with Step (e).

10. The method of claim 1, further comprising a procedure for creating auxiliary variables, said auxiliary variables representing conditions, parameters or values internal or external to the network, said procedure further enabling a user to include said auxiliary variables in expressions for specifying the costs, benefits, probabilities and utilities associated with any node of a probabilistic decision network.

11. The method of claim 1, wherein Step (e) further includes procedures for aiding a user in the construction of probability functions, said procedures including the use of a probability wheel in accordance with the principles and practices of modern decision theory.

12. The method of claim 1, further comprising a plurality of procedures for managing data and information received and generated while executing one or more of the steps recited in claims 1 through 11, said procedures comprising:

(a) a procedure for organizing said data and information into a collection of computer files, thereby creating a new project or updating an existing project,

(b) a procedure for appending a new project created in Step (a) to a stored list of previously created projects, and storing said new project in a computer memory location specified by a user,

(c) a procedure for retrieving any project on said stored list, thereby enabling a user to perform a plurality of operations on a retrieved project, said operations comprising;

(i) reviewing data associated with said retrieved project,

(ii) creating a copy of the retrieved project

(iii) deleting the retrieved project

(iv) executing the method in accordance with the data, information, parameters and models associated with the retrieved project

(v) in accordance with Steps (a) through (e) of claim 1, editing data and information associated with the retrieved project and executing the method using the edited data and information.

13. The method of claim 1, said method embodied as a controller for optimizing the global performance of a process by solving a plurality of decision problems associated with the operation of a plant comprised in said process, said controller functioning in conjunction and cooperation with conventional plant regulators and producing globally optimal plant commands derived from solutions to each of said plurality of decision problems in accordance with the steps of:

(a) receiving specifications in accordance with Steps (a), (c), (d) and (e) of claim 1, creating an objectives hierarchy in accordance with Step (b) of claim 1, and creating a sequential decision model in accordance with Step (e) of claim 1,

(b) receiving a model of the controlled process, said process model including a set of parameters for optimization by said controller,

(c) for each offspring choice of the principal node received in Step (a) of the current claim, collecting measurements from and about the process and updating the process model with the purpose of maintaining a faithful process model,

(d) for each offspring choice of the principal node received in Step (a) of the current claim, executing the process model, estimating, from the measurements collected during said execution, values for the performance measures received in Step (c) of claim 1 and for the probabilities associated with the parent-offspring chance node transitions received in Step (e) of claim 1, and including the results of said estimating in the sequential decision model,

(e) solving the decision model resulting from Step (d), thereby obtaining optimal plant commands and communicating said optimal commands to the controlled process,

(f) repeating Steps (c) through (e) at a rate determined by the system operator,

whereby a method is provided for maximizing the global performance of a process containing a plant whose operation involves sequential, probabilistic and multi-objective decisions.

14. The method of claim 13, further comprising procedures for performing one-way and two-way sensitivity analyses, said analyses providing the sensitivities of process performance and plant commands to variations in parameters associated with said sequential decision network.

15. The method of claim 13, further comprising procedures for performing a Tornado Analysis, said analysis providing a combination of one-way sensitivities of process performance indices and of plant commands to a limited number of extreme variations in a plurality of parameters associated with said sequential decision network.

16. The method of claim 13, wherein Step (a) further includes receiving a prioritization of said objectives and processing said prioritization with the method, comprising the steps of:

(a) eliciting a relative ranking of the objectives,

(b) further refining said relative ranking by eliciting an absolute ranking consistent with the relative ranking,

(c) numerically normalizing said absolute ranking, thereby obtaining an assignment of an absolute numerical priority to each objective, said assignment including the property that, for every parent objective, the sum of the normalized priorities assigned to its offspring equals unity.

17. The method of claim 13, further including procedures for performing a plurality of editing operations on the objectives hierarchy.

18. The method of claim 13, wherein said collection of performance measures may further include quantitative, discrete, continuous, qualitative or linguistic measures, and wherein Step (a) further comprises the steps of receiving for each in the collection of performance measures a specification of:

(a) upper bounds, lower bounds, and constraint values,

(b) a scale granularity and step size,

whereby an economical representation of said performance measures is obtained, said representation enabling efficient optimizations and a thorough management of constraint violations during said optimizations.

19. The method of claim 13, wherein Step (a) further includes procedures for aiding a system operator in the construction of utility functions, said functions not necessarily monotonic, linear or exponential, and said procedures including the application of the Mid-Value Splitting Method in accordance with the principles and practices of modern decision theory.

20. The method of claim 13, wherein Step (a) further includes procedures for aiding a system operator in the construction of utility functions, said functions not necessarily monotonic, linear or exponential, and said procedures including a point-by-point construction of piece-wise linear utility functions.

21. The method of claim 13, wherein step (a) further includes procedures for aiding a system operator using a pointing and selection device and a keyboard to specify explicit mathematical expressions for probability functions, cost and benefit functions and utility functions, said probability, cost, benefit and utility functions not necessarily monotonic, linear or exponential, and said expressions comprised in a list of expressions included in the invention and selected from said list by activating a command button from the second plurality of command buttons, said command button selected in accordance with Step (a).

22. The method of claim 13, further comprising a procedure for creating auxiliary variables, said auxiliary variables representing conditions, parameters or values internal or external to the network, said procedure further enabling a system operator to include said auxiliary variables in expressions specifying the costs, benefits, probabilities and utilities associated with any node of a probabilistic decision network.

23. The method of claim 13, wherein Step (a) further includes procedures for aiding a system operator in the construction of probability functions, said procedures including the use of a probability wheel in accordance with the principles and practices of modern decision theory.

24. The method of claim 13, further comprising a plurality of procedures for managing data and information received and generated while executing one or more of the steps recited in claims 13 and through 23, said procedures comprising:

(a) a procedure for organizing said data and information into a collection of computer files, thereby creating a new project or updating an existing project,

(b) a procedure for appending a new project created in Step (a) to a stored list of previously created projects, and storing said new project in a computer memory location specified by a user,

(c) A procedure for retrieving any project on said stored list, thereby enabling a user to perform a plurality of operations on a retrieved project, said operations comprising;

(i) reviewing data associated with said retrieved project,

(ii) creating a copy of the retrieved project

(iii) deleting the retrieved project

(iv) executing the method in accordance with the data, information, parameters and models associated with the retrieved project

(v) in accordance with Steps (a) through (e) of claim 1, editing data and information associated with the retrieved project and executing the method using the edited data and information.