Palo Alto Research Center Inc.

Other

Lukas Daniel Kuhn

Johan De Kleer

Rong Zhou

Robert Roy Price

Minh Binh Do

Tim C Schmidt

Juan Liu

Eric Saund

Hector Luis Palacios Verdes

A model-based control system for controlling a production system is disclosed. The production system provides jobs and objectives to the model-based control system. The model-based control system includes, a planner operative to provide the production system with a plan, the planner generating the plan using a depth-first target value search. The depth-first target value search being configured to generate the plan with a failure probability most closely approximating a target value. Also provided is a system model operative to model the behavior of the production system. Another aspect of the model-based control system is a diagnosis engine operative to estimate failure probabilities for plans and provide diagnostic guidance to the planner. The model-based control system including the planner, the system model and diagnosis engine are implemented via a digital processing device.

Depth-first search for target value problems

A computer based method and system for tiered inference multiple fault diagnosis is provided. The method includes using a computer processor to dissect a hypothesis space representing a production system having a plurality of production modules into tiers. Production modules in the current tier are partitioned into a group or a set of sub-groups. A fault diagnosis algorithm is applied to the group of each sub-group to identify an acceptable fault diagnosis. When no acceptable fault diagnosis is found, the process moves to the next tier to perform further investigations. The process continues to move to higher tiers until an acceptable fault diagnosis is obtained or the system instructs the process to end.

Computationally efficient tiered inference for multiple fault diagnosis

Methods and systems for solving a target value search problem using a multi-interval heuristic are presented. The methods and system identity a path, or paths, in a graph, whereby a connection graph is created and range sets are generated for each vertex in the connection graph. Range sets include one or more intervals. Thereafter, a best search is performed to identify a path, or paths, from a starting vertex to a goal vertex having a path value closest to a target value.

Multi-interval heuristics for accelerating target-value search

A method for determining a target path for a model-based control system. The model-based control system includes a directed acyclic graph, where the directed acyclic graph includes a plurality of vertices interconnected by a plurality of edges. The method includes the steps of performing a depth-first search of the directed acyclic graph for the target path. The depth-first search is operative to return an explicit solution or an implicit solution, wherein the implicit solution is determined using a heuristic. The method further includes determining if the depth-first search returned an explicit solution or an implicit solution, and if the depth-first search returned an implicit solution, constructing the target path from the implicit solution. The method may further include constructing a pattern database.

Diagnostic systems and methods are presented for determining the current condition of a production plant and the resources thereof, in which successively more complex diagnostic abstractions are used to determine the plant condition, with a more complex abstraction being selected when the most recently selected diagnostic abstraction is logically inconsistent with the current fault status indications.

Methods and systems for fault diagnosis in observation rich systems

Target value search methods and systems are presented for solving a target value path problem to identify a path or paths in a graph in which a connection graph is created and upper and lower bound values are determined for each node in the connection graph, and a first best search is performed to identify a path or paths from a starting node to a goal node having a path value closest to the target value.

Methods and systems for target value path identification

A computer-based method and system for pervasive model adaptation are provided. The method includes providing a production plan to a production system, executing the production plan in the production system, collecting observation data produced by at least one sensor in the production system during execution of the production plan into computer memory, calculating changes to be made to a system model in the computer memory based on the collected observation data, dynamically updating the system model according to the calculated changes, creating an updated production plan based at least partially on the dynamically updated system model, and storing the updated production plan within the computer memory.

Pervasive model adaptation: the integration of planning and information gathering in dynamic production systems

Systems and methods are presented for constructing production plans to achieve one or more production goals, in which a diagnosis engine determines the production plant condition based on a previously executed plan, observations from the plant, and a plant model, and formulates a heuristic based on a diagnostic objective, and with a planner to performing a partial-plan search using the heuristic to construct the plans that will concurrently achieve at least one production goal and facilitate the diagnostic objective.

Methods and systems for constructing production plans

Production control systems and methods are presented for estimation of production resource failure probabilities in which a set of four count values are maintained and updated for each resource including a first count value mindicating a number of plans where the resource m was used and failed, a second count value mindicating a number of plans where the resource m was used and succeeded, a third count value mindicating a number of plans where the resource m was not used and failed, and a fourth count value mindicating a number of plans where the resource m was not used and succeeded, and the current fault probability for each resource is estimated and stored for single or multiple, persistent or intermittent faults based on the corresponding count values.

Methods and systems for continuously estimating persistent and intermittent failure probabilities for production resources

A control systems and methods are presented for controlling a production system, in which a model-based planner includes a formulation, such as a SAT formulation representing possible actions in the production, with a solver being used to provide a solution to the formulation based at least partially on production and diagnostic goals and the current plant condition, and a translation component translates the solution into a plan for execution in the plant.

Methods and systems for active diagnosis through logic-based planning

Model-based production control systems and methods are presented for constructing plans for controlling operation of a production system with a plant having a plurality of resources to achieve one or more production goals, in which a planner constructs plans for execution in the plant based on production goals while balancing both production objectives (e.g., production cost, production time) and diagnostic objectives (e.g., diagnostic cost, information gained, repair cost), and a diagnosis engine determines a current plant condition based on a previously executed plan and corresponding observations from the plant, and provides expected information gain data to the planner, with the planner generating a plan that will achieve a given production goal and is improved for one or more diagnostic objectives and the expected information gain data.

Growing community of inventors

Palo Alto, CA, United States of America

Lukas Daniel Kuhn