Sensor placement and control design for distributed parameter systems

• US 6,772,044 B1
• Filed: 04/13/2000
• Issued: 08/03/2004
• Est. Priority Date: 04/13/2000
• Status: Expired due to Term

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Parthasarathy, Sanjay, Mathur, Anoop, Foslien-Graber, Wendy
• Primary Examiner(s): Picard, Leo
• Assistant Examiner(s): Rao, Sheela S.
• Application Number:
  US09/548,537
• Time in Patent Office: 1,573 Days
• Field of Search: 700/197, 700/204, 264/40.1
• US Class Current: 700/204
• CPC Class Codes: G05B 13/024 in which a parameter or coefficient is automatically adjusted to optimise the performance
• CPC Class Codes: G05B 13/0265 the criterion being a learning criterion
• CPC Class Codes: G05B 5/01 electric

Abstract

A sensor placement algorithm uses process data to determine the optimal distribution of sensors in a distributed parameter manufacturing system. An automatic classification procedure maps any problems in the process to a predetermined set of process disturbances. A control procedure uses process data to determine the best control action that will ensure good system response. Methods for sensor placement, automatic decision tree classification, corrective action control and the apparatus to effectuate these respective methods are integrated into a design methodology.

Claims

1. A computer-implemented method of controlling a manufacturing process, comprising:

identifying a set of possible disturbances in the process;

determining locations within the process for a set of sensors such that the data for particular disturbances are significantly different from the sensor data for other disturbances;

constructing a set of decision trees, each of the trees having inputs from multiple ones of the sensors and having an output indicating one of the disturbances;

determining a set of control actions for corresponding ones of the trees.

2. The method of claim 1, further includes determining locations within the process for a set of sensors such that any one sensor is maintained at a certain predetermined distance from other sensors.

3. The method of claim 1, wherein determining the sensor locations includes simulating the process both with at least one of the disturbances and without any disturbance.

4. The method of claim 1, further comprising placing the sensors at the determined locations.

5. A medium for holding representations of instructions and data for causing a suitably programmed computer to carry out the method of claim 1.

6. A computer program for controlling a manufacturing process, the program comprising:

an identification module for identifying a set of possible disturbances in the process;

a location-determining module for determining locations within the process for a set of sensors such that the data for particular disturbances are significantly different from the sensor data for other disturbances;

a decision-constructing module for constructing a set of decision trees, each of the trees having inputs from multiple ones of the sensors and having outputs indicating different ones of the disturbances; and

an action-determining module for determining at least one of a plurality of control actions, each control action ameliorating the one disturbance indicated by the outputs of the one decision tree.

7. The computer program of claim 6 where the identification module performs a number of simulations of the manufacturing process.

8. A method of determining optimal sensor locations in a distributed parameter process, comprising:

establishing a plurality of nodes in the process;

running M simulations for a plurality of disturbances, and a simulation with no disturbances;

establishing a process simulation function responsive to at least one process parameter; and

a number I that represents a particular iteration of the current simulation;

establishing an objective function equaling the summation, ranging from simulations 1 to M, of a weighting factor multiplied by the difference between the process simulation functions with the disturbances and the process simulation function for the simulation with no disturbances;

specifying an acceptable range for the objective function output;

eliminating those nodes which do not have an objective function output value within an acceptable range; and

assigning a plurality of sensors to the remaining nodes.

9. The method of claim 8 where the nodes have physical locations within the process.

10. The method of claim 9, further comprising providing that the sensors assigned to the remaining nodes are spaced a minimum distance from each other.

11. The method of claim 10, where the minimum distance is calculated using a Euclidean norm.