Virtual continuous emission monitoring system

• US 5,548,528 A
• Filed: 01/30/1995
• Issued: 08/20/1996
• Est. Priority Date: 08/05/1993
• Status: Expired due to Term

• Current Assignee: Rockwell Automation Technologies Incorporated (Allen-Bradley Co., Inc.)
• Original Assignee: Pavilion Technologies Incorporated (Rockwell Automation, Inc.)
• Inventors: Keeler, James D., Havener, John P., Ferguson, Ralph B., Godbole, Devendra
• Primary Examiner(s): Voeltz, Emanuel T.
• Assistant Examiner(s): SHAH, KAMINI S
• Application Number:
  US08/380,062
• Time in Patent Office: 568 Days
• Field of Search: 364/577, 364/571.04, 364/571.05, 364/492, 364/495, 364/497, 395/20-23, 73/31.02-31.03
• US Class Current: 702/22
• CPC Class Codes: F02D 2041/1433 using a model or simulation of the system
• CPC Class Codes: F02D 41/1401 characterised by the control or regulation method F02D41/1473, F02D41/1477 take precedence
• CPC Class Codes: F02D 41/1405 Neural network control
• CPC Class Codes: G01N 33/0034 comprising neural networks or related mathematical techniques
• CPC Class Codes: G01N 33/0075 for multiple spatially distributed sensors, e.g. for environmental monitoring transmission systems for measured values G08C
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: G05B 13/048 using a predictor
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks

Abstract

A continuous emission monitoring system for a manufacturing plant (10) includes a control system (16) which has associated therewith a virtual sensor network (18). The network (18) is a predictive network that receives as inputs both control values to the plant (10) and also sensor values. The network (18) is then operable to map the inputs through a stored representation of the plant (10) to output a predicted pollutant sensor level. This predicted pollutant sensor level is essentially the prediction of an actual pollutant sensor level that can be measured by a pollutant sensor (14). The network (18) therefore is a substitute for the pollutant sensor (14), thus providing a virtual sensor. The sensor values from the plant (10) are first processed through a sensor validation system (22). The sensor validation system (22) is operable to determine which of the sensors falls outside of acceptable limits and then substitutes therefor a predicted output, such that the inputs to the virtual sensor network (18) are within acceptable limits.

Claims

1. A method for monitoring select output parameters in a manufacturing plant that are generated as a byproduct of the operation of the plant, the manufacturing plant having controls operating in response to received control values to alter the operation of the plant, and sensors to measure the operating parameters of the plant and output associated sensor values, comprising the steps of:

determining at least once the level of the select output parameters output by the plant and the associated control values to the plant in addition to the associated sensor values, the step of determining at least once utilizing a predetermined determination process;

measuring on a substantially frequent basis the control values to the plant and the sensor values output by the sensors;

storing a representation of the plant in association with the determination process in a virtual sensor predictive network having as an output a predicted select output parameter level that corresponds to the actual select output parameter level determined by the determination process to be output by the manufacturing plant, the control values of the plant and the sensor values of the plant comprising inputs to the virtual sensor predictive network, the stored representation learned from the determined output parameter level by the determination process, the control values and the sensor values, the virtual sensor predictive network operable to be trained upon the determined level of the select output parameters, the measured control values and the sensor values to generate the stored representation; and

generating an error when the predicted output parameter level exceeds a predetermined threshold.

2. The method of claim 1, wherein the step of measuring the control values to the plant and the sensor values output by the sensors is done on a real time basis.

3. The method of claim 1, wherein the step of determining at least once comprises determining on a periodic basis the level of select output parameters output by the plant.

4. The method of claim 1, and further comprising the step of generating a training error when the difference between the determined level of the select output parameters for a given set of control values and sensor values differs from a predicted output parameter level for substantially the same control values and sensor values exceed a predetermined error threshold.

5. The method of claim 4, and further comprising the step of retraining the virtual sensor predictive network upon generation of the training error.

6. The method of claim 1, and further comprising the step of controlling the plant utilizing the predicted level in accordance with a predetermined control scheme to vary the control values to the plant.

7. The method of claim 1, wherein the select output parameters comprise the level of pollutants emitted by the manufacturing plant and the step of determining at least once the level of pollutants emitted by the plant comprises periodically disposing an external emissions sensor in a predetermined location within the manufacturing plant to measure the level of pollutants emitted by the plant and, after the level of pollutants is measured, removing the external emissions sensor.

8. A system for monitoring select output parameters in a manufacturing plant that are generated as a by-product of the operation of the plant, the manufacturing plant having controls operating in response to received control values to alter the operation of the plant and sensors to measure the operating parameters of the plant and output associated sensor values, comprising:

a system for measuring at least once the select operating parameters emitted by the plant and the associated control values to the plant with the associated sensor values, the determining system utilizing a predetermined determination process;

a plant information system for measuring the control values to the plant and the sensor values of the plant;

a virtual sensor predictive network for storing a representation of the plant in association with the determination process and having as an output a selected operating parameter level that corresponds to the actual select operating parameter level determined by the determination process to be output by the manufacturing plant, the control values of the plant and the sensor values of the plant from said plant information system comprising inputs to said virtual sensor predictive network, the stored representation learned from the determined select operating parameter level as determined by the determination process, the control values and the sensor values, said virtual sensor predictive network operable to be trained upon the determined level of the select operating parameters, the measured control values and the sensor values to generate the stored representation; and

an error generator for generating an error when the predicted pollutant level exceeds a predetermined threshold.

9. The system of claim 8, wherein the select operating parameters and the pollutants emitted by the plant and said measurement system comprises a pollutant sensor for measuring the level of pollutants emitted by the plant.

10. The system of claim 9, wherein:

said pollutant sensor is operable to measure the level of pollutants on a periodic basis and compare the measured level with the predicted pollutant level to determine the difference therebetween and generate a training error when the difference therebetween exceeds the predetermined threshold;

said virtual sensor predictive network operable to be trained upon the measured level of pollutants, the measured control values and the sensor values to generate the stored representation; and

further comprising;

a training system for retraining said virtual sensor predictive network upon the generation of said training error.

11. The system of claim 8, wherein said plant information system operate on a substantially real time basis.