Predictive network with graphically determined preprocess transforms

• US 6,002,839 A
• Filed: 08/21/1997
• Issued: 12/14/1999
• Est. Priority Date: 11/24/1992
• Status: Expired due to Term

• Current Assignee: Rockwell Automation Technologies Incorporated (Allen-Bradley Co., Inc.)
• Original Assignee: Pavilion Technologies Incorporated (Rockwell Automation, Inc.)
• Inventors: O'Hara, Steven A., Keeler, James D., Hartman, Eric J., Godbole, Devendra B., Kempf, Jill L.
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Dorvil, Richemond
• Application Number:
  US08/915,850
• Time in Patent Office: 845 Days
• Field of Search: 395/21, 395/22, 395/23, 395/24, 395/11
• US Class Current: 706/23
• CPC Class Codes: G06N 3/047 Probabilistic or stochastic networks
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs

Abstract

A predictive network is disclosed for operating in a runtime mode and in a training mode. The network includes a preprocessor (34'"'"') for preprocessing input data in accordance with parameters stored in a storage device (14'"'"') for output as preprocessed data to a delay device (36'"'"'). The delay device (36'"'"') provides a predetermined amount of delay as defined by predetermined delay settings in a storage device (18). The delayed data is input to a system model (26'"'"') which is operable in a training mode or a runtime mode. In the training mode, training data is stored in a data file (10) and retrieved therefrom for preprocessing and delay and then input to the system model (26'"'"'). Model parameters are learned and then stored in the storage device (22). During the training mode, the preprocess parameters are defined and stored in a storage device (14) in a particular sequence and delay settings are determined in the storage device (18). During the runtime mode, runtime data is derived from a distributed control system (24) and then preprocessed in accordance with predetermined process parameters and delayed in accordance with the predetermined delay settings. The preprocessed data is then input to the system model (26'"'"') to provide a predicted output, which is a control output to the distributed control system (24).

Claims

1. A system for generating inputs for controlling the operation of a plant, comprising:

a control system for controlling the plant, and having;

a control network for generating plant inputs to operate the plant,

a run time model for receiving inputs and mapping the received inputs through a stored representation of the plant trained on a given set of inputs and outputs to provide on the output predicted control inputs for input to the plant, and

a run time preprocessor for applying transforms to said inputs from the plant that are input to the run time model which transforms are operable to transform the data to said inputs in a predetermined manner, and

a transform modification device for modifying the transforms, and the associated predetermined manner in which they operate, in said run time preprocessor with said stored representation in said run time model being fixed, and in accordance with user determined transforms that are applied to said control system during running thereof.

2. The system of claim 1, and further comprising a simulation network for determining said transforms utilized by said run time preprocessor independent of the operation of said control system, and wherein said transform modification device modifies the transforms in said run time preprocessor such that said determined transforms by said simulation network can be applied to said control system during running thereof.

3. The system of claim 2, wherein said simulation network comprises:

an off-line preprocessor substantially identical to said run time preprocessor;

a database for storing data representative of data obtained from the plant during running thereof;

an off-line model substantially similar to said run-time model for receiving inputs from said off-line preprocessor and mapping the received inputs through a stored representation of the plant to an output, the output providing predicted control values representing the predicted control inputs output by said run-time model;

said off-line preprocessor operable to apply off-line transforms to said received data from said database prior to input to said off-line model; and

a control system for allowing a user to modify said transforms associated with said off-line preprocessor during operation thereof and wherein said transforms utilized by said off-line preprocessor can be accessed by said transform modification device for application of said off-line transforms to said run-time preprocessor.

4. The system of claim 3, wherein said control system includes a graphical interface for allowing a user to view the operation of said off-line model during operation thereof.

5. The system of claim 4, wherein said run-time model has an input space that is represented by data stored in said database, and said off-line model occupies substantially the same input space, wherein said control system is operable to modify said input space with said transforms, and said transform modification device is operable to translate modifications to the input space of said off-line model directly to the input space of said run-time model via said run-time preprocessor.

6. The system of claim 5, wherein said graphical interface includes a display device for graphically displaying said data in a predetermined format with said control system interfacing with said display device to allow a user to graphically manipulate said input data that represents said input space of said off-line model and to a transform conversion system for converting said graphical manipulation to a transform of said input data prior to input to said off-line model that is utilized by said off-line preprocessor.

7. The system of claim 3, wherein said database contains data obtained from the plant during running thereof.

8. A method for controlling the operation of a plant, comprising:

providing a control system for controlling the plant;

generating plant inputs to operate the plant;

storing a representation of the plant in a run-time model which has inputs for receiving data that corresponds to data input to the plant, which received data on the inputs is mapped through the stored representation of the plant to provide on the output of the run-time model predicted control inputs for input to the control system;

preprocessing in a run-time preprocessing step with a run-time preprocessor the data prior to input to the run-time model by applying transforms to the inputs to the plant that are input to the run-time model which transforms are operable to transform the data to said inputs in a predetermined manner; and

modifying the transforms, and the associated predetermined manner in which they operate, in the run-time preprocessing step with the stored representation in the run time model being fixed, and in accordance with user determined transforms, which modified transforms are applied to the control system during the running thereof.

9. The method of claim 8, and further comprising a simulation step of determining the transforms utilized by the run-time preprocessing step independent of the operation of the control system to provide simulated transforms, and wherein the step of modifying the transforms in the run-time preprocessing step comprises utilizing the simulated transforms determined by the simulation step, which transforms are applied to the control system during running thereof.

10. The method of claim 9, wherein the simulating step comprises:

providing an off-line model substantially similar to the run-time model which provides a stored representation of the plant and processes its inputs through the stored representation to provide on the output predicted control values representing the predicted control inputs output by the run-time model;

preprocessing data that is input to the off-line model with an off-line preprocessor in an off-line preprocessing step;

storing data representative of data obtained from the plant during running thereof, this data input to the off-line preprocessor; and

modifying the transforms associated with the off-line preprocessor during operation thereof and wherein the transforms utilized by the off-line preprocessor can be accessed by the step of modifying the transforms in the run-time preprocessor.

11. The method of claim 10, and further comprising the step of providing a graphical interface and allowing a user to view the operation of the off-line model during operation thereof.

12. The method of claim 11, wherein the step of providing the run-time model comprises providing a run-time model that has an input space that is represented by data stored in a database, and the step of providing the off-line model comprises providing an off-line model having an input space that occupies substantially the same input space as the run-time model, wherein the step of preprocessing is operable to modify the input space with the transforms, and is operable to translate modifications to the input space of the off-line model directly to the input space of the run-time model via the run-time preprocessor.

13. The model of claim 12, wherein the step of providing a graphical interface, comprises the step of graphically displaying the data on a display device in a predetermined format with the display device interfaced with the control system to allow a user to graphically manipulate the input data that represents the input space of the off-line model and to then convert the graphical manipulation to a transform of the input data prior to input to the off-line model that is utilized with the off-line preprocessor.