US 5,197,114 A · Computer neural network regulatory process control system and method

Abstract

A computer neural network regulatory process control system and method allows for the elimination of a human operator from real time control of the process. The present invention operates in three modes: training, operation (prediction), and retraining. In the training mode, training input data is produced by the control adjustment made to the process by the human operator. The neural network of the present invention is trained by producing output data using input data for prediction. The output data is compared with the training input data to produce error data, which is used to adjust the weight(s) of the neural network. When the error data is less than a preselected criterion, training has been completed. In the operation mode, the neutral network of the present invention provides output data based upon predictions using the input data. The output data is used to control a state of the process via an actuator. In the retraining mode, retraining data is supplied by monitoring the supplemental actions of the human operator. The retraining data is used by the neural network for adjusting the weight(s) of the neural network.

Claims44

Dependent claims

A computer neural network process control method adapted for predicting output data provided to an actuator used to control a process for producing a product having at least one product property, said method allowing the process to be controlled without a human operator, the computer neural network process control method comprising the steps of:

(1) configuring the neural network by specifying at least one input, at least one output, at least one training input, and at least one specified interval;
(2) training the neural network to produce a trained neural network comprising the substeps of:
(a) retrieving a first raw training input data;
(b) retrieving a second raw training input data;
(c) computing a corresponding first training input data based on said first raw training input data and said second raw training input data, said first training input data indicative of the action of a human operator of the process;
(d) retrieving a first input data;
(e) predicting a first output data using said first input data;
(f) computing a first error data in accordance with said first training input data and said first output data;
and
(g) training the neural network to produce said trained neural network in accordance with said first error data;
(3) at said at least one specified interval, retrieving a second input data and predicting, with said trained neural network weights, second output data using said second input data;
(4) retrieving said second output data;
and
(5) changing a state of the actuator in response to said second output data of step (4).

11 more elements

A computer neural network process control system adapted for predicting output data provided to an actuator used to control a process for producing a product having at least one product property, said system allowing the process to be controlled without a human operator, the computer neural network process control system comprising:

(a) a neural network, responsive to a first input data indicative of sensor or aim or lab data, comprising:
(1) training means, responsive to a first training input data indicative of the action of a human operator of the process, for training said neural network in accordance with a training step having said first input data and said first training input data to produce trained neural network weights, and
(2) predicting means using said trained neural network weights for predicting the output data in accordance with a second input data indicative of sensor or aim or lab data;
and
(b) an actuator, responsive to said predicting means, comprising,
(1) sending means for adjusting a state of the actuator in accordance with the output data without a setpoint adjustment by a controller, and
(2) adjustment means for accepting a change made by said human operator to said state of the actuator.

4 more elements

The computer neural network process control system adapted for predicting output data provided to an actuator used to control a process of producing a product having at least one product property, said system allowing the process to be controlled without a human operator, the computer neural network process control system comprising:

(a) a neural network, responsive to a first input data indicative of sensor or aim or lab data, comprising,
(1) training means, responsive to a first training input data indicative of the action of a human operator of the process, for training said neural network in accordance with a first training set having said first input data and said first training input data to produce trained neural weights, and
(2) predicting means using said trained neural network weights for predicting the output data in accordance with a second input data indicative of sensor or aim or lab data;
(b) an actuator, responsive to said predicting means, comprising,
(1) sending means for adjusting a state of the actuator in accordance with the output data without a setpoint adjustment by a controller, and
(2) adjustment means for accepting a change made by said human operator to said state of the actuator;
and
(c) an historical database comprising,
(1) storing means for storing said first training input data with an associated first timestamp, and for storing said first input data indicated by said associated first timestamp, and
(2) retrieving means, responsive to said storing means, for retrieving said first training set comprising said first training input data and said first input data indicated by said associated first timestamp, and connected to provide said first training set to said training means.

7 more elements

Dependencies are inferred from the claim text.

Computer neural network regulatory process control system and method

Life of this patent

Computer neural network regulatory process control system and method

Life of this patent

Abstract

Claims44