Control system with neural network trained as general and local models

• US 5,586,033 A
• Filed: 06/06/1995
• Issued: 12/17/1996
• Est. Priority Date: 09/10/1992
• Status: Expired due to Term

• Current Assignee: Deere & Company
• Original Assignee: Deere & Company
• Inventors: Hall, James W.
• Primary Examiner(s): PARK, COLLIN
• Application Number:
  US08/466,957
• Time in Patent Office: 560 Days
• Field of Search: 364/149, 364/150, 364/151, 364/424.07, 364/148, 395/22, 395/904, 395/23, 460/1, 460/4, 460/7
• US Class Current: 701/50
• CPC Class Codes: A01D 41/127 Control or measuring arrangements specially adapted for combines
• CPC Class Codes: G05B 13/027 using neural networks only

Abstract

A neural network is trained with a general set of data to function as a general model of a machine or process with local condition inputs set equal to zero. The network is then retrained or receives additional training on an extentd data set containing the general set of data, characterized by zero values for the local condition inputs, and data on specific local conditions, characterized by non-zero values for the local condition inputs. The result is a trained neural network which functions as a general model when the inputs for the local conditions inputs are set equal to zero, and which functions as a model of some specific local condition when the local condition inputs match the encoding of the some local data set contained within the training data. The neural network has an architecture and a number of neurons such that its functioning as the local model is partially dependent upon its functioning as the general model. This trained neural network is combined with sensors, actuators, a control and communications computer and with a user interface to function as combine control system.

Claims

1. A machine control system for controlling a machine which operates in a variety of locations and conditions and which produces an end result, the control system comprising:

a plurality of actuators, each for controlling a particular function of the machine in response to an actuator control signal;

a plurality of actuator sensors, each generating an actuator condition signal representing a condition of a corresponding one of the actuators;

a plurality of input condition sensors, each generating an input condition sensor signal representing an input condition which influences operation of the machine,

an actuator control unit for generating the actuator control signals as a function of the actuator condition signals and as a function of setpoint signals;

a neural network trained prior to and apart from normal production use of the machine with a set of general training data to function as a general model of the machine and trained to function as a submodel with respect to a set of local condition data together with the set of general training data, the neural network processing the input condition sensor signals and data collected prior to normal production use of the machine representing desired machine performance quality to produce a set of machine adjustments intended to produce the desired machine performance quality, the neural network generating the setpoint signals based upon predicted responses of the machine to varying conditions;

a data communication system comprising means for communicating the actuator signals to the actuator control unit, means for communicating the sensor signals to the neural network, and means for communicating the setpoint signals to the actuator control unit, the neural network and the actuator control unit cooperating to control operation of the machine without measuring the machine performance quality in connection with normal production use of the machine; and

operator controlled means for selectively causing the neural network to function as the general model or as the submodel.

2. A control system for a combine for operating in a variety of locations and conditions and which harvests a crop, the combine having a variable speed threshing cylinder, a concave having a variable clearance, a variable speed cleaning fan, a sieve, a chaffer, and a cleaner extension, the control system comprising:

a set of actuators, each for controlling a particular function of the combine in response to an actuator control signal;

a set of actuator sensors, each generating an actuator condition signal representing a condition of a corresponding one of the actuators;

a set of input condition sensors, each generating an input condition sensor signal representing an input condition which influences operation of the combine;

an actuator control unit for generating the actuator control signals as a function of the actuator condition signals and as a function of setpoint signals;

a neural network trained prior to and apart from normal production use of the combine with a set of general training data to function as a general model of the combine and trained to function as a submodel with respect to a set of local condition data together with the set of general training data, the neural network processing the input condition sensor signals and data collected prior to normal production use of the combine representing desired crop harvesting performance quality to produce a set of combine adjustments intended to produce the desired crop harvesting performance quality, the neural network generating the setpoint signals based upon predicted responses of the combine to varying conditions;

a data communication system comprising means for communicating the feedback signals to the actuator control unit, means for communicating the sensor signals to the neural network, and means for communicating the setpoint signals to the actuator control unit, the neural network and the actuator control unit cooperating to control operation of the combine without measuring the crop harvesting performance quality produced by the combine; and

operator controlled means for selectively causing the neural network to function as the general model or as the submodel.

3. A control system for a combine for operating in a variety of locations and conditions and which harvests a crop, the combine having a plurality of controllable components, the components comprising a variable speed threshing cylinder, a concave having a variable clearance, a variable speed cleaning fan, a sieve, a chaffer and a cleaner extension, the control system comprising:

a set of actuators, each for controlling a corresponding one of the components in response to a corresponding actuator control signal;

a set of actuator sensors, each generating an actuator condition signal representing a condition of a corresponding one of the actuators;

a set of input condition sensors, each generating an input condition sensor signal representing an input condition which influences operation of the combine;

an actuator control unit for generating the actuator control signals as a function of the actuator condition signals and as a function of setpoint signals;

a neural network trained prior to and apart from normal production use of the combine with a set of general training data to function as a general model of the combine and trained to function as a submodel with respect to a set of local condition data together with the set of general training data, the neural network processing the input condition sensor signals and data collected prior to normal production use of the machine representing desired crop harvesting performance quality to produce a set of combine adjustments intended to produce the desired crop harvesting performance quality, the neural network generating the setpoint signals as a function of predicted responses of the combine to varying conditions;

a data communication system comprising means for communicating the actuator signals to the actuator control unit, means for communicating the sensor signals to the neural network, and means for communicating the setpoint signals to the actuator control unit, the neural network and the actuator control unit cooperating to control operation of the combine without measuring the crop harvesting performance quality produced by the combine; and

operator controlled means for selectively causing the neural network to function as the general model or as the submodel.

4. The combine control system of claim 3, wherein the set of input condition sensors comprises:

a relative humidity sensor;

a grain moisture sensor; and

a grain temperature sensor.

5. The combine control system of claim 3, wherein the combine further comprises a set of combine parameter sensors compring:

a grain feedrate sensor;

a threshing cylinder speed sensor;

a concave clearance sensor;

a cleaning fan speed sensor;

a sieve position sensor;

a chaffer position sensor; and

a cleaner extension position sensor.

6. A machine control system for controlling a machine which operates in a variety of locations and conditions and which produces an end result, the control system comprising:

a plurality of actuators, each for controlling a particular function of the machine in response to an actuator control signal;

a plurality of actuator sensors, each generating an actuator condition signal representing a condition of a corresponding one of the actuators;

a plurality of input condition sensors, each generating an input condition sensor signal representing an input condition which influences operation of the machine;

an actuator control unit for generating the actuator control signals as a function of the actuator condition signals and as a function of setpoint signals;

a neural network trained prior to and apart from normal production use of the machine with a set of general training data to function as a general model of the machine and trained to function as a submodel with respect to a set of local condition data together with the set of general training data, the training data being obtained prior to normal production use of the machine and from sources other than the specific machine being controlled and not directly or indirectly from the specific machine being controlled, the neural network generating the setpoint signals based upon predicted responses of the machine to varying conditions;

a data communication system comprising means for communicating the actuator signals to the actuator control unit, means for communicating the sensor signals to the neural network, and means for communicating the setpoint signals to the actuator control unit, the neural network and the actuator control unit cooperating to control operation of the machine; and

operator controlled means for selectively causing the neural network to function as the general model or as the submodel.

7. A machine control system for controlling a machine which operates under a variety of conditions and which produces an end result, the control system comprising:

a plurality of actuators, each for controlling a particular function of the machine in response to an actuator control signal;

a plurality of actuator sensors, each generating an actuator condition signal representing a condition of a corresponding one of the actuators;

a plurality of input condition sensors, each generating an input condition sensor signal representing an input condition which influences operation of the machine;

an actuator control unit for generating the actuator control signals as a function of the actuator condition signals and as a function of setpoint signals;

a neural network trained prior to and apart from normal production use of the machine with a set of general training data to function as a general model of the machine and trained to function as a submodel with respect to a set of local condition data together with the set of general training data, the neural network processing the input condition sensor signals and data collected prior to normal production use of the machine representing desired machine performance quality to produce a set of machine adjustments intended to produce the desired machine performance quality, the neural network generating the setpoint signals based upon predicted responses of the machine to varying conditions;

a data communication system comprising means for communicating the actuator signals to the actuator control unit, means for communicating the sensor signals to the neural network, and means for communicating the setpoint signals to the actuator control unit, the neural network and the actuator control unit cooperating to control operation of the machine; and

operator controlled means for selectively causing the neural network to function as the general model or as the submodel.

8. A control system for a combine for operating in a variety of locations and conditions and which harvests a crop, the combine having a variable speed threshing cylinder, a concave having a variable clearance, a variable speed cleaning fan, a sieve, a chaffer, and a cleaner extension, the control system comprising:

a set of actuators, each for controlling a particular function of the combine in response to an actuator control signal;

a set of actuator sensors, each generating an actuator condition signal representing a condition of a corresponding one of the actuators;

a set of input condition sensors, each generating an input condition sensor signal representing an input condition which influences operation of the combine;

an actuator control unit for generating the actuator control signals as a function of the actuator condition signals and as a function of setpoint signals;

a neural network trained prior to and apart from normal production use of the combine with a set of general training data to function as a general model of the combine and trained to function as a submodel with respect to a set of local condition data together with the set of general training data, the neural network processing the input condition sensor signals and data collected prior to normal production use of the combine representing desired crop harvesting performance quality to produce a set of combine adjustments intended to produce the desired crop harvesting performance quality, the neural network generating the setpoint signals based upon predicted responses of the combine to varying conditions;

a data communication system comprising means for communicating the feedback signals to the actuator control unit, means for communicating the sensor signals to the neural network, and means for communicating the setpoint signals to the actuator control unit, the neural network and the actuator control unit cooperating to control operation of the combine; and

operator controlled means for selectively causing the neural network to function as the general model or as the submodel.