Stable adaptive neural network controller

Claim 2

A control system for controlling a non-linear plant having a plurality of outputs so that each of said plurality of plant outputs approaches a corresponding desired plant output, said plant outputs having a predetermined plurality of values constituting a normal operating range and a plurality of values outside of said normal operating range, said control system having means responsive to said plurality of plant outputs and to said plurality of desired plant outputs for generating a plurality of tracking error variables, said control system comprising: a first controller having a plurality of variable adaptation parameters and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating first control outputs for controlling said plant; an adaptive mechanism responsive to said plurality of tracking error variables for adjusting said variable adaptation parameters; a second controller having a plurality of fixed parameters and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating second control outputs for controlling said plant; and means responsive to said plurality of plant outputs for applying said first control outputs to said plant when said plurality of plant outputs are within said normal operating range and for applying said second control outputs to said plant when said plurality of plant outputs are outside of said normal operating range.

Claim 3

A control system according to claim 1 wherein said first controller comprises a neural network.

Claim 4

A control system according to claim 2 wherein said neural network comprises a plurality of nodes, each of said plurality of nodes comprising an interpolator circuit, said neural network further comprising means for generating a weighted set of tracking error variables and means for applying said weighted set of tracking error variables to each of said nodes to produce interpolated outputs.

Claim 5

A control system according to claim 3 wherein said means for generating a weighted set of tracking error variables comprises means for multiplying each of said plurality of tracking error variables by a predetermined, fixed weight value.

Claim 6

A control system according to claim 3 wherein said neural network further comprises means for generating a weighted set of interpolated outputs and means for summing said weighted set of interpolated outputs to generate said first control outputs.

Claim 7

A control system according to claim 5 wherein said means for generating a weighted set of interpolated outputs comprises means for multiplying each of said plurality of interpolated outputs by said variable adaptation parameters.

Claim 8

A control system according to claim 3 wherein each of said interpolator circuits multiplies one of said weighted set of tracking error variables by a predetermined interpolation function.

Claim 9

A control system according to claim 7 wherein said interpolation function is a radial Gaussian interpolation function.

Claim 10

A control system according to claim 1 wherein said second controller comprises a sliding controller.

Claim 11

A control system according to claim 1 wherein said applying means comprises a first modulator circuit responsive to said plurality of plant outputs for multiplying said first control outputs by a first modulation variable having a magnitude which decreases as said plant outputs move from said normal operating range to outside of said normal operating range to generate first modulated control outputs.

Claim 12

A control system according to claim 10 wherein said applying means comprises a second modulator circuit responsive to said plurality of plant outputs for multiplying said second control outputs by a second modulation variable having a magnitude which increases as said plant outputs move from said normal operating range to outside of said normal operating range to generate second modulated control outputs.

Claim 13

A control system according to claim 11 wherein said applying means further comprises means for summing said first modulated control outputs and said second modulated control outputs to generate a plant control signal which smoothly changes from said first control outputs to said second control outputs as said plant outputs move from said normal operating range to outside of said normal operating range.

Claim 14

A control system according to claim 1 further comprising a proportional/differential controller responsive to said plurality of tracking error variables for generating third control outputs and means for summing said first control outputs, said second control outputs and said third control outputs to generate a control sum and means for applying said control sum to said plant.

Claim 15

A control system for controlling a non-linear plant having a plurality of outputs so that each of said plurality of plant outputs approaches a corresponding desired plant output, said plant outputs having a predetermined plurality of values constituting a normal operating range, a plurality of values constituting an abnormal operating range outside of said normal operating range and a plurality of values in a transition operating range between said normal operating range and said abnormal operating range, said control system comprising: means responsive to said plurality of plant outputs and to said plurality of desired plant outputs for generating a plurality of tracking error variables; a neural network having a plurality of variable weighting factors and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating adaptive control outputs for controlling said plant; an adaptive mechanism responsive to said plurality of tracking error variables for adjusting said variable weighting factors; and a sliding controller having a plurality of fixed parameters and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating sliding control outputs for controlling said plant a modulator responsive to said plurality of plant outputs for applying said adaptive control outputs to said plant when said plurality of plant outputs are within said normal operating range, for applying said sliding control outputs to said plant when said plurality of plant outputs are in said abnormal operating range and for applying a linear combination of said adaptive control outputs and said sliding control outputs to said plant when said plurality of plant outputs are in said transition operating range.

Claim 16

A control system for controlling a non-linear plant having a plurality of outputs so that each of said plurality of plant outputs approaches a corresponding desired plant output, said plant outputs having a predetermined plurality of values constituting a normal operating range, a plurality of values constituting an abnormal operating range outside of said normal operating range and a plurality of values in a transition operating range between said normal operating range and said abnormal operating range, said control system comprising: means responsive to said plurality of plant outputs and to said plurality of desired plant outputs for generating a plurality of tracking error variables; a neural network having a plurality of variable weighting factors and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating adaptive control outputs for controlling said plant; an adaptive mechanism responsive to said plurality of tracking error variables for adjusting said variable weighting factors; and a sliding controller having a plurality of fixed parameters and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating sliding control outputs for controlling said plant a modulator responsive to said plurality of plant outputs for applying said adaptive control outputs to said plant when said plurality of plant outputs are within said normal operating range, for applying said sliding control outputs to said plant when said plurality of plant outputs are in said abnormal operating range and for applying a linear combination of said adaptive control outputs and said sliding control outputs to said plant when said plurality of plant outputs are in said transition operating range.

Claim 17

A control system according to claim 14 wherein said neural network comprises a plurality of nodes, each of said plurality of nodes comprising an interpolator circuit, said neural network further comprising means for generating a weighted set of tracking error variables and means for applying said weighted set of tracking error variables to each of said nodes to produce interpolated outputs.

Claim 18

A control system according to claim 15 wherein said means for generating a weighted set of tracking error variables comprising means for multiplying each of said plurality of tracking error variables by a predetermined, fixed weight value.

Claim 19

A control system according to claim 16 wherein said neural network further comprises means for generating a weighted set of interpolated outputs and means for summing said weighted set of interpolated outputs to generate said adaptive control outputs.

Claim 20

A control system according to claim 17 wherein said means for generating a weighted set of interpolated outputs comprises means for multiplying each of said plurality of interpolated outputs by said variable weighting factors.

Claim 21

A control system according to claim 18 wherein each of said interpolator circuits multiplies one of said weighted set of tracking error variable by a predetermined interpolation function.

Claim 22

A control system according to claim 19 wherein said interpolation function is a radial Gaussian interpolation function.

Claim 23

A control system according to claim 20 wherein said radial Gaussian interpolation function has a standard deviation and a mean which depend on the frequency chracteristics of said plant outputs.

Claim 24

A control system according to claim 21 wherein said neural network has a predetermined number of nodes which predetermined number depends on the frequency characteristics of said plant outputs.

Claim 25

A control system according to claim 14 wherein said modulator comprises a first modulator circuit responsive to said plurality of plant outputs for multiplying said adaptive control output by a first modulation variable to generate modulated adaptive control outputs, said first modulation variable having a magnitude which is "1" when said plant outputs are inside said normal operating range, which is between "0" and "1" when said plant outputs are in said transition operating range and which is "0" when said plant outputs are in said abnormal operating range.

Claim 26

A control system according to claim 23 wherein said modulator comprises a second modulator circuit responsive to said plurality of plant outputs for multiplying said sliding control outputs by a second modulation variable to generate modulated sliding control outputs, said second modulation variable having a magnitude which is "0" when said plant outputs are inside said normal operating range, which is between "0" and "1" when said plant outputs are in said transition operating range and which is "1" when said plant outputs are in said abnormal operating range.

Claim 27

A control system according to claim 24 wherein said first modulation variable and said second modulation variable sum to "1" for all plant outputs.

Claim 28

A control system according to claim 25 wherein said modulator further comprises means for summing said modulated adaptive control outputs and said modulated sliding control outputs to generate a plant control signal which smoothly changes from said adaptive control outputs to said sliding control outputs as said plant outputs move from said normal operating range through said transition operating range to said abnormal operating range.

Claim 29

A control system according to claim 14 further comprising a proportional/differential controller responsive to said plurality of tracking error variables for generating proportional/differential control outputs and means for summing said adaptive control outputs, said sliding control outputs and said proportional/differential control outputs to generate a control sum and means for applying said control sum to said plant.

Claim 30

In a control system for controlling a non-linear plant having a plurality of outputs so that each of said plurality of plant outputs approaches a corresponding desired plant output, said plant outputs having a predetermined plurality of values constituting a normal operating range and a plurality of values outside of said normal operating range, said control system having means responsive to said plurality of plant outputs and to said plurality of desired plant outputs for generating a plurality of tracking error variables, a first controller having a plurality of variable adaption parameters and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating control outputs for controlling said plant, and an adaptive mechanism responsive to said plurality of tracking error variables for adjusting said variable adaption parameters, the improvement wherein said first controller comprises: a neural network having a plurality of nodes, each of said plurality of nodes comprising a radial Gaussian interpolator circuit; means for generating a weighted set of tracking error variables; means for applying said weighted set of tracking error variables to each of said nodes to produce interpolated outputs; means responsive to said variable adaption parameters for generating a weighted set of interpolated outputs; and means for summing said weighted set of interpolated outputs to generate said control outputs.

Claim 31

In a control system for controlling a non-linear plant having a plurality of outputs so that each of said plurality of plant outputs approaches a corresponding desired plant output, said plant outputs having a predetermined plurality of values constituting a normal operating range and a plurality of values outside of said normal operating range, said control system having means responsive to said plurality of plant outputs and to said plurality of desired plant outputs for generating a plurality of tracking error variables, a first controller having a plurality of variable adaption parameters and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating control outputs for controlling said plant, and an adaptive mechanism responsive to said plurality of tracking error variables for adjusting said variable adaption parameters, the improvement wherein said first controller comprises: a neural network having a plurality of nodes, each of said plurality of nodes comprising a radial Gaussian interpolator circuit; means for generating a weighted set of tracking error variables; means for applying said weighted set of tracking error variables to each of said nodes to produce interpolated outputs; means responsive to said variable adaption parameters for generating a weighted set of interpolated outputs; and means for summing said weighted set of interpolated outputs to generate said control outputs.

Claim 32

In a control system, the improvement according to claim 28 wherein said means for generating a weighted set of interpolated outputs comprises means for multiplying each of said plurality of interpolated outputs by said variable adaption parameters.

Claim 33

In a control sytsem, the improvement according to claim 28 wherein said means for generating a weighted set of tracking error variables comprises means for multiplying each of said plurality of tracking error variables by a predetermined, fixed weight value.

Claim 34

In a control sytem, the improvement according to claim 30 wherein said means for generating a weighted set of interpolated outputs comprises means for multiplying each of said plurality of interpolated outputs by said variable adaptation parameters.

Claim 35

In a control system, the improvement according to claim 28 further comprising a second controller having a plurality of fixed parameters and being responsive to said plurality of tracking error variables, to said plurality of plant outputs and to said plurality of desired plant outputs for generating second control outputs for controlling said plant; and means responsive to said plurality of plant outputs for applying said control outputs to said plant when said plurality of plant outputs are within said normal operating range and for applying said second control outputs to said plant when said plurality of plant outputs are outside of said normal operating range.

Claim 36

In a control system, the improvement according to claim 32 wherein said second controller comprises a sliding controller.

Claim 37

In a control system, the improvement according to claim 32 wherein said applying means comprises a first modulator circuit responsive to said plurality of plant outputs for multiplying said control outputs by a first modulation variable having a magnitude which decreases as said plant outputs move from said normal operating range to outside of said normal operating range to generate first modulated control outputs.

Claim 38

In a control system, the improvement according to claim 34 wherein said applying means comprises a second modulator circuit responsive to said plurality of plant outputs for multiplying said second control outputs by a second modulation variable having a magnitude which increases as said plant outputs move from said normal operating range to outside of said normal operating range to generate second modulated control outputs.

Claim 39

In a control system, the improvement according to claim 35 wherein said applying means further comprises means for summing said first modulated control outputs and said second modulated control outputs to generate a plant control signal which smoothly changes from said first control outputs to said second control outputs as said plant outputs move from said normal operating range to outside of said normal operating range.

Claim 40

In a control system, the improvement according to claim 32 further comprising a proportional/differential controller responsive to said plurality of tracking error variables for generating third control outputs and means for summing said control outputs, said second control outputs and said third control outputs to generate a control sum and means for applying said control sum to said plant.