Active vibration control method and apparatus

• US 6,137,886 A
• Filed: 07/18/1994
• Issued: 10/24/2000
• Est. Priority Date: 07/18/1994
• Status: Expired due to Fees

• Current Assignee: Cooper Tire & Rubber Company (Goodyear Tire & Rubber Company)
• Original Assignee: Cooper Tire & Rubber Company (Goodyear Tire & Rubber Company)
• Inventors: Shoureshi, Rahmat A.
• Primary Examiner(s): Grant, II, Jerome
• Application Number:
  US08/276,551
• Time in Patent Office: 2,290 Days
• Field of Search: 381/71, 381/73.1, 381/94, 381/96, 381/71.2, 381/71.8, 381/94.1-94.9
• US Class Current: 381/71.2
• CPC Class Codes: F16F 15/002 characterised by the control method or circuitry control of mechanical oscillations per se G05D19/00
• CPC Class Codes: G10K 2210/10 Applications
• CPC Class Codes: G10K 2210/128 Vehicles
• CPC Class Codes: G10K 2210/3038 Neural networks
• CPC Class Codes: G10K 2210/3042 Parallel processing
• CPC Class Codes: G10K 2210/3045 Multiple acoustic inputs, single acoustic output
• CPC Class Codes: G10K 2210/3046 Multiple acoustic inputs, multiple acoustic outputs
• CPC Class Codes: G10K 2210/511 Narrow band, e.g. implementations for single frequency cancellation
• CPC Class Codes: G10K 2210/512 Wide band, e.g. non-recurring signals

Abstract

An improved active vibration control system using feedback and pseudo-feedforward sensor inputs is provided for solving the problem of random and repetitive active vibration control and noise cancellation in a system. In a first embodiment of the invention, an artificial neural network is used for learning the dynamics of a structure and for providing output signals that follow the state variables of the structure. In one implementation of the neural network, a plurality of neurons obtain biasing inputs derived from sensor inputs, as well as inputs from the other neurons in the network. Further, each neuron obtains a feedback input from itself. Each input to a neuron is weighted using a weighting function derived on-line. The neural network supplies structure parameters and state variables to an optimal controller which derives and provides a control signal to the actuators so as to counteract vibrations and/or noise sensed in the system. In a second embodiment an optimal controller utilizing a modified generalized predictive control algorithm is used to to consider the limitations on the physical characteristics of the actuator(s), on-line, in terms of the output level and the rate of change of the output in the system. Additional embodiments wherein an optimized control signal is sent to the actuator(s) to minimize vibration incident to the structure are provided.

Claims

1. An active vibration control system for controlling vibrations at a structure resulting from at least one excitation force acting upon the structure, comprising:

at least one actuator located at the structure for imparting a reaction force to the structure;

at least one sensor located away from said actuator, said at least one sensor producing a sensor output;

a controller connected between said at least one sensor and said at least one actuator, said controller including;

a system identifier for receiving said sensor output from said at least one sensor and deriving a relationship between said sensor output and said reaction force imparted to the structure by said at least one actuator; and

an optimal controller connected to said system identifier to receive said relationship and for developing control driving signals from said relationship for driving said at least one actuator;

said system identifier including a Hopfield based neural net-work for learning the dynamics of the structure represented in a state space form and for providing output signals that follow state variables of the structure.

2. The active vibration control system of claim 1, said system including a plurality of sensors and wherein said neural network comprises a plurality of artificial neurons, each neuron receiving a weighted input from every other neuron and a feedback input from itself, as well as a weighted input biasing current related to the output of said at least one sensor, said neural network providing an output from each neuron to said optimal controller.

3. The active vibration control system of claim 2, wherein a plurality of sensors includes at least one performance sensor located at said structure for providing an output indicative of the vibration sensed at said structure.

4. The active vibration control system of claim 2, wherein said plurality of sensors additionally includes at least one pseudo-feedforward sensor located near the at least one excitation source, said psuedo-feedforward sensor being in a protected location.

5. The active vibration control system of claim 2, wherein each weighting of said weighted inputs to said neuron is different for at least some of said inputs.

6. The active vibration control system of claim 5, wherein said feedback input is also weighted.

7. The active vibration control system of claim 5, wherein said neural network is embodied in software.

8. The active vibration control apparatus of claim 5, including a plurality of actuators.

9. The active vibration control system of claim 5, wherein the output of each neuron j is

10. The active vibration control system of claim 9, wherein the output of said neural network are at least a state matrix A, an input matrix B, and a state vector X.

11. The active vibration control system of claim 10, wherein said optimal controller receives the state matrix A, the input matrix B and the state vector X from said neural network and uses Kalman filtering to derive said control driving signals for said at least one actuator.

12. The active vibration control system of claim 11, wherein said optimal controller minimizes the following cost function:

##EQU14## wherein P and Q are diagonal matrices that carry the terms ##EQU15## and ##EQU16## and, wherein said optimal controller additionally derives the optimal control input, V, for said at least one actuator by solving;

space="preserve" listing-type="equation">R+RA+A.sup.T R+RBQ.sup.-1 B.sup.T R-C.sup.T PC=0

space="preserve" listing-type="equation">V=-Q.sup.-1 B.sup.T RX.

13. The active vibration control system of claim 10, wherein said optimal controller receives the state matrix A, the input matrix B and the state vector X from said neural network and uses Modified Generalized Predictive Control to derive said control driving signals for said at least one actuator.

14. The active vibration control system of claim 13, wherein said optimal controller minimizes the following cost function:

##EQU17## where y(t) is the performance signal from the performance sensor, w(t) is a defined threshold, and V(t) is the input effort to said at least one actuator, Γ

_j is a parameter used to represent a desired trade-off between the control effort V(t) and the resulting performance Y(t).

15. A method for controlling vibrations at a structure resulting from at least one excitation force acting upon the structure, the structure being part of a vibration control system comprising at least one actuator located at the structure for imparting a reaction force to the structure, at least one sensor located away from the at least one actuator, the at least one sensor producing a sensor output, a controller connected between the at least one sensor and the at least one actuator, the controller including, a system identifier for receiving the output from the at least one sensor and deriving a relationship between the sensor output and the reaction force imparted to the structure by said at least one actuator and an optimal controller connected to the system identifier to receive the relationship and for developing control driving signals from the relationship for driving the at least one actuator, wherein the system identifier includes a Hopfield based neural network for learning the dynamics of the structure represented in a state space form and for providing output signals that follow state variables of the structure, comprising the steps of:

(a) using the neural network for learning system dynamics using input data from said at least one sensor and output characteristics, and for deriving A, B and C matrices; and

state vector X, wherein said A matrix is a state matrix, said B matrix is an input matrix and said C matrix is an output matrix, and X represents modal characteristic of the system;

(b) providing at least the A, B, and C matrices and state vector X to the optimal controller;

(c) using at least the A and B matrices and state vector X to calculate the optimal control signals; and

(c) sending the resulting optimal control signals to the at least one actuator.

16. An active vibration control system for controlling vibrations at a structure resulting from at least one excitation force acting upon the structure, comprising:

at least one actuator located at the structure for imparting a reaction force to the structure;

at least one performance sensor located away from said actuator, said at least one performance sensor producing a performance output;

a psuedo-feedforward sensor producing a psuedo-feedforward output;

a controller connected between said sensors and said at least one actuator, said controller including;

a system identifier for receiving said performance and psuedo-feedforward outputs and deriving a relationship between said outputs and said reaction force imparted to the structure by said at least one actuator; and

an optimal controller connected to said system identifier to receive said relationship and for developing control driving signals from said relationship for driving said at least one actuator;

said optimal controller using modified generalized predictive control incorporating actuator output limitations and output rate limitations in combination with the performance output and the pseudo-feedforward output to derive said control driving signals for said at least one actuator.

17. The active vibration control system of claim 16, wherein said optimal controller receives a state matrix A, an input matrix B and a state vector X from said system identifier.

18. The active vibration control system of claim 17, wherein said optimal controller minimizes the following cost function:

##EQU18## where y(t) is the performance signal from the performance sensor, w(t) is a defined threshold, and V(t) is the input effort to said at least one actuator, Γ

_j is a parameter used to represent a desired trade-off between the control effort V(t) and the resulting performance Y(t).

19. The active vibration control system of claim 18, wherein matrices A and B and state vector X are identified using a recursive method.

20. The active vibration control system of claim 19, wherein said recursive method is a least means square (LMS) determination.

21. The active vibration control system of claim 19, wherein said recursive method is a recursive least means square (RLMS) determination.

22. The active vibration control system of claim 19 wherein said system identifier includes a neural network for learning the dynamics of the structure and for providing output signals that follow state variables of the structure.