What is the control system engineer’s favorite dance?

What is the control system What is the control system engineer’s favorite dance?engineer’s favorite dance?

The unit stepThe unit step

Non-Linear Internal Model Controller Non-Linear Internal Model Controller Design with Artificial Neural NetworksDesign with Artificial Neural Networks

By Vishal KumarBy Vishal KumarAdvisor: Gary L. DempseyAdvisor: Gary L. Dempsey

12/06/0712/06/07Bradley UniversityBradley University

Department of Computer and Electrical EngineeringDepartment of Computer and Electrical Engineering

Senior Project Proposal for

Senior Project ProposalSenior Project Proposal

1.1. Project DescriptionProject Description

2.2. Discussion of previous work Discussion of previous work

3.3. Project DetailsProject Details1.1. Functional Description and block diagramsFunctional Description and block diagrams

2.2. Functional Requirements and SpecificationsFunctional Requirements and Specifications

3.3. Fall ‘07 Lab WorkFall ‘07 Lab Work

4.4. Spring ’08 scheduleSpring ’08 schedule

Project DescriptionProject Description

This project is centered around controlling the This project is centered around controlling the Quanser Consulting Plant SRV-02Quanser Consulting Plant SRV-02 with a Non with a Non Linear Linear Internal Model ControllerInternal Model Controller implemented implemented with with Artificial Neural NetworksArtificial Neural Networks. Artificial Neural . Artificial Neural Networks with an adaptive transfer characteristic Networks with an adaptive transfer characteristic coupled with accurate disturbance detection of coupled with accurate disturbance detection of Internal Model Controller can help us design a Internal Model Controller can help us design a controller to manage the 4th order Quanser controller to manage the 4th order Quanser Plant despite its' non-linearity from friction and Plant despite its' non-linearity from friction and external disturbances due to the rotary flexible external disturbances due to the rotary flexible joint.joint.

Project DescriptionProject Description

Project DescriptionProject Description

Internal Model ControlInternal Model Control

Project DescriptionProject Description

Artificial Neural NetworksArtificial Neural Networks

Discussion of Previous WorkDiscussion of Previous Work

Virtual Control Workstation for Adaptive Virtual Control Workstation for Adaptive Controller Workstation - Joseph Faivre, Controller Workstation - Joseph Faivre, Kain Osterholt, and Adam Vaccari, 2006Kain Osterholt, and Adam Vaccari, 2006

Design of a Simulink based 2-DOF robot Design of a Simulink based 2-DOF robot arm control workstation – Chris Edwards arm control workstation – Chris Edwards and Emberly Smith, 2007and Emberly Smith, 2007

Discussion of Previous WorkDiscussion of Previous Work

Using a Neural Network Model for a robot Using a Neural Network Model for a robot arm to design conventional and neural arm to design conventional and neural controllers – Thuong D. Le, 2003controllers – Thuong D. Le, 2003

Implementation of Conventional and Implementation of Conventional and Neural Networks using position and Neural Networks using position and velocity feedback - Christopher Spevacek, velocity feedback - Christopher Spevacek, and Manfred Meissner, 2000and Manfred Meissner, 2000

PrespectivePrespective

What makes this project different?What makes this project different?

New ToolsNew ToolsSimulink/Real Time Execution WorkshopSimulink/Real Time Execution WorkshopUpdated WinCon Client and WinCon Server Updated WinCon Client and WinCon Server

interfaceinterface

Implementing an advanced controller – IMC Implementing an advanced controller – IMC with ANNs with ANNs

Exploring project worthExploring project worth

Functional DescriptionFunctional Description

Individual ComponentsIndividual Components1.46 GHz Windows Based PC1.46 GHz Windows Based PCData Acquisition and Capture BoardData Acquisition and Capture BoardPower Module PAO103Power Module PAO103Quanser Plant SRV-02 with embedded Quanser Plant SRV-02 with embedded

position sensors, gripper and motorposition sensors, gripper and motor

Functional DescriptionFunctional Description

Acquisition Board Port InterfaceAcquisition Board Port Interface

Functional DescriptionFunctional Description

Power ModulePower Module

High Level System Block DiagramHigh Level System Block Diagram

Functional DescriptionFunctional Description

Software Interface – Discuss on Previous Software Interface – Discuss on Previous SlideSlide

Examples on next 2 slidesExamples on next 2 slides

Example Simulink DiagramExample Simulink Diagram

Example Simulink DiagramExample Simulink Diagram

Functional RequirementsFunctional Requirements

1.1. Single Loop – Proportional , Single Loop – Proportional , Proportional–Derivative ControllerProportional–Derivative Controller

2.2. Single Loop – Feed ForwardSingle Loop – Feed Forward

3.3. Feed Forwards with Artificial Neural Feed Forwards with Artificial Neural NetworksNetworks

4.4. Internal Model Control with Artificial Internal Model Control with Artificial Neural NetworksNeural Networks

Performance SpecificationsPerformance Specifications

Percent Overshoot Percent Overshoot 5% max5% max Time to PeakTime to Peak 50ms max50ms max Time to settle Time to settle 200ms max200ms max Closed Loop Bandwidth Closed Loop Bandwidth 2Hz min2Hz min Closed Loop Frequency Resp.Closed Loop Frequency Resp. 3dB max 3dB max Gain Margin Gain Margin 5.0 min5.0 min Phase Margin Phase Margin 60 degrees 60 degrees

minmin Steady State Error Steady State Error 1 degree max1 degree max Controller Execution Time Controller Execution Time 1ms max1ms max

Fall ’07 WorkFall ’07 Work

Proportional Controller Design without armProportional Controller Design without armGc(s) = K = .3Gc(s) = K = .3

1.5 1.55 1.6 1.65 1.7 1.75-5

0

5

10

15

20

25

30

35

Time

Deg

rees

Proportional ControlGc(s) = .3

Fall ’07 WorkFall ’07 WorkProportional – Derivative Controller Design Proportional – Derivative Controller Design

without armwithout armGc(s) = .61(s + 61.5)/(s+120)Gc(s) = .61(s + 61.5)/(s+120)

1.5 1.55 1.6 1.65 1.7 1.75 1.8 1.85 1.9 1.95 2-5

0

5

10

15

20

25

30

Time

Degre

es

Proportional Derivative ControllerGc(s) = .35 (s + 61.5)/(s + 120)

Fall ’07 WorkFall ’07 Work

Comparison of ResultsComparison of Results

Fall ’07 WorkFall ’07 Work

System Identification without armSystem Identification without arm

)150/(

69)(

)35(

ss

esGp

mss

Fall ’07 WorkFall ’07 Work

Spring ’07 ScheduleSpring ’07 Schedule

Week - TaskWeek - Task 00 - - System Identification with ArmSystem Identification with Arm 1 1 - - Single Loop Feed Forward Design Single Loop Feed Forward Design 2 2 - - Internal Model Controller with approximate Internal Model Controller with approximate

Linear ModelLinear Model 3 3 - - Train Adaline with Linear modelTrain Adaline with Linear model 4 4 - - Implement Adaline in Internal Model Control Implement Adaline in Internal Model Control 5-6 5-6 - - Train Adaline with real plant offline Train Adaline with real plant offline 7 7 - - Implement Adaline in Internal Model Implement Adaline in Internal Model

Controller Controller 8 8 - - Performance testing, comparison with Performance testing, comparison with

conventional methodsconventional methods 9-14 -9-14 - Left open for finalization, additional work, Left open for finalization, additional work,

presentations and reports presentations and reports

Questions? Comments?Questions? Comments?