What is the control system engineer’s favorite dance?

The unit step

Non-Linear Internal Model Controller Design with Artificial Neural Networks By Vishal Kumar Advisor: Gary L. Dempsey 12/06/07 Bradley University Department of Computer and Electrical Engineering Senior Project Proposal for

Senior Project Proposal 1. Project Description 2. Discussion of previous work 3. Project Details 1. Functional Description and block diagrams 2. Functional Requirements and Specifications 3. Fall ‘07 Lab Work 4. Spring ’08 schedule

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

Project Description

Internal Model Control Internal Model Control

Project Description Artificial Neural Networks

Discussion of Previous Work Virtual Control Workstation for Adaptive Controller Workstation - Joseph Faivre, Kain Osterholt, and Adam Vaccari, 2006 Virtual Control Workstation for Adaptive Controller Workstation - Joseph Faivre, Kain Osterholt, and Adam Vaccari, 2006 Design of a Simulink based 2-DOF robot arm control workstation – Chris Edwards and Emberly Smith, 2007 Design of a Simulink based 2-DOF robot arm control workstation – Chris Edwards and Emberly Smith, 2007

Discussion of Previous Work Using a Neural Network Model for a robot arm to design conventional and neural controllers – Thuong D. Le, 2003 Using a Neural Network Model for a robot arm to design conventional and neural controllers – Thuong D. Le, 2003 Implementation of Conventional and Neural Networks using position and velocity feedback - Christopher Spevacek, and Manfred Meissner, 2000 Implementation of Conventional and Neural Networks using position and velocity feedback - Christopher Spevacek, and Manfred Meissner, 2000

Prespective What makes this project different? What makes this project different? New Tools Simulink/Real Time Execution Workshop Simulink/Real Time Execution Workshop Updated WinCon Client and WinCon Server interface Updated WinCon Client and WinCon Server interface Implementing an advanced controller – IMC with ANNs Exploring project worth

Functional Description Individual Components 1.46 GHz Windows Based PC 1.46 GHz Windows Based PC Data Acquisition and Capture Board Data Acquisition and Capture Board Power Module PAO103 Power Module PAO103 Quanser Plant SRV-02 with embedded position sensors, gripper and motor Quanser Plant SRV-02 with embedded position sensors, gripper and motor

Functional Description Acquisition Board Port Interface

Functional Description Power Module

High Level System Block Diagram

Functional Description Software Interface – Discuss on Previous Slide Software Interface – Discuss on Previous Slide Examples on next 2 slides Examples on next 2 slides

Example Simulink Diagram

Functional Requirements 1. Single Loop – Proportional, Proportional–Derivative Controller 2. Single Loop – Feed Forward 3. Feed Forwards with Artificial Neural Networks 4. Internal Model Control with Artificial Neural Networks

Performance Specifications Percent Overshoot 5% max Percent Overshoot 5% max Time to Peak 50ms max Time to Peak 50ms max Time to settle 200ms max Time to settle 200ms max Closed Loop Bandwidth 2Hz min Closed Loop Bandwidth 2Hz min Closed Loop Frequency Resp. 3dB max Closed Loop Frequency Resp. 3dB max Gain Margin 5.0 min Gain Margin 5.0 min Phase Margin 60 degrees min Phase Margin 60 degrees min Steady State Error 1 degree max Steady State Error 1 degree max Controller Execution Time 1ms max Controller Execution Time 1ms max

Fall ’07 Work Proportional Controller Design without arm Proportional Controller Design without arm Gc(s) = K =.3 Gc(s) = K =.3

Fall ’07 Work Proportional – Derivative Controller Design without arm Proportional – Derivative Controller Design without arm Gc(s) =.61(s )/(s+120) Gc(s) =.61(s )/(s+120)

Fall ’07 Work Comparison of Results Comparison of Results

Fall ’07 Work System Identification without arm System Identification without arm

Fall ’07 Work

Spring ’07 Schedule Week - Task Week - Task 0 - System Identification with Arm 0 - System Identification with Arm 1 -Single Loop Feed Forward Design 1 -Single Loop Feed Forward Design 2 -Internal Model Controller with approximate Linear Model 2 -Internal Model Controller with approximate Linear Model 3 -Train Adaline with Linear model 3 -Train Adaline with Linear model 4 -Implement Adaline in Internal Model Control 4 -Implement Adaline in Internal Model Control 5-6 -Train Adaline with real plant offline 5-6 -Train Adaline with real plant offline 7 -Implement Adaline in Internal Model Controller 7 -Implement Adaline in Internal Model Controller 8 -Performance testing, comparison with conventional methods 8 -Performance testing, comparison with conventional methods Left open for finalization, additional work, presentations and reports Left open for finalization, additional work, presentations and reports

Questions? Comments?