Teleoperation and System Health Monitoring Mo-Yuen Chow, Ph.D.

Size: px

Start display at page:

Download "Teleoperation and System Health Monitoring Mo-Yuen Chow, Ph.D."

Lynette Crawford
5 years ago
Views:

1 Teleoperation and System Health Monitoring Mo-Yuen Chow, Ph.D. Advanced Diagnosis and Control (ADAC) Lab Department of Electrical and Computer Engineering North Carolina State University Raleigh, NC USA

2 Teleoperation Potential applications in NASA Remote robotics manipulator control and teleoperation. E.g.: SPDM (Special Purpose Dexterous Manipulator). Teleoperation and System Health Monitoring, Mo-Yuen Chow 2

Main challenge in teleoperation Challenge Network delay during data transfer among sensors, controllers, and actuators Effect of of network delays System performance degradation System

3 Main challenge in teleoperation Challenge Network delay during data transfer among sensors, controllers, and actuators Effect of of network delays System performance degradation System instability What are the causes of of network delays Bandwidth constraint Multiple packet transmissions Packet loss Packet collision Etc. Teleoperation and System Health Monitoring, Mo-Yuen Chow 3

4 Delays in teleoperation r ( k ) Controller τ c k u( k ) τ ca k u ( kt τ ca k ) Plant y ( t) y ( kt τ sc k ) τ sc k y ( k ) Z.O.H T Computational delay Controller-to-actuator delay Sensor-to-controller delay Time skew : k r u y y ( t) ( k) r( kt) ( k) u( kt) ( k) y( kt) Network : Continuous system output : Discrete reference signal : Discrete control signal : Discrete system output Teleoperation and System Health Monitoring, Mo-Yuen Chow 4

5 Gain Scheduler Middleware (GSM) Middleware gain scheduler Feedback preprocessor Network traffic estimator Probing Controller Gain scheduler Control signal Network Remote system Feedback signal Gain Scheduler Middleware (GSM) is a novel methodology to utilize middleware to enable an existing non-network-based controller so they can be used for networked control. The proposed methodology applies middleware to modify the controller output with respect to the current network traffic condition in addition to providing appropriate network conditions. Teleoperation and System Health Monitoring, Mo-Yuen Chow 5

6 Optimal gain [K(i) surface] with respect to different curvature and time delay UV position Y J v Y B A v X path for the UV to follow v UV position X τ Optimal K(i) A() i 8 1 τ ( i) Jˆ ( i + 1) ε τ A ( i ) ( i ) 5 10 K ( i ) = 4 K ( i ) = 1.4 K ( i ) = 0.1 Teleoperation and System Health Monitoring, Mo-Yuen Chow 6

Illustration: Tele-operation of an Unmanned Vehicle (UV) The actual networked mobile robot is setup with the following configuration: The speed of both wheels are controlled by two PI controllers

7 Illustration: Tele-operation of an Unmanned Vehicle (UV) The actual networked mobile robot is setup with the following configuration: The speed of both wheels are controlled by two PI controllers using a SK-515C microcontroller board. The path tracking controller and GSM are implemented on a notebook computer as RTLinux processes. Data transfers between the tracking controller (or GSM) are delayed by an RTLinux process, using the internal hardware timer and delays from ADAC to KU. This is a scenario to focus specifically on delay effects and to avoid packet loss effects. Teleoperation and System Health Monitoring, Mo-Yuen Chow 7

Experimental results in virtual reality world 2.5 2 Robot tracks No delay With delay, no GSM With delay and GSM 1.5 y(m) 1 0.5 0-0.5-0.5 0 0.5 1 1.5 2 2.5 x(m) 0.

8 Experimental results in virtual reality world Robot tracks No delay With delay, no GSM With delay and GSM 1.5 y(m) x(m) 0.12 No delay With delay, no GSM 0.1 With delay and GSM 0.08 D(m) Time(s) Teleoperation and System Health Monitoring, Mo-Yuen Chow 8

Web-based remote access real-time Mechatronics Teleoperation Lab COMMAND + - PID

network LAB server wireless unmanned vehicle netmeeting LAB server WEB cam

The lab server can communicate with the UV wirelessly PID controller gains

9 Web-based remote access real-time Mechatronics Teleoperation Lab COMMAND + - PID CONTROLLER PROCESS or PLANT RESPONSE PID controller WEB browser Communication network LAB server wireless unmanned vehicle netmeeting LAB server WEB cam Unmanned vehicle Use web browser to access the lab remotely. The lab server can communicate with the UV wirelessly PID controller gains adjustment through wireless communication. WEB browser Teleoperation and System Health Monitoring, Mo-Yuen Chow 9

10 System health monitoring + Fault Management Module Domain Expertise (e.g. Shuttle Experts) Invasive vs. non-invasive methods Model-Based Paradigms (Less suitable for system with high nonlinearity) Parameter estimation Kalman filter Knowledge Engineering (e.g.: Neural Network and Fuzzy Logic Technologies) Model-Free Paradigms (More suitable for many real world power engineering application) DSP Expert system Artificial Intelligence Neural Networks Fuzzy Logic Hybrid Neural/Fuzzy Fault Detection System Teleoperation and System Health Monitoring, Mo-Yuen Chow 10

11 Challenges Millions and millions of data Data stream constantly flowing into the database from sensors Working environments are dynamics have many unexpected events Need to make proper decision on Fault Detection, Diagnosis, Prognosis, {D,D,P} and Mitigation Control Actions in real time Need tools to: Provide meaningful information {D,D,P} by mining the million and million of data Able to react properly to unseen data of the dynamics operating environments Teleoperation and System Health Monitoring, Mo-Yuen Chow 11

12 Hybrid neural/fuzzy fault detection system Pure NN Approach (Disadvantages) "Black-box" characteristics Does not give heuristic reasoning Pure FZ Approach (Disadvantages) Difficult to Give an Exact Solution to the Problems Hybrid NN/FZ System (Synergy) Give Exact Solution by Training Data Module 2 Module 1 Ι GOOD FAIR BAD Ι ω Provide Valuable Information of the Fault Detection in a Heuristic Manner Current Sensor ω MOTOR Tachometer Teleoperation and System Health Monitoring, Mo-Yuen Chow 12

13 Immune system inspired {D,D,P} technologies Natural immune system has distinguished features such as: Memory and learning; Self/non-self discrimination; Responding to unknown patterns. These features form the bases of a new emerging research field, Artificial Immune Systems (AIS), and have been used in application areas such as: Anomaly/Fault Detection and Diagnosis; Control; Robotics. What is our objective? Data mining Incorporate with NN/FZ system for {D,D,P} Teleoperation and System Health Monitoring, Mo-Yuen Chow 13

What is our objective? Artificial immune systems for motor fault detection, diagnosis, and prognosis Many motor incipient fault detection schemes are insufficient to discover and identify new faults.

14 What is our objective? Artificial immune systems for motor fault detection, diagnosis, and prognosis Many motor incipient fault detection schemes are insufficient to discover and identify new faults. Our objective is to investigate the feasibility of developing an Intelligent Fault Detection, Diagnosis and Prognosis System using Artificial Immune Systems on top of a Neural Network Fuzzy Logic (NN-FZ) structure to actively detect, diagnose motor incipient faults and make an estimation of the remaining lifetime. Motor fault database Motor system measurement, waveform analyses and feature extraction AIS algorithm Neural-Fuzzy Incipient fault detection and diagnosis Estimated motor condition Actual motor condition (verified at a later time) Environmental information Teleoperation and System Health Monitoring, Mo-Yuen Chow 14

15 Biologically inspired intelligent fault diagnosis for power distribution systems Distribution fault database (knowledge base) data mining and feature extraction actual fault cause (found and verified at a later time) wooded area Web geographic information house mall update database with new data feature extraction AIS algorithm continuously guide the NN-FZ system to learn and absorb new information to improve its performance (including district, spatial information, etc.) Web weather channel geographic weather conditions Neural / Fuzzy fault root cause identification estimated distribution fault causes feature extraction on-line distribution system fault alarms including protection device status (e.g. fuse) and/or fault recorder waveforms waveform analyses and feature extraction Teleoperation and System Health Monitoring, Mo-Yuen Chow 15

ADAC lab and research projects Mechatronics and Control Distributed Network Based Control and Applications Distributed Network-Based Mobile Robots (Unmanned Vehicles) control with Network QoS

Development Motors Intelligent Fault Detection Diagnosis and Prognosis of Induction Motors DC Motor Acoustic Noise Analysis using Signal Processing Techniques Fast Prototype Motor System Simulation

16 ADAC lab and research projects Mechatronics and Control Distributed Network Based Control and Applications Distributed Network-Based Mobile Robots (Unmanned Vehicles) control with Network QoS (Quality of Service) Constraints Adaptive Fuzzy Modulation for Network-Based PI Control (Completed) Distance Learning Delta Project: Web-Based Remote Access Real-Time Mechatronics Laboratory Development Motors Intelligent Fault Detection Diagnosis and Prognosis of Induction Motors DC Motor Acoustic Noise Analysis using Signal Processing Techniques Fast Prototype Motor System Simulation (MotorSIM II) A Neural/Fuzzy Approach for Motor Incipient Fault Detection (Completed) Incipient Fault Detection of Rotating Machines Using Neural Networks (Completed) Teleoperation and System Health Monitoring, Mo-Yuen Chow 16

Distribution System Load Management (Completed) Power Quality Assessment of Power Distribution Systems (Completed) Intelligent Energy Control (Completed) Networking Proactive

17 ADAC Lab and research projects cont. Power Distribution Systems Biologically Inspired Intelligent Fault Diagnosis for Power Distribution Systems Power Distribution Fault Location and Diagnosis (Completed) Distribution System Load Management (Completed) Power Quality Assessment of Power Distribution Systems (Completed) Intelligent Energy Control (Completed) Networking Proactive Intelligent Network Fault Management (Completed) A Novel Set Theoretic Based Neural/Fuzzy Network Traffic Feature Extraction and Modeling Methodology (Completed) Communication System Network Control Software Performance Modeling and Fault Detection (Completed) Teleoperation and System Health Monitoring, Mo-Yuen Chow 17

Performance Characterization of IP Network-based Control Methodologies for DC Motor Applications Part II

Performance Characterization of IP Network-based Control Methodologies for DC Motor Applications Part II Tyler Richards, Mo-Yuen Chow Advanced Diagnosis Automation and Control Lab Department of Electrical