1. PrefaceAccording to the original definition of mechatronics proposed by the Yasakawa Electric Company andthe definitions that have appeared since, many of the engineering products designed and manufacturedin the last 30 years integrating mechanical and electrical systems can be classified as mechatronic systems.Yet many of the engineers and researchers responsible for those products were never formally trained inmechatronics per se. The Mechatronics Handbook, 2nd Edition can serve as a reference resource for thosevery same design engineers to help connect their everyday experience in design with the vibrant field ofmechatronics.The Handbook of Mechatronics was originally a single-volume reference book offering a thoroughcoverage of the field of mechatronics. With the need to present new material covering the rapid changesin technology, especially in the area of computers and software, the single-volume reference book quicklybecame unwieldy. There is too much material to cover in a single book. The topical coverage in theMechatronics Handbook, 2nd Edition is presented here in two books covering Mechatronic Systems, Sensors,and Actuators: Fundamentals and Modeling and Mechatronic System Control, Logic, and Data Acquisition.These two books are intended for use in research and development departments in academia, government,and industry, and as a reference source in university libraries. They can also be used as a resource forscholars interested in understanding and explaining the engineering design process.As the historical divisions between the various branches of engineering and computer science becomeless clearly defined, we may well find that the mechatronics specialty provides a roadmap for nontradi-tionalengineering students studying within the traditional structure of most engineering colleges. It isevident that there is an expansion of mechatronics laboratories and classes in the university environmentworldwide. This fact is reflected in the list of contributors to these books, including an internationalgroup of academicians and engineers representing 13 countries. It is hoped that the books comprisingthe Mechatronics Handbook, 2nd Edition can serve the world community as the definitive reference sourcein mechatronics.

2. OrganizationThe Mechatronics Handbook, 2nd Edition is a collection of 56 chapters covering the key elements ofmechatronics:a. Physical Systems Modelingb. Sensors and Actuatorsc. Signals and Systemsd. Computers and Logic Systemse. Software and Data AcquisitionPhysical system modelingSensors and actuatorsMECHATRONICSSignals and systemsComputers andlogic systemsSoftware anddata acquisitionKey Elements of MechatronicsMechatronic System Control, Logic, and Data AcquisitionAn overview of signals and system control, computers, logic systems, software, and data acquisition ispresented in this book. These are most rapidly changing areas of mechatronics. 3. Section IMechatronic System ControlSince there is a significant body of readily available material to the reader on the general subject of signalsand systems, there is no overriding need to repeat that material here. Instead, the goal of this book is topresent the relevant aspects of signals and systems of special importance to the study of mechatronics.The book begins with chapters on the role of control in mechatronics and on the role of modeling inmechatronic design. These chapters set the stage for the more fundamental discussions on signals andsystems comprising the bulk of the material in this section. Modern aspects of control design usingoptimization techniques from H2 theory, adaptive and nonlinear control, neural networks, and fuzzysystems are also included as they play an important role in modern engineering system design. The bookincludes chapters on design optimization for mechatronic systems, and real-time monitoring and control.The chapters, listed in order of appearance, are1. The Role of Controls in Mechatronics2. The Role of Modeling in Mechatronics Design3. Signals and Systems3.1 Continuous- and Discrete-Time Signals3.2 z Transforms and Digital Systems3.3 Continuous- and Discrete-Time State-Space Models3.4 Transfer Functions and Laplace Transforms4. State Space Analysis and System Properties5. Response of Dynamic Systems6. The Root Locus Method7. Frequency Response Methods8. Kalman Filters as Dynamic System State Observers9. Digital Signal Processing for Mechatronic Applications10. Control System Design via H2 Optimization11. Adaptive and Nonlinear Control Design12. Neural Networks and Fuzzy Systems13. Advanced Control of an Electrohydraulic Axis14. Design Optimization of Mechatronic Systems15. Motion Control16. Real-Time Monitoring and Control17. Micromechatronics and Microelectromechanical Motion DevicesSection IIComputers and Logic SystemsThe development of the computer, and then the microcomputer, embedded computers, and associatedinformation technologies and software advances, has impacted the world in a profound manner. This isespecially true in mechatronics where the integration of computers with electromechanical systems hasled to a new generation of smart products. The future is filled with promise of better and more intelligentproducts resulting from continued improvements in computer technology and software engineering. Inthis section, the focus is on computer hardware and associated issues of logic, communication, network-ing,architecture, fault analysis, embedded computers, and programmable logic controllers. The chapters,listed in order of appearance, are18. Introduction to Computers and Logic Systems19. Digital Logic Concepts and Combinational Logic Design20. System Interfaces21. Communications and Computer Networks22. Fault Analysis in Mechatronic Systems23. Logic System Design 4. 24. Architecture25. Control with Embedded Computers and Programmable Logic Controllers26. Graphical System Design for Embedded Systems27. Field-Programmable Gate Arrays28. Graphical Programming for Field-Programmable Gate Arrays: Applications in Control andMechatronicsSection IIISoftware and Data AcquisitionGiven that computers play a central role in modern mechatronics products, it is very important tounderstand how data is acquired and how it makes its way into the computer for processing and logging.The final section of this book is devoted to the issues surrounding computer software and data acquisition.The chapters, listed in order of appearance, are29. Introduction to Data Acquisition30. Measurement Techniques: Sensors and Transducers31. A/D and D/A Conversion32. Signal Conditioning33. Virtual Instrumentation Systems34. Software Design and Development35. Data Recording and Logging 5. AcknowledgmentsI wish to express my heartfelt thanks to all the contributing authors. Taking time in otherwise busy andhectic schedules to author the excellent chapters appearing in this book is much appreciated.This handbook is a result of a collaborative effort expertly managed by CRC Press. My thanks to theeditorial and production staff:Nora Konopka Acquisitions EditorTheresa Delforn Project CoordinatorJoette Lynch Project EditorThanks to my friend and collaborator Professor Richard C. Dorf for his continued support andguidance. And finally, a special thanks to Lynda Bishop for managing the incoming and outgoing draftmanuscripts. Her organizational skills were invaluable to this project. 6. EditorRobert H. Bishop is a professor of aerospace engineeringand engineering mechanics at The University of Texas atAustin and holds the Joe J. King Professorship. He receivedhis BS and MS from Texas A&M University in aerospaceengineering, and his PhD from Rice University in electricaland computer engineering. Prior to coming to The Uni-versityof Texas at Austin, he was a member of the technicalstaff at the MIT Charles Stark Draper Laboratory.Dr. Bishop is a specialist in the area of planetary explora-tionwith emphasis on spacecraft guidance, navigation andcontrol. He is a fellow of the American Institute of Aero-nauticsand Astronautics. Currently, Dr. Bishop is cur-rentlyworking with the NASA Johnson Space Center ontechniques for achieving precision landing on the moonand Mars. He is an active researcher authoring and co-authoring over 100 journal and conference papers.He was twice selected a faculty fellow at the NASA Jet Propulsion Laboratory and as a Welliver facultyfellow by The Boeing Company. Dr. Bishop co-authors Modern Control Systems with Professor R. C. Dorf,and he has authored two other books entitled Learning with LabView and Modern Control System Designand Analysis Using Matlab and Simulink. He received the John Leland Atwood Award by the AmericanSociety of Engineering Educators and the American Institute of Aeronautics and Astronautics that isgiven periodically to a leader who has made lasting and significant contributions to aerospace engineeringeducation. Dr. Bishop is a member of the Academy of Distinguished Teachers at The University of Texasat Austin. 7. List of ContributorsMaruthi R. AkellaDepartment of AerospaceEngineering and EngineeringMechanicsThe University of Texas at AustinAustin, TexasCraig AndersonNational Instruments, Inc.Austin, TexasDragos ArotariteiDepartment of ComputerScience and EngineeringAalborg UniversityEsbjerg, DenmarkBrian BettsData Acquisition and SignalConditioningNational Instruments, Inc.Austin, TexasTomas BrezinaTechnical University of BrnoBrno, Czech RepublicGeorge I. CohnCalifornia State UniversityLos Angeles, CaliforniaDaniel A. ConnorsDepartment of Electrical andComputer EngineeringUniversity of Colorado at BoulderBoulder, ColoradoKevin C. CraigDepartment of Mechanical,Aerospace, and NuclearEngineeringRensselaer Polytechnic InstituteTroy, New YorkTimothy P. Crain IIDepartment of AerospaceEngineering and EngineeringMechanicsNASA Johnson Space CenterHouston, TexasRaymond A. de CallafonDepartment of Mechanical andAerospace EngineeringUniversity of California,San DiegoLa Jolla, CaliforniaDarcy DementNational Instruments, Inc.Austin, TexasSantosh DevasiaDepartment of MechanicalEngineeringUniversity of WashingtonSeattle, WashingtonC. Nelson DornyMoore School of ElectricalEngineeeringUniversity of PennsylvaniaPhiladelphia, PennsylvaniaStephen A. DyerElectrical and ComputerEngineeringKansas State UniversityManhattan, KansasJeannie Sullivan FalconNational Instruments, Inc.Austin, TexasDaniel R. FayUniversity of Colorado atBoulderBoulder, ColoradoGerardo GarciaNational Instruments, Inc.Austin, Texas 8. Shelley GretleinNational Instruments, Inc.Austin, TexasMargaret H. HamiltonHamilton Technologies, Inc.Tucson, ArizonaCecil HarrisonUniversity of SouthernMississippiHattiesburg, MississippiBonnie S. HeckSchool of Electrical andComputer EngineeringGeorgia Institute of TechnologyAtlanta, GeorgiaWen-Mei W. HwuUniversity of IllinoisUrbana, IllinoisMohammad IlyasFlorida Atlantic UniversityBoca Raton, FloridaFlorin IonescuUniversity of Applied SciencesKonstanz, GermanyHugh JackProduct Design andManufacturing EngineeringGrand Valley State UniversityGrand Rapids, MichiganJeffrey A. JalkioUniversity of St. ThomasSt. Paul, MinnesotaRolf JohanssonDepartment of Automatic ControlLund Institute of TechnologyLund, SwedenJayantha KatupitiyaThe University ofNew South WalesSydney, New South Wales,AustraliaCtirad KratochvilTechnical University of BrnoBrno, Czech RepublicRahul KulkarniIndustrial Data Acquisition andControlNational Instruments, Inc.Austin, TexasThomas R. KurfessDepartment of MechanicalEngineeringClemson UniversityClemson, South CarolinaKam K. LeangDepartment of MechanicalEngineeringUniversity of WashingtonSeattle, WashingtonSergey Edward LyshevskiDepartment of ElectricalEngineeringRochester Institute ofTechnologyRochester, New YorkThomas N. MooreDepartment of MechanicalEngineeringQueens UniversityKingston, Ontario, CanadaLeila NotashDepartment of MechanicalEngineeringQueens UniversityKingston, Ontario, CanadaCestmir OndrusekTechnical University of BrnoBrno, Czech RepublicHitay zbayDepartment of ElectricalEngineeringThe Ohio State UniversityColumbus, OhioM. K. RamasubramanianDepartment of Mechanical andAerospace EngineeringNorth Carolina State UniversityRaleigh, North CarolinaArmando A. RodriguezDepartment of ElectricalEngineeringArizona State UniversityTempe, ArizonaMomoh-Jimoh EyiomikaSalamiInternational Islamic Universityof MalaysiaKuala Lumpur, MalaysiaMario E. SalgadoDepartamento de ElectrnicaUniversidad Tecnica FedericoSanta MariaValparaso, ChileJyh-Jong SheenDepartment of MechanicalEngineering and MarineEngineeringNational Taiwan OceanUniversityKeelung, TaiwanAndrew SterianPadnos College of Engineeringand ComputingGrand Valley State UniversityGrand Rapids, MichiganFred StolfiXerox Mechanical EngineeringSciences LaboratoryRensselaer PolytechnicInstitute,Troy, New YorkMichael J. TordonThe University of New SouthWalesSydney, New South Wales,Australia 9. Michael TrimbornNational Instruments, Inc.Austin, TexasJob van AmerongenUniversity of TwenteEnschede, The NetherlandsCrina VladPolitehnica University of BucharestBucharest, RomaniaBogdan M. WilamowskiDepartment of Electrical andComputer EngineeringAuburn UniversityAuburn, AlabamaJuan I. YuzUniversidad Tecnica FedericoSanta MariaValparaso, ChileQingze ZouUniversity of WashingtonSeattle, Washington 10. ContentsSECTION I Mechatronic System Control1 The Role of Controls in MechatronicsJob van Amerongen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-12 The Role of Modeling in Mechatronics DesignJeffrey A. Jalkio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-13 Signals and Systems3.1 Continuous- and Discrete-Time Signals Momoh-Jimoh Eyiomika Salami. . . . . . . . . . . 3-13.2 z Transforms and Digital Systems Rolf Johansson . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-293.3 Continuous- and Discrete-Time State-Space Models Kam K. Leang, Qingze Zou, andSantosh Devasia. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-403.4 Transfer Functions and Laplace Transforms C. Nelson Dorny . . . . . . . . . . . . . . . . . . 3-544 State Space Analysis and System PropertiesMario E. Salgado and Juan I. Yuz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-15 Response of Dynamic SystemsRaymond A. de Callafon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-16 The Root Locus MethodHitay zbay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-17 Frequency Response MethodsJyh-Jong Sheen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-1 11. 8 Kalman Filters as Dynamic System State ObserversTimothy P. Crain II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-19 Digital Signal Processing for Mechatronic ApplicationsBonnie S. Heck and Thomas R. Kurfess . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9-110 Control System Design Via H2 OptimizationArmando A. Rodriguez. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-111 Adaptive and Nonlinear Control DesignMaruthi R. Akella . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-112 Neural Networks and Fuzzy SystemsBogdan M. Wilamowski . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-113 Advanced Control of an Electrohydraulic AxisFlorin Ionescu, Crina Vlad, and Dragos Arotaritei. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-114 Design Optimization of Mechatronic SystemsTomas Brezina, Ctirad Kratochvil, and Cestmir Ondrusek. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-115 Motion ControlRahul Kulkarni . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-116 Real-Time Monitoring and ControlGerardo Garcia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16-117 Micromechatronics and Microelectromechanical Motion DevicesSergey Edward Lyshevski . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-1SECTION II Computers and Logic Systems18 Introduction to Computers and Logic SystemsKevin C. Craig and Fred Stolfi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-119 Digital Logic Concepts and Combinational Logic DesignGeorge I. Cohn. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19-120 System InterfacesMichael J. Tordon and Jayantha Katupitiya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20-1 12. 21 Communications and Computer NetworksMohammad Ilyas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21-122 Fault Analysis in Mechatronic SystemsLeila Notash and Thomas N. Moore . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22-123 Logic System DesignM. K. Ramasubramanian. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23-124 ArchitectureDaniel A. Connors and Wen-Mei W. Hwu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24-125 Control with Embedded Computers and Programmable Logic ControllersHugh Jack and Andrew Sterian. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25-126 Graphical System Design for Embedded SystemsShelley Gretlein . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26-127 Field-Programmable Gate ArraysDaniel R. Fay and Daniel A. Connors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27-128 Graphical Programming for Field-Programmable Gate Arrays: Applications inControl and MechatronicsJeannie Sullivan Falcon and Michael Trimborn. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28-1SECTION III Software and Data Acquisition29 Introduction to Data AcquisitionCraig Anderson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29-130 Measurement Techniques: Sensors and TransducersCecil Harrison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30-131 A/D and D/A ConversionBrian Betts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31-132 Signal ConditioningStephen A. Dyer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32-133 Virtual Instrumentation SystemsDarcy Dement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33-1 13. 34 Software Design and DevelopmentMargaret H. Hamilton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34-135 Data Recording and LoggingCraig Anderson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35-1Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I-1 14. IMechatronic SystemControl1 The Role of Controls in MechatronicsJob van Amerongen ................................................................................................................ 1-1Introduction Key Elements of Controlled Mechatronic Systems Integrated Modeling,Design and Control Implementation Modern Examples of Mechatronic Systems in Action Special Requirements of Mechatronics that Differentiate from Classic Systems andControl Design2 The Role of Modeling in Mechatronics DesignJeffrey A. Jalkio ............................................................................................................................... 2-1Modeling as Part of the Design Process The Goals of Modeling Modeling of Systemsand Signals3 Signals and SystemsMomoh-Jimoh Eyiomika Salami, Rolf Johansson, Kam K. Leang,Qingze Zou, Santosh Devasia, and C. Nelson Dorny ................................................................ 3-1Continuous- and Discrete-Time Signals z Transforms and Digital Systems Continuous-andDiscrete-Time State-Space Models Transfer Functions and Laplace Transforms4 State Space Analysis and System PropertiesMario E. Salgado and Juan I. Yuz ......................................................................................... 4-1Models: Fundamental Concepts State Variables: Basic Concepts State SpaceDescription for Continuous-Time Systems State Space Description for Discrete-Time andSampled Data Systems State Space Models for Interconnected Systems SystemProperties State Observers State Feedback Observed State Feedback5 Response of Dynamic SystemsRaymond A. de Callafon .......................................................................................................... 5-1System and Signal Analysis Dynamic Response Performance Indicators for DynamicSystems6 The Root Locus MethodHitay zbay ................................................................................................................................... 6-1Introduction Desired Pole Locations Root Locus Construction ComplementaryRoot Locus Root Locus for Systems with Time Delays Notes7 Frequency Response MethodsJyh-Jong Sheen ................................................................................................................................ 7-1Introduction Bode Plots Polar Plots Log-Magnitude Versus Phase Plots Experimental Determination of Transfer Functions The Nyquist StabilityCriterion Relative Stability 15. 8 Kalman Filtersas Dynamic System State ObserversTimothy P. Crain II ..................................................................................................................... 8-1The Discrete-Time Linear Kalman Filter Other Kalman Filter Formulations Formulation Summary and Review Implementation Considerations9 Digital Signal Processing for Mechatronic ApplicationsBonnie S. Heck and Thomas R. Kurfess ...................................................................................... 9-1Introduction Signal Processing Fundamentals Continuous-Time to Discrete-TimeMappings Digital Filter Design Digital Control Design10 Control System Design Via H2OptimizationArmando A. Rodriguez .............................................................................................................. 10-1Introduction General Control System Design Framework H2Output FeedbackProblem H2State Feedback Problem H2Output Injection Problem Summary11 Adaptive and Nonlinear Control DesignMaruthi R. Akella ...................................................................................................................... 11-1Introduction Lyapunov Theory for Time-Invariant Systems Lyapunov Theory forTime-Varying Systems Adaptive Control Theory Nonlinear Adaptive Control Systems Spacecraft Adaptive Attitude Regulation Example Output Feedback Adaptive Control Adaptive Observers and Output Feedback Control Concluding Remarks12 Neural Networks and Fuzzy SystemsBogdan M. Wiliamowski ..................................................................................................... 12-1Neural Networks and Fuzzy Systems Neuron Cell Feedforward Neural Networks Special Feedforward Networks Recurrent Neural Networks FuzzySystems Genetic Algorithms13 Advanced Control of an Electrohydraulic AxisFlorin Ionescu, Crina Vlad, and Dragos Arotaritei .......................................................... 13-1Introduction Generalities Concerning ROBI_3, a Cartesian Robot with ThreeElectrohydraulic Axes Mathematical Model and Simulation of Electrohydraulic Axes Conventional Controllers Used to Control the Electrohydraulic Axis Control ofElectrohydraulic Axis with Fuzzy Controllers Neural Techniques Used to Control theElectrohydraulic Axis Neuro-Fuzzy Techniques Used to Control the ElectrohydraulicAxis Software Considerations Conclusions14 Design Optimization of Mechatronic SystemsTomas Brezina, Ctirad Kratochvil, and Cestmir Ondrusek ...................................................... 14-1Introduction Optimization Methods Optimum Design of Induction Motor TheUse of a Neuron Network for the Identication of the Parameters of a Mechanical DynamicSystem15 Motion ControlRahul Kulkarni .......................................................................................................................... 15-1Introduction to Motion Control Components of a Typical Motion ControlSystem Functions of a Motion Controller Motion Controller Hardware Summary16 Real-Time Monitoring and ControlGerardo Garcia ..................................................................................................................16-1Introduction to Real-Time Systems Real-Time Development Tools Real-TimeSoftware Architecture Deterministic Timing Implementing Real-Time Control Monitoring Systems Summary17 Micromechatronics and Microelectromechanical Motion DevicesSergey Edward Lyshevski ................................................................................................... 17-1Micromechatronic Systems Design Tracking Control of Micromechatronic Systems Synthesis of Microelectromechanical Motion Devices Micromechatronic System withan Axial Topology Motion Device Synchronous Micromachines Fabrication Aspects 16. 1The Role of Controlsin Mechatronics1.1 Introduction ............................................................... 1-11.2 Key Elements of ControlledMechatronic Systems ................................................. 1-31-11.3 Integrated Modeling, Designand Control Implementation .................................... 1-3Modeling Control System Design Methodologies Servo System Design Design of a Mobile Robot1.4 Modern Examples of MechatronicSystems in Action ........................................................ 1-12Rudder Roll Stabilization of Ships Compensationof Nonlinear Effects in a Linear Motor1.5 Special Requirements of Mechatronicsthat Differentiate from ClassicSystems and Control Design ...................................... 1-15References ................................................................................. 1-16Job van AmerongenUniversity of Twente1.1 IntroductionMechatronic design deals with the integrated and optimal design of a mechanical system and itsembedded control system. This definition implies that the mechanical system is enhanced with electroniccomponents in order to achieve a better performance, a more flexible system, or just reduce the cost ofthe system. In many cases the electronics are present in the form of a computer-based embedded (control)system. This does not imply that every controlled mechanical system is a mechatronic system becausein many cases the control is just an add-on to the mechanical system in a sequential design procedure.A real mechatronics approach requires that an optimal choice be made with respect to the realization ofthe design specifications in the different domains. In control engineering the design of an optimal controlsystem is well understood and for linear systems standard methods exist. The optimization problem isformulated as: given a process to be controlled, and given a performance index (cost function), findoptimal controller parameters such that the cost function is minimized. With a state feedback controllerand a quadratic cost function, solutions for the optimal controller gains can be found with standardcontroller design software, such as MATLAB1 (Figure 1.1).Mechatronic design on the contrary requires that not only the controller be optimized. It requiresoptimization of the system as a whole. In the ideal case all the components in the system: the processitself, the controller, as well as the sensors and actuators, should be optimized simultaneously (Figure 1.2).In general this is not feasible. The problem is ill defined and has to be split into smaller problems thatcan be optimized separately. Later on the partial solutions have to be combined and the performance ofthe complete system has to be evaluated. After eventually readjusting some parts of the system this leadsto a sub-optimal solution. 17. 1-2 Mechatronic System Control, Logic, and Data AcquisitionR +ControllerInterfaceFIGURE 1.1 Optimization of the controller.ProcessSensor(s)R +ControllerInterfaceOptimizationProcessSensor(s)FIGURE 1.2 Optimization of the all system components simultaneously.CCIn the initial conceptual design phase it has to be decided which problems should be solved mechan-icallyand which problems electronically. In this stage decisions about the dominant mechanical propertieshave to be made, yielding a simple model that can be used for controller design. Also a rough idea aboutthe necessary sensors, actuators, and interfaces has to be available in this stage. When the different partialdesigns are worked out in some detail, information about these designs can be used for evaluation ofthe complete system and be exchanged for a more realistic and detailed design of the different parts.Although the word mechatronics is new, mechatronic products have been available for some time. Infact, all electronically controlled mechanical systems are based on the idea of improving the product byadding features realized in another domain. Good mechatronic designs are based on a real systemsapproach. But mostly, control engineers are confronted with a design in which major parameters arealready fixed, often based on static or economic considerations. This prohibits optimization of the systemas a whole, even when optimal control is applied.In the last days of gramophones, the more sophisticated designs used tacho feedback in combinationwith a light turntable to achieve a constant number of revolutions. But a really new design was thecompact disc player. Instead of keeping the number of revolutions of the disc constant, it aims for aconstant speed of the head along the tracks of the disc. This means that the disc rotates slower whentracks with a greater diameter are read. The bits read from the CD are buffered electronically in a bufferthat sends its information to the DA converter, controlled by a quartz crystal. This enables the realizationof a very constant bit rate and eliminates all audible speed fluctuations. Such a performance could neverbe obtained from a pure mechanical device only, even if it were equipped with a good speed controlsystem. In fact, the control loop for the disc speed does not need to have very strict specifications. Itshould only prevent overflow or underflow of the buffer. The high accuracy is obtained in an open loopmode, steered by a quartz crystal (Figure 1.3).The flexibility introduced by the combination of precision mechanics and electronic control hasallowed the development of CD-ROM players, running at speeds more than 50 times faster than theoriginal audio CDs. A new way of thinking was necessary to come to such a new solution. On the other 18. The Role of Controls in Mechatronics 1-3TrackspeedDataflowQuartzclockBuffer DataflowDA-converterFIGURE 1.3 Combination of closed-loop and open-loop control in a CD player.SetpointgeneratorInformation domainFIGURE 1.4 Mechatronic system.+_FeedforwardFeedbackcontrollerSensorfusionDAAD+ +ActuatorSensor(s)ProcessPower domainhand, the CD player is still a sophisticated piece of precision mechanics. No solid-state electronic memorydevice can compete yet economically with the opto-mechanical storage capabilities of the CD and itssuccessor the DVD. But this may change rapidly.1.2 Key Elements of Controlled Mechatronic SystemsA mechatronic system consists by definition of a mechanical part that has to perform certain motionsand an electronic part (in many cases an embedded computer system) that adds intelligence to the system.In the mechanical part of the system power plays a major role. This in contrast to the electronic part ofthe system where information processing is the main issue. Sensors convert the mechanical motions intoelectrical signals where only the information content is important or even into pure information in theform of numbers (if necessary, through an AD converter). Power amplifiers convert signals into modulatedpower. In most cases the power supply is electrical, but other sources such as hydraulic and pneumaticpower supplies are possible as well. A controlled mechanical motion system thus typically consists of amechanical construction, one or more actuators to generate the desired motions, and a controller thatsteers the actuators based on feed-forward and sensor-based feedback control (Figure 1.4).1.3 Integrated Modeling, Design and ControlImplementation1.3.1 ModelingDuring the design of mechatronic systems it is important that changes in the construction and thecontroller be evaluated simultaneously. Although a proper controller enables building a cheaper con-struction,a badly designed mechanical system will never be able to give a good performance by addinga sophisticated controller. Therefore, it is important that during an early stage of the design a properchoice can be made with respect to the mechanical properties needed to achieve a good performance ofthe controlled system. On the other hand, knowledge about the abilities of the controller to compensatefor mechanic imperfections may enable that a cheaper mechanical construction be built. This requiresthat in an early stage of the design a simple model is available that reveals the performance limitingfactors of the system. Still there is a gap between modeling and simulation software used for evaluationof mechanical constructions and software used for controller design. Mechanical engineers are used to 19. 1-4 Mechatronic System Control, Logic, and Data Acquisitionfinite element packages to examine the dynamic properties of mechanical constructions. It is only afterreduction to low-order models (modal analysis) that these models can be used for controller design. Onthe other hand, typical control-engineering software does not directly support the mechatronic designprocess either; in the modeling process the commonly used transfer functions and state space descriptionsoften have lost the relation with the physical parameters of the mechanical construction. Tools are requiredthat allow modeling of mechanical systems in a way that the dominant physical parameters (like massand dominant stiffness) are preserved in the model and simultaneously provide an interface to thecontroller design and simulation tools control engineers are used to (Coelingh;2 Coelingh et al.3).Simulation is an important tool to evaluate the design of mechatronic systems. Most simulation pro-gramslike Simulink1 use block diagram representations and do not support physical modeling in a waythat direct tuning of the physical parameters of the mechanical construction and those of the controlleris possible as required in the design of mechatronic systems. Recently, programs that allow physicalmodeling in various physical domains became available. They use an object-oriented approach that allowshierarchical modeling and reuse of models. The order of computation is only fixed after combining thesubsystems. Examples of these programs are 20-sim,4 described by Broenink5 as CAMAS and Dymola.6In this section the modeling and simulation program 20-sim (pronounced Twente Sim) will be usedto illustrate the simultaneous design of construction and controller in a mechatronic system. 20-sim sup-portsobject-oriented modeling. Power and signal ports to and from the outside world determine eachobject7. Inside the object there can be other objects or, on the lowest level, equations. Various realizationsof an object can contain different or more detailed descriptions as long as the interface (number andtype of ports) is identical. Modeling can start by a simple interconnection of (empty) submodels. Laterthey can be filled with realistic descriptions with various degrees of complexity. De Vries8 refers to thisas polymorphic modeling. Submodels can be constructed from other submodels in hierarchical structures.Proper physical modeling is achieved by coupling the submodels by means of the flow of energy, ratherthan by signals such as voltage, current, force, and speed. This way of modeling is well suited formechatronics system design. It will be illustrated with an example. We want to consider the design of asimple servo system, considering the use of a voltage source, a DC motor, and a mechanical load driventhrough a transmission (Figure 1.5).The transmission is disregarded for the time being. The belt is considered as infinitely stiff and thetransformation ratio is taken care of by changing the motor constant. If a power amplifier driven by asignal generator describes the voltage source, we can draw the iconic diagram of Figure 1.6. At this stagethe different components in this model are still empty. But all components have electrical and/or mechan-icalports. With the proper interfaces (ports) defined, the components can be connected to each other.FIGURE 1.5 Simple DC-servo system. 20. The Role of Controls in Mechatronics 1-5FIGURE 1.6 Iconic diagram of the simple servo system.IconELResistor 1 Inductor 1DC motor 1Ideal physical elementsGround 1FIGURE 1.7 Icon of the motor expanded to ideal physical elements.Inertia 1MECHIn the next step we can detail the description of the DC motor. One solution could be the descriptiongiven in Figure 1.7. The motor is now described by a number of ideal physical elements, each representinga basic physical relation. The motor has an electrical (EL) as well as a mechanical port (MECH).Each of the elements in this figure can be described as an element with an electrical and/or mechanicalport. The idea of ports is made more explicit in so-called bond graphs.912 For the electrical elementsthese are the voltage difference over the element and the current through the element. For the mechanicalelements these are the torque and the (angular) velocity. The products of these conjugated variables (P = uior P = T) represent power.If we go down a step further into the hierarchy, we arrive at the level of equations. For instance, anelectrical resistor can be described by the equation:(1.1)p u = R p iwhere the variables p u and p i indicate the conjugated variables u and i of the electrical port p. Notethat this is an equation and not an assignment statement. It could have been written equally well in theform:(1.2)p i = 1/R p uIn a similar way the inductance can be described by the equations:(1.3)p u = L ddt(p i) or (p i ) = 1/L int(p u)where ddt(p i) denotes di/dt and int(p u) denotes u dt. In case of an R-element there is no preferencefor one of the two forms. For the I-element the integral form is preferred in the simulations. 20-simdetermines the preferred causal form and derives the equations automatically.The energy flow or power P is the product of two conjugated signals, called effort (e) and flow ( f ):P = ef (1.4)Examples of this expression in the mechanical and electrical domain areP = Fv or P = T (1.5)P = ui (1.6)where F is force, v is velocity, T is torque, is angular velocity, u is voltage, and i is current. 21. 1-6 Mechatronic System Control, Logic, and Data AcquisitionResistor 1 Inductor 1DC motor 1Inertia 1FIGURE 1.8 Complete model in the form of ideal physical elements.Resistor 1 Inductor 1DC motor 1Inertia 1 Belt pulley 1FIGURE 1.9 Model extended with transmission.Bearing 2Inertia 2Spring 1Belt pulley 2Inertia 2When we expand the complete Figure 1.6 we obtain Figure 1.8. When this model is processed amessage pops up that indicates that inertia 2 has a dependent state. The two inertias in this modelalways have the same speed, and therefore, they are dependent. They cannot have independent initialconditions. The message indicates that this element can only be written in derivative form:T = J d /dt (1.7)There are several ways to deal with this problem.1. The two inertias can be combined into one inertia (the program will do this automatically). Amessage pops up that the dependency of the two inertias has been solved symbolically.2. Dealing with the derivative causality by means of an implicit integration algorithm.3. The transmission can be added, including some flexibility in the belt.If the flexibility is negligible, solution 1 leads to the simplest model. On the other hand, the warningraises the question whether the flexibility of the belt can be disregarded indeed. If not, the model has tobe extended with a spring element. It should be noted that this should not be done for numerical reasonsonly. If the transmission were very stiff, this would result in high-frequency dynamics and lead to unnec-essaryslow simulations. On the other hand, if the flexibility is important, as it is in this system, the warningdraws the designers attention to the fact that the model may be oversimplified. In Figure 1.9 the trans-mission,including a spring element, has been added. Processing of this model does not produce anywarnings.This example illustrates how modern software can help to come up with a model that has the com-plexitythat is needed for a particular problem. Physical models, in the form of an iconic diagram, basedon connecting elements by means of power ports, may help in this modeling process. The user can selectthe preferred view, whether this is a bond graph, an iconic diagram with ideal physical element, or a viewusing higher lever submodels, like in Figure 1.6. In the next section it will be shown how to use thismodel for the design of controllers.1.3.2 Control System Design MethodologiesMany processes can be reasonably well controlled by means of PID controllers. This is due to the fact thatthese processes can be more or less accurately described by means of a second-order model. Tuning rules,like those of Ziegler Nichols, enable less experienced people to tune such controllers. Relatively simple 22. The Role of Controls in Mechatronics 1-7models can also describe many mechatronic systems. A mechatronic system mostly consists of an actuator,some form of transmission, and a load. A fourth-order model can properly describe such a system. Theperformance-limiting factor in these systems is the resonance frequency. A combination of position andtacho feedback (basically a PD controller) can be applied here as well. But due to the resonant poles properselection of the signals to be used in the feedback is essential. Efforts have been made2,3,13 to derive recipesfor tuning such systems, in addition to selecting the proper feedback signals. Computer support toolsare essential to enable less experienced designers to use these recipes (Van Amerongen, Coelingh, andDe Vries14). Coelingh2 and Coelingh et al.3 describe a structural design method for mechatronic systems.The method starts with reducing the conceptual design to a fourth-order model that represents thedominant properties of the system in terms of the total mass to be moved and the dominant stiffness.This model still has physical meaningful parameters. In this model appropriate sensors are chosen, aswell as a path generator. In the conceptual design phase a simple controller is developed and mechanicalproperties are changed, if necessary. Then a more detailed design phase follows where also parameteruncertainties are taken into account.1.3.3 Servo System DesignHere we will consider some simple aspects of the design of a servo system in order to illustrate theadvantage of the use of physical models and to illustrate the need for an integrated design approach. Weconsider the model discussed before, a load driven by an electric motor, through a flexible transmission.The iconic diagram of this model was given in Figure 1.9. In this example a current amplifier has replacedthe voltage amplifier allowing the removal of the electrical resistor and the inductance. In the stepresponses of Figure 1.10 the resonance due to the flexible transmission is clearly visible.From the equations used for the simulation, 20-sim can automatically derive a model in a form suitablefor controller design, such as a state-space description, a transfer function, or poles and zeros. An interfaceis provided to MATLAB1 enabling, for instance, to use MATLAB algorithms to compute the gains ofadvanced controllers like an LQR (optimal state feedback) or LQG controller (with a Kalman filter forstate estimation and optimal state feedback). The diagram of the process together with an LQG controlleris given in Figure 1.11 and some responses in Figure 1.12.A properly designed P(I)D controller is able to perform almost similarly, especially when the amountof noise is small. A first attempt could be to use only measurements of the load angle and load speed.1 150 101 52 03 54 100 1 2 3 4 5FIGURE 1.10 Open loop responses.62.55037.52512.50Load angleStepLoad angular velocityOpen loop response, step, load angular velocity and load angleStepLoad angular velocityLoad angleTime (s) 23. 1-8 Mechatronic System Control, Logic, and Data AcquisitionK1+KFIGURE 1.11 Process with Kalman filter and state feedback.32.521.510.5000.5LOG controlController output (A)Phl KalmanPhi load1 2 3 4 5Phi KalmanPhi loadController output (A)FIGURE 1.12 Response of the LQG-controlled system.qL+CX=Ax+Bu+LTime (s)This attempt fails, because feedback of the load speed leads almost immediately to an unstable systemas can be seen from the root locus for variations in the gain of the velocity feedback. From the responsesof Figure 1.10, 20-sim can easily determine the transfer function between the motor current and the loadspeed and plot the root locus (Figure 1.13).Figure 1.14 clearly shows that even a small amount of velocity feedback will lead to an unstable system.It is well known that feedback of the motor speed is a better solution. Using again the model of Figures1.9 and 1.10 to determine the transfer from input current to motor speed yields the root locus of Figure 1.15.Complex zeros now accompany the complex poles and because they are close together their influenceon the response will be almost negligible. The branch of the root locus on the real axis now shows thedesired behavior: moving the dominant pole to the left in the s-plane. Combining the feedback of the motorspeed with feedback of the load angle yields the PD-controller structure of Figure 1.15 and the responsesof Figure 1.16. Except for the noise there is not much difference in the responses of the system with theKalman filter, although the PD-controlled system is simpler. The observations made here are generallyapplicable. A system with two resonant (complex) poles and no zeros, such as in Figure 1.13, is difficultto control by means of a simple controller. If complex zeros accompany the resonant poles with animaginary part smaller than that of the poles, stable control is easily achieved. In the frequency domainthis is seen as an anti-resonance, followed by a resonance (type AR). On the contrary a type RA system,where the resonance frequency is lower than the anti-resonance frequency (the imaginary part of thepoles is smaller than that of the zeros), is just as difficult to control as in the case of only resonant poles.The existence and location of resonant zeros is completely determined by the (geometrical) location of 24. The Role of Controls in Mechatronics 1-940302010010203040Root locus for feedback of the load speed50 40 30 20 10 0 10 20 30ReImFIGURE 1.13 Root locus for velocity feedback of load axis.40302010010203040Root locus for feedback of the motor speed60 50 40 30 20 10 0 10 20ReImFIGURE 1.14 Root locus for velocity feedback of motor axis.the sensors in the mechanical system. A careful choice of these sensor locations is therefore crucial forthe successful application of a controller. It should be noted that using a properly designed set-pointgenerator could prevent resonance, as seen in Figure 1.10. The set point generator should not excite theresonance frequencies, for instance, by using a low pass filter with bandwidth lower than the resonancefrequencies. However, such a set-point generator does not solve the above-mentioned stability problems. 25. 1-10 Mechatronic System Control, Logic, and Data Acquisition+K Kp+FIGURE 1.15 Servo system with PD-controller.KKd32.5SetpointPhi loadController output (A)21.510.500.51PD controlController output {A}SetpointPhi load0 1 2 3 4 5FIGURE 1.16 Responses of the system of Figure 1.16.402002040601000100200300400Bode Plot type RA0.01 0.1 1 10 100 1000Frequency (rad/sec)0.01 0.1 1 10 100 1000Frequency (rad/sec)Magnitude (dB)Phase (deg)FIGURE 1.17 Bode plots of type RA and AR systems.f qTime (s)50301010305010050050100Bode Plot type AR0.01 0.1 1 10 100 1000Frequency (rad/sec)0.01 0.1 1 10 100 1000Frequency (rad/sec)Phase (deg) Magnitude (dB)1.3.4 Design of a Mobile RobotA typical example of the early design procedure is the conceptual design of a mobile assembly robot.Already in a very early stage of the design conflicting demands have to be resolved. Such a robot shouldbe able to collect parts all around a production facility and do the assembly while driving. Because ahigh accuracy is required between the gripper of the robot and the surface where the parts are located,it is important that floor irregularities and vibration modes of the structure do not prevent properassembly. On the other hand the path controller, partly based on dead reckoning (i.e., measuring of thewheel speed and orientation), requires that the wheels be very stiff. Damping of disturbances has to berealized by another means of suspension. This has led to the concept of an upper frame and a lowerframe, connected by means of springs (Figure 1.18). 26. The Role of Controls in Mechatronics 1-11Manipulator Z-tipUpper frameLower frameFIGURE 1.18 Conceptual design of the mobile robot.Z-upper frameZ-lower frameTipUpper framemmm Lower frameParameters:mtipctipdtipmuppercupperduppermlowercwheelsdwheels15 kg0.3105 N m11000 N s m1 kg200 N m11870 N s m1500 kgN s m1107 N m11100 N s m1FIGURE 1.19 Simple model with ideal physical elements to compute the error etip.The robot can be mounted at the upper frame and should have sufficient bandwidth such that theposition error (etip = ztip zupper frame) between the tip of the robot (ztip) and the upper frame (zupper frame)is small enough.The next step is to derive a simple model, in order to have some parameters for the weight distributionand the stiffness and damping of the springs. In the model of Figure 1.22 the robot is confronted witha bump in the floor at a speed of 1 m/s.Based upon the payloadmainly the weight of the batteriesthe total mass of the vehicle was estimatedto be 500 kg. Stiffness and damping of the wheels follow from the demands for the accuracy of the positionestimation. The mass and bandwidth of the controlled manipulator were already known from other studies,yielding the effective stiffness and damping for the robot tip. When also initial estimates of the stiffnessand damping of the springs between the upper and lower frame are made, the only parameter to be variedis the weight distribution between the upper frame and lower frame. By using the optimization feature of20-sim, the optimal weight distribution can easily be found. In order to minimize the error between thetip of the robot and the upper frame (Figure 1.19), the weight has to be placed as much as possible in the 27. 1-12 Mechatronic System Control, Logic, and Data AcquisitionError of the tip before and after optimization0.010.00500.0050.01ErrorBeforeAfter0 0.2 0.4 0.6 0.8 1Time (s)FIGURE 1.20 Error of the tip before and after optimization of the weight distribution between upper and lowerframe.FIGURE 1.21 The mobile robot (MART) after completion.upper frame (Figure 1.20). This example illustrates how the mechanical configuration of the system isdetermined by the requirements for good path control and accurate control of the assembly task.A next step could be to optimize the properties of the suspension between upper and lower frame.This will further improve the error. This decision made in a very early stage of the design directed otherdesign decisions. After completion of the project it appeared that the different parameters of the finalconstruction were close to these early estimates (Figure 1.21).1.4 Modern Examples of Mechatronic Systems in ActionA few examples have already been treated in the previous sections. In this section two more exampleswill be given.1.4.1 Rudder Roll Stabilization of ShipsNowadays most ships use an autopilot to control the heading of the ship. A rudder is the most commonlyused actuator. Some ships, like ferries and naval ships, need also roll stabilization. This can be achievedpassively by means of two connected tanks filled with water that generate stabilizing forces that should 28. The Role of Controls in Mechatronics 1-13be in counter phase with the forces of the waves. In order to make the system effective for varyingfrequencies of the waves, the water flow between the two tanks should be controlled. For fast ships mostlystabilizing fins are used. These are a kind of actively controlled wings that generate the moments neededto counteract the moments of the waves. The fins not only influence the roll motions but also haveinfluence on the heading. On the other hand, the rudder not only influences the heading but also inducesroll. In control engineering terms this leads to a multivariable system that requires a multivariablecontroller design for optimum performance. In practice such a multivariable system is seldom seen andtwo separate control systems are used.Another approach is to use only one of the actuators (rudder or fins) to achieve course control androll reduction. Because the frequencies of the roll motions are outside the bandwidth of the course-controlsystem this is possible. The rudder is most suited as actuator. An additional advantage for navalships is that removing the fins will reduce the underwater noise of the vessel.Redesigning the course controller in order to stabilize the roll as well, demonstrates the feasibility ofthis approach, but also makes clear that the processthe shipshould be modified. The most impor-tantmodification is needed for the steering machine. The maximum speed of the steering machineappears to be the limiting factor for such a system (it should increase from the commonly used valuesof 37/s to 2025/s). By means of dynamic simulations the demands for the steering machine can befound in terms of the maximum speed of the steering machine and the maximum time constant that isallowed for reaching this speed. This requires reengineering of the hydraulic steering machine. A stepfurther would be to consider also changes in the shape of the ship, in order to optimize the parametersthat determine the effectiveness of the rudder roll stabilization system.In order to decide whether this new solution is better, it should be evaluated whether the redesignedsteering machine is less expensive than the original rudder and fin actuators. These design issues haveto be solved in a very early stage of the design. Rudder roll stabilization has been successfully applied onnaval as well as merchant marine ships.151.4.2 Compensation of Nonlinear Effects in a Linear MotorMany mechanical systems suffer from nonlinear effects that limit the accuracy that can be achieved.Friction and cogging are two examples. A (linear) feedback controller can diminish the influence of non-linearities,but complete compensation may be difficult. For systems that perform repetitive motions, anIterative Learning Controller can help to further improve the performance.16,17 The basic idea is explainedin Figure 1.22.When only the feedback loop is present and under the assumption that there are no disturbances, theerror signal and thus the controller signal UC will be the same for each repetitive motion. It is obviousthat the accuracy can be improved when in the next motion the controller signal from the former cycleis used as a feed-forward signal, UF. The feedback will generate a signal that further compensates for theremaining error by updating the feed-forward signal UF with the formula(1.8)k+1 UFUF= k + LEkLearningFIGURE 1.22 Principle of iterative learning control.++MemoryController Processr+eUcUF+ + y 29. 1-14 Mechatronic System Control, Logic, and Data AcquisitionB-splinenetworkt+r e UcControllerUF++ yProcessFIGURE 1.23 Learning feed-forward controller for repetitive motions.where L is the transfer function of the learning filter. The superscript k denotes the kth repetitive motion.The signal UF should converge to a feed-forward signal that compensates for all repetitive errors. Anexample of a situation where such errors are present is, for instance, a CD player that has to compensatefor the eccentricity of the disk.A variation on this idea and even more straightforward is the learning feed-forward controller (LFFC)setup of Figure 1.23. When the feed-forward signal would be perfect, the output of the controller wouldbe zero. This implies that this output can be used as a training signal for a neural network. An adaptiveB-spline network enables learning of complex nonlinear characteristics. Also support vector machineshave been used to implement the learning feed forward.18 The input of the B-spline network is thetime t. It is reset each time a new motion starts. This is called a time-indexed LFFC. Instead of the time,also the reference signal and its derivativesobtained from a path generatorcould be used as indexfor the network (path-indexed LFFC). The advantage of this structure is that after proper training theLFFC can successfully be used for nonrepetitive motions as well. Velthuis has given a stability analysisfor time-indexed as well as path-indexed LFFC.19 The stability analysis is relatively easy for the time-indexedcase. For the path-indexed case it is more complex and some heuristics are required to guaranteea stable system. The main issue is that the number of B-splines should not be too large. On the otherhand a sufficiently dense B-spline distribution is desired for an accurate approximation of the nonlinearprocess. LFFC has successfully been applied to compensate for cogging in an industrial Linear Motor 20and for compensation of (Coulomb) friction of a linear motor used in a flight simulator.19 It has alsobeen applied to the tracking control of the mobile robot described in the section Design of a MobileRobot.21The application to cogging compensation of a linear motor will be described in a little bit more detail.Such a motor is a commonly used element in assembly machines. Even with the best magnets and accurateassembly the error could not be made smaller than 100 , with a PID controller in combination withnonlearning feed-forward control. The design goal was to improve the maximally achievable accuracyfrom 100 to less than 10 . Figure 1.24 shows a picture of a linear motor.According to the structure of Figure 1.23 the linear motor is controlled by means of a PID controller,while a B-spline neural network is present to learn the inverse motor model, including the nonlinearity dueto cogging. Cogging occurs in DC motors with permanent magnets. It causes more or less sinusoidal shapedforces that depend on the position of the translator with respect to the stator. If these forces really had asinusoidal shape, they would be easy to compensate for by means of a feed-forward compensator. However,this would require magnets with completely similar magnetic properties and very accurate spacing of themagnets. An alternative is to design a controller that learns the disturbance pattern and compensates it bymeans of a learning feed-forward compensator. An additional advantage is that such a system can also beused to compensate for other nonlinear effects, such as friction. This has also been demonstrated in a partof a flight simulator (a control stick) where friction forces spoil the feeling of a realistic simulation especiallyat almost zero speed. Figure 1.25 shows that learning is almost completed after six training cycles.Learning feed-forward control is an attractive method to compensate for nonlinearities that are presentin mechatronic systems, such as cogging and friction. The use of B-spline neural networks results in fastconvergence, relatively low computational effort, and a good generalizing ability. Because of recentlyobtained results with respect to the stability of such systems, robust control systems can be designed. 30. The Role of Controls in Mechatronics 1-15FIGURE 1.24 Linear motor. The magnets that cause the cogging are clearly visible.4e-0053e-0052e-0051e-00501e-0052e-0050.50.30.10.10.30.50.7Learning behavior of the linear motorPositionPosition error0 10 20 30 40 50 60 70 80 90 100Time (s)Positional errorFIGURE 1.25 Position and error signal during learning of the LFFC.PositionA mechatronic view on this design problem raises the question whether it is possible to use the sametechniques to build a less expensive linear motor, when maximum accuracy is not the main goal of thedesign. It has been demonstrated that a motor constructed with less expensive components and lessdemanding assembly specifications but with LFFC can compete well with the more expensive construc-tion.The accuracy can typically be improved by a factor 10.1.5 Special Requirements of Mechatronics that Differentiatefrom Classic Systems and Control DesignThe main difference between ordinary controller design and mechatronic system design is that the latterdeals with the design of the system as a whole. This approach can be considered as optimization of allcomponents of the system simultaneously, although there are no algorithms to do this automatically. Inpractice the problem is often split into smaller problems that can be optimized. After integration of allthe partial solutions a suboptimal system is achieved that can be further optimized by retuning the differentparts, taking into account the already available intermediate design of the overall system. In order toachieve optimization of the system as a whole, it is desired that the mechanical part, where power plays arole, and the information processing part (the controller) can be simulated and adjusted simultaneously.This requires that mechanical parameters like masses and compliances be available in simulations of the 31. 1-16 Mechatronic System Control, Logic, and Data Acquisitioncontrolled system. Examples have been given of modeling and simulation with 20-sim that allows for suchan approach.Mechatronic designers should constantly be aware of the fact that solutions can be found in differentdomains. Not every mechanical deficiency can easily be solved by control. A good mechanical designmay be easier and cheaper to achieve. On the other hand, a good controller may be able to achieve thedesired performance much easier and cheaper than a complex mechanical construction. In some casesthe combination can even achieve performances that would never have been possible without a mecha-tronicdesign.The same holds for the design of sensors. Each sensor could be fitted with a filter to remove noisefrom the measurements. But if several sensors are being combined, sensor fusion in a Kalman filteralgorithm will benefit from the availability of the raw data.Communication between all the designers involved and transparency of the design decisions in thevarious domains are essential for the success of a true mechatronic design.References1. Matworks, The Mathworks: Developers of MATLAB and Simulink, 2000, www.mathworks.com2. Coelingh, H.J., Design Support for Motion Control Systems, Ph.D. thesis, University of Twente, 2000,also www.rt.el.utwente.nl/clh/3. Coelingh, H.J., de Vries, T.J.A., van Amerongen, J., Design support for motion control systemsapplication to the Philips fast component mounter, in Mechatronics Forum 7th Int. Conf., Mecha-tronics2000, Atlanta, Ga, USA.4. Controllab Products, 20-sim, www.20sim.com5. Broenink, J.F., Computer-Aided Physical-Systems Modeling and Simulation: a Bond-Graph Approach,Ph.D. thesis, University of Twente, 1990.6. Dynasim, Dymola, www.dynasim.se/7. Weustink, P.B.T., de Vries, T.J.A., Breedveld, P.C., Object Oriented Modeling and Simulation of Mecha-tronicSystems with 20-sim 3.0, in Mechatronics 98, J. Adolfson and J. Karlsn (Eds.), Elsevier Science,1998.8. De Vries, T.J.A., Conceptual Design of Controlled Electro-Mechanical Systems, Ph.D. thesis, Universityof Twente, 1994.9. Breedveld, P.C., Fundamentals of bond graphs, in IMACS Annals of Computing and Applied Math-ematics,Vol. 3: Modelling and Simulation of Systems, Basel, 1989, pp. 714.10. Cellier, F.E., Elmqvist, H., Otter, M., Modeling from physical principles, in The Control Handbook,W.S. Levine (Ed.), CRC Press, 1996, pp. 99108.11. Gawthrop, P., Lorcan Smith, L., Metamodelling: Bond Graphs and Dynamic Systems, Prentice-Hall,NJ, 1996.12. Van Amerongen, J., Modelling, simulation and controller design for mechatronic systems with 20-sim3.0, in Proc. 1st IFAC Conf. on Mechatronic Systems, Darmstadt, Germany, September 2000, pp. 831836.13. Groenhuis, H., A Design Tool for Electromechanical Servo Systems, Ph.D. thesis, University of Twente,1991.14. Van Amerongen, J., Coelingh, H.J., de Vries, T.J.A., Computer support for mechatronic controlsystem design, Robotics and Autonomous Systems, vol. 30, no. 3, pp. 249260, PII: SO921-8890(99)00090-1, 2000.15. Van Amerongen, J., van der Klugt, P.G.M., van Nauta Lemke, H.R., Rudder roll stabilization forships, Automatica, vol. 26, no. 4, pp. 679690.16. Arimoto, S., A brief history of iterative learning control, in Iterative Learning Control: Analysis, Design,Integration and Applications, Kluwer Academic Publishers, 1988, pp. 37.17. De Vries, T.J.A., Velthuis, W.J.R., van Amerongen, J., Learning feed-forward control: a survey andhistorical note, in 1st IFAC Conf. on Mechatronic Systems, Darmstadt, Germany, September 2000,pp. 949954. 32. The Role of Controls in Mechatronics 1-1718. De Kruif, Bas J., de Vries, T.J.A., On using a support vector machine in learning feed-forward control,in Proc. 2001 IEEE/ASME Int. Conf. on Advanced Intelligent Mechatronics, Como, Italy, 812 July, 2001.19. Velthuis, W.J.R., Learning Feed-Forward ControlTheory, Design and Applications, Ph.D. thesis,University of Twente, 2000, also http://www.rt.el.utwente.nl/vts/20. Otten, G., de Vries, T.J.A., van Amerongen, J., Rankers, A.M., Gaal, E., Linear motor motion controlusing a learning feedforward controller, IEEE/ASME Transactions on Mechatronics, vol. 2, no. 3, ISSN1083-4435, 1997, pp. 179187.21. Starrenburg, J.G., van Luenen, W.T.C., Oelen, W., van Amerongen, J., Learning feed-forward con-trollerfor a mobile robot vehicle, Control Engineering Practice, vol. 4, no. 9, 1996, pp. 12211230. 33. 2The Role of Modelingin Mechatronics Design2.1 Modeling as Part of the Design Process ................... 2-12-1Phase 1 Phase 2 Phase 3 Phase 42.2 The Goals of Modeling .............................................. 2-6Documentation and Communication HierarchicalFramework Insights Analogies Identificationof Ignorance2.3 Modeling of Systems and Signals ............................. 2-9Analytical versus Numerical Models Partial versusOrdinary Differential Equations Stochastic versusDeterministic Models Linear versus NonlinearReferences ................................................................................ 2-11Jeffrey A. JalkioUniversity of St. ThomasIf mechatronics design is more than just the combination of electronic, software, and mechanical design,the additional feature must lie in the ability of the mechatronic designer to optimize a design solutionacross these disparate fields. This requires a sufficient understanding of each of these fields to determinewhich portions of an engineering problem are best solved in each of these domains given the currentstate of technology. In turn, this requires the ability to model the problem and potential solutions usingtechniques that are domain independent or at least permit easy comparison of solutions and tools fromdifferent domains.For example, the optical inspection system shown in Figure 2.1 depends on optical components inprecise alignment, mechanical elements capable of precise motion, transducers for sensing and providingmechanical power, electrical systems to control motion and filter sensor signals, and software for imageanalysis and motion control. Only by dividing these tasks appropriately among electronics, mechanicalcomponents, and software can the system be optimized. This requires an understanding of all the systemrequirements and limitations as well as the capabilities of each component in the various domains.Modeling of requirements and systems is crucial in determining whether a proposed solution is acceptableas well as in documenting these determinations for future use. In this article we shall examine the varietiesof models used at different points in the design process, the diverse roles of these models and their relativestrengths and weaknesses in each of these roles, and finally the specific tradeoffs involved in choosingdynamic models for signals and systems analysis.2.1 Modeling as Part of the Design ProcessModels serve different purposes at different points in the design process; so to decide which modelingtools are most effectively employed in different phases we must examine the design process itself. Manydescriptions of the design process are available that have been developed by researchers around theworld.13 Typically these descriptions serve to systematize the process to improve the productivity of 34. 2-2 Mechatronic System Control, Logic, and Data AcquisitionFIGURE 2.1 An optical inspection system for printed circuit boards. (Used by permission CyberOptics Cor-poration2001, all rights reserved.)designers or to describe techniques that provide improved product quality, lower cost, or other benefits.However, since our purpose is to examine the modeling needs of the design process, we can consider asimple model that distinguishes phases of the design process in terms of types of design activity ratherthan a more complex model that may be preferable for other purposes. For this purpose, we can considera four-phase process consisting of requirements analysis, concept generation, analysis and selection, anddetailed design. In the first phase of this process the designer focuses on analysis of the problem withoutconsidering possible solutions. In the second phase, conceptual solutions are generated with the hopethat an acceptable solution can be found from these initial concepts via combination or modification ofconcepts or by variation of parameters present in one of the conceptual solutions. In the third phase,these concepts are evaluated and a design is chosen for implementation. The fourth phase consists ofidentifying design problems that need to be solved to implement the chosen concept and applying thedesign process to those smaller problems. We shall consider the activities of each of these phases in detail.2.1.1 Phase 1The requirements analysis phase consists in obtaining a sufficient understanding of the problem to besolved. The difficulty of this process varies with the scale of the problem, the designers familiarity withthe problem domain, the variability of market needs, and the presence of hidden requirements that arepoorly articulated in the initial problem statement. Depending on the nature of the design problem, therequirements identified in this phase may be the needs of a single customer, the common needs of agroup of potential customers identified via a market survey, or societal needs identified by governmentregulations. Most design problems include some combination of these as well as internal requirementssuch as design guidelines and company policies. The key objective of this phase is to obtain enough detailto know when a design has solved the problem satisfactorily. Models in this phase serve primarily ascommunication and documentation tools since the primary problem in this phase is the clear commu-nicationand documentation of the criteria for design success. Examples of models that aid in this processinclude specification listings, use case diagrams, sequence diagrams, and context diagrams. Many of thesemodeling tools have now been standardized as parts of the Unified Modeling Language (UML), whichis becoming increasingly important as an analysis tool.4Use case diagrams model the interactions between a system and its users at a very high level of abstractionin terms of purpose of the interaction. Figure 2.2 gives an example of a use case diagram used to documentthe various operations required of a network printer. The use case diagram helps us avoid overlookingimportant but rare use cases such as maintenance. It is important to note that use case diagrams do not 35. The Role of Modeling in Mechatronics Design 2-3Network printerMaintenanceuserSystemadministratorFIGURE 2.2 Use case diagram for a network printer.PrintSetup