Parallel Processing for Jet Engine Control

Parallel Processing Applications for Jet Engine Control is a volume in the new Advances in Industrial Control series, edited by Professor M.J. Grimble and Dr. M.A. Johnson of the Industrial Control Unit, University of Strathclyde. The book describes the mapping and load balancing of gas turbine engine and controller simulations onto arrays of transputers. It compares the operating system for transputers and the Uniform System upon the Butterfly Plus computer. The problem of applying formal methods to parallel asychronous processors is addressed, implementing novel fault tolerant systems to meet real-time flight control requirements. The book presents real-time closed-loop results highlighting the advantages and disadvantages of Occam and the transputer. Readers will find that this book provides valuable material for researchers in both academia and the aerospace industry.

1 Preliminaries.- 1.0 Introduction.- 1.1 Outline of Chapters.- 1.2 Contributions to Current Research.- 2 Controller and Engine Simulation Implementation.- 2.0 Introduction.- 2.1 Engines — An Introduction.- 2.1.1 Compressor.- 2.1.2 Combustion Chamber.- 2.1.3 Turbine.- 2.2 Gas Turbine Types.- 2.3 Afterburning or Reheat.- 2.4 Requirements and Current Trends in Engine and Controller Design.- 2.5 Compressor Surge.- 2.6 Engine Modelling.- 2.6.1 Fuel System.- 2.6.2 Gas Generator Dynamics.- 2.6.3 Turbine Blade Temperature (TBT).- 2.6.4 Engine Thrust.- 2.7 The Controller.- 2.8 Engine Starting.- 2.8.1 Temperature Limiting During Starting.- 2.9 Normal Operation.- 2.9.1 Governor Control.- 2.9.2 Acceleration Control Loop.- 2.9.3 Deceleration Control Loop.- 2.9.4 Overselling and Underselling Limiter Tasks.- 2.9.5 TBTLimiter.- 2.9.6 Low Pressure Turbine Control.- 2.10 Reheat and Nozzle Control.- 2.10.1 Nozzle Control.- 2.10.2 Fuel Flow Demand Control.- 2.10.3 Reheat Sequencing Control.- 2.11 Baseline Simulation.- 2.11.1 Controller Simulation.- 2.12 TSIM Simulation Implementation.- 2.13 Pole-Zero Mapping.- 2.14 Matlab Simulation Implementation.- 2.15 Concluding Remarks.- 2.15.1 Engine Modelling.- 2.15.2 Baseline Simulation.- 2.15.3 Digital Simulation.- 3 Controller and Engine Simulation on the Inmos Transputer.- 3.0 Introduction.- 3.1 Occam and the Transputer.- 3.1.1 Von Neumann Computers.- 3.1.2 Inmos Transputer.- 3.1.2.1 Inmos Serial Links.- 3.1.2.2 Memory.- 3.2 Occam.- 3.2.1 The Occam Constructs.- 3.2.2 Further Aspects of Occam.- 3.3 Occam Simulation of Engine Model.- 3.3.1 Operating Modes of Simulation.- 3.3.2 Simulation Initialisation.- 3.3.3 Program Options.- 3.3.4 Results from Simulation.- 3.4 Multi-Processor Implementation.- 3.4.1 Engine Simulation Load Balancing.- 3.4.2 Controller Simulation.- 3.4.3 Modified Controller Simulation.- 3.5 Load Balancing.- 3.5.1 Mapping Considerations.- 3.5.2 Controller Mapping onto 2 Processors.- 3.5.3 Controller Mapping onto 3 and 4 Processors.- 3.6 Model Enhancements.- 3.6.1 Addition of Reheat.- 3.6.2 Reheat Mapping onto 2 Processors.- 3.6.3 Reheat Mapping onto 3 and 4 Processors.- 3.6.4 Start-up Control and In-flight Relighting.- 3.7 Concluding Remarks.- 3.7.1 The Transputer and Occam.- 3.7.2 Mapping Issues.- 3.7.3 Performance Estimation and Improvements.- 3.7.4 Architectural Issues.- 3.7.5 Achievements.- 4 Alternative Approaches to Parallel Processing — The Butterfly Plus and Helios Systems.- 4.1 The Helios Operating System.- 4.1.1 Helios Shell Interface.- 4.1.2 I/O Server.- 4.1.3 Helios C Compiler.- 4.1.4 Static and Dynamic Load Balancing.- 4.2 Interprocess Communication.- 4.2.1 Communication.- 4.2.2 Communication Failure.- 4.2.3 Synchronisation.- 4.2.4 Implementation of Communication.- 4.3 Programming in Helios.- 4.3.1 Modifications Needed to C Code.- 4.3.2 Debugging.- 4.3.3 Assignment of Parallel Tasks.- 4.3.4 Task Force Manager (TFM).- 4.4 Analysis of System Performance.- 4.4.1 Timing Using the Real-Time Clock.- 4.4.2 Execution Times.- 4.4.3 Task Force Manager Overheads.- 4.4.4 Consistency of Timings.- 4.5 General Comments on the Helios System.- 4.6 Conclusions.- 4.6.1 Suitability for Real-Time Control.- 4.6.2 Fault Tolerance.- 4.6.3 General Comments.- 4.6.4 Reliability.- 4.6.5 Farm Construct.- 4.7 The Butterfly Plus Computer.- 4.8 Overview of Hardware.- 4.8.1 Processor Node Card.- 4.8.2 Switch Interface and Deadlock Handling.- 4.8.2.1 Receiver.- 4.8.2.2 Transmitter.- 4.8.3 Switch Card.- 4.8.3.1 Example of Butterfly Switch Efficiency.- 4.8.3.2 Switch Protocol.- 4.8.3.3 Switch Protocol Example.- 4.8.3.4 Switch Contention Handling.- 4.9 Programming the Butterfly Plus.- 4.9.1 Introduction.- 4.9.2 Utilities Available.- 4.9.3 Communication Synchronisation.- 4.9.4 Programming Alternatives.- 4.9.5 Uniform System Approach.- 4.10 Summary of Software Design.- 4.11 Comparison of Sun 3 and Butterfly Plus.- 4.12 GIST Analysis.- 4.12.1 Introduction.- 4.12.2 Typical GISTTrace.- 4.12.3 Parallel Mapping.- 4.13 GIST Analysis of Program.- 4.13.1 System Performance: 1–5 Processors.- 4.13.2 Detailed Analysis of System Performance on 4 and 5 Processors.- 4.13.2.1 Four Processors — Engine Simulation.- 4.13.2.2 Four Processors — Controller Simulation.- 4.13.2.3 Five Processors — Engine Simulation.- 4.13.2.4 Five Processors — Controller Simulation.- 4.13.3 Detailed Analysis of System Performance on 4 and 5 Processors — Concluding Remarks.- 4.14 Task Length Modification.- 4.14.1 Modified Simulation Results.- 4.15 Summary of Butterfly Plus Results.- 4.15.1 Program Organisation.- 4.15.2 Task Generators.- 4.15.3 Real-Time Scheduler.- 4.16 Comparison of Helios and Butterfly Plus.- 4.17 Concluding Remarks.- 4.17.1 Summary.- 4.17.2 Helios Operating System.- 4.17.3 Butterfly Plus and Uniform System.- 4.17.4 Automatic Load Balancing.- 5 Formal Methods and System Specifications.- 5.0 Introduction.- 5.1 System Requirements.- 5.2 The Need for Formal Methods.- 5.2.1 The Advantages of Formal Methods.- 5.2.2 The Need for a Mathematical Specification.- 5.2.3 The Structure of Formal Methods.- 5.2.3.1 The Vienna Development Method (VDM).- 5.2.3.2 The Z Specification Language.- 5.3 Summary of Disadvantages of Formal Methods.- 5.4 System Validation and Verification.- 5.4.1 System Validation.- 5.4.2 System Verification.- 5.5 Conventional Testing.- 5.5.1 Static Code Analysis.- 5.5.2 Automated Tools for Analysis.- 5.6 Formal Methods for Hardware Specification.- 5.6.1 ELLA.- 5.6.2 VIPER.- 5.6.3 Formal Methods Applied to the T800 Transputer Floating Point Unit.- 5.7 Application of Formal Methods to Transputers.- 5.7.1 Problems of Asynchronous Processors.- 5.7.2 The Trace Model of CSP.- 5.8 Use of Formal Methods in Fault Tolerant Systems.- 5.8.1 Fail-Safe or Shut-Down Systems.- 5.8.2 Control and Protection Systems.- 5.8.3 Formal Specification of N-Lane Replication (N>1).- 5.8.4 Reconfigurable Systems.- 5.9 Specifications for Gas Turbine Systems.- 5.9.1 Top-Level Specification of Gas Turbine Controller Software.- 5.9.2 Review of CSAN 1454 Draft Guidelines Document with Regard to Parallel Processing and in Particular to the use of Transputer Arrays.- 5.9.3 Intercommunication.- 5.9.4 Input-Output Conditioning.- 5.9.5 Cycle Segregation.- 5.9.6 Interrupts.- 5.9.7 Fault Detection and System Monitoring Principles.- 5.9.8 Precision of Monitoring.- 5.9.9 Independence of Comparison.- 5.9.10 Scope of Self-Check Program.- 5.9.11 Protection Against Dormant Faults.- 5.9.12 Computer Automony.- 5.9.13 Fault Modes and Effects.- 5.9.13.1 Fault Categories.- 5.9.13.2 Spurious Faults.- 5.9.14 Final Failure Case.- 5.10 Concluding Remarks.- 5.10.1 Summary.- 5.10.2 Perceived Benefits of Formal Methods.- 5.10.3 Design of Gas Turbine Controller Software.- 6 Failure Management and its Application in Gas Turbine Engine Control.- 6.0 Summary.- 6.1 Introduction.- 6.2 The History of Fault Tolerance in Computers.- 6.3 Characterisation of Faults.- 6.3.1 Design Faults.- 6.3.2 Operational Faults.- 6.3.3 Byzantine Generals Disagreement.- 6.3.4 Power Supply Failure.- 6.3.5 Fault Categories Addressed.- 6.4 Desired Response from a Fault Tolerant System.- 6.5 Fault Tolerant Techniques.- 6.5.1 Fault Tolerant Techniques.- 6.6 Hardware Redundancy.- 6.6.1 Active Replication.- 6.6.2 Passive Replication.- 6.7 Software Redundancy.- 6.8 Designing Fault Tolerance into the Process.- 6.9 Evaluation Measures.- 6.10 Application of Fault Tolerance to Systems.- 6.10.1 Long-Life Applications.- 6.10.2 Critical Computations.- 6.10.3 Maintenance Avoidance.- 6.10.4 Availability.- 6.11 Fault Tolerant Processor Topologies.- 6.11.1 Loop Topology.- 6.11.2 Tree Architectures.- 6.11.3 Array Processors.- 6.12 Review of Fault Tolerant Designs Already Implemented.- 6.13 Design and Implementation of Fault Tolerant Gas Turbine Engine Controllers.- 6.13.1 Aim of the Investigation.- 6.14 Design Constraints of Aero-Engines.- 6.14.1 Design Philosophy.- 6.15 Backward Error Recovery/TMR Scheme (BER/TMR).- 6.15.1 Normal Mode.- 6.15.2 Voter/Identifier Mode.- 6.15.3 Conclusions From BER/TMR Method.- 6.15.4 Summary of BER/TMR System.- 6.16 Method of Overlapping Triads (DTMR System).- 6.16.1 Evolution of the Processor Topology.- 6.16.2 Reliability Analysis of the Configuration.- 6.16.3 Input Data Validation.- 6.16.4 Controller Task Calculation.- 6.16.5 Data Output Validation.- 6.16.6 Relay Selection of Output DACs.- 6.16.7 Link Adaptor Shutdown Logic.- 6.16.8 System Testing.- 6.17 Communication Channel Failure.- 6.17.1 System Synchronisation.- 6.17.2 Communication Failure Detection and Accommodation.- 6.17.3 Cascade Approach.- 6.17.4 “Lemming” Approach.- 6.17.5 Self-Adjusting Time Frame Technique.- 6.17.6 Link Failure Handling Software.- 6.17.7 Limitations Imposed by Link Failure Handling.- 6.17.8 Modified Input Data Validation.- 6.17.9 Conclusions of “Overlapping Triads” Method.- 6.18 “Hot Sparing” Technique.- 6.18.1 Data Input.- 6.18.2 Controller Task Implementation.- 6.18.3 Data Output.- 6.18.4 Byzantine Disagreement Handling.- 6.18.5 Problems Emcountered with the Inmos C004.- 6.18.6 Voted Deselect.- 6.18.7 Link Relays.- 6.18.8 Conclusions of “Hot Sparing” Implementation.- 6.19 Concluding Remarks.- 6.19.1 Summary.- 6.19.2 BER/TMR Backward Error Recovery System.- 6.19.3 Overlapping Triad System.- 6.19.4 Hot Sparing System.- 6.19.5 Transputer Architectural Restrictions.- 7 Concluding Remarks.- 7.0 Project Motivation.- 7.1 Gas Turbine Engine Modelling.- 7.2 Transputer Implementation.- 7.3 Comparison Using Diverse Architectures.- 7.4 System Requirements and Formal Methods.- 7.5 Fault Tolerant Systems.- 7.6 Areas for Further Research.- 7.6.1 T9000 Series of Transputers.- 7.6.2 Future Software Developments for the T9000.- Appendices.- A Hardware Development.- A.1 Introduction.- A.2 Single Eurocard Development Rack.- A.3 Inmos Strategy.- A.4 Overview of Transputer Products.- A.4.1 T414 Transputer.- A.4.2 T800 Transputer.- A.4.3 T212 Transputer.- A.4.4 M212 Transputer.- A.4.5 C004 Programmable Link Switch.- A.4.6 The Inmos Link Adaptor.- A.5 The Transputer Architecture.- A.5.1 Clock.- A.5.2 Analyse and Error.- A.5.3 Event Request and Event Acknowledge.- A.5.4 External Memory Interface.- A.5.5 External Memory Interface Program.- A.5.6 Reset.- A.5.7 Booting.- A.6 Development Transputer Card.- A.6.1 Static RAM Board.- A.6.2 Dynamic RAM Board.- A.6.3 Software Testing.- A.7 The Single Eurocard Transputer Card.- A.8 Standalone Network Loader Card.- A.8.1 Design Overview.- A.8.2 EPROM Programming.- A.9 Reset/Analyse Debugging.- A.9.1 Debugging Facilities.- A.9.2 Debugger Operation.- A.10 Programmable Link Switch Card.- A.10.1 Performance of Link Switch.- A. 11 12-Bit Transputer Multi-Channel ADC Card.- A. 12 12-Bit Transputer Multi-Channel DAC Card.- A.13 VME Bus Interface Card.- A.13.1 VME Bus Transputer Interface (VMETI).- A.14 Summary.- B Helios Default Maps.- C Text of Section 21 of DS 00–55.- D Evaluation Measures of Fault Tolerant Techniques.- D.1 Terminology.- D.1.1 Coverage.- D.1.2 Reliability.- D.1.3 Mission Time.- D.1.4 Availability.- D.I.5 Testability and Verifiability.- D.1.6 Other Important Factors.- D.2 Quantitative Measures.- D.2.1 Combinatorial Modelling.- D.2.2 Markov Modelling.- D.2.3 Reliability Analysis of Passive Replication.- D.2.4 Series Reliability.- D.2.5 Parallel Reliability.- D.2.6 Simplex System.- D.2.7 Duplex System.- D.2.8 M of N Systems.- D.2.9 Reliability Comparison.- D.2.10 MTTF Comparison.- D.2.11 Mission Time Comparison.- E. Overview of Fault Tolerant Designs Already Realised.- E.1 Tandem 16 Nonstop.- E.2 Augusta A129.- E.3 The General Purpose Digital Controller (GPDC).- E.4 SIFT.- E.5 FTMP.- E.6 MAFT.- E.7 FASP.- E.8 Fault Tolerance in VLSI (The Intel iAPX 432).- F. The Fault Integrator.- G. Formal Expression of Overlapping Triads Technique.- G.1 The “Overlapping Triads” Technique.- G.1.1 Voting.- G.1.2 Voting Action.- G.2 Monitor Action.- G.2.1 Fault Categories.- G.3 Example.- G.3.1 Reconfiguration.- G.3.2 Failure of Monitor Transputers.- H. Self-Test Procedure.- I. Formal Expression of Extension of Overlapping Triads to Hot Sparing System.- 1.1 The “Hot Sparing” Technique.- 1.1.1 Voting.- 1.1.2 Voting Action.- 1.2 Monitoring Action.- 1.2.1 Fault Categories.- 1.3 Example.- 1.3.1 Reconfiguration.- 1.3.2 Failure in Monitoring of Transputers.- References.