A Systems Engineering Approach to Designing Complex Systems Dr. Michael Winter, Mr. Randy Skelding, & Dr. Ravi Rajamani Pratt & Whitney, United Technologies Corporation This document contains no technical data. United Technologies Business units aerospace systems commercial power solutions commercial building systems 2 PRATT & WHITNEY Leading industry change Commercial Engines and Global Services P&W Rocketdyne Power Systems P&W Canada Military Engines System Engineering Process Driven by Product Needs Propulsion System Complexity Driving Need for More Robust Systems Engineering Process and Tools Variant-common F135 Turbomachinery Integrated Integrated TEC TEC & & Augmentor Augmentor Lift Fan, Clutch, & Driveshaft 3-Bearing 3-Bearing Swivel Swivel Duct Duct Roll Roll Control Control Ducts Ducts and and Nozzles Nozzles LO LO Axi-symmetric Axi-symmetric Nozzle Nozzle Controls & externals, engine gearbox Pratt Pratt&&Whitney Whitney Rolls-Roy Rolls-Royce ce Hamilt Hamilton onSSundstrand undstrand System of Systems 4 Modern Gas Turbine Optimization is an Exercise in Managing Complexity ~ 80,000 PARTS ~5000 PART NUMBERS ~ 200 MAJOR PART NUMBERS REQUIRING 3D FEA/CFD ANALYSIS ~ 5000-10,000 PARAMETRIC CAD VARIABLES DEFINE MAJOR PART NUMBERS ~ 200 MAN-YEAR ANALYTICAL DESIGN EFFORT ~ 200 MAN-YEARS DRAFTING / ME EFFORT 5 Requirements Management Output Company Program Job Ticket System SRD Module CRD Part PRD Requirements Flow to 3 Levels Job Ticket or Contract => Requirements Job Ticket measures compliance to requirements Program Job Ticket System Parameter Product / Service Solution Performance System Weight Efficiency Reliability Roles “Activities” Deliverables System = Part I Operability Augmentor Module Fan = Part II Observability Cost Maintenance Cost Durability Part Fan Blade = Part III System Analysis & Optimization Execution THERMALS Optimization Simulation CAD/CAM Y1 Computational Systems Engineering DRAFTING MFG A C E B D Complex Designs are Inherently Iterative & Bounded CAD MODEL PHYSICS MODEL DESIGN SPACE WORK INSTRUCTIONS - NON LINEAR - MULTI MODAL - DISCONTINUOUS - NOISEY - HIGHLY CONSTRAINED DESIGN STANDARD WORK CRITERIA VALIDATED ANALYSIS DECISION Manual Iteration 100s-1000s of Times PREFERRED CONFIGURATIONS 9 Sophisticated Simulation Based Design Systems 10 …and Complex Designs are Iterated Across Disciplines & Organizations… FEED FORWARD IPT PROCESS CYCLE FEEDBACK 1D AERO COOLING FLOWS 3D STAR AERO PROSTAR 3.00 23-SEP-97 VIEW 1.000 1.000 1.000 ANGLE 0.000 DISTANCE 6.549 CENTER 10.138 -0.554 0.776 EHIDDEN PLOT HEAT X DESIGN PLATFORM NECK ATTACHMENT DISK & SEALS X Y Z 11 . . . And Iterations Can Take Place Across the Globe • OUTSOURCING • INTER-DIVISIONAL • PARTNERSHIPS • CUSTOMERS 12 Gains are Being Made by Shifting from “Human” to “Computer” Based MDO “HUMAN” BASED “COMPUTER” BASED Program IPMT Program Standard Work Flow 0 Business Plan Concept & Venture Definition I Integrated Business & Project Plan Product/ Industrial Plan Execution & FETT II III Validate, Certify, Deliver IV EIS, Operational Service & Support WORKFLOW, RULES, And DESIGN ITERATIONS AUTOMATED WITHIN And ACROSS SYSTEMS & DISCIPLINES V System CIPT CIPT CIPT Select Concept Program Launch After FETT Release to Production 0 Concept Optim ization I Prelim inary Design 2 Product Design Procurement & Initial Validation (FETT) 3 Validation/ Certification 4 Airplane Validation Service & Support 5 IPT Concept Optim ization I Prelim inary Design 2 Product Design Procurement & Initial Validation (FETT) 3 Validation/ Certification 4 Airplane Validation Service & Support 5 I Prelim inary Design I2 Product Design Procurement & Initial Validation (FETT) 3 Validation/ Certification Airplane Validation Service & Support Concept Initiation 3 4 IPT 1 1 1 1 1 2 2 2 2 2 3 3 3 3 3 4 4 4 4 4 IPT 1 1 1 1 1 2 2 2 2 2 3 3 3 3 3 4 4 4 4 4 IPT NON-ANALYTICAL 2 IPT NON-ANALYTICAL 1 IPT NON-ANALYTICAL LEVELS OF FIDELITY Module Concept Initiation Part 1 1 1 1 Concept Initiation Concept Optim ization 2 2 2 2 2.5 4 5 Engineering Standard Work Flow 3 3 3 3 4 4 4 4 -- LABOR INTENSIVE -- SYSTEM SUB-SYSTEM A SUB-SYSTEM B MANUAL WORK FLOW per PROCESS MAPS AUTOMATE WORKFLOW MANUAL CAD/CAE MODEL BUILDING AUTOMATE MODEL BUILDING & EXECUTION MANUAL EXPLORATION TO FIND OPTIMAL DESIGNS AUTOMATE DESIGN EXPLORATION 13 Large Scale Computer Based MDO is Already Practical 3D Aero-Vibratory Shape Optimization Of A Cooled Turbine Airfoil (Single Row RANS CFD, Cooled UG Parametric Model, 3D ANSYS Vibes) AIRFOIL SHAPE 3D AERO CFD PARAMETRIC CAD MESH OPTIMIZER VIBRATORY ANALYSIS CAMP BELL DI AGRAM STRESS MODE 1 EFFICI ENCY MODE 3 MODE 4 Aero Xsections 0.5 0.4 Area MODE 2 0.3 Initial 0.2 Iteration 515 0.1 0 AreaS1 AreaS2 AreaS3 AreaS4 AreaS5 14 Large Scale Computer Based MDO is Already Becoming Practical 3D Shape Optimization Based On Hybrid Genetic Algorithm & Rule System (3D RANS Multi Row CFD, Population Size 80, Total Runs 2400, Run Time 48 hrs on 40CPUs) DELTA TURBINE EFFICIENCY LOSS CONTOURS 150 VARIABLES 15 CONSTRAINTS 0.6 0.4 LOSS CONTOURS Discovered “bowed” rotor To control tip leakage Vortex 0.2 0.0 0 10 20 GENERATION 30 15 Will Enable New Design Paradigms CUSTOMER NEEDS UNDERSTAND THE FUTURE CREATE TECHNOLOGY ENGINEERS COMPUTER BASED DESIGN RUN 24/7 365 DAYS A YEAR CONTINUOUS DETAILED DESIGN IMPROVE MODELS RE-FORMULATE PROBLEM SOLVE ALL POSSIBLE APPLICATIONS @ TECHNOLOGY READINESS LEVEL UPGRADE COMPUTER BASED DESIGN “MACHINE” CUSTOMER REQ. EXCEED TECHNOLOGY LESS TIME FEWER PEOPLE 16 Verification - Convergence to Requirements Value (e.g.. Weight) Technical Performance Measurement Tracking Chart Tolerance Band Planned Profile) (±x s) Actual Profile NTE (Not to Exceed) Goal Milestones Time Convergence to Requirements Dry engine weight (lbs) 5450 5400 5424 Commitment 5400 lb 5350 5338 -29 lb 5318 5300 5286 5289 Average Compliance EIS Projected 5250 Compliance 1 Compliance 2 Status Entry Into Service >100 lbs Below Commitment 18 Generic 2 Spool Gas Turbine Engine Diagram ITT N1 N2 PB Source: Wikipedia commons EGT Putting Rigor into System Requirements with Requirements Modeling The classical “paper” based method for Systems Requirements Picture Source: Dr Peter Hoffman – IBM / Rational The classical “paper” based method for Systems Requirements Picture Source: Dr Peter Hoffman – IBM / Rational Requirements Requirements are explicit contracts between the system element that consumes a product feature and the system element that provides it. • There are two major types of requirements. – Product Requirements “the system shall” – Statement of Work (SOW) Requirements “the contractor shall” • Product Requirements specify something the product must do or a quality the product must have. – “The engine shall generate up to 20000 pounds of thrust during engine operation.” Focus on Product Requirements Product Requirements • Product Requirements further classified as: • Functional Requirements specify a task or activity the system must perform & its duty cycle. • Performance Requirements specify a constraint on how the system should perform a functional task. Performance Requirements linked to Functional Requirements Modeling Overview • Models are abstractions that allow us to focus on a solution to a particular problem. Abstractions are essential to managing complexity. • Abstractions can be layered • accurately represent essential content • high fidelity and still remain simple. – The key to managing layers is to control the complexity of both the layer and its interfaces to other layers. – Push the details as low as possible but keep the essential meaning at all levels. Keep each layer simple & push the details down Models have different purposes • Functional Modeling – Logical relationships between activities and sequences in time • Parametric Modeling – Extends Functional Modeling to include equations or models of constraints on physical and functional elements. Data/Results may be collected by repeated computer runs in time-domain or thru a separate Monte Carlo analysis. • Dynamic Modeling – Focus is on mathematical representations of physical behavior of system or subsystem components. This may or may not be time domain. • Business / Economic Modeling – Focus is on cost and schedule Connect the network of models together Functional Modeling – Activity Diagrams / Tasks and Control Flow Activity Diagram Start, Stop, Operate Engine Functional Modeling • Requirements Modeling elucidate functional product requirements and their inter-relationships • It is designed to catch situations like the following • Page 257 states “The valve shall be on” when yyy. • Page 5205 states “The valve shall be off” when zzz • But the yyy and zzz conditions overlap, so the valve has to be both on and off at the same time. • State space of the system based on an analysis of the system requirements. ON or OFF but not both Modeling – Overview - Parametric Modeling Distiller - SYSML Parametric Model 28 Modeling – Overview - Dynamic Modeling Rocket Engine – SIMULINK model 29 Modeling – Form of model should match purpose – – – – – SYSML is ideal for functional modeling UML is ideal for Software Architecture and Design MATLAB / SIMULINK is ideal for Control System work. NPSS is ideal for Aerodynamic simulations. Mathmatica is ideal for symbolic calculations and mathematics. – Minitab is ideal for statistical calculations. – Microsoft Excel is also a modeling tool! Pick the model to match the problem What is SYSML? UML not required by SysML UML reused by SysML INCOSE slide – from tutorial by “The Aerospace Corporation” UML • SysML: • • Reuses a subset of UML 2.0 Uses UML 2.0 profile mechanisms to specify extensions for SysML UML4SysML SysML SysML extensions to UML (Have no counterpart in UML or place UML constructs) SysML tailored for Systems Engineering 31 Example Drawings for a Functional Modeling using SYSML • Use Case Diagram – captures system or subsystem scope • Activity Diagram – captures tasks and control flow • Internal Block Diagram – captures system structure and interfaces • Sequence Diagram – captures details of interactions between system and external actors. • State Diagram – details states and modes of system. To be shown in demo The Harmony Mini Cycle – as we use it Draw / Modify Use Case Diagram Generate/Modify Internal Block Diagram Draw High Level Activity Diagrams Add parameters, attributes, and messaging Annotate and finalize message sequences Create Ports, Interfaces, and Links Draw Detailed Activity Diagrams Draw State Charts Generate Sequences Animate and Execute To be shown in demo Family of use cases – highest level system description Use Case Diagram – Gas Turbine Engine – family of use cases What does the engine need to do? Lets zoom in so we can read the diagram Pick a particular use case – Operate Engine Engine Startup and Shutdown – Problem Overview • To start a gas turbine engine: – Turbine Rotation established by Air Starter Subsystem – driving generator for power and pressure for pumps – Fuel flow is enabled. – Proper Fuel / Air mix is established in combustion chamber. – Electrical spark from Ignition - Subsystem starts combustion. – Conditions monitored for automatic restart if necessary. – Controlled ramp increases fuel flow per schedule to achieve stable idle • Cockpit switch semantics rationalized with standard signals & start sequence Generic 2 Spool Gas Turbine Engine Diagram ITT N1 N2 PB Source: Wikipedia commons EGT Use Case Diagram for “How to Start a Jet Engine” Hit a button and… Internal Block Diagram – formal system interfaces FADEC Control System Engineer draw connections Functional Modeling – Activity Diagrams / Tasks and Control Flow Level 1 Activity Diagram Start, Stop, Operate Engine Level 2 Activity Diagram – Start_Engine Draw flow chart Sequence Diagram – sequence and content of interactions Wizard reads flow chart and assists in developing sequence State Diagram – states, modes, detailed logic End-result is executable code* Formal Methods can be applied * Simulation or control logic So… Where are the Requirements? • Model first from concept-of-operations information. – the model becomes the requirements! • The model then guides the writing of the requirements document. • Key model elements – activities, dialogs, states, will trace to explicit requirements paragraphs. Next Step - Verification • After creating an executable model, and writing requirements based on the model, the next step is to create formal test sequences • One way to do this is to create another “actor”, and connect this actor to the external actors of the model. The test sequencer drives particular tests by setting states • Animated sequence diagrams capture the results of the test • Book keep your work, linking requirement to test Systems Engineering Requirements models start with pictures Parametric Structural Functional 48 SysML models are visual Details of “Control_Is_Active” Details of Engine Startup