Discussions on electrification, air pollution control and driving bans in inner cities bring major challenges for powertrain development. Real Driving Emissions (RDE), Worldwide Harmonized Light-Duty Test Procedures (WLTP) and the next level of CO2 reduction enforce new development methods. At the same time, new measurement technology and better IT infrastructure mean that ever larger amounts of data are available. Thereby, methods of digitization, e. g. Machine Learning, may be used in automotive development. Another challenge arises from the ever-increasing number of vehicle variants. Many OEMs reduce the number of their engines to reduce costs. However, the basic engines are then installed with little hardware customization in numerous vehicle models. As a result, the application of derivatives and the systematic validation of an application play an important role.
The book will expand on the topics discussed in the precursors entitled „DoE in Powertrain Development“ with the related areas of „machine learning“ and „big data“. Now it its ninth outing, it will thus be a forum on which to critically engage with the future challenges of the digital revolution. Real driving emissions (RDE), worldwide harmonized light-duty test procedures (WLTP) and the next round of CO2 guidelines all demand ongoing technical refinement of the drive train. The combination of changed environmental requirements, stricter limit values and new measurement techniques additionally require changes to existing processes and the development of new methods. To reduce costs, many OEMs are scaling down the size of their engine ranges. A small number of standard engines are then installed in numerous vehicle models with minor hardware modifications. The result is an increased focus on the use of derivatives and the systematic validation of an application. Contents: Machine learning and artificial intelligence for engine calibration - Big Data and Machine Learning Made Easy - Automated Calibration Using Simulation and Robust Design Optimization Improving Shift and Launch Quality of Automatic Transmissions - Development of a Simulation Platform for Validation and Optimisation of Real-World Emissions - Implementation of data-based models using dedicated machine learning hardware (AMU) and ist impact on function development and the calibration processes - The Global DoE Model Based Calibration and the Test Automation of the Gasoline Engine - Optimization of ECU Map Sampling Point Values and Positions with Model-Based Calibration - Dynamic Route-Based Design of Experiments (R-DoE) - System for Real-time Evaluation of Real Drive Emission (RDE) Data - System optimization for automated calibration of ECU functions - Dynamic MBC Methodology for Transient Engine Combustion Optimization - Implementing a real time exhaust gas temperature model for a Diesel engine with ASC@ECU - Dynamic Modelling for Gasoline Direct Injection Engines - Excitation Signal Design for Nonlinear Dynamic Systems - Application of a DoE based robust design process chain for system simulation of engine systems - Application of Emulator Models in Hybrid Vehicle Development - Fast response surrogates and sensitivity analysis based on physico-chemical engine simulation applied to modern compression ignition engines - The Connected Car and ist new possibilities in ECU calibration - Processing vehicle-related measurement data - On the selection of appropriate data from routine vehicle operation for system identification of a diesel engine gas system - Data Plausibility at the Engine Test Bench: How im-portant is the Human Factor in the Process? - Non-Convex Hulls for Engineering Applications - Modern Online DoE Methods for Calibration: Constraint Modeling, Continuous Boundary Estimation, and Active Learning - Model-based iterative DoE in highly constrained spaces - Approach for Automated Adjusting of the Road Load and Tire Simulation on Powertrain Test Beds
Developing complex powertrains without model-based processes is inconceivable today. But which methods can be used to address future challenges in powertrain development such as real driving emissions (RDE) and the diversity of derivatives? Contents: A strategy to employ criteria for online location selection when using Gaussian Processes - Efficient online test plan calculation with integrated boundary detection for modern model based engine calibration - Benchmark Problem for Near Boundary Operation Control for Automotive Engine - Lookup Table Optimisation for Engineering Applications - A Dynamometer Dynamic Calibration Method for the Diesel Air-Path - Application of DoE Method on Synthetic Gas Bench for SCR studies in Early Stages of the Model-Based System Development Process - Application of Model Based Calibration to Mass Production Diesel Engine Development for Indian Market - Application of global model based calibration methodology to optimize a 2.3 litre diesel engine with SCR on WLTC cycle - Advanced Gaussian Process Modeling Techniques - Fast Engine Modelling Using On-Line Calibration Data - Steady State Calibration for Catalyst Heat-up Optimization on Gasoline Direct Injection Engine - Efficient variant calibration by automation using rules and dependencies - Efficient Calibration Process for Series Programme with Multiple Engine, Vehicle and Market Variants - Computing Optimized Calibration Maps including Model Prediction and Map Smoothness - DoE and beyond: the evolution of the model based development approach how legal trends are changing methodology - Vehicle and OEM generic HIL / SIL Model in the transmission development - Data-based Models on the ECU - Pattern recognition for classifying degradation states of lithium ion batteries - Simulation-Error Based Identification of Dynamic Calibration Models
The method Design of Experiments is well-established in today’s engine development process. This thesis is confirmed in the conference proceedings at hand, where experts of international automotive manufacturers and supply industry as well as various research institutes and universities present their current results and experiences concerning the topic. Contents: Direction of Model Based Engine Calibration – Challenges during the Broad Implementation of Model Based Methods in the Development and Calibration of ECUs – Use of Model-Based Calibration in the Test Cell – Design of Dynamic Experiments – Modelling of Transient Diesel Engine Emissions – Quasi-Stationary Measurement Strategies for Cylinder Charge Determination of a 6 Cylinder Gasoline Engine – Transient Measurement for Steady-State Calibration – Intelligent Calibration Tool: An Approach to Cover the Whole Calibration Process – Modelling Engine Operating Space for DoE Calibration Methods – Utilization of the Slow Dynamic Slope Methodology for the Calibration of the ECU-Functions »Air Charge Determination« and »Torque Prediction« in the Series Production – Automated ECU-Calibration. Example: Torque Structure of Gasoline Engine – Multi-Layer Global DoE for Automated Optimization of Shift Quality for Automatic Transmissions – Automatic Optimisation Strategies for Diesel Engine Calibration – Model Based Controller Calibration on Powertrains – Application of DoE to the Optimisation of Engines with Selective Catalytic Reduction (SCR) Systems – The Optimisation of Camshaft Bearing Friction as an Example of DoE Application in Engine Design – A Design of Experiments Approach to the Control of Chassis Dynamometer Testing Error – Demonstration of the DoE Process with Software Tools – Multi-Objective Constrained Optimization of Engine Maps – Developing a New Procedure for the Determination of Target-Aimed Mathematical Models – A Method to Combine Physical and Statistical Modelling Implemented for a Dynamic Air Charge Compensation – Methods for the Global Dynamic Measurement and Modelling of Combustion Engines – A Model-Independent Test Planning Method for Iterative Data Acquisition Combined with Structured Process Modelling – Efficient Test Bed Automation – Proposal for a Generic Interface between Test Bench Automation System and DoE Modelling Application
Bei der Qualitätskontrolle von Universalmotoren, die z. B. für Haushaltsgeräte verwendet werden, wird jeder Motor auf Fehlersymptome überprüft. Dazu wird eine Klassifikation durchgeführt, d. h. der geprüfte Motor wird derjenigen Klasse zugeordnet, der er am ähnlichsten ist. Der Klassifikator wird dazu mit einer sogenannten Entwurfsstichprobe angelernt. Sie besteht aus repräsentativen Vertretern der Fehlerklassen und der fehlerfreien Klasse. Der Bildung dieser Klassen kommt daher besondere Bedeutung zu, weil die Qualität der automatischen Prüfung entscheidend von ihrer Definition abhängt. Heutzutage wird die Klasseneinteilung von Prüfexperten durchgeführt. Sie fassen die auftretenden Fehler in Klassen zusammen, wobei diese Zusammenfassung weitgehend von ihrer subjektiven Einschätzung abhängt. Grundlage für ihre Einteilung sind meistens Geräusch und Vibration der Motoren. Ziel dieser Arbeit ist die Objektivierung der Klassenbildung. Wegen seiner besseren Reproduzierbarkeit wird als Diagnosesignal ausschließlich der Motorstrom verwendet. Zur Bildung der Fehlerklassen werden zwei Verfahren untersucht. Zum einen werden multivariate statistische Verfahren verwendet, und zwar die Faktorenanalyse, Visualisierungsalgorithmen und die Clusteranalyse. Mit diesen Verfahren kann eine Klasseneinteilung ganz ohne a-priori Wissen durchgeführt werden. Die zweite Möglichkeit ist die Erklärung der Signalveränderung im Fehlerfall. Für die Fehler „loser Kommutatorhaken“, „Lamellen-“ bzw. „Windungsschluß“ und die Luftspaltfehler „Exzentrizität“ und „unrunder Rotor“ werden mit Hilfe einer Modellbildung deterministische Anteile der Stromsignale erklärt. Dadurch werden aussagekräftige Merkmale ermittelt, mit denen die Klasseneinteilung verbessert werden kann. Die theoretischen Untersuchungen werden im sechsten Kapitel anhand einer Meßreihe an einem seriengefertigten Waschmaschinenmotor auf ihre praktische Tauglichkeit überprüft.
Design and Modelling: Methodology for an Outlier Detection in a Test Matrix – Rapid Hull Determination: A New Method to Determine the Design Space for Model Based Approaches – Using Distributed Computing to Improve Model-Based Calibration – An Efficient Way of Building a Global Model of a System at Equilibrium – Template Functions for On-line Test Control and Multistage Modelling – Dynamic Powertrain Calibration: Using Transient DoE and Modelling Techniques Applications Base Engine: Bowl Geometry Optimization Using a Combustion CFD and Design of Experiments (DoE) for Diesel Combustion and Emissions – DoE in the Early Stage of Base Engine Development – DoC: CFD-Planning for Designing a Pilot Injection System Applications Diesel Engine: A Rule-based Calibration Method for Diesel Engines – Online Adaptive DOEs and Global Modelling for Diesel Engines and DPF Calibration – A DoE Approach to Find the Effect of Biodiesel Oxidation on Engine Performance and Emissions – Model-Based Design of Experiments for Static and Dynamic Measurements of Combustion Engines Applications Gasoline Engine: Automated Gas Exchange Model Calibration Using an Online Optimizer Tool – D-Optimal Global Models for Gasoline Direct Injection Engines – An Optimal Efficiency Seeking Spark Advance Control Strategy – DoE Application within the Analytical Valve Train Development – Application of Advanced Modelling Techniques to the Calibration of Gasoline Engines with Direct Injection and Variable Valve Timing Requirements for the Engine Test Bench: A Framework for the Integration of the Test Bench Environment and the Engine Application Software into Matlab/Simulink – Model Based Calibration of ECUs Using a Highly Dynamic HiL Test Bench System – DoE: A Knowledge Based Approach Development Process: Future Calibration Processes and Methodologies at Toyota – Components for the Successful Engine Calibration at the Test Rig Using DoE Methods – DoE for the Calibration Engineer: Applications for Beginners to Experts – Integrated Powertrain Optimisation: A Case Study – Model Based Calibration Process at FEV – Rapid Calibration(R): The Efficient Development Process for the Parameterization of ECUs – Future Use of Computer-Assisted Engine Map Optimization Systems in Combustion Process Development DoE Influences on Function Development: DoE Model Driven Alternatives to Map-Based ECU Software Structures – Applying the ATI No-Hooks Technology to Improve ECU Software Testing
