Plant Modeling for Powertrain Control Design

Modelica Automotive WorkshopDearborn, MI

November 19, 2002

Dr. Larry MichaelsGM Powertrain Controls Engineering

Challenges in PT Control Design

• Control System Design starts late• Rush to develop algorithm and write code• Minimum opportunity to influence hardware• Sensors & actuators may already be determined• Cost is always a major factor

Model-Based PT Control Design Process

HWIL tests (Simucar) CalibrationProcedures

Control SystemConceptual Design

SoftwareEngineering

Testing&

Calibration

Concept

SoftwareSystem Engineering Analysis Control System Design Algorithm Models Plant Models Diagnostic modelsOther DeliverablesCalibration MethodsTest ScenariosTest CalibrationsRPC to validate concept

Use models to understand algorithmValidate S/W vs. algo. modelSoftware-in-the-loop tests

Control Strategy Enhancements

Control System Architecture Design

Implementation IssuesElaborated algo modelRefined PT modelsSensor/Actuator modelsDiagnosticsCalibration Procedure Cost modelsRobustness

Controller H/WEngineering

Sensor/ActuatorDesign

PowertrainDesignPT Plant

Models

ExistingAlgorithm Models

NewIdeas

SSTS

CTS

CTR

ADD

Control Algorithm Design Process

BuildPlantModel

Analyzeand

ChooseControlTheory

BuildPlant

ControlDesignModel

CalculateControl

Algorithm

Build Algorithm

Model

Types of Models• Plant - physical system that is a major PT element of interest in a

control problem, e.g., engine, transmission, EGR subsystem. It typically includes actuators, sensors, controller circuitry, wiring and environment

• Plant Model - physics-based or behavioral math model built to understand the problem and represent the dynamics of the system being controlled

• Plant Control Design Model - simplified plant model used to develop control algorithm via Control Theory, usually linearized model, possibly created from experimental data (System Identification)

• Algorithm Model - executable representation of an algorithm - equations, data flow diagram, state charts

Plant Modeling Issues• Plant models may take long to develop• Many different models needed

– Problem dependent– Fidelity and complexity

• No standard plant model architecture• Need to validate model vs. hardware• Desire to reuse models• Different tools being used• Sensors, actuators, signal delivery system important

Controller Hardware and SoftwareActuators Sensors

Powertrain Engine and

Transmission

Platform Interface

Instrumentation Service ToolsAssembly Plant

Powertrain Control System Block Diagram

Input Circuits

and HWIO

Compute Engine

Output Circuits

and HWIO

On/Off High Side

Pulse Width Modulated

On/Off Low Side

Operating System

Application SW

HWIO SW

Math Library

Crank Sensor

Throttle Position

Coolant Temp

O2 Sensor

Force Motor

Injectors

Ignition signals

Fuel Pump

ECM or TCM

Controller Hardware & Software

A/D Converter

Pulse Counter

Frequency measure

Plant Modeling Strategies

• Specific model for problem-at-hand• One-size-fits-all model• Reconfigurable models for levels of complexity• Causal vs. a-causal models

– Data flow diagrams– Topological models

Plant Modeling Tool Requirements• Co-simulation with other tools

– detailed engine & transmission models

– algorithm models

– sensor & actuator models

– calibration tools

• Capable of real-time simulation for HIL• Support for various levels of fidelity

– e.g., model order reduction

• Modular structure• Easily validated and correlated

Plant Modeling Tool Requirements

• Compatible with CM and version control tools

• Support for libraries

• Parametric configuration facilities

• Enterprise support– documentation, readability, portability

• Supports software-in-the-loop

• Scripting capability

• Web interface