Gearbox-Oil-Conditioning-System

This master thesis performed data-driven development-cycle transformation and optimisation for Scania powertrain test rig department. More specifically:

• Developed systematic framework for testbed model estimation and gearbox oil controller design; Collaborated team decreased controller design timespan from 2 weeks to 1 week (0.25 FTE)
• Created testbed numeric model which was previously impossible
• Greatly improved controller performance with data and Machine Learning

Table of contents

• Introduction
• JupyterLab Folder
• Matlab_Folder
• Simulink_Folder
• TwinCAT_Folder
• Practice_Folder

Introduction

This master thesis contains two research questions:

RQ1: Is about the investigation of system identification of a Gearbox Oil Conditioning System.

RQ2: Is about the further controller implementation using pole placement methods via TwinCAT3.

5 different linear system identification models were used: ARX, ARMAX, BJ, TF, SS.

JupyterLab_Folder

The original recoreded data from the testbed are cleaned and the useful information are extracted into three separate dataset: (D_47414, D_48414, D_49404), which represents the system data at different valve operating range. In each dataset, it is then further splitted into training and testing dataset with size ratio 1:1.

SI_47414, SI_48414, SI_49404 are the system identification programs developed with JupyterLab in language Matlab for easier reading.

• DPP_System_Identification
• DPP_Torque
• DPP_Flow
• SI_47414 (± 30% operating range )
• SI_48404 (± 40% operating range )
• SI_49404 (± 50% operating range )

Matlab_Folder

Best_30, Best_40, Best_50 contains all the best model structure for each system identification models at ± 30%, ± 40%, ± 50% operating range. The best 3 system identification models based on simulation accuracy are selected for research question 2. (ARMAX, SS, TF)

• Best_30
• Best_40
• Best_50

Controller_ARMAX, Controller_SS, Controller_TF contains the controllers developed upon the top 3 models from research question 1.

• Controller_ARMAX
• Controller_SS
• Controller_TF

NRMSE_calculation calculates the NRMSE of the simulation

• NRMSE_calculation

Simulink_Folder

The simulink files are used for initial simulation of the system identification models as well as inspecting the controller performances.

• Simulink_ARMAX
• Simulink_SS
• Simulink_TF

TwinCAT_Folder

The developed controllers from Matlab are transformed into compatible version for TwinCAT3 environment.

• PI
• PI + Pole placement
• PID + Low Pass
• PID + Low Pass + Pole placement

Practice_Folder

The purpose for this folder is mainly for archiving past practicing files.

Contact

Chieh-Ju Wu (Jeremy) - [email protected]