Thesis worker - Physics-Aware Multiclass Root-Cause Classification of AC System Faults

30hp - Physics-Aware Multiclass Root-Cause Classification of AC System Faults 

Introduction

Thesis work is an excellent way to get closer to Scania and build relationships for the future. Many of today's employees began their Scania career with their degree project.

Background 

Reliable AC (Active Cooling) operation affects system availability, component life, energy consumption, cabin comfort and service cost.
This thesis focuses on refrigerant-loop faults whose physical symptoms can overlap:

•    low refrigerant charge and expansion-valve underfeeding can both starve the chiller;
•    expansion-valve overfeeding can create low superheat and liquid-return risk;
•    compressor degradation can alter pressures, discharge temperature and electrical power in ways that resemble charge or metering faults; and
•    normal pressures and temperatures vary with compressor speed, refrigerant state and thermal boundary conditions.

A classifier based only on absolute sensor thresholds may learn test conditions rather than fault physics. The observer-based and direct-sensor methods provide two ways to obtain information that can separate low refrigerant charge from expansion-valve underfeeding.

The observer reconstructs pre-expansion-valve state and refrigerant mass flow from coolant-side energy balances; the direct-sensor method measures pre-expansion-valve pressure and temperature. Because TXV (Thermostatic eXpansion Valve ) and EXV (Electrically controlled thermostatic eXpansion Valve) hardware also differ in control dynamics, fault mechanisms and available signals, the study evaluates both factors and their interaction.

Existing project work provides quantized refrigerant-property calculations and healthy references for discharge temperature and steady-state compressor power. These references express each deviation as the difference between measured behaviour and the expected healthy value at the same operating condition.

Objective 

1.    Establish an auditable label and data-quality protocol for all five classes on both TXV and EXV hardware.
2.    Build synchronized, qualified operating windows from laboratory and operational data.
3.    Implement common raw and physics-derived features plus valid, uncertainty-aware features specific to each state-information method and valve technology.
4.    Compare interpretable low-order baselines, nonlinear tree ensembles, flat classification and physics-guided hierarchical classification.
5.    Compare all four cases with matched methods and data partitions, using a common-base-signal model for each valve type as the lower-information baseline.
6.    Calibrate class probabilities and implement an operating-domain and confidence-based inconclusive result.
7.    Validate performance and diagnostic coverage by complete hardware unit and test session, and define the next-stage production-validation plan.

Job description 

The aim is to develop and rigorously evaluate a physics-aware multiclass classification framework for five refrigerant-loop health states across four study cases formed by two state-information methods (observer-based and direct-sensor) and two expansion-valve technologies (TXV and EXV).
Particular emphasis is placed on:

•    separation of low refrigerant charge from expansion-valve underfeeding;
•    separation of refrigerant-side faults from abnormal compressor performance;
•    diagnostic accuracy, availability and uncertainty of observer-based versus directly measured pre-expansion-valve information;
•    the effect of TXV versus EXV hardware and its interaction with the state-information method;
•    generalization to unseen hardware units and test sessions; and
•    calibrated uncertainty and abstention when a root cause is not identifiable.

Education/program/focus

Number of students: 1
Start date for the thesis work: 1st of February 2027
Estimated time required: 5 month

Contact persons and supervisors

Contact Name: Göran Lissel
Function: Head of TGRBTEA – Electrification Management control – Thermal & Aux
Phone: +46 8 553 51937  
Mobile: +46 70 086 0984
e-Mail: goran.lissel@scania.com

Supervisor Name: Karim Mohammed Razaul
Function: Software Development Engineer – Thermal & Aux
Mobile: +46767225137
e-Mail: mohammed.razaul.karim@scania.com

Application:

Your application must include a CV, personal letter and transcript of grades 
A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.     

Publication date from - to. 
2026-10-22 – 2026-11-22