Chalmers Open Digital Repository
Välkommen till Chalmers öppna digitala arkiv!
Här hittar du:
- Studentarbeten utgivna på lärosätet, såväl kandidatarbeten som examensarbeten på grund- och masternivå
- Digitala specialsamlingar, som t ex Chalmers modellkammare
- Utvalda projektrapporter
Forskningspublikationer, rapporter och avhandlingar hittar du i research.chalmers.se
Enheter i Chalmers ODR
Välj en enhet för att se alla samlingar.
Senast publicerade
- Retrieval-Augmented Generation for Sustainable Material Data Handling in Automotive Value Chain(2026) Tran, John; Larsson, DanielApplying large language models (LLMs) to industrial material data workflows has the potential to improve efficiency. However, conventional LLMs are limited by hallucinations, depend on proprietary training data, and are costly to update. This thesis explores Retrieval-Augmented Generation (RAG) as an alternative approach, in which an LLM generates responses grounded in a restricted, domain-specific corpus of documents and databases and provides source citations for them. The study is carried out in an industrial setting with two main data domains: an internal SQL materials database with tabular material properties, and a corpus of unstructured textual documents, including supplier documents, corporate standards, and environmental product declarations. A RAG system is developed that (1) indexes both textual and tabular data, (2) retrieves relevant chunks via dense vector search, and (3) generates source-grounded responses. The work investigates whether a RAG model that explicitly integrates both domains can outperform a baseline tuned for unstructured text and explores which tabular serialization format yields the most semantically informative embeddings for pretrained embedding models. To achieve this, we first constructed LLM-based pipelines to generate documentand table-based test sets with ground-truth chunk annotations, and implemented a modular RAG pipeline with separate indices for textual and tabular data. Then, we experimented with multiple retrieval strategies, ranging from concatenating the retrieval results to using cross-encoders to weigh them. In addition, several fusion strategies were tested to evaluate whether they could improve retrieval accuracy when operating across different domains. Experiments are conducted comparing nine tabular serialization strategies, studying performance as a function of index size, chunk size, and top-k, and evaluating different fusion modes and embedding models. The evaluation metrics used are Hit Rate, Recall, Precision, F1-score, and Mean Reciprocal Rank. Results show that enriched serialization, which converts tabular rows into natural-language statements, yields stronger tabular retrieval performance than a standard key-value-based format, without degrading performance on document retrieval. Larger chunk sizes and higher top-k values systematically improve retrieval metrics, highlighting both the difficulty of relying solely on similarity search and the benefits of cross-encoder reranking on larger candidate sets. A domain-aware weighted fusion retriever further improves overall retrieval performance over the optimized baseline with only moderate computational overhead. These findings demonstrate that semantically rich tabular representations and domain-aware fusion can enhance RAG performance on heterogeneous industrial material data.
- Drive Cycle Analysis for the Electric Powertrain in Terminal Tractors: A Data-Driven Approach to Performance Evaluation(2026) Magnusson, Fabian; Marjanovic, EdwinThe transition toward electrified powertrains in industrial vehicles places increased demands on understanding real world operational behavior. Terminal tractors operate under highly variable and transient conditions, making traditional standardized drive cycles insufficient for accurate performance evaluation and optimization. This thesis presents a data driven framework for analyzing, categorizing, and simulating operational drive cycles of electric terminal tractors based on real world field data. Multivariate time series data collected from electric terminal tractors were preprocessed and segmented into individual drive cycles using application specific operational signals. A set of interpretable features capturing both steady state and dynamic behavior was extracted for each cycle. Dimensionality reduction and feature selection were performed using Principal Component Analysis and Principal Feature Analysis to retain the most informative characteristics while maintaining interpretability. The results were that only 17 principal components out of the original 24 were needed to describe 95% of the explained variance. Multiple clustering techniques, including Hierarchical Agglomerative Clustering, K-Means, K-Medoids, and a convolutional autoencoder based clustering, were applied and evaluated using internal validation metrics. The resulting clusters revealed two distinct operational regimes, representative usage patterns, and outlying behaviors across the fleet. These two operational regimes were defined as low load and high load, where the low load cluster is defined by its lower variance and more stable values, while the high load cluster is defined by higher variance and a broader range of values in torque for example. Representative and atypical drive cycles from each cluster were subsequently integrated into a simulation model of the electric power train to evaluate component behavior, energy usage, and engine efficiency operation under different load configurations. The results demonstrate that data driven cycle analysis can effectively characterize real world usage patterns and provide valuable insights for how powertrain components, such as the engine and battery, are affected during operation. The proposed methodology offers a scalable framework for leveraging operational data to identify dominant operating regimes and guide structured evaluation of the electrical powertrain in terminal tractors.
- Cage-Encapsulated Metal Nanoclusters for Electrocatalysis(2026) Tingström, Ellen
- Din boendemiljö - dröm eller verklighet(1978) Enkel, NewtonUppsatsen analyserar relationen mellan arkitektur, samhällsstruktur och människors välbefinnande, med särskilt fokus på hur planering och bostadsutformning påverkar sociala relationer och livskvalitet. Den lyfter behovet av mer människocentrerade och hållbara miljöer som främjar gemenskap och funktionalitet.