Project Abstract: This project aims to determine if unsupervised machine learning models, trained solely on sounds of normally operating equipment, can effectively detect mechanical failures. The research utilizes the ToyADMOS and MIMII datasets to train k-Nearest Neighbors (k-NN) and Autoencoder models using Mel-frequency cepstral coefficients (MFCCs) for feature extraction. DMP Content: This document outlines the data management strategy for the ASDTM project, covering: Data Description: Details on reused datasets (ToyADMOS, MIMII) and produced datasets (Anomaly Detection Scores, Source Code). Documentation: Metadata standards, file naming conventions, and version control via GitLab. Storage & Backup: Usage of TUcloud and TUgitLab for secure storage during the research process. Preservation: Plans for long-term preservation (10+ years) in the TU Wien Research Data repository.