Filters

Settings

Now showing1 - 3 of 3
  • Some of the metrics are blocked by your 
    consent settings
    Item type:Publication, Access status: with files ,
    Generalization of LSTM and CNN autoencoders for anomaly detection across orthogonal and longitudinal turning
    (Springer, 2026-07-14)
    Wu, Ya-Jing
    ;
    Kopp, Justin
    ;
    Zemke, Jens
    ;
    Götschel, Sebastian
    ;
    Volke, Pascal
    Machine learning models enable automated process monitoring in manufacturing and support adaptive quality assurance and predictive maintenance systems. Autoencoders have emerged as a promising approach to unsupervised anomaly detection in processes. However, their application to turning processes, particularly their generalizability across different process variants such as varying machine kinematics, cutting geometries, and process variability, remains largely unexplored. Existing studies mainly focus on controlled laboratory environments using single machine tools and standardized parameters, limiting industrial scalability. This work addresses this gap by systematically comparing long short-term memory (LSTM) and convolutional (CNN) based autoencoders developed in collaboration with two research projects. The models are trained and evaluated on two harmonized datasets derived from orthogonal and longitudinal turning processes. The datasets differ with respect to machine tools, kinematics, and cutting tools. Each model is trained on both its native process data and on the respective other dataset to evaluate cross-domain performance. The results on normalized reconstruction errors show that LSTM-based autoencoders yield higher temporal sensitivity, whereas CNN-based autoencoders exhibit greater reconstruction stability but reduced sensitivity to transient dynamics. Moreover, relative reconstruction errors are additionally evaluated in the original physical force units to improve process interpretability. This analysis reveals larger relative deviations for LSTM models due to their heightened sensitivity. Furthermore, the variability of the training dataset is identified as a key factor influencing out-of-distribution generalization performance. The findings demonstrate the generalizability of autoencoder based monitoring approaches across different cutting conditions and provide insights for data-driven and grey-box modeling frameworks in manufacturing.
    Publicationtype: Journal Article
      31  7
  • Some of the metrics are blocked by your 
    consent settings
    Item type:Publication, Access status: with files ,
    A novel data concept for cutting processes through comprehensive experimental setup enabling grey-box models
    (Elsevier, 2025-09-17)
    Schibsdat, Sebastian
    ;
    Wu, Ya-Jing
    ;
    Keunecke, Martin
    ;
    Baron, Sarah
    ;
    Höche, Daniel
    This paper presents a novel experimental setup for recording the cutting process of external longitudinal turning in situ, capturing data such as cutting force, vibration, acoustic emission (AE), rake face temperature, surface quality, and tool wear. The primary objective of this setup is to generate reliable, repeatable data for developing a comprehensive grey-box model. To achieve this, the measurements were partly automated. Additionally, a complete post-processing pipeline has been introduced to combine all relevant data. To validate the setup, external longitudinal turning experiments were conducted and two of these were analysed for sensitivity to tool wear. This was achieved by analysing the mean and variance of the resultant force FZ, resultant vibrations aZ and AE signal, in order to demonstrate the impact of tool wear on the measurement. The experiments were also used to train an autoencoder to analyse the process data.
    Publicationtype: Journal Article
      55  44
  • Some of the metrics are blocked by your 
    consent settings
    Item type:Publication, Access status: with files ,
    Improved machine learning based flank wear identification by incorporating 3D-height maps
    (Springer, 2025-09)
    Schibsdat, Sebastian
    ;
    Dege, Jan H.
    Coated cutting tools are very important in the field of machining, but predicting the temporal evolution of tool wear is still a major challenge due to the high complexity of the wear process. Although machine learning (ML) shows promise in this respect, traditional supervised learning models rely on labour-intensive and error-prone manual measurement of tool wear to generate training data, which is a process that is time-consuming and inefficient. This paper uses a machine vision approach that automates the identification of flank wear on coated indexable inserts for turning applications and extends these techniques by integrating 3D tool surface data via height maps into the input. For this purpose, images of indexable inserts are taken during external longitudinal turning of C45 + N and X5CrNi18–10 using a focus variation microscope. This enables a height map of the wear area to be generated. Image annotation is performed on the 2D images to maintain comparability with traditional methods. Furthermore, the images and height maps are automatically cropped to focus on the flank wear region. This reduces the class imbalance and allows the model to maintain higher image resolution, despite the limited input. Three different U-Net models are trained on this dataset: two with height maps and one without. A comparison of the three models focusing on prediction accuracy and learning efficiency shows that including height maps improves prediction performance and reduces the required dataset size for comparable accuracy.
    Publicationtype: Conference Paper
      45  15