Using machine learning (ML) techniques in general and deep learning techniques in specific needs a certain amount of data often not available in large quantities in technical domains. The manual inspection of machine tool components and the manual end-of-line check of products are labor- intensive tasks in industrial applications that companies often want to automate. To automate classification processes and develop reliable and robust machine learning-based classification and wear prognostics models, one needs real-world datasets to train and test the models. The presented dataset consists of images of defects on ball screw drive spindles showing the progression of the defects on the spindle surface. The dataset is analysed via an initial object detection model available under: https://github.com/2Obe?tab=repositories. The reuse potential of the dataset lays in the development of failure detection and failure forecasting models for the purpose of condition monitoring and predictive maintenance. The dataset is available under https://doi.org/10.5445/IR/1000129520.

To realize autonomous production machines it is necessary that machines are able to automatically and autonomously predict their condition. Although many classical as well as Deep Learning based approaches have shown the ability to classify faults, so far there are no approaches that go beyond the basic detection of faults. A complete, image based predictive maintenance approach for machine tool components has to the best of our knowledge not been investigated so far. In this paper it is shown how defects on a Ball Screw Drive (BSD) can be automatically detected by using a machine learning based detection module, quantified by using an intelligent defect quantification module and finally forecasted by a prognosis module in a combined approach. To optimize the presented method, it is shown how existing domain knowledge can be formalized in an expert system and combined with the predictions of the machine learning model to aid quality of the prediction system. This enables the practitioner to forecast the severity of failures on BSD and prevent machine breakdowns. The work is intended to set new accents for the development of practical predictive maintenance systems and bridging the fields of machine learning and production engineering. The code is available under: https://github.com/2Obe/Pitting_Pred_Maintenance.

Machine learning methods have widely been applied to detect anomalies in machine and cutting tool behavior during lathe or milling. However, detecting anomalies in the workpiece itself have not received the same attention by researchers. In this article, the authors present a publicly available multivariate time series dataset which was recorded during the milling of 16MnCr5. Due to artificially introduced, realistic anomalies in the workpiece, the dataset can be applied for anomaly detection. By using a convolutional autoencoder as a first model, good results in detecting the location of the anomalies in the workpiece were achieved. Furthermore, milling tools with two different diameters where used which led to a dataset eligible for transfer learning. The objective of this article is to provide researchers with a real-world time series dataset of the milling process which is suitable for modern machine learning research topics such as anomaly detection and transfer learning.

Ensembles of convolutional neural networks have shown remarkable results in learning discriminative semantic features for image classification tasks. However, the models in the ensemble often concentrate on similar regions in images. This work proposes a novel method that forces a set of base models to learn different features for a classification task. These models are combined in an ensemble to make a collective classification. The key finding is that by forcing the models to concentrate on different features, the classification accuracy is increased. To learn different feature concepts, a so-called feature distance loss is implemented on the feature maps. The experiments on benchmark convolutional neural networks (VGG16, ResNet, AlexNet), popular datasets (Cifar10, Cifar100, miniImageNet, NEU, BSD, TEX), and different training samples (3, 5, 10, 20, 50, 100 per class) show the effectiveness of the proposed feature loss. The proposed method outperforms classical ensemble versions of the base models. The Class Activation Maps explicitly prove the ability to learn different feature concepts. The code is available at: https://github.com/2Obe/Feature-Distance-Loss.git.

A common challenge in real-world classification scenarios with sequentially appending target domain data is insufficient training datasets during the training phase. Therefore, conventional deep learning and transfer learning classifiers are not applicable especially when individual classes are not represented or are severely underrepresented at the outset. Domain Generalization approaches reach their limits when domain shifts become too large, making them occasionally unsuitable as well. In many (technical) domains, however, it is only the defect/ worn/ reject classes that are insufficiently represented, while the non-defect class is often available from the beginning. The proposed classification approach addresses such conditions. Following a contrastive learning approach, a CNN encoder is trained with a modified triplet loss function using two datasets: Besides the non-defective target domain class (= 1st dataset), a state-of-the-art labeled source domain dataset that contains highly related classes (e.g., a related manufacturing error or wear defect) but originates from a (highly) different domain (e.g., different product, material, or appearance) (= 2nd dataset) is utilized. The approach learns the classification features from the source domain dataset while at the same time learning the differences between the source and the target domain in a single training step, aiming to transfer the relevant features to the target domain. The classifier becomes sensitive to the classification features and – by architecture – robust against the highly domain-specific context. The approach is benchmarked in a technical and a non-technical domain and shows convincing classification results. In particular, it is shown that the domain generalization capabilities and classification results are improved by the proposed architecture, allowing for larger domain shifts between source and target domains.