Machine Tool Component Health Identification with Unsupervised Learning

Abstract

Unforeseen machine tool component failures cause considerable losses. This study presents a new approach to unsupervised machine component condition identification. It uses test cycle data of machine components in healthy and various faulty conditions for modelling. The novelty in the approach consists of the time series representation as features, the filtering of the features for statistical significance, and the use of this feature representation to train a clustering model. The benefit in the proposed approach is its small engineering effort, the potential for automation, the small amount of data necessary for training and updating the model, and the potential to distinguish between multiple known and unknown conditions. Online measurements on machines in unknown conditions are performed to predict the component condition with the aid of the trained model. The approach was exemplarily tested and verified on different healthy and faulty states of a grinding machine axis. For the accurate classification of the component condition, different clustering algorithms were evaluated and compared. The proposed solution demonstrated encouraging results as it accurately classified the component condition. It requires little data, is straightforward to implement and update, and is able to precisely differentiate minor differences of faults in test cycle time series.

International Conference on Advanced and Competitive Manufacturing Technologies milling tool wear prediction using unsupervised machine learning

“…The underlying principle used is density-based spatial clustering of applications with noise (DBSCAN), respectively a modification called hierarchical DBSCAN (HDBSCAN). Despite the encouraging results by Zhang et al [9], there are some inherent limitations. Their solution is capable of detecting deviations from a pre-defined state only.…”

“…Moreover, supervised approaches struggle with the detection and labeling of component behavior outside of the learned cases, as long as the model is not used exclusively for anomaly detection. Hence, noise, outliers, and inaccurate data have an unfavorable impact due to the inherent input-output relationship of supervised models [9], raising the requirements either for model engineering or for training data set size.…”

“…In this study, we describe an unsupervised clustering approach capable of separating wear data in such a way that meaningful estimations of the part condition are possible. It extends an approach previously proposed for the qualitative diagnosis of actual failure states [9] to the quantitative estimation of intermediate degradations. It uses time series signals of multiple sensors as an input, decomposes the time series into characteristic features, and uses the multidimensional space This study focuses on the steps related to data preprocessing, model building, and testing.…”

“…Additionally, the binary distinction between healthy and faulty states does not consider the various faults and their severity regarding the usability of the machine. In practice, not only the presence of anomalies but also the different kinds and their severity levels on machine tool components need to be captured, as described by Gittler et al [9]. Condition, failure mode or wear identification (2), and RUL models (3) manage to indicate status and reasons for deteriorations.…”

“…Run-to-failure data is usually not available or expensive to obtain, as faults are often the result of multiple simultaneous degradations and therefore not unidimensional. Besides that, other challenges are the data gathering and modeling effort to reconstruct the different faulty states, as well as the transfer of the derived model to different machine and component types [9]. This holds especially true for multiple operating conditions, each requiring an individual and corresponding data set for training, which is generally not available and cumbersome to acquire in appropriate quantities.…”