Boundary conditions for the application of machine learning based monitoring systems for supervised anomaly detection in machining

Aug 3, 2026

Upload by Ivy

Monitoring systems improve machine tool availability and enable timely detection of process deviations. While machine learning has been applied across many machining monitoring tasks, boundary conditions that define when supervised anomaly detection approaches are principally applicable have not been comprehensively documented. The paper extracts key objectives and shortcomings of existing literature and details factors that affect monitoring quality. It then derives practical boundary conditions and discusses implementation challenges for successful deployment in industrial settings.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages6

Reading Time15 min

AccessFree to download

Abstract

Introduction

Monitoring system concepts

Monitoring approach types

Data acquisition, processing, and feature generation

Fault data and anomaly detection assumptions

Supervised anomaly detection in machining

[Available online at www.sciencedirect.com](Available online at www.sciencedirect.com) ScienceDirect Procedia CIRP 118 (2023) 519–524 16th CIRP Conference on Intelligent Computation in Manufacturing Engineering, CIRP ICME ‘22, Italy Boundary conditions for the application of machine learning based monitoring systems for supervised anomaly detection in machining B. Denkenaa, M. Wichmanna, H. Noskea,*, D. Stoppela aInstitute of Production Engineering and Machine Tools, An der Universität 2, 30823 Garbsen, Germany * Corresponding author. Tel.: +49(0)511-762-5997; [E-mail address:](E-mail address: noske@ifw.uni-hannover.de)[ noske@ifw.uni-hannover.de](E-mail address: noske@ifw.uni-hannover.de) Abstract Monitoring systems may contribute increasing the availability of machine tools and detecting process deviations in time. In the past, machine learning has been used to solve a variety of monitoring problems in machining. However, boundary conditions for the assessment of the principal applicability of machine learning approaches for supervised anomaly detection in machining have not been exhaustively described in the literature. In this paper, objectives as well as deficits of literature approaches are identified and influencing factors on the monitoring quality are described. As a result, we derive boundary conditions and discuss challenges for successful implementation of machine learning based monitoring systems for supervised anomaly detection in industrial practice. © 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ([https://creativecommons.org/licenses/by-nc-nd/4.0](https://creativecommons.org/licenses/by-nc-nd/4.0)) Peer-review under responsibility of the scientific committee of the 16th CIRP Conference on Intelligent Computation in Manufacturing Engineering Keywords: Machine learning; Machining; Monitoring; Quality assurance 1. Introduction Monitoring systems are used to prevent machine component damages during machining and to ensure compliance with quality requirements [1] . Achieving of the defined quality requirements is necessary for the subsequent functionality of the manufactured workpiece [2] . A distinction is made between process and condition monitoring. In process monitoring, the tool condition, the workpiece surface quality and the chip condition are monitored. In addition, chatter vibrations are detected [3-6] . The subject of condition monitoring is the detection of damages of machine components such as linear drives or spindles, which in turn impact the workpiece quality [1] . Monitoring approaches are divided into continuous or intermittent. Continuous monitoring takes place in parallel with the manufacturing process. In contrast, intermittent monitoring is carried out at intervals [10] . Furthermore, direct and indirect monitoring approaches are distinguished. In the case of direct monitoring, physical condition variables are monitored directly using a suitable sensor. In indirect monitoring, auxiliary variables such as machine-internal signals from the machine control system and external sensors (dynamometer, acoustic emission, acceleration) are utilized to evaluate conditions [3- 5] . Sensory components such as sensory spindles, sensory tool holders and sensory workpiece holders have also been developed for process monitoring [7,8] . Indirect monitoring is used in applications where an explicit determination of machine and process conditions is time consuming or impossible due to the nature of the process [2,9] . After data acquisition and signal processing, segmentation and feature generation are performed [3-5] . Feature generation is often necessary to perform evaluations to cope with high sampling rates of sensors like acoustic emission [6] . When selecting the monitoring approach, the presence of fault data must be taken into account. In the context of semisupervised anomaly detection, it is assumed that o

FAQs

Why are boundary conditions important for machine learning monitoring systems in machining?

Because monitoring quality depends on when machine learning approaches are principally applicable. Existing literature does not exhaustively describe these boundary conditions for supervised anomaly detection.

What does the paper identify regarding current literature approaches?

It identifies objectives and deficits in literature approaches and explains influencing factors that affect monitoring quality.

How do supervised anomaly detection methods relate to labeled data in machining?

They rely on labeled datasets containing anomaly information; in machining, labels correspond to discrete condition classes or measured quantities describing tool or workpiece condition.

Aug 3, 2026

Upload by Ivy