Dynamic classifier auditing by unsupervised anomaly detection methods: an application in packaging industry predictive maintenance

TL;DR

This work tackles predictive maintenance for packaging machines by augmenting a pre-trained Random Forest work-order classifier with a dynamic, unsupervised anomaly-detection audit in a streaming framework. It compares One-Class SVM and Minimum Covariance Determinant, including a voting ensemble, and implements a sliding-window online auditing scheme that uses the classifier output to detect anomalies without requiring extensive labeling. Across 23 machines with long histories, all AD methods improve the baseline F1 score, with the ensemble achieving the highest average improvement (≈192.5% relative to baseline), demonstrating strong practical value for ERP-integrated maintenance scheduling. The approach offers a simple, fast, and deployable mechanism to enhance precision and recall in predictive maintenance, with potential applicability to other industrial settings where labeled data are scarce.

Abstract

Predictive maintenance in manufacturing industry applications is a challenging research field. Packaging machines are widely used in a large number of logistic companies' warehouses and must be working uninterruptedly. Traditionally, preventive maintenance strategies have been carried out to improve the performance of these machines. However, this kind of policies does not take into account the information provided by the sensors implemented in the machines. This paper presents an expert system for the automatic estimation of work orders to implement predictive maintenance policies for packaging machines. The key idea is that, from a set of alarms related to sensors implemented in the machine, the expert system should take a maintenance action while optimizing the response time. The work order estimator will act as a classifier, yielding a binary decision of whether a machine must undergo a maintenance action by a technician or not, followed by an unsupervised anomaly detection-based filtering stage to audit the classifier's output. The methods used for anomaly detection were: One-Class Support Vector Machine (OCSVM), Minimum Covariance Determinant (MCD) and a majority (hard) voting ensemble of them. All anomaly detection methods improve the performance of the baseline classifer but the best performance in terms of F1 score was obtained by the majority voting ensemble.

Figures (6)

• Figure 1: Automatic packaging machine.
• Figure 2: Predictive maintenance expert system. The main contribution of this work is the improvement of the existing classification module by means of unsupervised AD techniques (AD module).
• Figure 3: Anomaly detection streaming procedure for on-line auditing of the classifer.
• Figure 4: Proposed architecture for the AD module.
• Figure 5: Auditing example for one of the machines using OCSVM. The output from the classifier and the AD model output are represented as different lines and the shaded part of each signal indicates if the threshold to call for a work order has been exceeded.