A Comparison of Anomaly Detection Methods for Industrial Screw Tightening

Diogo Ribeiro,Luís Miguel Matos,André Pilastri,Guilherme Moreira,Paulo Cortez

doi:10.1007/978-3-030-86960-1_34

Diogo Ribeiro, Luís Miguel Matos + Show 3 more

Open Access

https://doi.org/10.1007/978-3-030-86960-1_34

Copy DOI

Publication Date: Jan 1, 2021
Citations: 5	License type: cc-by

Affiliation: University of Minho, Centro de Computação Gráfica

Abstract

Within the context of Industry 4.0, quality assessment procedures using data-driven techniques are becoming more critical due to the generation of massive amounts of production data. In this paper, we address the detection of abnormal screw tightening processes, which is a relevant industrial task. Since labeling is costly, requiring a manual effort, we focus on unsupervised approaches. In particular, we assume a low-dimensional input screw fastening approach that is based only on angle-torque pairs. Using such pairs, we explore three main unsupervised Machine Learning (ML) algorithms: Local Outlier Factor (LOF), Isolation Forest (iForest) and a deep learning Autoencoder (AE). For benchmarking purposes, we also explore a supervised Random Forest (RF) algorithm. Several computational experiments were held by using recent industrial data with 2.8 million angle-torque pair records and a realistic and robust rolling window evaluation. Overall, high quality anomaly discrimination results were achieved by the iForest (99%) and AE (95% and 96%) unsupervised methods, which compared well against the supervised RF (99% and 91%). When compared with iForest, the AE requires less computation effort and provides faster anomaly detection response times.

Highlights

The current competition market increases the pressure for industrial companies to improve their productive processes
The operator is presented with a “Good Or Fail” (GOF) result, which is calculated automatically by the assembly machine based on its internal configuration
As for the average AUC values, Isolation Forest (iForest) is significantly better than other methods, while AE reuse is significantly better than Local Outlier Factor (LOF) and AE reset

Summary

Introduction

The current competition market increases the pressure for industrial companies to improve their productive processes (e.g., increasing efficiency and reducing costs). Within this context, a key aspect is the reduction of assembly errors during the production of products. Following the Industry 4.0 revolution, most modern factories make use of automation and robots that are interconnected with data sensors. While fully autonomous robots are used in some production plants, there are still industrial tasks that require a human operator. Several companies use assembly machines that combine the flexibility of robotic arms with the guidance of human operators. During the screw tightening process, data is collected in real-time, generating several instances with multiple features, including angle-torque pairs. The defect is detected too late down the production chain, which results in extra production times and costs

Methods

Results

Conclusion