Predictive Maintenance: An Autoencoder Anomaly-based Approach for a 3 DoF Delta Robot

Abstract

Performing predictive maintenance (PdM) is challenging for many reasons. Dealing with large datasets which may not contain run-to-failure data (R2F) complicates PdM even more. When no R2F data are available, identifying condition indicators (CIs), estimating the health index (HI), and thereafter, calculating a degradation model for predicting the remaining useful lifetime (RUL) are merely impossible using supervised learning. In this paper, a 3 dof delta robot used for pick and place task is studied. In the proposed method, autoencoders (AEs) are used to predict when maintenance is required based on the signal sequence distribution and anomaly detection, which is vital when no R2F data is available. Due to the sequential nature of the data, non-linearity of the system, and correlations between parameter time series, convolutional layers are used for feature extraction. Thereafter, a sigmoid function is used to predict the probability of having an anomaly given CIs acquired from AEs. This function can be manually tuned given the sensitivity of the system or optimized by solving a minimax problem. Moreover, the proposed architecture can be used for fault localization for the specified system. Additionally, the proposed method is capable of calculating RUL using Gaussian process (GP), as a degradation model, given HI values as its input.

Data mining in predictive maintenance systems: A taxonomy and systematic review

“…These records are extracted later or in real time to apply PdM. Examples of this case are found on the one hand in industrial processes (Bampoula et al, 2021; Bekar et al, 2020; Chang et al, 2021; Kim et al, 2021; Kolokas et al, 2020; Lepenioti et al, 2020; Li et al, 2017; Morariu et al, 2020; Rafique et al, 2018; Ruiz‐Sarmiento et al, 2020; Susto et al, 2015; Uhlmann et al, 2018; Zschech et al, 2019), in production lines (Ayvaz & Alpay, 2021; Azab et al, 2021; Cerquitelli et al, 2021; Fathi et al, 2021; Giordano et al, 2021; Liu et al, 2021), in power plants (de Carvalho Chrysostomo et al, 2020; Khodabakhsh et al, 2018; Sun et al, 2021; Zhang, Liu, et al, 2018), in wind turbines (Chen, Hsu, et al, 2021; Leahy et al, 2018; Santolamazza et al, 2021), in ventilation systems (Fernandes et al, 2020), cryogenic pumps (Crespo Márquez et al, 2020), heat meters (Pałasz & Przysowa, 2019), press machines (Serradilla et al, 2021), or water treatment plants (Srivastava et al, 2018). Another common scenario for PdM is related to transportation: different types of land vehicles (Chen et al, 2020; Patil et al, 2021; Prytz et al, 2015; Shafi et al, 2018), aircrafts (Baptista et al, 2021; Basora et al, 2021; Ning et al, 2021; Savitha et al, 2020; Yang et al, 2017), and naval ships (Berghout et al, 2021; Fernández‐Barrero et al, 2021; Gribbestad et al, 2021) have been monitored through their electronic control units.…”

“…In a second stage, the detected anomalous points are merged by a low-pass filter for subsequence outlier detection. Fathi et al (2021) also explored the CNN-based AE to perform novelty detection over the time series, previously fragmented for batch processing. The reconstruction error obtained from the AE is passed to a minimax optimized sigmoid function to infer the health index of the monitored system, that depends on if an anomaly has been detected or not.…”