Machine Learning for Predictive Maintenance: A Multiple Classifier Approach

Abstract

In this paper a multiple classifier machine learning methodology for Predictive Maintenance (PdM) is presented. PdM is a prominent strategy for dealing with maintenance issues given the increasing need to minimize downtime and associated costs. One of the challenges with PdM is generating so called 'health factors' or quantitative indicators of the status of a system associated with a given maintenance issue, and determining their relationship to operating costs and failure risk. The proposed PdM methodology allows dynamical decision rules to be adopted for maintenance management and can be used with high-dimensional and censored data problems. This is achieved by training multiple classification modules with different prediction horizons to provide different performance trade-offs in terms of frequency of unexpected breaks and unexploited lifetime and then employing this information in an operating cost based maintenance decision system to minimise expected costs. The effectiveness of the methodology is demonstrated using a simulated example and a benchmark semiconductor manufacturing maintenance problem.

Machine learning in manufacturing: advantages, challenges, and applications

“…ML has been successfully utilized in various process optimization, monitoring and control applications in manufacturing, and predictive maintenance in different industries (Alpaydin, 2010;Gardner & Bicker, 2000;Kwak & Kim, 2012;Pham & Afify, 2005;Susto, Schirru, Pampuri, McLoone, & Beghi, 2015). ML techniques were found to provide promising potential for improved quality control optimization in manufacturing systems (Apte, Weiss, & Grout, 1993), especially in 'complex manufacturing environments where detection of the causes of problems is difficult' (Harding, Shahbaz, & Kusiak, 2006).…”

Home Automation Oriented Gesture Classification From Inertial Measurements

“…2) DA [27] provides a linear or quadratic combination of features that separates the classes by exploiting the assumption that the conditional probability distributions are Gaussian [28]. 3) SVMs [29], [30] are binary classifiers that identify the classification decision boundaries by maximizing the distance, called margin, between the two classes; the objective function of SVM is a tradeoff between margin maximization and penalty on errors in the classification, which is governed by a parameter . SVM can also provide nonlinear solutions, by employing kernel methods [31]; a popular choice for defining the kernel transformation is the radial basis function (RBF)…”

A Bayesian network based learning system for modelling faults in large-scale manufacturing

“…Recent studies [2], [3], [8], [9] have investigated machine learning methods for modelling manufacturing domains with particular emphasis on predictive maintenance and fault detection. Rare event predictions and creation of fault prognostic systems using alternative methods have been covered in [9], [10].…”