IFAC-PapersOnLine 49-12 (2016) 544–549
ScienceDirect
Available online at www.sciencedirect.com
2405-8963 © 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2016.07.692
© 2016, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Keywords: Maintenance actions guidance, unscheduled maintenance duration, Bayesian network, FDC,
maintenance procedure.
1. INTRODUCTION
The semiconductor industry (SI) is characterized by high-mix
low-volume and a multi technology production environment,
challenged by short product life cycles, where demand comes
from end-user markets (Ballhaus et al., 2009). This results in a
highly competitive production environment where success
requires improved and stable production capacities. Literature
suggests that such complex environment often results in more
unscheduled breakdowns with longer durations (Bode et al.,
2007) that is an evidence of misdiagnosis by the technicians
(Abu-Samah et al., 2014). Consequently, inappropriate choice
of maintenance procedure results in poor preventive actions.
The FMEA (failure mode effect analysis) approach is used in
the SI to manage experts' knowledge about equipment failure
and causes. This serves as the basis for defining diagnostic and
maintenance procedures for unknown and known failures
respectively. The diagnostic procedures are merely guidelines
to carryout maintenance actions; however, choice depends on
the experience and intuitiveness of technicians. Misdiagnosis,
at this stage, can result in longer failure durations and impact
the future equipment behaviour. We must provide technicians
the decision support to choose appropriate procedures based
on the knowledge extracted from historical diagnosis efforts.
We argue that the FMEA with static knowledge must evolve
dynamically with emerging equipment behaviour (Mili et al.,
2009). This must be updated based on the lessons learned by
technicians in the diagnosis during unscheduled breakdown
events.
(Ison and Spons, 1996), (Moore et al., 2006) focused research
on characterizing and controlling variability in semiconductor
manufacturing equipment through advanced process control
(APC) methods like fault detection and classification (FDC).
The multiagent based approaches have emerged as one of the
most interesting techniques to schedule maintenance actions,
dynamically (Aissani et al., 2009). These approaches are well
suited to select appropriate procedure for the known failures
and are static in nature. They also do not offer any support to
the technicians to select the procedure during diagnosis upon
unknown failures.
The FDC (fault detection and classification) system monitors
the equipment through sensors and generates warnings and
alarms. These are used by the automation system to trigger the
Abstract: The Semiconductor Industry (SI) is one of the fastest growing manufacturing domains
challenged by the high-mix low-volume production and short product life cycles. This results an increase
in unscheduled equipment breakdowns that often result in decreasing and unstable production capacities.
The success in the SI depends on our ability to quickly recover from unplanned events. It is reported (Abu-
Samah et al, 2014) that misdiagnosis is one of the key reason for the extended failure durations. This is
because of the fact that existing procedures to support maintenance decisions for an equipment recovery
are often based on FMEA approach that represents static experts' knowledge. In this paper, we present a
methodology based on Bayesian network (BN) to advise technicians on the choice of maintenance
procedure in case of unscheduled breakdowns. We argue that the sequence of patterns and alarms as
generated by the equipment during production can be associated to the choice of maintenance procedure;
therefore, BN is learned as a function of these alarms and warnings to predict the choice of maintenance
procedure from unscheduled breakdown historical data. The set of warnings and alarms are grouped
together in the proposed methodology using hybrid approach where these are first clustered based on
distribution similarity followed by an experts' intervention to fine tune initial clusters. The main
contribution of the proposed methodology is to support technicians with advise on the choice of most likely
effective maintenance procedure that will reduce unscheduled breakdown period and help in improving
and stabilizing the production capacities. The proposed methodology is validated with a case study, from
the world reputed semiconductor manufacturer, using historical data. The results show 49% of gain in
terms of productive time from unscheduled breakdown periods.
Anis BEN SAID*’ **, Muhammad-Kashif SHAHZAD**, Eric ZAMAI**,
Stephane HUBAC*, Michel TOLLENAERE**
* STMicroelectronics, 850 Rue Jean Monnet 38926 Crolles, France
(anis.bensaid@st.com, stephane.hubac@st.com)
** Univ. Grenoble Alpes, G-SCOP, F-38000 Grenoble, France CNRS, G-SCOP, F-38000 Grenoble, France
(muhammadkashif.shahzad@g-scop.inpg.fr, eric.zamai@grenoble-inp.fr, michel.tollenaere@grenoble-inp.fr)
Towards proactive maintenance actions scheduling in the
Semiconductor Industry (SI) using Bayesian approach