Professional Documents
Culture Documents
and Control
International Federation of Automatic Control
June 19-21, 2013. Saint Petersburg, Russia
* LCFC, Arts et Métiers ParisTech, Centre de Metz, F-57070 Metz France (e-mail: Liaqat.Shah@ensam.eu).
** LGIPM, University of Metz, Ile du Saulcy, F-57070 Metz France (e-mail:
Francois.Vernadat@eca.europa.eu)
Abstract: A process-oriented risk assessment methodology is proposed. Risks involved in a process and the
corresponding risk factors are identified through an objectives-oriented risk identification approach and evaluated
qualitatively in the Process FMEA. The critical risks of the PFMEA are then incorporated in the process model for
further quantitative analysis employing simulation technique. Using the proposed methodology as a decision-making
tool, alternative scenarios are developed and evaluated against the developed risk measures. The risk measures values
issues out of simulation are normalized and aggregated to form a global risk indicator to rank the alternative
processes on the basis of desirability. The methodology is illustrated with a case study issued from parts
manufacturing but is applicable to a wide range of other processes.
Keywords: Risk Assessment, Decision Support, Manufacturing Processes, Process Modeling and Simulation
*
"# $ /#& %#& ' #&
01&
2. RISK MEASUREMENT MODEL
(4)
In the process perspective, uncertainty lies “internal” and &+,
“external” to the activity in the form of risk factors/sources Therefore, the global risk of the process path2 is given by
(Elmaghraby, 2005), which may trigger the occurrence of any Equation 5:
"34 56#789 " # =56#789 5&+, . %#& ' #&
*
undesirable events that ultimately affect the activity progress
towards objective attainment (cf. Figure 1). (5)
And probability : 2 of the path set (or scenario) is such
1*… 1*… 1*… 1*… that:
;
Risk factor Risk event Objective
$ : 2
Causes Influences
(6)
4+,
So, the expected risk of the processmade of < path sets is:
Figure 1: Risk factors and objectives relationship
; *
Therefore, it is essential to identify and assess the impact of
To quantify the process risks of the same nature, the
Figure 2: Activity exposure to risk factors Context is about defining external and internal parameters to
be considered when managing risk and setting the scope and
likelihood of all risk events multiplied by their corresponding risk criteria for the risk management policy (ISO Guide
consequences (same effect) on an individual activity are first 73:2009). In the framework of the current methodology, the
modeled using Equation 2: context establishment phase identifies application domain,
(2)
the stakeholders of the process, their roles and responsibilities
as well as the upper and lower bounds for each stakeholder’s
217
2013 IFAC MIM
June 19-21, 2013. Saint Petersburg, Russia
Spec.
Performance risk Schedule risk Cost overrun Scrap
Raw Ok
Activity i Activity i Inspection X
part
C
Order processing delay Mfg. delay Logistic risk
Rework
Resources
Figure 5: Excerpt from risk-embedded process plan model
Quality failures Disruptions Time estimation error
To analyze risk events of the process model quantitatively,
Figure 3: Risk events hierarchy risk measurement model described in section 2 is employed
Because risk events originate from risk factors, it is necessary in simulation experiment.
to find risk factors which are required in mitigation plans. To
this end, a simple question: “What causes the risk event to 3.3. Risk analysis via simulation phase
occur” brings forward the sources or factors of risk event
Quantitative risk evaluation requires quantitative data
under question. For risk events relevant to manufacturing
regarding risk event parameters which can be obtained from
delay, the possible risk factors are shown in Figure 4.
the historical records employing statistical analysis (Ahmed
Manufacturing delay et al. 2007). However, in the a priori evaluation of processes,
such data does not usually exist. If it exists, it is either not
enough or in the shape to be used. In such circumstances,
Quality failures Time estimation error Disruption simulation is the right tool that can generate sufficient data
regarding risk parameters that the simulation model use later
to estimate risk measures. To carry out simulation
Unskilled Improper Product/ Process Process Resource experiments for this purpose, input data for simulation is
operator machine/tool novelty complexity unavailability generated.
Figure 4: Risk events and corresponding risk factors Input data generation: The input data required for risk
analysis in the simulation environment can be divided into
Process-based risk analysis: The identified risks are then three categories: functional data (activities), parameters to
linked via risk/activity matrix to the activities whose make the conceptual process model executable and risk
objectives are influenced by the identified risks (cf. Table 1). evaluation methods. The former data is obtained in the
For instance, R5 represents the risk event “time estimation context establishment phase by process modeling of the
error” and is related to A1 and A3 due to their uncertain scenario using IDEF3 method. For the parameters, it is
execution times. Similarly, all identified risks are linked to advisable to divide them into several categories such as run
responsible activities of the process scenario. In the next step, parameters (job arrival law, arrival type: batch or single
they are analyzed qualitatively using Process FMEA entity arrival, warm-up period and so on) process plan
technique as shown in Table 2. parameters (i.e. operation times) and objectives related
parameters (i.e. order quantity) as described in (Shah et al.,
2012; Shah, 2012). Concerning the evaluation methods (see
218
2013 IFAC MIM
June 19-21, 2013. Saint Petersburg, Russia
Table 3), they are mathematical expressions for measures and attractiveness such as {null, very weak, weak, moderate,
are required to analyze the scenario. strong, very strong, extreme}.
Risk Measures Calculation Method commensurability issue and to map them on a OP Q scale. As
I
MACBETH relies on the weighted mean to aggregate risk
2 $ D EF G H
Cost overrun expressions, which is often not the case in real-life examples
+, where criteria may interact, the Choquet integral has been
I
chosen as the operator for the aggregation of risk expressions.
2 $ D .J. H
Schedule Risk It can handle interdependencies among different risk
+, expressions through Choquet capacity.
I
Choquet Integral: To aggregate the risk expressionsN , the
2 $ K L M G
Quality Risk following 2-additive CI mathematical model is used:
+, X X
R S $ T# U# V $ Y#& ZU# V U& Z
W
Simulation experimentations and output data collection: (8)
#+, #+,
Having defined the input data, the risk-embedded process
Where [ models vectors of risk expressions N , denotes a
model (conceptual model) is transformed into simulation
Shapely index with 5I+, that represents importance of
model. The simulation parameters and evaluation method are
risk expression N relative to all other risks expressions and
fed into the simulation model. In the probabilistic setup, the
\ , interaction between risk expressions (N N , ranging in [-
simulation model generates random data for each parameter
The situations where only one N = 1 and all others are equal
risk can only be established when compared against its target two risk expressions simultaneously (Clivillé et al., 2007).
value, upper and lower bounds. In addition, risk measures
to zero, the aggregated value is as follows:
I
out of simulation experiment are heterogeneous in nature and
"]^ V $+, \
W
most often large in numbers. Therefore, it is desirable to
consider multiple criteria decision-making technique(s) (9)
_
"]^ V V $ \
the other hand, set the stage for aggregation of the normalized
W +,_
risk measures (henceforth, risk expressions) to facilitate (10)
decision-making process.
4. APPLICATION
In the multi-criteria model of the current methodology, risk
measures are normalized using the MACBETH method. The methodology is illustrated on a case study about a
Determination of interaction among risk measures is carried manufacturing company that designs and fabricates product
out using Choquet capacity while aggregation is performed on make-to-order (MTO) basis.
using the 2-additive Choquet integral (2-additive CI). Loaded workpiece
clamped
MACBETH: MACBETH (Measuring Attractiveness by a Axle
Categorical Based Evaluation TecHnique) is a multi-criteria
decision analysis approach used to determine value functions Cap
219
2013 IFAC MIM
June 19-21, 2013. Saint Petersburg, Russia
The reference product for this case study is a mechanical Table 4: Excerpt of PFMEA for quality risks
locator (Figure 6), a work holding device used for centering
of job on a machine-tool. To illustrate the use of the proposed Proces Reqs. Failure Causes Effects P C D RPN
methodology within the company, a manufacturing scenario -s mode
is defined.
Op 02 ` cone not
Axle Improper Unable
4.1. Case study: Manufacturing Scenario assembly to hold
5 7 6 210
locate tolerance the WP
The company under study receives an order for 200 high
the WP allocation rightly
quality mechanical locators from a customer with a lead time
High RPN risks are integrated into the process plan model for
of two weeks (10 working days). The price of the product is
further analysis in the simulation environment. To calculate
kept at $14 /per unit. So, the most critical objective in the
the probability of a risk event, it is sufficient to observe and
scenario is technical performance of the product, measured in
record at a critical activity (whose risk is of concern) in the
terms of satisfaction index (q) (Anselmetti, 2008).
simulation model; the number of times the activity objective
The company has sufficient resources at its disposal, is not achieved divided by the total number of observations
P
f W
ji j
W
decomposed into geometrical features. To manufacture each
feature, process candidates are selected using inquiries to the Mfg. process plan 2
appropriate knowledge base of the manufacturing processes. Mfg. process plan 3 g
i P
je g
hj
Once the process plans are generated, they are then modelled
using IDEF3 modeling method. To obtain the risk expressions, first, two fictive situations
To identify risk events and the corresponding risk factors, representing upper and lower bounds for each measure are
objectives-oriented risk identification approach is used. A defined. Next, process plans and the two fictive situations are
preferences
i as follow:
global risk: “failure to satisfy customer order” for the ranked in the desirability order along with strength of
To determine risk parameters for the risk events, they are first Risk expressions N as well as the CI parameters ( & \ )
contextualized using activity/risk matrix. Once linked, they identified via Equations (9) & (10)are put in Equation (8)
are then analyzed in the process FMEA Table (cf. Table 4) and global risk for each process plan is calculated (Table 7).
220
2013 IFAC MIM
June 19-21, 2013. Saint Petersburg, Russia
Table 6: Risk expressions (normalized risk measures) Elmaghraby, S. E. (2005). On the fallacy of averages in
project risk management. European Journal of
Schedule Cost overrun Performance Operational Research, 165(2), 307-313.
risk risk Grabisch, M., Labreuche, C. (2009). A decade of application
P
W P
hd P
ii
of the Choquet and Sugeno integrals in multi-criteria
Mfg. process plan 1
decision aid. Annals of Op. Research, 175 (1), 247–286.
Mfg. process plan 2 P
gd P
ef P
fd ISO Guide 31000:2009, Risk management: Guidelines on
P
dd P
j P
ei
principles and implementation of risk management (ISO,
Mfg. process plan 3 Geneva, Switzerland).
ISO/IEC Guide 73:2002 Risk Management - Vocabulary -
Table 7: global risk indicator for mfg. process plans Guidelines for use in standards (ISO, Geneva,
q
rs \,b P
Ph , P
gj
process design in concurrent engineering projects. CIRP
Mfg. process plan 1 Annals - Manufacturing Technology, 56 (1), 167–170.
q
rt \,c P
g b P
d
Keeney, R. (1996). Value-focused thinking: A path to
Mfg. process plan 2 creative decisionmaking. Harvard University Press.
q
ru \bc P
Ph c P
jd
Larson, N., Kusiak, A. (1996). Managing design processes: A
Mfg. process plan 3 risk assessment approach. IEEE Transactions on
From Table 7, it can be concluded that manufacturing process Systems, Man, and Cybernetics- Part A: Systems and
plan 1 involves the least risk (P
eg) followed by process plan Humans, 26, 749–759.
Martinez, V. (2003). Understanding value creation: the value
P
eh
Furthermore, the technical performance being the
3 (P
ei) and process plan 2 with a maximum risk score of
matrix and the value cube. University of Strathclyde,
critical risk in the scenario carries more weight with a Glasgow, Ph.D. Thesis.
Shapley index of (P
jd followed by schedule risk (P
gj) and Moeller, R. (2007). COSO Enterprise Risk Management:
cost overrun (P
d). understanding the new integrated ERM framework, 1st
ed. Wiley& Sons, New-York.
5. CONCLUSIONS AND FUTURE WORK Mun, J., 2006. Modeling Risk: Applying Monte Carlo
Simulation, Real Options Analysis, Forecasting, and
In this paper, a process-oriented risk assessment methodology
Optimization Techniques. Wiley & Sons, New-York.
for manufacturing processes is presented. The methodology
Neiger, D., Rotaru, K., Churilov, L. (2009). Supply chain risk
establishes a risk assessment process to develop a global risk
identification with value-focused process engineering.
indicator that can be used a decision-making tool.
Journal of Operations Management 27(2), 154–168.
The methodology will be extended to model the causality Shah, L.A. (2012). Value-Risk based evaluation of industrial
relationship between risk factors and risk events via Bayesian systems. Arts et Métiers ParisTech, PhD Thesis.
network. In addition, cost effective risk mitigation strategy Shah, L., Etienne, A., Siadat, A., & Vernadat, F. (2012).
will be incorporated. A risk acceptability zone vis-à-vis value (Value, Risk)-Based Performance Evaluation of
will also be defined to make the proposed performance Manufacturing Processes. In Information Control
framework more robust for decision-making. Problems in Manufacturing, 14 (1), 1586-1591.
Sienou, A., Lamine, E., Pingaud, H. (2008). A method for
REFERENCES
integrated management of process-risk. Proceedings of
Ahmed, A., Kayis, B., and Amornsawadwatana, S. (2007). A GRCIS 17.
review of techniques for risk management in projects. Sormaz, D.N., Khoshnevis, B., 2003. Generation of
Benchmarking: An International Journal, 14(1), 22-36. alternative process plans in integrated manufacturing
Anselmetti, B. (2008). Manuel de tolérancement: Volume 2, systems. Journal of Intelligent Mfg. 14(6), 509–526.
Bases de la cotation fonctionnelle. Hermès Science Tixier, J.G., Dusserre, O. Salvi, D. Gaston. (2002). Review of
Publications, Paris (in French). 62 risk analysis methodologies of industrial plants.
Aven, T. (2008). Risk analysis: assessing uncertainties Journal of Loss Prevention in the Process Industries, 15
beyond expected values and probabilities. Wiley& Sons, (4), 291–303.
New-York. Parnell, G., Hughes, D.W., and others (2013). Invited
Costa, C. A. B. E. and Vansnick, J. C. (1999). The Review—Survey of Value-Focused Thinking:
MACBETH approach: basic ideas, software, and an Applications, Research Developments and Areas for
application (pp. 131-157). Springer Netherlands. Future Research. Journal of Multi-Criteria Decision
Cliville, V., Berrah, L. and Mauris, G. (2007). Quantitative Analysis, 20 (1). 49–60.
expression and aggregation of performance Vernadat, F., Shah, L., Etienne, A., and Siadat, A. (2013).
measurements based on the MACBETH multi-criteria VR-PMS: a new approach for performance measurement
method. International Journal of Production Economics, and management of industrial systems. International
105 (1), 171-189. Journal of Production Research, 51, doi: 10.1080/
00207543.2012.752593.
221