Received: 30 June 2017 Revised: 30 April 2018 Accepted: 16 July 2018

DOI: 10.1002/qre.2385

RESEARCH ARTICLE

Monitoring nonlinear profile data using support vector

regression method

Chung‐I Li | Jeh‐Nan Pan | Chun‐Han Liao

Department of Statistics, National Cheng

Kung University, Tainan, Taiwan, ROC
Correspondence
Jeh‐Nan Pan, Department of Statistics,
National Cheng Kung University, Tainan
70101, Taiwan, ROC.
Email: jnpan@mail.ncku.edu.tw
Funding information
Ministry of Science and Technology of
Taiwan, ROC, Grant/Award Number:
MOST 106-2410-H-006-007
Abstract
In today's manufacturing industries, if the quality characteristic of a product or
a process is assumed to be represented by a functional relationship between the
response variable and one or more explanatory variables, then the data gener-
ated from such a relationship are called profile data. Generally speaking, the
functional relationship of the profile data rarely occurs in linear form, and
the real data usually do not follow normal distribution. Thus, in this paper,
the functional relationship of profile data is described via a nonparametric
regression model and a nonparametric exponentially weighted moving average
(EWMA) control chart is developed for detecting the process shifts for nonlin-
ear profile data in the Phase II monitoring.
We first fit the nonlinear profile data via a support vector regression model and
use the fitted values to calculate the five metrics. Then, the nonparametric
EWMA control chart with the five metrics can be constructed accordingly.
Moreover, a simulation study is conducted to evaluate the detecting perfor-
mance of the new control chart under various process shifts using the out‐of‐
control average run length. Finally, a realistic nonlinear profile example is used
to demonstrate the usefulness of our proposed nonparametric EWMA control
chart and its monitoring schemes. It is expected that the proposed nonparamet-
ric EWMA control chart can enhance the monitoring efficiency for nonlinear
profile data in the phase II study.

KEYWORDS
metrics, nonlinear profile data, nonparametric EWMA control chart, support vector regression

1 | INTRODUCTION profile. The illustration of profile data collection process

In statistical process control, we usually apply control
charts to monitor quality of a process or a product by
using one or multiple quality characteristics. Recently,
rapid development of distributed sensing and computer
technology has facilitated a wide collection of data during
system operations. Generally speaking, the quality of a
process or a product characterized by a functional rela-
tionship between the response variable (y) and one or
more explanatory variables (x) is usually referred to as a
is shown in Figure 1.
Based on the data types, statistical methods for moni-
toring product and process quality profiles are commonly
classified into linear profile or nonlinear profile monitor-
ing. The process changes occurred in a functional rela-
tionship of the profile can be detected and classified by
the profile monitoring and control. During the past
decade, different methods have been proposed for profile
monitoring in both Phases I and II. In the Phase I study,
the parameters of the process are estimated based on a set

Qual Reliab Engng Int. 2018;1–9. wileyonlinelibrary.com/journal/qre © 2018 John Wiley & Sons, Ltd. 1
2 LI
of historical data and used to establish control limits for
Phase II monitoring. In Phase II, the data are sequentially
collected over time to assess whether the parameters of
the process have changed from the estimated values in
the Phase I study. In this paper, we focus our research
on the Phase II study for timely detecting the shifts/
changes in the process parameters. Due to the complexity
of parameters estimation involved in nonlinear profile
monitoring, Williams et al1 considered the use of non-
parametric monitoring methods and the use of metrics
to measure deviations from a baseline profile. After
applying spline smoothing techniques to model the verti-
cal density profile (VDP) data, they calculated the five
metrics and employed an individual Shewhart control
chart based on the moving range (I‐MR chart) to establish
control limits. However, according to Montgomery2 and
Pan et al,3 exponentially weighted moving average
(EWMA) and cumulative sum control charts are proved
to be more efficient than Shewhart control chart for
detecting the small sustained shift in process mean. Thus,
instead of using I‐MR chart, we propose a nonparametric
rank‐based EWMA control chart in Phase II monitoring.
Moreover, a more efficient support vector regression
(SVR) method is used to detect and classify the process
shifts in nonlinear profiles.
2 | L I T E R A T U R E RE V I E W
2.1 | Linear profile monitoring
Linear profile monitoring was studied by several
researchers, such as Kang and Albin,4 and Khedmati
and Niaki.5 Kang and Albin4 proposed two control chart
methods for Phase II monitoring of simple linear profiles
process. Their first method used a multivariate T2 control
chart to monitor the coefficients of regression. The sec-
ond one used EWMA and R control chart to monitor
the mean and variance of the regression residuals. Kim
et al6 proposed a method based on the combination of
three EWMA charts. They showed that their proposed
method is superior to those proposed by Kang and Albin.4
Khedmati and Niaki5 considered the effect of between‐
ET AL.
FIGURE 1 Profile data collection
process
profile autocorrelation of error terms in Phase II monitor-
ing. First, they proposed U statistic to eliminate autocor-
relation. Then, based on the adjusted parameter
estimation, a multivariate T2 control chart was designed
to monitor the parameters of the model. Finally, they
used average run length (ARL) to evaluate the detecting
performance of their proposed method. Woodall et al7
and Woodall8 provided a detailed review on the use of
control charts for monitoring process and product quality
profiles. Zi et al9 developed a nonparametric method (a
multivariate sign EWMA control chart) for monitoring
linear profile data in case where the normality assump-
tion is doubtful. Recently, Chiang et al10 investigated
another operational and sensitive Phase‐II monitoring
method for the linear profile model that exhibits within‐
profile autocorrelation.
2.2 | Nonlinear profile monitoring
Nonlinear profile monitoring, on the other hand, was
considered by other researchers, for instance, Walker
and Wright,11 Williams et al,1 Moguerza et al,12 and
Hung et al.13 Williams et al1 proposed using three differ-
ent T2 statistics for phase I analysis to monitor the coeffi-
cients resulting from a parametric nonlinear regression
model that was used to fit profile data. They also consid-
ered the use of nonparametric regression method and the
use of metrics to measure deviations from a reference
profile. Hung et al13 proposed using SVR to fit in‐control
profiles. Then, they employed the moving block bootstrap
method to generate correlated samples for each in‐control
profile and obtain a simultaneous confidence region for
the underlying functional relationship. The obtained con-
fidence region was used to monitor the real AIDS data
collected from hospitals in Taiwan. In practice, the func-
tional relationship of the profile data rarely occurs in lin-
ear form and the real data usually do not follow normal
distribution. Thus, in this paper, the functional relation-
ship of profile data is described via a nonparametric
regression model and a nonparametric rank‐based
EWMA control chart is developed for detecting the
LI ET AL. 3

process shifts for nonlinear profile data in Phase II denotes the mean absolute deviation, M4 denotes the
monitoring. absolute value of maximum deviation (disregarding the
direction of dissimilarity), and M5 denotes the sum of
squared differences between the predicted values for the
3 | RESEARCH METHODOLOGY
measurement yij and the corresponding baseline profiles.
Let Mj, p, j = 1, 2, …, be the pth metric for the corre-
3.1 | Developing a nonparametric EWMA
sponding jth profile in the Phase II study. The rank of
control chart for phase II monitoring
M with respect to the reference metric
n j, p o
Let yij be the measurement of the ith observations in the M 1;p ; M 2;p ; ; …; M *g−1;p
* *
can be calculated as
jth profile and xij be the vector of the corresponding  
R*j;p ¼ 1 þ ∑k¼1 I M j;p > M *k;p , where the indicator
g−1
explanatory variables such that i = 1, 2, …, nj for each  
j = 1, 2, …. When the process is in statistical control, function I M j;p > M *k;p ¼ 1 if M j;p > M *k;p ; otherwise,
the underlying model is assumed to be the indicator function = 0, and g − 1 is the sample

 size of referencen metric. The detailed o calculation of
yij ¼ f x ij þ εij ; i ¼ 1; 2; …; nj ; j ¼ 1; 2; …; (1) reference metric M *1;p ; M *2;p ; ; …; M *g−1;p is explained in
Section 3.2. Then, the standardized rank of R*j;p can be
where f (x) is a known function with certain degree of
calculated as
smoothness and the random errors εij are generally
assumed to follow some distributions. Without loss of  
2 * gþ1
generality, the nj are taken to be equal and the explana- Rj;p ¼ Rj;p − : (2)
g 2
tory variables are assumed to be fixed for different j's.
The predicted values byij of the measurement yij for the Replacing the individual observation by the standard-
corresponding explanatory variables are calculated based ized rank Rj, p, our proposed EWMA statistics can be
on the SVR model. Then, the five metrics for the jth written as Equation (3).
profile suggested by Williams et al1 can be calculated
as follows: EWMAj;p ¼ ð1 − λÞEWMAj−1;p þ λ Rj;p ; j ¼ 1; 2; …; (3)

 n  where EWMA0, p = 0 and λ is a smoothing parameter

o
M j1 ¼ sign max byij − e
yi with 0 < λ ≤ 1. Because the standardized ranks in
i¼1; …; n
Equation (2) follow a discrete uniform distribution on

1 3 1
n   the points − 1; − 1; … ; 1 − with mean zero
 
Mj2 ¼ ∑ byij − e
y i g g g
i¼1 g2 − 1
and variance , the control limits of our proposed
3g2
1 n  
 nonparametric EWMA control chart are defined as
Mj3 ¼ ∑ b yij − e
y i
n i¼1 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
λ g2 − 1
  ±LR (4)
Mj4 ¼ Mj1  2 − λ 3g2

where LR is determined based on Monte Carlo simulation

 n 2
Mj5 ¼ ∑ byij −e
yi to achieve a specified in‐control ARL.
i¼1

where e y i is the reference profile for the ith observation. 3.2 | Constructing the reference profile in
The detailed calculation of e y i is explained in Section 3.2. Phase I study
In conducting the Phase II study, we adopt the concept
In this paper, we use SVR to fit profiles for describing the
of residual control chart and employ a nonparametric
functional relationship shown in Equation (1). The SVR
EWMA control chart proposed by Hackel and Ledolter14
is a supervised statistical learning algorithm for regres-
to monitor these five metrics. In other words, we consider
sion problem. In SVR, the decision variables are mapped
the metric for profile j as an individual observation and
onto a feature space, and a linear model in Equation (5) is
use a nonparametric EWMA control chart is to monitor
constructed in the feature space.
the mean change of these five metrics with n observations
over j time period, where M1 denotes the maximum devi- gðx; wÞ ¼ wT ϕðx Þ þ b; (5)
ation, M2 denotes the sum of absolute deviations, M3 where w is normal vector, ϕ(.) is a nonlinear
4 LI ET AL.

transformation function, b is bias term, and x is the 4 | T H E SI M U L A T I O N ST U D Y

explanatory variables, which are also called vector of
decision variables in SVR model. Suppose that we have In the simulation study, once the in‐control ARL (ARL0)
the Phase I in‐control data {yij, i = 1, …, n, j = 1…, is fixed at 370, the out‐of‐control average run length
(g − 1)}. For a given j, the quality of estimation is mea- (ARL1) is used to evaluate the detecting performance of
sured by the loss function L(yij, g(xi, w)), and the SVR our proposed nonparametric EWMA control chart after
model is formulated as a minimization problem as listed conducting 5000 simulation runs. As the exponential
in Equation (6). form is frequently occurred in relating y to x in nonlinear
profile data, we first assume that the in‐control nonlinear
1 n  
min kw k2 þ C ∑ ξ i þ ξ *i profile model follows an exponential form:
2 i¼1


subject to: yij ¼ 0:9 exp β0 þ β1 x ij þ εij ;

8
>
> y − gðx i ; w Þ ≤ ϵ þ ξ i where β0 = 0.5, β1 = 2, and the explanatory variable are
< ij
gðx i ; w Þ − yij ≤ ϵ þ ξ *i (6) fixed as xij = 0.01i, i = 1, …, 300 for each profile j. For ran-
>
> dom error terms εij, we consider the following two scenar-
:
ξ i ; ξ *i ≥ 0 ios: (1) a standard normal distribution N(0, σ) with
parameter σ = 1; (2) an exponential distribution with rate
where ϵ > 0 is a certain threshold, constant C > 0 is a
parameter θ = 0.5. Since a reference metric
penalization that can be viewed as a way to control n o
M *1;p ; M *2;p ; ; …; M *g−1;p obtained from Phase I study
over‐fitting, and ξi and ξ *i are slack variables. The loss
function is given by (see Sections 3.1 and 3.2 for details) is needed to construct
a Phase II control chart, an SVR model for obtaining
(  
    the reference metric is built based on Phase I profile
 0  ; if yij − gðx i ; wÞ < ϵ
L yij ; gðx i ; w Þ ¼   : training data generated from the SVR model shown in
yij − gðx i ; wÞ − ϵ ;
otherwise Equation (6). In the parameter optimization process with
5‐fold cross validation, the optimal values of the parame-
The esp‐regression machine in R package e1071 (Mayer ters C and ϵ in Equation (6) are set as 512 and 0.0001,
et al15) is used to build the above SVR model based on respectively. Then, the reference metrics can be obtained
the training dataset collected from the Phase I study. accordingly. In this paper, the sample size of the refer-
Through the SVR model, the reference profile for the ence metrics is set as g = 21. The simulation results for
g−1
ith observation is calculated as yei ¼ ∑j¼1 y_ ij =ðg − 1Þ, various control parameter LR values under different λ
i = 1, …, n, where y_ ij is the predicted value for the corre- are summarized in Table 1.
sponding observation yij. Thus, the following reference For the random error terms following a standard nor-
metrics can be calculated accordingly: mal distribution or an exponential distribution, the simu-
lation results of ARL1 values with λ = 0.05 and λ = 0.1
 n o  under different shifts in parameter β0, β1, and standard
 
M *j;1 ¼ sign max y_ ij − e
y i  ; j ¼ 1; …; ðg − 1Þ derivation σ are shown in Figures 2 and 3, respectively.
i¼1; …; n
Note that the shifts in parameters β0 and β1 are defined
as β0 + δ and β1 + δ. The shifts in parameters σ and θ
n  
  are defined as 1 × δ and 0.5 + δ for a standard normal
M *j;2 ¼ ∑ y_ ij − e
y i ; j ¼ 1; …; ðg − 1Þ
i¼1 and an exponential distribution, respectively.
As one can see from Figure 2, the ARL1 decreases as δ
increases and ARL1 increases as λ increases. The
1 n  

M *j;3 ¼ ∑ y_ ij − e
y i ; j ¼ 1; …; ðg − 1Þ
n i¼1
TABLE 1 Various control parameter LR values under different λ
with g = 21. for achieving ARL0=370
 
 
M *j;4 ¼ M *j;1 ; j ¼ 1; …; ðg − 1Þ λ 0.05 0.1 0.2 0.3 0.4
LR 1.37 1.77 2.10 2.18 2.17
λ
n  2 0.5 0.6 0.7 0.8 0.9
M *j;5 ¼ ∑ y_ ij −e
y i ; j ¼ 1; …; ðg − 1Þ: LR 2.13 2.05 1.96 1.87 1.77
i¼1
LI ET AL.
simulation results show that the detecting ability of the
proposed nonparametric EWMA chart increases when
the magnitudes of shift in parameter δ increase or the
smoothing parameter λ decreases. In most cases, the
ARL1 of the nonparametric EWMA control chart with
the metric M1 and M4 for detecting the shifts in various
parameters are larger than those of M2, M3, and M5,
ie, the detecting performances of the nonparametric
EWMA control chart with metrics M1 and M4 are worse
than those of M2, M3, and M5. Note that the detecting
performances of the nonparametric EWMA control chart
with metrics M2 and M3 are the same; thus, only four
curves are shown in Figure 2. The results are expected
since the statistical forms of the two metrics M2 and M3
differ only by a constant. Moreover, Figure 2 shows
that the detecting performances of the nonparametric
EWMA control chart with metrics M2, M3, and M5
FIGURE 2 ARL1 values under different shifts in parameter: A, β0; B, β1; or C, σ for the given λ = 0.05 (left panel) and λ = 0.1 (right panel)
when the random error terms follow a standard normal distribution
5
are almost the same under different shifts in various
parameters when the random error terms follow a
standard normal distribution.
As one can see from Figure 3, the ARL1 decreases as
δ increases and ARL1 increases as λ increases. The
simulation results show that the detecting ability of
the proposed nonparametric EWMA chart increases
when the magnitudes of shift in parameter δ increase
or the smoothing parameter λ decreases. Note that the
detecting performances of the nonparametric EWMA
control chart with metrics M2 and M3 are the same;
thus, only four curves are shown in Figure 3. In other
words, the detecting performances of the control chart
with metrics M2, M3, and M5 are almost the same under
different shifts in various parameters when the random
error terms follow an exponential distribution. More-
over, in most cases, the ARL1 of the EWMA control
6 LI ET AL.

chart with the metric M1 and M4 for detecting shifts in
various parameters except β0 are larger than metrics M2,
M3, and M5. Therefore, one can conclude that the
detecting performances of the nonparametric EWMA
control chart with metrics M1 and M4 are worse than
those of M2, M3, and M5. These results are expected
since the metric M1 and M4 did not use all the informa-
tion in the data set.
5 | A NUMERICAL EXAMPLE
In order to demonstrate the practical application of our
proposed control chart, the numerical example with
VDP data given in Walker and Wright11 is used for illus-
tration purpose. In manufacturing the particle boards, the
density property of finished boards is the quality charac-
teristic required to be closely monitored. Each resulting
profile consists of 314 density measurements, and the dis-
tance between two consecutive measures is 0.002 inch.
The original VDP data with 24 profiles are illustrated in
Figure 4, where the depth of thickness of the particle
boards ranges from 0 to 0.626 inch.
To construct a Phase II control chart, we first employ
the SVR technique as mentioned in Section 3 to model
the VDP data. Note that the parameter C and ϵ are set
as 8 and 0.001, respectively, in Equation (6). Then, the
mean squared error (MSE) and mean absolute percent
error (MAPE) for VDP data can be obtained and summa-
rized in Table 2.
According to Table 2, the average MAPE can be
approximated to 0.49%, which indicates its percent

FIGURE 3 ARL1 values under different shifts in parameter: A, β0; B, β1; or C, θ for given the λ = 0.05 (left panel) and λ = 0.1 (right panel)
when the random error terms follow an exponential distribution
LI ET AL. 7

FIGURE 4 The original VDP data with 24 particle boards FIGURE 5 The 24 reference profiles after using SVR model to fit
[Colour figure can be viewed at wileyonlinelibrary.com] VDP data [Colour figure can be viewed at wileyonlinelibrary.com]

accuracy = 99.51%. Thus, we conclude that the fitted SVR

model for VDP data is adequate. The 24 reference profiles
obtained from the fitted model are shown in Figure 5.
Since the predicted values can be calculated from the
SVR model, the corresponding values of five (5) metrics
suggested by Williams et al1 are obtained accordingly.
Then, the nonparametric EWMA control chart with five
(5) metrics is proposed to determine which board is out‐
of‐control. Its upper and lower control limits are calcu-
lated based on Equation (4) with λ = 0.05, g = 21, and
LR = 1.37. In Figure 6, the nonparametric EWMA control
chart with five metric curves is plotted against the control
limits according to their sample IDs. Figure 6 shows that
all the VDP profiles using metric curves M1 and M4 are
in‐control. This result is consistent with the findings of

TABLE 2 The MSE and MAPE for VDP data

Profile MSE MAPE Profile MSE MAPE

FIGURE 6 The nonparametric EWMA control chart with five
1 0.0467 0.48% 13 0.0649 0.48% metric curves [Colour figure can be viewed at wileyonlinelibrary.
2 0.0635 0.53% 14 0.0601 0.53% com]

3 0.0654 0.48% 15 0.0469 0.48%

4 0.0589 0.50% 16 0.0410 0.50% Williams et al.1 It is worthy to note that Williams et al1
5 0.0610 0.48% 17 0.0505 0.48% also pointed out the third profile is consistently higher
than all other profiles. But, the nonparametric I‐MR con-
6 0.0577 0.49% 18 0.0630 0.49%
trol chart proposed by Williams et al1 could not detect
7 0.0529 0.53% 19 0.0656 0.53% this an abnormal profile until the sixth profile, whereas
8 0.0525 0.49% 20 0.0501 0.49% our proposed nonparametric EWMA control chart with
9 0.0498 0.47% 21 0.0547 0.47% the metrics M2, M3, and M5 signals an out‐of‐control
10 0.0471 0.48% 22 0.0533 0.48% at the fourth profile, and these metric curves have the
similar patterns. Apparently, the detecting ability of our
11 0.0617 0.47% 23 0.0513 0.47%
proposed EWMA control chart outperforms the I‐MR
12 0.0509 0.49% 24 0.0533 0.49%
control chart proposed by Williams et al.1
8 LI
6 | C O N C L U S I O N S AN D
DISC USS I ONS
In real applications, we often do not know the distribution
of the studied quality characteristics. Thus, a nonparamet-
ric method is proposed for monitoring the nonlinear
profile data. In this paper, we first employ SVR model to
smooth out the nonlinear profile data (ie, to remove the
noise of data). Then, the five metrics proposed by Williams
et al1 can be obtained, and the nonparametric rank‐based
EWMA control chart with five metrics is used to monitor
the quality characteristic of the nonlinear profile data.
Both the simulation results and numerical example show
that our proposed nonparametric EWMA control chart
with the metrics M2, M3, and M5 has better detecting per-
formances than those of the I‐MR control chart. Hence,
the monitoring efficiency for nonlinear profile data in the
Phase II study can significantly be enhanced by using our
proposed nonparametric EWMA control chart with the
metrics M2, M3, and M5. Now that the nonparametric
EWMA control chart with M2, M3, and M5 has better
detecting performances and the statistical forms of metrics
M2 and M3 differ only by a constant, the future researchers
may consider to develop a nonparametric rank‐based mul-
tivariate method by incorporating the metrics M2 (or M3)
and M5 simultaneously into a nonparametric rank‐based
multivariate EWMA control chart. Moreover, in the
simulation study, we consider the nonlinear profile with
an exponential form. Similar research can be extended to
other nonlinear profiles, such as power, reciprocal,
hyperbolic, or Weibull, etc. Hopefully, the results of this
research can serve as a useful reference for quality practi-
tioners when monitoring and controlling the process
quality for the nonlinear profile data in the Phase II study.
ACK NO WLE DGE MEN TS
We are grateful to the two anonymous reviewers for their
helpful comments. The second author would like to grate-
fully acknowledge financial support (MOST 106‐2410‐H‐
006‐007) from the Ministry of Science and Technology of
Taiwan, ROC.
R EF E RE N C E S
1. Williams JD, Woodall WH, Brich JB. Statistical monitoring of
nonlinear product and process quality profiles. Qual Reliab
Eng Int. 2007;23(7):925‐941.
2. Montgomery DC. Introduction to Statistical Quality Control. 6th
ed. New York: John Wiley & Sons; 2009.
3. Pan JN, Li CI, Wu JJ. A new approach to detecting the process
changes for multistage systems. Expert Systems with
Applications. 2016;62(11):293‐301.
ET AL.
4. Kang L, Albin SL. On‐line monitoring when the process yields a
linear profile. J Qual Technol. 2000;32(4):418‐426.
5. Khedmati M, Niaki STA. Phase II monitoring of general linear
profiles in the presence of between‐profile autocorrelation. Qual
Reliab Eng Int. 2015;32(2):443‐452.
6. Kim K, Mahmoud MA, Woodall WH. On the monitoring of
linear profiles. J Qual Technol. 2003;35(3):317‐328.
7. Woodall WH, Spitzner DJ, Montgomery DC, Gupta S. Using
control charts to monitor process and product quality profiles.
J Qual Technol. 2004;36(3):309‐320.
8. Woodall WH. Current research on profile monitoring. Revista
Producão. 2007;17(3):420‐425.
9. Zi X, Zou C, Tsung F. A distribution‐free robust method for
monitoring linear profile using rank‐based regression. IIE
Transactions. 2012;44(11):949‐963.
10. Chiang JY, Lio YL, Tsai TR. MEWMA control chart and process
capability indices for simple linear profiles with within‐profile
autocorrelation. Qual Reliab Eng Int. 2017;33(5):1083‐1094.
11. Walker E, Wright SP. Comparing curves using additive models.
J Qual Technol. 2002;34(1):118‐129.
12. Moguerza JM, Munoz A, Psarakis S. Monitoring nonlinear
profiles using support vector machines. Progress in Pattern Rec-
ognition, Image Analysis and Applications. 2007;4756:574‐583.
13. Hung YC, Yang SF, Chuang SC, Tseng YK. Nonparametric
profile monitoring in multi‐dimensional data spaces. J Process
Control. 2012;22(2):397‐403.
14. Hackel P, Ledolter J. A control chart based on ranks. J Qual
Technol. 1991;23(2):117‐124.
15. Mayer D, Dimitriadou E, Hornik K, Weingessel A, Leisch F.
e1071: Misc Functions of the Department of Statistics (e1071),
2014 TU Wien. R Package Version 1.6–3.
Dr Chung‐I Li is an assistant professor in the Depart-
ment of Statistics, National Cheng Kung University.
His current research interest is in the fields of Statisti-
cal Quality Control and Applied Statistics.
Dr Jeh‐Nan Pan is a full professor in the Department
of Statistics, National Cheng Kung University
(NCKU). Professor Pan' expertise is in the fields of
quality management and reliability engineering,
industrial statistics, and decision methods. He has
published 78 papers in both domestic and renowned
SCI/SSCI international journals such as IEEE Trans-
action on Reliability, European Journal of Operational
Research, Quality and Reliability Engineering Interna-
tional, Quality Engineering, and International Journal
of Production Research. He holds MS and PhD degrees
in Industrial Engineering from Texas Tech University,
Lubbock, Texas. He is ASQ‐certified Quality and Reli-
ability Engineer. He has received numerous distin-
guished awards including three (3) Research and
Teaching Excellence awards from NCKU, four (5) best
LI ET AL. 9

paper awards from Chinese Society for Quality (CSQ)

How to cite this article: Li C‐I, Pan J‐N, Liao C‐H.
and Asian Network for Quality (ANQ). He has also
Monitoring nonlinear profile data using support
been elected as a CSQ fellow in 2016.
vector regression method. Qual Reliab Engng Int.
2018;1–9. https://doi.org/10.1002/qre.2385
Mr Chun‐Han Liao is a graduate student in the
Department of Statistics, National Cheng Kung
University.