• Structural Engineering and Mechanics
• /
• v.81 no.5
• /
• pp.647-664
• /
• 2022

The powerful data mapping capability of computational deep learning methods has been recently explored in academic works to develop strategies for structural health monitoring through appropriate characterization of dynamic responses. In many cases, these studies concern laboratory prototypes and finite element models to validate the proposed methodologies. Therefore, the present work aims to investigate the capability of a deep learning algorithm called Sparse Autoencoder (SAE) specifically focused on detecting structural alterations in real-case studies. The idea is to characterize the dynamic responses via SAE models and, subsequently, to detect the onset of abnormal behavior through the Shewhart T control chart, calculated with SAE extracted features. The anomaly detection approach is exemplified using data from the Z24 bridge, a classical benchmark, and data from the continuous monitoring of the San Vittore bell-tower, Italy. In both cases, the influence of temperature is also evaluated. The proposed approach achieved good performance, detecting structural changes even under temperature variations.

• Journal of the Korean Professional Engineers Association
• /
• v.20 no.3
• /
• pp.15-25
• /
• 1987

Statistical control charts are useful tools to monitor and control the manufacturing processes and are widely used in most Korean industries. Many Korean companies, however, do not always obtain desired results from the traditional control charts by Shewhart such as the X-chart, X-chart, X-chart, etc. This is partly because the quality charterstics of the process are not distributed normally but are skewed due to the intermittent production, small lot size, etc. In Shewhart X-chart, which is the most widely used one in Korea, such skewed distributions make the plots to be inclined below or above the central line or outside the control limits although no assignable causes can be found. To overcome such shortcomings in nonnormally distributed processes, a distribution-free type of confidence interval can be used, which should be based on order statistics. This thesis is concerned with the design of control chart based on a sample median which is easy to use in practical situation and therefore properties for nonnormal distributions may be easily analyzed. Control limits and central lines are given for tile more famous nonnormal distributions, such as Gamma, Beta, Lognormal, Weibull, Pareto, Truncated-normal distributions. Robustness of the proposed median control chart is compared with that of the X-chart, the former tends to be superior to the latter as the probability distribution of the process becomes more skewed. The average run length to detect the assignable cause is also compared when the process has a Normal or a Gamma distribution for which the properties of X are easy to verify, the proposed chart is slightly worse than the X-chart for the normally distributed product but much better for Gamma-distributed products. Average Run Lengths of the other distributions are also computed. To use the proposed control chart, the probability distribution of the process should be known or estimated. If it is not possible, the results of comparison of the robustness force us to use the proposed median control chart based on a normal distribution. To estimate the distribution of the process, Sturge's formula is used to graph the histogram and the method of probability plotting, $X^2$-goodness of fit test and Kolmogorov-Smirnov test, are discussed with real case examples. A comparison of the propose4 median chart and the X chart was also performed with these examples and the median chart turned out to be superior to the X-chart.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.6
• /
• pp.1353-1366
• /
• 2015

These days Shewhart control chart for evaluating stability of the process is widely used in various field. But it must follow strict assumption of distribution. In real-life problems, this assumption is often violated when many quality characteristics follow non-normal distribution. Moreover, it is more serious in multivariate quality characteristics. To overcome this problem, many researchers have studied the non-parametric control charts. Recently, SVDD (Support Vector Data Description) control chart based on RBF (Radial Basis Function) Kernel, which is called K-chart, determines description of data region on in-control process and is used in various field. But it is important to select kernel parameter or etc. in order to apply the K-chart and they must be predetermined. For this, many researchers use grid search for optimizing parameters. But it has some problems such as selecting search range, calculating cost and time, etc. In this paper, we research the efficiency of selecting parameter regions as data structure vary via simulation study and propose a new method for determining parameters so that it can be easily used and discuss a robust choice of parameters for various data structures. In addition, we apply it on the real example and evaluate its performance.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.10 no.16
• /
• pp.101-106
• /
• 1987

Statistical control charts are useful tools to monitor and control the manufacturing processes and are widely used in most Korean industries. Many Korean companies, however, do not always obtain desired results from the traditional control charts by Shewhart such as the $\bar{X}$-chart, $\bar{X}$-chart, $\bar{X}$-chart, etc. This is partly because the quality charterstics of the process are not distributed normally but are skewed due to the intermittent production, small lot size, etc. In Shewhart $\bar{X}$-chart. which is the most widely used one in Kora, such skewed distributions make the plots to be inclined below or above the central line or outside the control limits although no assignable causes can be found. To overcome such shortcomings in nonnormally distributed processes, a distribution-free type of confidence interval can be used, which should be based on order statistics. This thesis is concerned with the design of control chart based on a sample median which is easy to use in practical situation and therefore properties for nonnormal distributions may be easily analyzed. Control limits and central lines are given for the more famous nonnormal distributions, such as Gamma, Beta, Lognormal, Weibull, Pareto, Truncated-normal distributions. Robustness of the proposed median control chart is compared with that of the $\bar{X}$-chart; the former tends to be superior to the latter as the probability distribution of the process becomes more skewed. The average run length to detect the assignable cause is also compared when the process has a Normal or a Gamma distribution for which the properties of X are easy to verify, the proposed chart is slightly worse than the $\bar{X}$-chart for the normally distributed product but much better for Gamma-distributed products. Average Run Lengths of the other distributions are also computed. To use the proposed control chart, the probability distribution of the process should be known or estimated. If it is not possible, the results of comparison of the robustness force us to use the proposed median control chart based oh a normal distribution. To estimate the distribution of the process, Sturge's formula is used to graph the histogram and the method of probability plotting, $\chi$$^2$-goodness of fit test and Kolmogorov-Smirnov test, are discussed with real case examples. A comparison of the proposed median chart and the $\bar{X}$ chart was also performed with these examples and the median chart turned out to be superior to the $\bar{X}$-chart.

• Journal of Radiation Protection and Research
• /
• v.44 no.2
• /
• pp.64-71
• /
• 2019

Background: It is very difficult to distinguish between a radioactive contamination source and background radiation from natural radionuclides in the marine environment by means of online monitoring system. The objective of this study was to investigate a statistical process for triggering abnormal level of count rate data measured from our on-line seawater radioactivity monitoring. Materials and Methods: Count rate data sets in time series were collected from 9 monitoring posts. All of the count rate data were measured every 15 minutes from the region of interest (ROI) for $^{137}Cs$ ($E_{\gamma}=661.6keV$) on the gamma-ray energy spectrum. The Shewhart ($3{\sigma}$), CUSUM, and Bayesian S-R control chart methods were evaluated and the comparative analysis of determination methods for count rate data was carried out in terms of the false positive incidence rate. All statistical algorithms were developed using R Programming by the authors. Results and Discussion: The $3{\sigma}$, CUSUM, and S-R analyses resulted in the average false positive incidence rate of $0.164{\pm}0.047%$, $0.064{\pm}0.0367%$, and $0.030{\pm}0.018%$, respectively. The S-R method has a lower value than that of the $3{\sigma}$ and CUSUM method, because the Bayesian S-R method use the information to evaluate a posterior distribution, even though the CUSUM control chart accumulate information from recent data points. As the result of comparison between net count rate and gross count rate measured in time series all the year at a monitoring post using the $3{\sigma}$ control charts, the two methods resulted in the false positive incidence rate of 0.142% and 0.219%, respectively. Conclusion: Bayesian S-R and CUSUM control charts are better suited for on-line seawater radioactivity monitoring with an count rate data in time series than $3{\sigma}$ control chart. However, it requires a continuous increasing trend to differentiate between a false positive and actual radioactive contamination. For the determination of count rate, the net count method is better than the gross count method because of relatively a small variation in the data points.

• Journal of Korean Institute of Industrial Engineers
• /
• v.31 no.1
• /
• pp.87-98
• /
• 2005

This research investigates economic-statistical characteristics of variable sampling size and interval (VSSI)$\bar{X}$charts under two assignable causes. A Markov chain approach is employed in order to calculate average run length (ARL) and average time to signal (ATS). Six transient states are derived by carefully defining the state. A steady state cost rate function is constructed based on Lorenzen and Vance(1986) model. The cost rate function is optimized with respect to six design parameters for designing the VSSI $\bar{X}$ charts. Computational experiments show that the VSSI $\bar{X}$ chart is superior to the Shewhart $\bar{X}$ chart in the economic-statistical sense, even under two assignable causes. A comparative study shows that the cost rate may increase up to almost 30% by overlooking the second cause. Critical input parameters are also derived from a sensitivity study and a few guideline graphs are provided for determining the design parameters.

• Journal of Korean Society for Quality Management
• /
• v.27 no.4
• /
• pp.143-152
• /
• 1999

In Shewhart control chart, the average run length(ARL) is calculated using the mean of a conventional geometric distribution(CGD) assuming a sequence of identical and independent Bernoulli trials. In this, the success probability of CGB is the probability that any point exceeds the control limits. When the process is in-control state, there is no problem in the above assumption since the probability that any point exceeds the control limits does not change if the in-control state continues. However, if the out-of-control state begins and continues during the process, the probability of exceeding the control limits may take two forms. First, once the out-of-control state begins with exceeding probability p, it continues with the same exceeding probability p. Second, after the out-of-control state begins, the exceeding probabilities may very according to some pattern. In the first case, ARL is the mean of CGD with success probability p as usual. But in the second case, the assumption of a sequence of identical and independent Bernoulli trials is invalid and we can not use the mean of CGD as ARL. This paper concentrate on that point. By adopting one generalized binomial distribution(GBD) model that allows correlated Bernoulli trials, generalized geometric distribution(GGD) is defined and its mean is derived to find an alternative ARL when the process is in out-of-control state and the exceeding probabilities take the second form mentioned in the above. Small-scale simulation is performed to show how an alternative ARL works.

• Journal of Korean Institute of Industrial Engineers
• /
• v.39 no.5
• /
• pp.422-428
• /
• 2013

In many cases, an $\bar{X}$ control chart which is based on the performance variable is used in industrial fields. However, if the performance variable is too costly or impossible to measure and a less expensive surrogate variable is available, the process may be more efficiently controlled using surrogate variables. In this paper, we propose a model for the economic design of a VSI (Variable Sampling Interval) $\bar{X}$ control chart using a surrogate variable that is linearly correlated with the performance variable. The total average profit model is constructed, which involves the profit per cycle time, the cost of sampling and testing, the cost of detecting and eliminating an assignable cause, and the cost associated with production during out-of-control state. The VSI $\bar{X}$ control charts using surrogate variables are expected to be superior to the Shewhart FSI (Fixed Sampling Interval) $\bar{X}$ control charts using surrogate variables with respect to the expected profit per unit cycle time from economic viewpoint.

• Proceedings of the KIEE Conference
• /
• 2006.07a
• /
• pp.22-24
• /
• 2006

The installations of power quality monitoring system have increased drastically over the past several decades. These systems have been effectively used to monitor, analyze and diagnose the conditions of power system, and furthermore can be used to improve the present asset maintenance policy, scheduled (time-based) method, into the advanced, cost-effective and labor-effective maintenance methods, such as condition-based maintenance, predictive maintenance and reliability centered maintenance. As an approach to this, this paper introduces the statistical methods, three kinds of control charts (Shewhart chart, CUSUM chart and EWMA chart), and discusses the applicability of these methods to recognize the changing trends of power quality indices and to estimate the system's condition, using Matlab.

• The Korean Journal of Applied Statistics
• /
• v.19 no.1
• /
• pp.81-96
• /
• 2006

This paper proposes a simple control chart method which can be practically used for asymmetric process data where the distribution is unknown. If we use the Shewhart type control charts which are based on normality assumption for the asymmetric process data, the type I error could increase as the asymmetry increases and the effectiveness of control chart to control variation decreases. To solve such problems, this paper suggests to calculate the control limits based on the quartiles. If we obtain the control limits by such quartile method, the type I error could decrease and it looks much more practical for asymmetric distributed process data.