Gaussian Mixture Model을 이용한 선박 메인 엔진 빅데이터의 이상치 탐지

김동현(Dong Hyun Kim),이상봉(Sang Bong Lee),이지환(Ji Hwan Lee) 한국자료분석학회 2020 Journal of the Korean Data Analysis Society Vol.22 No.4

선박 메인 엔진은 선박의 추진 동력을 제공하는 핵심 장비로써 고장 예방을 위한 주기적인 정비와 모니터링이 요구된다. 최근 선박에 부착된 센서를 이용하여 선박 메인 엔진의 상태를 실시간으로 측정할 수 있게 되면서, 수집된 빅데이터를 바탕으로 엔진의 이상 징후를 조기에 발견할 수 있는 체계가 마련되었다. 본 연구는, 7개월 이상 수집된 선박 메인 엔진 데이터를 학습하여 선박 메인 엔진의 이상 징후를 조기에 파악할 수 있는 방법론을 제안한다. 학습을 위해 전문가 인터뷰 및 상관계수 분석을 통해 중요 변수를 추출하였고, 데이터 전처리 기법을 통해 학습에 필요한 변수만을 추출하였다. 다수의 정상적인 메인 엔진 상태 데이터로부터 이상치를 분리하기 위하여 비지도 학습(unsupervised learning) 기반의 이상치 탐지 기법인 Gaussian Mixture Model을 적용하였다. 또한, 이상치 값들에 대한 사후적인 분석을 하여 선박 메인 엔진 이상치의 특성과 관련된 정보들을 파악했다. 이를 통해 선박 빅데이터를 활용하여 실시간으로 이상치를 탐지할 수 있는 의사결정 체계를 구축하고, 선박 유지 보수의 효율성과 경제성을 높이는데 이바지할 수 있을 것으로 예상한다. The ship s main engine is a key equipment that provides power for propulsion of the ship, and requires regular maintenance and monitoring to prevent failure. Recently, the health state of the ship s main engine can be measured in real time using a sensor attached to the ship, enabling an user to detect abnormal signs of the engine early on the basis of the collected big data. This study proposes a methodology for early identification of abnormal signs of the ship main engine by learning the ship main engine data collected for more than 7 months. For learning, important variables were extracted through expert interviews and correlation coefficient analysis, and only important variables were extracted through data pre-processing techniques. To separate outliers from a number of normal main engine state data, the Gaussian Mixture Model, an unsupervised learning-based outlier detection technique, was applied. In addition, by comparing the distribution of the outliers and the normal distribution, the characteristics of the ship s main engine outlier was analyzed. Through this, it is expected to build a decision-making system that can detect outliers in real time using ship big data and contribute to improving the efficiency and economics of ship maintenance.

Skewness of Gaussian Mixture Absolute Value GARCH(1, 1) Model

Lee, Taewook The Korean Statistical Society 2013 Communications for statistical applications and me Vol.20 No.5

This paper studies the skewness of the absolute value GARCH(1, 1) models with Gaussian mixture innovations (Gaussian mixture AVGARCH(1, 1) models). The maximum estimated-likelihood estimator (MELE) employed (a two- step estimation method in order to estimate the skewness of Gaussian mixture AVGARCH(1, 1) models. Through the real data analysis, the adequacy of adopting Gaussian mixture innovations is exhibited in reflecting the skewness of two major Korean stock indices.

A Comparative Study of Foreground Detection using Gaussian Mixture Models- Novice to Novel

Ajmal Shahbaz,Laksono Kurnianggoro,Kang-Hyun Jo 제어로봇시스템학회 2016 제어로봇시스템학회 국제학술대회 논문집 Vol.2016 No.10

Foreground detection is the classical computer vision task of segmenting out motion information from a particular scene. Foreground detection using Gaussian Mixture Models (GMM) is the famous choice. Since first time proposed, many researchers tried to improve GMM. This paper focuses on the comparative evaluation of three most famous improvements in the algorithm. The improved methods are compared both qualitatively and quantitatively using standard datasets available online.

Fault Detection Method Using Multi-mode Principal Component Analysis Based on Gaussian Mixture Model for Sewage Source Heat Pump System

유영준 제어·로봇·시스템학회 2019 International Journal of Control, Automation, and Vol.17 No.8

This paper presents an algorithm for fault detection of a sewage heat pump system by designing multimode principal component analysis with Gaussian mixture model. If the heat pump system fails, the loss of energy and time is enormous, therefore the fault detection of the system is important. For this purpose, this study proposes a fault detection method using multi-mode principal component analysis with Gaussian mixture model. The data were clustered into multi-mode of Gaussian on principal component subspace. Based on the multi-model, the values of Hotelling’s T2 and SPE were calculated and used for the fault detection as indexes that are compared performance with clustering model using k-means and k-medoids algorithm as well as conventional PCA. Actual data of the sewage heat pump were used to verify the proposed method. The results of the fault detection performance show that the proposed model shows the best performance of fault detection among the conventional, k-means, and kmedoids PCA models.

A spatial downscaling of soil moisture from rainfall, temperature, and AMSR2 using a Gaussian-mixture nonstationary hidden Markov model

Kwon, Moonhyuk,Kwon, Hyun-Han,Han, Dawei Elsevier 2018 Journal of hydrology Vol.564 No.-

<P><B>Abstract</B></P> <P>A multivariate stochastic soil moisture (SM) estimation approach based on a Gaussian-mixture nonstationary hidden Markov model (GM-NHMM) is introduced in this study to spatially disaggregate the AMSR2 SM data for multiple locations in the Yongdam dam watershed in South Korea. Rainfall and air temperature are considered as additional predictors in the proposed modeling framework. In GM-NHMM, a six-state model is constructed with three predictors representing an unobserved state associated with SM. It is clearly seen that the rainfall predictor plays a substantial role in achieving the overall predictability. Using weather variables (i.e., rainfall and temperature) can be effective in picking up some of the predictability of local SM that is not captured by the AMSR2 data. On the other hand, larger scale dynamic features identified from the AMSR2 data seem to facilitate the identification of regional spatial patterns of SM. The efficiency of the proposed model is compared with that of an ordinary regression model (OLR) using the same predictors. The mean correlation coefficient of the proposed model is about 0.78, which is significantly greater than that of the OLR at about 0.49. The proposed GM-NHMM method not only provides a better representation of the observed SM than the OLR model but also preserves the spatial coherence across all stations reasonably well.</P> <P><B>Highlights</B></P> <P> <UL> <LI> A nonstationary HMM model is employed to spatially downscale the soil moisture data. </LI> <LI> Rainfall predictor plays a substantial role in achieving the overall predictability. </LI> <LI> Climate data are effective in picking up the predictability of local soil moisture. </LI> <LI> The proposed model preserves the spatial coherence across stations reasonably well. </LI> </UL> </P>

Color Image Segmentation Based on Morphological Operation and a Gaussian Mixture Model

이명은,박순영,조완현,Lee Myung-Eun,Park Soon-Young,Cho Wan-Hyun The Institute of Electronics and Information Engin 2006 電子工學會論文誌-SP (Signal processing) Vol.43 No.3

본 논문에서는 수학적 모폴로지 연산과 가우시안 혼합 모형에 기초한 새로운 칼라 영상 분할 알고리즘을 제안한다. 우리는 혼합 모형에서 구성 성분의 수를 결정하고, 각 구성 성분의 중심값을 계산하는데 모폴로지의 연산과 라벨링 연산을 이용한다. 그리고 칼라 특징 벡터의 확률 모형으로 가우시안 혼합 모형을 사용하고, 이들의 모수 값들을 추정하는데 결정적 어닐링 EM알고리즘을 사용한다. 최종적으로 혼합 모형으로부터 계산된 사후 확률을 이용하여 칼라 영상을 분할한다. 실험 결과를 통하여 모폴로지 연산이 혼합모형의 수를 자동으로 결정하고 각 성분의 모드를 계산하는데 아주 효율적인 방법임을 보였고, 또한 결정적 어닐링 EM 알고리즘에 의하여 추정된 가우시안 혼합 모형을 사용하여 계산된 사후 확률에 의한 영상 분할 방법이 기존의 분할 알고리즘보다 정확한 분할 방법임을 보였다. In this paper, we present a new segmentation algorithm for color images based on mathematical morphology and a Gaussian mixture model(GMM). We use the morphological operations to determine the number of components in a mixture model and to detect their modes of each mixture component. Next, we have adopted the GMM to represent the probability distribution of color feature vectors and used the deterministic annealing expectation maximization (DAEM) algorithm to estimate the parameters of the GMM that represents the multi-colored objects statistically. Finally, we segment the color image by using posterior probability of each pixel computed from the GMM. The experimental results show that the morphological operation is efficient to determine a number of components and initial modes of each component in the mixture model. And also it shows that the proposed DAEM provides a global optimal solution for the parameter estimation in the mixture model and the natural color images are segmented efficiently by using the GMM with parameters estimated by morphological operations and the DAEM algorithm.

A Study on Tool Breakage Detection During Milling Process Using LSTM-Autoencoder and Gaussian Mixture Model

Jun Sik Nam,Won Tae Kwon 한국정밀공학회 2022 International Journal of Precision Engineering and Vol.23 No.6

In the milling process, a rotating cutting tool is used to cut the raw material into the desired shape. Since tool breakage adversely affects productivity, real-time tool breakage detection is required. In this study, a tool breakage monitoring system using AE signals and a deep learning model was investigated. First, LSTM-Autoencoder was constructed and trained using the AE signal, cutting speed, spindle speed, and depth of cut as input data. In order to distinguish between tool normality and anomalies, the largest value among the normal cutting data set was determined as the threshold to determine if the tool was broken. As the result of the experiment, we obtained the accuracy of 82.1% during normal cutting, but the accuracy was significantly reduced to 63.1% and 63.6% at the time of entry/exit. This is because the AE value that occurs during normal entry/exit is so large that it is mistaken for breakage. To overcome this problem, a combined model that uses both LSTMAutoencoder and Gaussian Mixture Model was developed. First LSTM-Autoencoder was used to determine the breakage, and then Gaussian Mixture Model was used to determine the authenticity of the breakage. As a result of the experiment using the developed model, 52 out of 57 cuttings including entry/exit cutting were detected as failures, showing a high reliability of 91.2%, proving the superiority of the combined model.

가우시안 혼합모델 기반 탄종별 K2 소화기의 약실압력 모델링

김종환,이병학,김경민,신규용,이원우 한국군사과학기술학회 2019 한국군사과학기술학회지 Vol.22 No.1

This paper presents a chamber pressure model development of K2 rifle by applying Gaussian mixture model. In order to materialize a real recoil force of a virtual reality shooting rifle in military combat training, the chamber pressure which is one of major components of the recoil force needs to be investigated and modeled. Over 200,000 data of the chamber pressure were collected by implementing live fire experiments with both K100 and M193 of 5.56 mm bullets. Gaussian mixture method was also applied to create a mathematical model that satisfies nonlinear, asymmetry, and deviations of the chamber pressure which is caused by irregular characteristics of propellant combustion. In addition, Polynomial and Fourier Regression were used for comparison of results, and the sum of squared errors, the coefficient of determination and root-mean-square errors were analyzed for performance measurement.

Anomalous Event Detection in Traffic Video Based on Sequential Temporal Patterns of Spatial Interval Events

( Ashok Kumar P. M ),( Vaidehi. V ) 한국인터넷정보학회 2015 KSII Transactions on Internet and Information Syst Vol.9 No.1

Detection of anomalous events from video streams is a challenging problem in many video surveillance applications. One such application that has received significant attention from the computer vision community is traffic video surveillance. In this paper, a Lossy Count based Sequential Temporal Pattern mining approach (LC-STP) is proposed for detecting spatio-temporal abnormal events (such as a traffic violation at junction) from sequences of video streams. The proposed approach relies mainly on spatial abstractions of each object, mining frequent temporal patterns in a sequence of video frames to form a regular temporal pattern. In order to detect each object in every frame, the input video is first pre-processed by applying Gaussian Mixture Models. After the detection of foreground objects, the tracking is carried out using block motion estimation by the three-step search method. The primitive events of the object are represented by assigning spatial and temporal symbols corresponding to their location and time information. These primitive events are analyzed to form a temporal pattern in a sequence of video frames, representing temporal relation between various object`s primitive events. This is repeated for each window of sequences, and the support for temporal sequence is obtained based on LC-STP to discover regular patterns of normal events. Events deviating from these patterns are identified as anomalies. Unlike the traditional frequent item set mining methods, the proposed method generates maximal frequent patterns without candidate generation. Furthermore, experimental results show that the proposed method performs well and can detect video anomalies in real traffic video data.

Online nonparametric Bayesian analysis of parsimonious Gaussian mixture models and scenes clustering

Zhou, Ri-Gui,Wang, Wei Electronics and Telecommunications Research Instit 2021 ETRI Journal Vol.43 No.1

The mixture model is a very powerful and flexible tool in clustering analysis. Based on the Dirichlet process and parsimonious Gaussian distribution, we propose a new nonparametric mixture framework for solving challenging clustering problems. Meanwhile, the inference of the model depends on the efficient online variational Bayesian approach, which enhances the information exchange between the whole and the part to a certain extent and applies to scalable datasets. The experiments on the scene database indicate that the novel clustering framework, when combined with a convolutional neural network for feature extraction, has meaningful advantages over other models.