Hierarchical Attention Network를 이용한 복합 장애 발생 예측 시스템 개발

박영찬(Youngchan Park),안상준(Sangjun An),김민태(Mintae Kim),김우주(Wooju Kim) 한국지능정보시스템학회 2020 지능정보연구 Vol.26 No.4

The data center is a physical environment facility for accommodating computer systems and related components, and is an essential foundation technology for next-generation core industries such as big data, smart factories, wearables, and smart homes. In particular, with the growth of cloud computing, the proportional expansion of the data center infrastructure is inevitable. Monitoring the health of these data center facilities is a way to maintain and manage the system and prevent failure. If a failure occurs in some elements of the facility, it may affect not only the relevant equipment but also other connected equipment, and may cause enormous damage. In particular, IT facilities are irregular due to interdependence and it is difficult to know the cause. In the previous study predicting failure in data center, failure was predicted by looking at a single server as a single state without assuming that the devices were mixed. Therefore, in this study, data center failures were classified into failures occurring inside the server (Outage A) and failures occurring outside the server (Outage B), and focused on analyzing complex failures occurring within the server. Server external failures include power, cooling, user errors, etc. Since such failures can be prevented in the early stages of data center facility construction, various solutions are being developed. On the other hand, the cause of the failure occurring in the server is difficult to determine, and adequate prevention has not yet been achieved. In particular, this is the reason why server failures do not occur singularly, cause other server failures, or receive something that causes failures from other servers. In other words, while the existing studies assumed that it was a single server that did not affect the servers and analyzed the failure, in this study, the failure occurred on the assumption that it had an effect between servers. In order to define the complex failure situation in the data center, failure history data for each equipment existing in the data center was used. There are four major failures considered in this study: Network Node Down, Server Down, Windows Activation Services Down, and Database Management System Service Down. The failures that occur for each device are sorted in chronological order, and when a failure occurs in a specific equipment, if a failure occurs in a specific equipment within 5 minutes from the time of occurrence, it is defined that the failure occurs simultaneously. After configuring the sequence for the devices that have failed at the same time, 5 devices that frequently occur simultaneously within the configured sequence were selected, and the case where the selected devices failed at the same time was confirmed through visualization. Since the server resource information collected for failure analysis is in units of time series and has flow, we used Long Short-term Memory (LSTM), a deep learning algorithm that can predict the next state through the previous state. In addition, unlike a single server, the Hierarchical Attention Network deep learning model structure was used in consideration of the fact that the level of multiple failures for each server is different. This algorithm is a method of increasing the prediction accuracy by giving weight to the server as the impact on the failure increases. The study began with defining the type of failure and selecting the analysis target. In the first experiment, the same collected data was assumed as a single server state and a multiple server state, and compared and analyzed. The second experiment improved the prediction accuracy in the case of a complex server by optimizing each server threshold. In the first experiment, which assumed each of a single server and multiple servers, in the case of a single server, it was predicted that three of the five servers did not have a failure even though the actual failure occurred. However, assuming multiple servers

암반절개사면의 붕괴위험도 판정

이창우 ( Lee Chang U ) 한국산림과학회 2004 한국산림과학회지 Vol.93 No.1

This study was carried out to make the score table for prediction of slope failure on cutting slope through investigating effect factors on slope failure at 150 slopes in the whole country. As a result, The effective factors of slope failure are showed in order as follows; technical level of slope control works, fill condition of discontinuity surface, slope form, direction of discontinuity surface to slope direction, record of slope failure, slope degree, height of slope, and weathering degree of bedrock. The result suggested that the cutting slope will be cut under 16m slope height and maintained under 62 degree of slope as much as possible in order to reduce slope failure. The score table for prediction of slope failure was made by the Quantification theory( I ), which is well corresponded with dangerous degree of slope failure (R²=0.48), and the verification ratio of the score table shows 62.5%. The score table for prediction of slope failure was evaluated to stable prediction method and will be useful for judgement of slope stability.

수산기업의 부실화 요인 및 예측에 관한 연구

이윤원,장창익,홍재범 한국수산경영학회 2007 추계학술발표회 Vol.2007 No.-

The objectives of this paper are to identify the causes of the corporate distress and to develop a distress prediction model with the financial information in fishery industry. In this study, the corporate distress is defined as economic failure and technical insolvency. Economic failure occurs by reduction, shut-down, or change of the business and technical insolvency results from failure to pay the financial debt of companies. The 33 distressed firms from 1991 to 2003 were composed by 14 economic failure companies, 15 technical insolvency companies. 4 companies applied to the both cases. The analysis of distress prediction of fishery companies were accomplished according to the distress definition. The analysis was carried out as two steps. The first step was the univariate analysis, which was used for checking the prediction power of individual financial variable. The t-test is used to identify the differences in financial variables between the distressed group and the non-distressed group. The second step was to develop distress prediction model with logistic regression. The variables showed the significant difference in univariate analysis were selected as the prediction variables. The financial ratios, used in the logistic regression model, were selected by backward elimination method. To test stability of the distress prediction model, the whole sample was divided as three sub-samples, period l(1990~1993), period 2(1994~1997), period 3(1998~2002). The final model built from whole sample appled each three sub-samples. The results of the logistic analysis were as follows. the growth, profitability, stability ratios showed the significant effect on the distress. the some different result was found in the sub-sample (economic failure and technical insolvency). The growth and the profitability were important to predict the economic failure. The profitability and the activity were important to predict technical insolvency. It means that profitability is the really important factor to the fishery companies.

단기간 고장 데이터를 활용한 1차원 합성곱 오토인코더 기반 장기간 고장 수 예측

김성목(Seong-Mok Kim),이정오(Jeong-Oh Lee),김용수(Yong Soo Kim) 한국신뢰성학회 2024 신뢰성응용연구 Vol.24 No.2

Failure simulation of nuclear pressure vessel under severe accident conditions: Part II – Failure modeling and comparison with OLHF experiment

박의균,박준원,김윤재,Takahashi Yukio,임국희,김응수 한국원자력학회 2023 Nuclear Engineering and Technology Vol.55 No.11

This paper proposes strain-based failure model of A533B1 pressure vessel steel to simulate failure, followed by application to OECD lower head failure (OLHF) test simulation for experimental validation. The proposed strainbased failure model uses simple constant and linear functions based on physical failure modes with the critical strain value determined either using the lower bound of true fracture strain or using the average value of total elongation depending on the temperature. Application to OECD Lower Head Failure (OLHF) tests shows that progressive deformation, failure time and failure location can be well predicted.

신뢰도 예측 모델 개선방안에 대한 연구

김소정(So Jung Kim),서양우(Yang Woo Seo),이승상(Seung Sang Lee),김정태(Jung Tae Kim) 한국산학기술학회 2021 한국산학기술학회논문지 Vol.22 No.10

본 연구에서는 신뢰도 예측의 정확성을 향상시키기 위해 개선된 신뢰도 예측 모델을 제안한다. 예측의 정확성을 높이기 위해, 예측하고자 하는 대상의 BOM과 기술자료를 확보한 후, 첫째, 필드 데이터와 유사 장비 데이터를 사용하여 고장률 값을 예측한다. 필드 데이터가 가장 우선시 되어야 하며 없는 경우 유사 장비 데이터를 사용한다. 둘째, 앞선 데이터가 존재하지 않는 경우, 예측 규격을 활용한다. 전자 부품은 복합스트레스의 경우, 고장률 혼합방법 사용을 가장 우선시하며 단일스트레스인 경우, MIL-HDBK-217F를 우선시하여 사용한다. 기계 부품은 NSWC를 우선시하여 사용한다. 신뢰도 예측 후 commercial일 경우 ANSI/VITA를 적용해 팩터값을 보정해준다. 셋째, 고장의 특성을 고려해주기 위해 FMD, 환경조건보정(MIL-HDBK-338B), NOC(TO 00-20-2)를 이용하여 신뢰도 예측값을 보정하여 MTBF를 산출한다. 제시한 신뢰도 예측 모델과 A 시스템 분석보고서의 데이터를 활용하여 정확성을 검증하였다. A 시스템 분석보고서의 데이터에 대하여 검증한 결과, 기존 분석값보다 더 정확한 값을 확인하였다. This paper proposes an improved reliability prediction model to improve the accuracy of a reliability prediction. The BOM and technical data of the target to predict were obtained to improve the accuracy of the prediction. Field data and similar equipment data were used to predict the failure rate values. Field data should be a top priority, and if not, similar equipment data should be used. Second, the prediction specification was utilized if the preceding data did not exist. For electronic components, failure rate mixing models were prioritized in the presence of complex stress, and MIL-HDBK-217F was prioritized in the presence of single stress. Mechanical components used the NSWC. After performing a reliability prediction, ANSI/VITA was used to correct the factor value if the part was commercial. Third, to consider the failure properties, the reliability prediction value was corrected using the FMD, environment condition correction, and NOC. The proposed reliability prediction model and the data from the A-system analysis report were used to verify its accuracy. The data in the A-system analysis report confirmed a more accurate value than the current analysis value.

감사의견, 감사법인 및 기업부실리스크의 예측

김경철,김용덕 한국경영컨설팅학회 2022 경영컨설팅연구 Vol.22 No.1

Corporate failure internally affects shareholders, creditors, or managers, and externally, it has a large negative impact on society as a whole, such as a series of bankruptcies in related industries, inducing mass unemployment, and extinguishing accumulated technology and knowledge. Thus, the main purpose of this study is to analyze whether audit opinions of accounting firms provide useful information for predicting corporate failure risks. Previous studies predicted corporate bankruptcy based on the model mostly using financial information. But this study defines corporate bankruptcy and aims to analyze the prediction model of corporate failure including audit opinions and audit firm as well as financial variables through the logistic model. The study found that the audit opinion had a statistically significant negative effect on corporate failure one year before failure. On the other hand, the size of the audit firm did not significantly affect corporate failure. This suggests the audit opinion is judged to have an information effect about corporate bankruptcy and the results of the study would provide useful information to the investors. 기업의 부실화는 내부적으로는 주주, 채권자 또는 경영자에게 영향을 미치며 외부적으로는 관련업계의 연쇄부도, 대량실업의 유발, 축적된 기술과 지식의 소멸 등 사회전반에 미치는 부정적 영향이 크다. 이에 따라 본 연구는 감사의견에 관한 정보가 기업의 부실리스크를 예측하는데 유용한 정보를 제공하는 지를 분석하고자 하는 것이 주된 목적이다. 선행연구들이 재무적 정보를 기초로 한 기업부실리스크 모형을 예측하였으나 본 연구에서는 도산기업에 대한 정의를 내리고 이를 토대로 비재무적 정보인 감사의견 및 감사법인의 도산예측가능성을 중심으로 투자자에게 유용한 정보를 제공함을 시도한다. 본 연구에서는 우선 35개의 재무비율 변수 중 통계적으로 부실기업과 건전기업 간의 유의한 차이가 있는 재무비율들을 선정하고 이들을 기초로 기업의 부실예측모형을 구축하였으며, 이를 통하여 로짓회귀분석을 실시하였다. 본 연구에서 감사의견은 부실 1년 전에 기업부실 예측에 유의한 음(-)의 영향을 미치는 것으로 나타났다. 반면에 감사법인의 규모는 기업부실예측에 유의한 영향을 미치지 못한 것으로 나타났다. 결론적으로, 본 연구에서는 감사의견은 정보효과가 존재하는 것으로 판단된다.

EHMM-CT: An Online Method for Failure Prediction in Cloud Computing Systems

( Weiwei Zheng ),( Zhili Wang ),( Haoqiu Huang ),( Luoming Meng ),( Xuesong Qiu ) 한국인터넷정보학회 2016 KSII Transactions on Internet and Information Syst Vol.10 No.9

The current cloud computing paradigm is still vulnerable to a significant number of system failures. The increasing demand for fault tolerance and resilience in a cost-effective and device-independent manner is a primary reason for creating an effective means to address system dependability and availability concerns. This paper focuses on online failure prediction for cloud computing systems using system runtime data, which is different from traditional tolerance techniques that require an in-depth knowledge of underlying mechanisms. A `failure prediction` approach, based on Cloud Theory (CT) and the Hidden Markov Model (HMM), is proposed that extends the HMM by training with CT. In the approach, the parameter ω is defined as the correlations between various indices and failures, taking into account multiple runtime indices in cloud computing systems. Furthermore, the approach uses multiple dimensions to describe failure prediction in detail by extending parameters of the HMM. The likelihood and membership degree computing algorithms in the CT are used, instead of traditional algorithms in HMM, to reduce computing overhead in the model training phase. Finally, the results from simulations show that the proposed approach provides very accurate results at low computational cost. It can obtain an optimal tradeoff between `failure prediction` performance and computing overhead.

코스닥기업의 실패요인과 주가반응에 관한 연구

남삼현(Sam-Hyun Nam),오성배(Sung-Bea Oh),이화득(Hwa-Deuk Yi) 대한경영학회 2008 大韓經營學會誌 Vol.21 No.6

BIS비율과 부채비율

강선민(Sun Min Kang),황인태(In Tae Hwang),Shun Ji Jin 한국경영학회 2013 經營學硏究 Vol.42 No.1

In 2011, many mutual savings banks suspended their operations. The suspension of mutual savings banks was a critical decision that did serious damage to the national economy and various interested parties, such as depositors. The BIS capital ratio divides the capital of a savings bank into risk-weighted assets, and if the ration of a savings bank is more than 8%, the bank is classed as a highly successful savings bank. However, the BIS capital ratio for many suspended savings banks in the previous year was higher than 8%. The purpose of this study is to analyze whether the BIS capital ratio, which measures the soundness of savings banks, is a suitable indicator for predicting their failure. Moreover, the study examines whether the debt ratio used for predicting corporate bankruptcy can be used as an indicator for predicting the failure of savings banks, comparing it with the BIS capital ratio. This study investigates 36 savings banks that suspended their operations from 2004 to 2012, conducting a logit analysis for predicting failure using 73 savings banks as the paired sample from among the 82 that were in full operation. The results of the study are as follows. The failure prediction model for the BIS capital ratio was only significant in the previous year before the savings banks suspended their operations. However, the debt ratio was statistically significant over two-year periods. The failure prediction information must be timely, so that it helps catch signs of failure at an early stage, as well as accurate. Thus, using the debt ratio with the BIS capital ratio is suitable for predicting the failure of savings banks. Financial regulators and investors need to consider the debt ratio, as well as the BIS capital ratio, when they predict and measure the suspension of savings banks. This study suggests, according to the results of empirical study, that the debt ratio is a valid variable for catching signs of potential suspension at an early stage. Moreover, the results from this study have universality for all savings banks, as the study considered all savings banks and included almost all the savings banks in the sample. Thus, I expect that the study will help financial regulators and other interested parties respond to the possible failure of savings banks at an early stage, and thus contribute to minimizing the social costs of suspension.