• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.77-91
• /
• 2022

Various monitoring systems have been implemented in civil infrastructure to ensure structural safety and integrity. In long-term monitoring, these systems generate a large amount of data, where anomalies are not unusual and can pose unique challenges for structural health monitoring applications, such as system identification and damage detection. Therefore, developing efficient techniques is quite essential to recognize the anomalies in monitoring data. In this study, several machine learning techniques are explored and implemented to detect and classify various types of data anomalies. A field dataset, which consists of one month long acceleration data obtained from a long-span cable-stayed bridge in China, is employed to examine the machine learning techniques for automated data anomaly detection. These techniques include the statistic-based pattern recognition network, spectrogram-based convolutional neural network, image-based time history convolutional neural network, image-based time-frequency hybrid convolution neural network (GoogLeNet), and proposed ensemble neural network model. The ensemble model deliberately combines different machine learning models to enhance anomaly classification performance. The results show that all these techniques can successfully detect and classify six types of data anomalies (i.e., missing, minor, outlier, square, trend, drift). Moreover, both image-based time history convolutional neural network and GoogLeNet are further investigated for the capability of autonomous online anomaly classification and found to effectively classify anomalies with decent performance. As seen in comparison with accuracy, the proposed ensemble neural network model outperforms the other three machine learning techniques. This study also evaluates the proposed ensemble neural network model to a blind test dataset. As found in the results, this ensemble model is effective for data anomaly detection and applicable for the signal characteristics changing over time.

• 한국정보통신학회논문지
• /
• 제25권11호
• /
• pp.1649-1654
• /
• 2021

본 논문에서는 고령자의 낙상상황을 감지할 수 있는 텐서플로우 장단기 메모리 기반 낙상감지 시스템의 정규화에 대하여 소개한다. 낙상감지는 고령자의 몸에 부착한 3축 가속도 센서 데이터를 사용하며, 총 7가지의 행동 패턴들에 대하여 학습하며, 각각 4가지는 일상생활에서 일어나는 패턴이고, 나머지 3가지는 낙상에 대한 패턴이다. 학습시에는 손실함수(loss function)를 효과적으로 줄이기 위하여 정규화 과정을 진행하며, 정규화 과정은 데이터에 대하여 최대최소 정규화, 손실함수에 대하여 L2 정규화 과정을 진행한다. 3축 가속도 센서를 이용하여 구한 다양한 파라미터에 대하여 정규화 과정의 최적의 조건을 제시한다. 낙상 검출율면에서 SVM을 이용하고 정규화 127과 정규화율 λ 0.00015일 때 Sensitivity 98.4%, Specificity 94.8%, Accuracy 96.9%로 가장 좋은 모습을 보였다.

• Smart Structures and Systems
• /
• 제33권2호
• /
• pp.93-103
• /
• 2024

Despite the rapid development of sensors, structural health monitoring (SHM) still faces challenges in monitoring due to the degradation of devices and harsh environmental loads. These challenges can lead to measurement errors, missing data, or outliers, which can affect the accuracy and reliability of SHM systems. To address this problem, this study proposes a classification method that detects anomaly patterns in sensor data. The proposed classification method involves several steps. First, data scaling is conducted to adjust the scale of the raw data, which may have different magnitudes and ranges. This step ensures that the data is on the same scale, facilitating the comparison of data across different sensors. Next, informative features in the time and frequency domains are extracted and used as input for a deep neural network model. The model can effectively detect the most probable anomaly pattern, allowing for the timely identification of potential issues. To demonstrate the effectiveness of the proposed method, it was applied to actual data obtained from a long-span cable-stayed bridge in China. The results of the study have successfully verified the proposed method's applicability to practical SHM systems for civil infrastructures. The method has the potential to significantly enhance the safety and reliability of civil infrastructures by detecting potential issues and anomalies at an early stage.

• 한국운동역학회지
• /
• 제15권1호
• /
• pp.75-89
• /
• 2005

D. K. LEE, J. S. LEE, B. J. LEE, H. S. LEE, Y. J. KIM, S. B. PARK, J. P. JOO. Plantar foot pressure analysis during golf swing motion using plantar foot pressure measurement system. Korean Journal of Sport Biomechanics, Vol. 15, No. 1, pp. 75-89, 2005. In this study, weight carrying pattern analysis and comparison method of four foot region were suggested. We used three types of club(driver, iron7, pitching wedge). This analysis method can compare between top class golfer and beginner. And the comparison data can be used to correct the swing pose of trainee. If motion analysis system, which can measure the swing speed and instantaneous acceleration at the point of hitting a ball, is combined with this plantar foot force analysis method, new design development of golf shoes to increase comfort and ball flight distance will be available. 1. Address acting, forces concentrated in rare foot regions and lateral foot of right foot. Back swing top acting, relatively high force occurred in medial forefoot region of left foot and forefoot region of right foot. Impact acting, high force value observed in the lateral rarefoot region of left foot and medial forefoot region of right foot. Finish acting, force concentration observed on the lateral region and rarefoot region of left foot. 2. Forces were increased in address of right foot with clubs length increased. All clubs, back swing top acting, high force value observed in the lateral forefoot region of right foot. All clubs, in impact, high force value observed in the lateral rarefoot region of left foot and medial forefoot region of right foot. Finish acting, force concentration observed on the rarefoot region in driver and lateral foot region in iron on left foot. 3. Right foot forces distribution were increased in address, back swing top and left foot force distribution were increased in impact, finnish

• 한국전기전자재료학회논문지
• /
• 제30권12호
• /
• pp.768-775
• /
• 2017

Energy harvesting characteristics of trapezoidal piezoelectric cantilever generator, which has a lead zirconate titanate (PZT) laminate film, were compared by longitudinal (3-3) and transverse (3-1) modes. The PZT laminate film, fabricated by a conventional tape casting process, was cofired with Ag electrode at $850^{\circ}C$ for 2 h. A multi-layered Ag electrode by a planar pattern and an interdigitated pattern was applied to the PZT laminate to implement the 3-3 and 3-1 modes, respectively. The energy harvesting performance of the 3-3 mode trapezoidal piezoelectric cantilever generator was better than that of the 3-1 mode. An extremely high output power density of $26.7mW/cm^3$ for the 3-3 mode was obtained at a resonant frequency of 145 Hz under a load resistance of $50{\Omega}$ and acceleration of 1.3 G, which is ~3-times higher than that for the 3-1 mode. Therefore, the 3-3 mode is considered significantly efficient for application to high-performance piezoelectric cantilever generator.

• 한국운동역학회지
• /
• 제25권3호
• /
• pp.323-333
• /
• 2015

Objective : The purpose of this study was to investigate the changes of gait patterns and muscle activations with increased loads during stair walking. Also, it can be used as descriptive data about continuous stair walking in a real life setting. Method : Twelve sedentary young male adults(Age: $27.0{\pm}1.8yrs$, Weight: $65.8{\pm}9.9kg$) without any lower extremity injuries participated in this study. Participants performed stair walking up 7 floors and their ascending and descending motion on each floor was analyzed. A wireless electromyography(EMG) were attached on the Rectus Femoris(RF), Biceps Femoris(BF), Gastrocnemius(GN), Tibialis Anterior(TA) muscle to calculate integrated EMG(iEMG), median frequency(MDF) and co-contraction index(CI). Chest and left heel accelerometer signal were recorded by wireless accelerometer and those were used to calculate approximate entropy(ApEn) for analyzing gait pattern. All analyses were performed with SPSS 21.0 and for repeated measured ANOVA and Post-hoc was LSD. Results : During ascending stairs, there were a statistically significant difference in Walking time between 1-2nd and other floors(p=.000), GN iEMG between 2-3th and 6-7th(p=.043) floor, TA MDF between 1-2nd and 5-6th(p=.030), 6-7th(p=.015) floor and TA/GN CI between 2-3th and 6-7th(p=.038) floor and ApEn between 1-2nd and 6-7th(x: p=.003, y: p=.005, z: p=.006) floor. During descending stairs, there were a statistically significant difference in TA iEMG between the 6-5th and 3-2nd(p=.026) floor, and for the ApEn between the 1-2nd and 6-7th(x: p=.037, y: p=.000, z: p=.000) floor. Conclusion : Subjects showed more regular pattern and muscle activation response caused by regularity during ascending stairs. Regularity during the first part of stair-descending could be a sign of adaptation; however, complexity during the second part could be a strategy to decrease the impact.

• 대한토목학회논문집
• /
• 제30권3B호
• /
• pp.269-276
• /
• 2010

최근 지구온난화가 가속화되면서 전 세계적으로 기록적인 기상재해가 급증하고 있다. 특히 강우패턴의 변화로 인하여 강우강도가 증가하여 집중호우의 발생빈도가 높아지고 있다. 본 연구에서는 기후변화 시나리오에 따른 강우패턴의 변화를 반영하여 목표연도 확률강우량을 산정하는 비정상성 강우빈도해석법을 제안하였다. BCM2 모형(A2 시나리오)과 NCEP 자료를 K-NN 축소기법을 사용하여 축소시킨 연 총 강우량을 이용하여 연 최대 강우량 평균, 연 최대 강우량 평균과 매개변수 간 통계학적 관계를 분석하여 목표연도 확률강우량을 산정하였다. 분포형은 Gumbel 분포를 사용하였으며 매개변수 추정법은 확률가중모멘트법을 사용하였다. 국내에서 가장 긴 관측 강우자료를 가진 서울지점을 대상으로 모형의 적합성 검증을 실시하였으며, 2006년 현재 통계학적으로 증가경향성을 가진 7개의 강우관측지점에 적용한 결과를 분석하였다. 기후변화 시나리오에 따른 지역적 연 총 강우량의 변화는 미래 확률강우량의 증감에 영향을 미치는 것으로 나타났다.

• 한국운동역학회지
• /
• 제16권3호
• /
• pp.149-163
• /
• 2006

The purpose of this study was to compare the lower extremity's joint and segment coupling patterns between forward and backward running in subjects who were twelve healthy males. Three-dimensional kinematic data were collected with Qualisys system while subjects ran to forward and backward. The thigh internal/external rotation and tibia internal/external rotation, thigh flexion/extension and tibia flexion/extension, tibia internal/external rotation and foot inversion/eversion, knee internal/external rotation and ankle inversion/eversion, knee flexion/extension and ankle inversion/eversion, knee flexion/extension and ankle flexion/extension, and knee flexion/extension and tibia internal/external rotation coupling patterns were determined using a vector coding technique. The comparison for each coupling between forward and backward running were conducted using a dependent, two-tailed t-test at a significant level of .05 for the mean of each of five stride regions, midstance(1l-30%), toe-off(31-50%), swing acceleration(51-70%), swing deceleration(71-90), and heel-strike(91-10%), respectively. 1. The knee flexion/extension and ankle flexion/extension coupling pattern of both foreward and backward running over the stride was converged on a complete coordination. However, the ankle flexion/extension to knee flexion/extension was relatively greater at heel-strike in backward running compared with forward running. At the swing deceleration, backward running was dominantly led by the ankle flexion/extension, but forward running done by the knee flexion/extension. 2. The knee flexion/extension and ankle inversion/eversion coupling pattern for both running was also converged on a complete coordination. At the mid-stance. the ankle movement in the frontal plane was large during forward running, but the knee movement in the sagital plane was large during backward running and vice versa at the swing deceleration. 3. The knee flexion/extension and tibia internal/external rotation coupling while forward and backward run was also centered on the angle of 45 degrees, which indicate a complete coordination. However, tibia internal/external rotation dominated the knee flexion/extension at heel strike phase in forward running and vice versa in backward running. It was diametrically opposed to the swing deceleration for each running. 4. Both running was governed by the ankle movement in the frontal plane across the stride cycle within the knee internal/external rotation and tibia internal/external rotation. The knee internal/external rotation of backward running was greater than that of forward running at the swing deceleration. 5. The tibia internal/external rotation in coupling between the tibia internal/external rotation and foot inversion/eversion was relatively great compared with the foot inversion/eversion over a stride for both running. At heel strike, the tibia internal/external rotation of backward running was shown greater than that of forward(p<.05). 6. The thigh internal/external rotation took the lead for both running in the thigh internal/external rotation and tibia internal/external rotation coupling. In comparison of phase, the thigh internal/external rotation movement at the swing acceleration phase in backward running worked greater in comparison with forward running(p<.05). However, it was greater at the swing deceleration in forward running(p<.05). 7. With the exception of the swing deceleration phase in forward running, the tibia flexion/extension surpassed the thigh flexion/extension across the stride cycle in both running. Analysis of the specific stride phases revealed the forward running had greater tibia flexion/extension movement at the heel strike than backward running(p<.05). In addition, the thigh flexion/extension and tibia flexion/extension coupling displayed almost coordination at the heel strike phase in backward running. On the other hand the thigh flexion/extension of forward running at the swing deceleration phase was greater than the tibia flexion/extension, but it was opposite from backward running. In summary, coupling which were the knee flexion/extension and ankle flexion/extension, the knee flexion/extension and ankle inversion/eversion, the knee internal/external rotation and ankle inversion/eversion, the tibia internal/external rotation and foot inversion/eversion, the thigh internal/external rotation and tibia internal/external rotation, and the thigh flexion/extension and tibia flexion/extension patterns were most similar across the strike cycle in both running, but it showed that coupling patterns in the specific stride phases were different from average point of view between two running types.

• 한국운동역학회지
• /
• 제15권1호
• /
• pp.237-257
• /
• 2005

It was to study as a following-research of "A Case Study on Center of Gravity(COG) Analysis when Performing Uchimata(inner thigh reaping throw) by Posture and Voluntary Resistance Levels(VRL) of Uke in Judo[I]". The purpose of this study was to analyze the COG variables when performing uchimata(inner thigh reaping throw) by two postures and voluntary resistance levels(VRL) of uke(reciver) in Judo. The subjects, who were one male judoka(YH) for 1992 Barcelona Olympic Games Olympian(silver medalist), and one male trainee; Y.I.University representative member (SDK), and were filmed on two S-VHS 16mm video cameras(60fields/sec.) through 3-dimensional motion analysis methods, that postures of uke were shizenhontai (straight natural posture) and jigohontai(straight defensive posture), VRL of uke were 0% and 100%, respectively. The kinematical variable was COG variable, distance of COG, and distance of resultant COG between uke and tori(the thrower), velocity and acceleration of COG. The data of this study collection were digitized by SIMI Motion Program computed the mean values and the standard deviation calculated for each variables. When performing uchinmata according to each posture and VRL of uke and classifying. From the data analysis and discussion, the conclusions were as follows : 1. Displacement of COG Subject YH, COG was the highest in kuzushi(balance -breaking), vertical COG was low when following in tsukuri(positioning; set-up), kake(application; execution), and COG was pattern of same character each postures and resistance, respectively. Subject SDK, COG was low from kumikata(engagement positioning) to kake, and COG was that each postures and resistance were same patterns, respectively. Subject YH, SDK, each individual, postures and resistance, vertical COG was the lowest in kake phase, when performing. 2. Distance of COG between uke and tori The distance of COG between uke and tori when performing, subject YH was $0.64{\sim}0.70cm$ in kumikata, $0.19{\sim}0.28cm$ in kake, and SDK was $0.68{\sim}0.72cm$ in kumikata, $0.30{\sim}0.42\;cm$ in kake. SDK was wider than YH. 3. Distance of resultant COG between uke and tori The distance of resultant COG between uke and tori when performing, subject YH was $0.27{\sim}0.73cm$ from kumikata to kake. and SDK was $0.14{\sim}0.34cm$ in kumikata, $0.28{\sim}0.65cm$ in kake. Jigohontai(YH:$0.43{\sim}0.73cm$,SDK:$0.59{\sim}0.65cm$) was more moved than shizenhontai(YH:$0.27{\sim}0.53cm$, SDK: $0.28{\sim}\;0.34cm$). 4. Velocity of COG The velocity of COG when performing uchimata, subject YH was fast anterior-posterior direction in kuzushi, ant.-post. and vertical direction fast in tsukuri and kake. SDK was lateral, ant.-post. and vertical direction in kuzushi, ant.-post. and vertical direction in tsukuri and ant.-post. direction in take, respectively. 5. Acceleration of COG The acceleration of COG when performing uchimata, The trend of subject YH was showed fast vertical direction in kuzushi and tsukuri, ant.-post. and vertical direction fast in kake. The trends of SDK showed lateral direction in kuzushi, lateral and ant.-post. direction in tsukuri and ant.-post. direction in kake, respectively.

• 한국산림과학회지
• /
• 제37권1호
• /
• pp.35-40
• /
• 1978

1975년(年) ${\times}$ P. albaglandulosa(은수원사시나무) 0/1묘(苗)를 식재(植栽)하고 방사성동위체(放射性同位體) $P^{32}$-표식(標識) 인산질비료(燐酸質肥料)의 흡수율(吸收率)과 유산고토(硫酸苦土)의 첨가(添加) 시용(施用)에 의(依)한 인산질비료(燐酸質肥料)의 흡수이용촉진효과(吸收利用促進効果)를 시시(試試)한바 다음과 같이 간추릴수가 있었다. 1. ${\times}$ P. albaglandulosa(은수원사시나무)는 구내(構內) 적색계(赤色系) 중점토(重粘土)에서 시용인산질비료(施用燐酸質肥料)의 평균(平均) 13%정도만이 흡수이용(吸收利用)하고 있었다. 2. 흡수(吸收)된 고토분(苦土分)은 줄기보다 주(主)로 잎에 많이 축적되었다. 3. 유산고토(硫酸苦土)를 첨가시용(添加施用)하므로써 시용인산(施用燐酸)의 흡수율(吸收率)은 16~33%선(線)까지 증진(增進)시킬수 있었다. 4. 유산고토(硫酸苦土)를 인산질비료(燐酸質肥料)와 병용(倂用)하므로써 인산(燐酸)의 흡수(吸收), 이용율(利用率)을 증진(增進)시킬뿐만이 아니라 엽록소(葉綠素)의 구성성분(構成成分)인 고토분(苦土分)의 엽내(葉內) 축적량증대(蓄積量增大)는 동화량촉진(同化量促進)과 연결(連結)되어 생장량(生長量)의 증대(增大)도 고려(考慮)될수가 있겠다.