Monitoring, diagnostic and prognostic methods - VI
Chair: Prof. Yun Seok HEO (Keimyung University)
Development of Real-Time Driver's Health Detection System by Using Smart Handle Prof. Yun Seok HEO, Hao LIU, Jae-Cheon LEE and Yoon Nyun KIM (Keimyung University)
The number of car accidents due to driver inattention continues to increase, and driver's drowsiness and fatigue have become one of the major causes of serious traffic accidents. Moreover, some serious disease, such as heart attack, of drivers would lead to fatal traffic accidents. Therefore， prompt and effective detection for driver's healthy while driving is crucial to improvement of traffic safety. A set of real-time health detection system built-in a smart handle for drivers is proposed in the research. The proposed detection system of the smart handle is composed of three biological sensors, an interface circuit, a data acquisition (DAQ) board, and a personal computer (PC), which can monitor driver’s biological signals, including respiration, the gripping force, a photoplethysmogram (PPG), and electrocardiogram (ECG). The respiration and gripping force signals are obtained from pressure sensors attached to the seat belt and steering wheel, respectively. The PPG signal is acquired from a PPG sensor that consists of a light-emitting diode (LED) and a phototransistor (PT) attached on the steering wheel. And the ECG signal is obtained two the probes attached on the steering wheel and a probe on backrest. Based on these monitored signals, driver's healthy condition is able to be detected via the developed algorithm. The driver's drowsiness, or driver’s state of vigilance, is firstly judged by respiration, gripping force, and PPG signals. Meanwhile the driver's state of arrhythmia is also can be detected by analyzing R-wave duration and R-R interval of ECG signal. The test results indicate that the developed real-time driver’s health detection system can effectively monitor not only the state of vigilance but also the state of arrhythmia of a driver. It is expected that the results could be advanced to the automotive driver's prognostic health management in near future.
Unsupervised Fault Diagnosis of Journal Bearing Rotor System with Heterogeneous Data Prof. Hyunseok OH, Byung Chul JEON, Joon Ha JUNG and Byeng D. YOUN (Gwangju Institute of Science and Technology, Air Force Logistic Command, Republic of Korea, Seoul National University)
In general, it is extremely difficult to obtain failure data of real systems in the field such as power plant rotor systems. To accommodate the dearth of field failure data, conventional approaches employ data collected from a testbed that emulates the normal and faulty conditions of the real systems. Nevertheless, it is obvious that approaches developed with failure data solely collected from a testbed may not be ideal to diagnose the real systems. To this end, this paper proposes a unsupervised fault diagnostic approach for journal bearing systems that incorporates heterogeneous data from the testbed and real field systems. To demonstrate the validity of the proposed approach, a case study is conducted with the RK4 rotor kit and the power plant journal bearing system. The combination of vibration image generation with deep learning helps us use data from systems with an identical working principle but different scales. We anticipate that, by incorporating the heterogeneous data, the proposed approach can diagnose the conditions of actual journal bearing systems in the field more accurately.
An online hybrid prognostics ANFIS-PF method with an application to gearbox for RUL prediction Ms. Atefeh GOVAHIANJAHROMI, Jaehoon KIM, Moussa HAMADACHE and Dongik LEE (Kyungpook National University, Seoul National University)
Rotary components are dealing with performance degradation phenomenon, which contains the massage of unexpected damages. Therefore, prognostics and health management (PHM), has been introduced to calculate and predict the remaining useful life (RUL) in order to prevent costly damages or repairs. Data-driven, model-based and hybrid-based techniques are three main categories of PHM techniques. From the health monitoring view, the main idea is to use the experimental run-to-failure data as an intelligence-based model for our gearbox and predict the RUL with the probability (model) based method (Particle Filtering). First place, to perform our prognostics technique, we require the degradation information from gearbox, then in the period 10-days, we conduct a run-to-failure experiment for a test bench with initiative fault injection in day 7th. Period of last-three-days is considered as run-to-failure signal for proposed algorithm. After preprocessing the data, we apply a combined prediction method ANFIS-PF, using Adaptive Neuro Fuzzy Inference System (ANFIS) and Particle Filtering (PF). ANFIS used as a prediction model tool, while the particle filter method was used to find a step-ahead behavior of the gear. ANFIS as a powerful data-driven method will model the prediction of degradation data and finally this model is applied to particle filtering to predict a-step-ahead of the gear behavior until failure will happen. Meanwhile, some important signal characteristics known as condition indicators (CIs) have been extracted from the residual, energy, frequency based data processing. Then, the energy-based health index (Health Index) is calculated using threshold and sum of distributions, to show the degradation trend of tested gearbox. The online prediction results properly demonstrate the performances of the proposed ANFIS-PF algorithm, to predict the RUL of gearbox system with a 95% confidence boundary distribution.
State Estimation of Degradation Process Subject to Random Change of Mode Dr. Yan-Hui LIN, Xiao-Yang LI and Rui KANG (Beihang University)
Due to changes of the surrounding environment, the dynamic of one degradation process may change at random time and it follows different modes before and after change occurs. For solving on-line degradation state estimation problems subject to random change of mode, a novel state estimation method is proposed in this paper based on the degradation models and related monitored data. The proposed method employs sequential probability ratio test based on log-likelihood ratio to detect the unknown change time of degradation mode, and particle filtering to estimate the degradation states given observations and also to evaluate the decision functions of the sequential probability ratio test. A case study is presented to illustrate the effectiveness of the proposed method.