• The Journal of the Convergence on Culture Technology
• /
• v.9 no.3
• /
• pp.845-850
• /
• 2023

In this paper, we propose a method for diagnosing ball bearing vibration using transfer learning. STFT, which can analyze vibration signals in time-frequency, was used as input to CNN to diagnose failures. In order to rapidly learn CNN-based deep artificial neural networks and improve diagnostic performance, we proposed a transfer learning-based deep learning learning technique. For transfer learning, the feature extractor and classifier were selectively learned using a VGG-based image classification model, the data set for learning was publicly available ball bearing vibration data provided by Case Western Reserve University, and performance was evaluated by comparing the proposed method with the existing CNN model. Experimental results not only prove that transfer learning is useful for condition diagnosis in ball bearing vibration data, but also allow other industries to use transfer learning to improve condition diagnosis.

• Journal of Sensor Science and Technology
• /
• v.29 no.4
• /
• pp.249-254
• /
• 2020

It is difficult to control children who exhibit negative behavior in dental clinics. Various methods are used for preventing pediatric dental patients from being afraid and for eliminating the factors that cause psychological anxiety. However, when it is difficult to apply this routine behavioral control technique, sedation therapy is used to provide quality treatment. When the sleep anesthesia treatment is performed at the dentist's clinic, it is challenging to identify emergencies using the current breath detection method. When a dentist treats a patient that is under the influence of an anesthetic, the patient is unconscious and cannot immediately respond, even if the airway is blocked, which can cause unstable breathing or even death in severe cases. During emergencies, respiratory instability is not easily detected with first aid using conventional methods owing to time lag or noise from medical devices. Therefore, abnormal breathing needs to be evaluated in real-time using an intuitive method. In this paper, we propose a method for identifying abnormal breathing in real-time using an intuitive method. Respiration signals were measured using a 3M Littman electronic stethoscope when the patient's posture was supine. The characteristics of the signals were analyzed by applying the signal processing theory to distinguish abnormal breathing from normal breathing. By applying a short-time Fourier transform to the respiratory signals, the frequency range for each patient was found to be different, and the frequency of abnormal breathing was distributed across a broader range than that of normal breathing. From the wavelet transform, time-frequency information could be identified simultaneously, and the change in the amplitude with the time could also be determined. When the difference between the amplitude of normal breathing and abnormal breathing in the time domain was very large, abnormal breathing could be identified.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.11 no.1
• /
• pp.71-83
• /
• 2009

Shotcrete is an important primary support for tunnelling in rock. The quality control of shotcrete is a core issue in the safe construction and maintenance of tunnels. Although shotcrete may be applied well initially onto excavated rock surfaces, it is affected by blasting, rock deformation and shrinkage and can debond from the excavated surface, causing problems such as corrosion, buckling, fracturing and the creation of internal voids. This study suggests an effective non-destructive evaluation method of the tunnel shotcrete bonding state applied onto hard rocks using the impact-echo (IE) method and ground penetration radar (GPR). To verify previous numerical simulation results, experimental study carried out. Generally, the bonding state of shotcrete can be classified into void, debonded, and fully bonded. In the laboratory, three different bonding conditions were modeled. The signals obtained from the experimental IE tests were analyzed at the time domain, frequency domain, and time-frequency domain (i.e., the Short- Time Fourier transform). For all cases in the analyses, the experimental test results were in good agreement with the previous numerical simulation results, verifying this approach. Both the numerical and experimental results suggest that the bonding state of shotcrete can be evaluated through changes in the resonance frequency and geometric damping ratio in a frequency domain analysis, and through changes in the contour shape and correlation coefficient in a time-frequency analysis: as the bonding state worsens in hard rock condition, the autospectral density increases, the geometric damping ratio decreases, and the contour shape in the time-frequency domain has a long tail parallel to the time axis. The correlation coefficient can be effectively applied for a quantitative evaluation of bonding state of tunnel shotcrete. Finally, the bonding state of shotcrete can be successfully evaluated based on the process suggested in this study.

• Science of Emotion and Sensibility
• /
• v.16 no.4
• /
• pp.527-534
• /
• 2013

In this study, we analyzed spatial-frequency relationship related brain function for change of the time during attentional visual stimulus through the analysis of Coherence. With experimentation about ERP(Event Related Potential)data, it revealed that change of the phase synchronization between different scalp locations at ${\theta}$, ${\alpha}$ band. ERP between left and right frontal lobes, between the frontal and central lobes showed the phase synchronization at the P100, N200, ERP between the frontal and occipital lobes showed the phase synchronization at the P300 related information of visual stimulus. Compared to STFT using the window of a fixed length, CWT is able to multi-resolution analysis with the adjustment of parameters of mother wavelet. Thus, coherence results with CWT was found to be effective for analysis of time-varying spatial-frequency relationship in ERP. The phase synchronization for inattentional visual stimulus was not observed.

• Journal of Electrical Engineering and Technology
• /
• v.12 no.6
• /
• pp.2378-2387
• /
• 2017

Violinists learn to make better sounds by hearing and evaluating their own playing though numerous practice. This study proposes a new method of auditory feedback, which mimics this violinists' step and verifies its efficiency using experiments. Making the desired sound quality of a violin is difficult without auditory feedback even though an expert violinist plays. An algorithm for controlling a robot arm of violin playing robot is determined based on correlations with bowing speed, bowing force, and sound point that determine the sound quality of a violin. The bowing speed is estimated by the control command of the robot arm, where the bowing force and the sound point are recognized by using a two-axis load cell and a photo interrupter, respectively. To improve the sound quality of a violin playing robot, the sounds information is obtained by auditory feedback system applied Short Time Fourier Transform (STFT) to the sounds from a violin. This study suggests Gaussian-Harmonic-Quality (GHQ) uses sounds' clarity, accuracy, and harmonic structure in order to decide sound quality, objectively. Through the experiments, the auditory feedback system improved the performance quality by the robot accordingly, changing the bowing speed, bowing force, and sound point and determining the quality of robot sounds by GHQ sound quality evaluation system.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.33 no.3
• /
• pp.264-269
• /
• 2013

Since MFC(macro-fiber composite) transducer has been developed, many researchers have tried to apply this transducer on SHM(structural health monitoring), because it is so flexible and durable that it can be easily embedded to various kinds of structures. The objective of this paper is to figure out the benefits and feasibility of applying MFC transducers to guided wave technique. For this, we have experimentally tested the performance of MFC patches as transmitter and sensors for excitation and reception of guided waves on the thin aluminum alloy plate. In order to enhance the signal accuracy, we applied the FIR filter for noise reduction as well as used STFT(short-time Fourier transform) algorithm to image the guided wave characteristics clearly. From the results, the guided wave generated based on MFC showed good agreement with its theoretical dispersion curves. Moreover, the ultrasonic Lamb wave techniques based on MFC patches in pitch-catch manner was tested for detection of surface notch defects of which depths are 10%, 20%, 30% and 40% of the aluminum plate thickness. Results showed that the notch was detectable well when the notch depth was 10% of the thickness or greater.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.32 no.7
• /
• pp.51-59
• /
• 2004

An experimental procedure to identify failure modes of impact damage using sensor signals and to analyze their general features is examined. A series of low-velocity impact tests from low energy to damage-induced high energy were performed on the instrumented drop weight impact tester to monitor the stress wave signals due to failure modes such as matrix cracking, delamination, and fiber breakage. The wavelet transform(WT) and Short Time Fourier Transform(STFT) are used to decompose the piezoelectric sensor signals in this study. The extent of the damage in each case was examined by means of a conventional ultrasonic C-scan. The PVDF sensor signals are shown to carry important information regarding the nature of the impact process that can be extracted from the careful signal processing and analysis.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.2
• /
• pp.348-369
• /
• 2024

With the advancement of Industry 4.0 and Industrial Internet of Things (IIoT), manufacturing increasingly seeks automation and intelligence. Temperature and vibration monitoring are essential for machinery health. Traditional abnormal state detection methodologies often overlook the intricate frequency characteristics inherent in vibration time series and are susceptible to erroneously reconstructing temperature abnormalities due to the highly similar waveforms. To address these limitations, we introduce synergistic, end-to-end, unsupervised Frequency-Time Domain Memory-Enhanced Autoencoders (FTD-MAE) capable of identifying abnormalities in both temperature and vibration datasets. This model is adept at accommodating time series with variable frequency complexities and mitigates the risk of overgeneralization. Initially, the frequency domain encoder processes the spectrogram generated through Short-Time Fourier Transform (STFT), while the time domain encoder interprets the raw time series. This results in two disparate sets of latent representations. Subsequently, these are subjected to a memory mechanism and a limiting function, which numerically constrain each memory term. These processed terms are then amalgamated to create two unified, novel representations that the decoder leverages to produce reconstructed samples. Furthermore, the model employs Spectral Entropy to dynamically assess the frequency complexity of the time series, which, in turn, calibrates the weightage attributed to the loss functions of the individual branches, thereby generating definitive abnormal scores. Through extensive experiments, FTD-MAE achieved an average ACC and F1 of 0.9826 and 0.9808 on the CMHS and CWRU datasets, respectively. Compared to the best representative model, the ACC increased by 0.2114 and the F1 by 0.1876.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.25 no.5
• /
• pp.765-771
• /
• 2001

Elastic-plastic fracture toughness JIC can be used a s an effective design criterion in elastic-plastic fracture mechanics. Among the JIC test methods approved by ASTM, unloading compliance method was used in this study. In order to examine the relationship between fracture behavior of JIC test and AE signals, the post processing of AE signals has been carried out by Short Time Fourier Transform(STFT), one of the time-frequency analysis methods. The objective of this study is to evaluate the application of characterization of AE signals for unloading compliance method of JIC test. As a result of time-frequency analysis, we could extract the AE from the raw signal and analyze the frequencies in AE signal at the same time. AE signal generated by elastic-plastic fracture of material has some different aspects at elastic and plastic ranges, or the first portion of crack growth by fracture. First of all, increased energy recorded and detected by using AE count method increase rapidly from the start of ductile fracture. The variation of main frequency range with time-frequency analysis method could be confirmed. We could know fracture behavior of interior material by examination AE characteristics generated in real-time when elastic-plastic fracture occurred in material under loading.

• Composites Research
• /
• v.18 no.6
• /
• pp.26-33
• /
• 2005

The mechanical properties of composite materials may severely degrade in the presence of damage. Especially, the high-velocity impact such as bird strike, a hailstorm, and a small piece of tire or stone during high taxing, can cause considerable damage to the structures and sub-system in spite of a very small mass. However, it is not easy to detect the damage in composite plates using a single technique or any conventional methods. In this paper, the PVDF(polyvinylidene fluoride) film sensors were used for monitoring high-velocity impact damage initiation and propagation in composite laminates. The WT(wavelet transform) and STFT(short time Fourier transform) are used to decompose the sensor signals. A ultrasonic C-scan and a digital microscope are also used to examine the extent of the damage in each case. This research shows how various sensing techniques, PVDF sensor in particular, can be used to characterize high-velocity impact damage in advanced composite.