Search

This study focuses on evaluating several machine learning algorithms for the condition assessment of utility timber poles. An efficient feature extraction technique combining Hilbert Huang transform (HHT) and wavelet packet transform (WPT) is adopted from the authors' previous work and implemented to determine damage-sensitive features from vibration data related to five serviceable and eight unserviceable in-situ timber poles. Then, these features are pre-processed using correlation heat map analysis for feature selection. Principal component analysis (PCA) is adopted as the final step of pre-processing for reducing noise interference and enhancing classification accuracy. Afterwards, a feature matrix is formed, which is fed into the selected classifiers for pattern recognition. In addition, the information gain method is also implemented and compared against PCA to examine the effect of feature selection. Finally, selected classifiers are employed using those dominant features, and their performance is evaluated based on six parameters – accuracy, precision, recall, F1-score, confusion matrix and Receiver Operating Characteristic (ROC) curve. It was found that choosing the best feature set related to the health state helps to improve the performance of the pattern recognition algorithms to a great extent. [ABSTRACT FROM AUTHOR]

Bridge structural health monitoring (BSHM) is extensively employed to assess whether bridge damage exceeds maintenance limits and helps maintenance decision-making. Frequency identification is an essential part of BSHM, which utilizes the first-order modal frequency as an evaluation indicator of bridge health status. This paper proposes a method for bridge damage identification based on extracting the first-order modal frequency of the bridge from dynamic responses of vehicle bogies. Firstly, the feasibility of extracting bridge modal frequency from bogie accelerations is theoretically deduced via a vehicle–bridge interaction model. The result indicates that the bogie accelerations encompass three frequency components, namely vehicle driving frequency, vehicle pitching frequency and bridge first-order modal frequency. Then, a multi-domain conjoint analysis method incorporating the time domain, frequency domain and time–frequency domain is proposed to extract the first-order frequency of damaged bridge from bogie accelerations. By defining a damage sensitivity index of the sliding window, the time-domain sensitive boundaries of bogie acceleration to bridge damage are pinpointed. Leveraging prior knowledge of the bridge modal frequency range, the frequency-domain sensitive boundaries bogie acceleration to bridge frequency are determined. On this basis, the sensitive region of bogie acceleration to bridge frequency can be precisely localized in the Hilbert Huang transform spectrum, ultimately identifying bridge frequency through searching local maximum instantaneous energy points within this region. Finally, comprehensive experiments, considering different vehicle speeds and track spectra scenarios, are carried out to investigate the effectiveness and robustness of the proposed method. The results affirm the performance of the method to accurately identify the first-order modal frequency of the damaged bridge, further underscoring the promising prospect of on-board BSHM. [ABSTRACT FROM AUTHOR]

To address the shortage of damage identification methods for modal bridge expansion joints (MBEJs). Two damage indexes, DI(IPH) and DI(HS) were constructed using the instantaneous phase (IP) and Hilbert spectrum of Hilbert Huang transform (HHT). Three specimens were designed and fabricated to identify the damage to the concrete in the center beam, support bar and anchorage zone of MBEJs. The impact vibration of MBEJs was excited by the hammer impulses and acceleration sensors were used to obtain the vibration response. The damage of the center beam, support bar and concrete in anchorage zone was all classified into five levels (level 1–level 5). DI(IPH) and DI(HS) increase gradually with increasing damage levels. After the center beam was completely broken, the DI(IPH) and DI(HS) improved by 40.31 and 40.61, respectively. The DI(IPH) and DI(HS) increased by 25.25 and 55.91 after the complete breakage of the support bar. The DI(IPH) and DI(HS) of the concrete in the anchorage zone at Level 5 showed more minor numerical changes than the center beam and support bar, with an increase of 7.49 and 35.45, respectively. The experimental results showed that HHT could effectively identify the damage in various parts of MBEJs with high accuracy. There is a clear indication of the degree of damage, which is suitable for the identification of damage in MBEJs. [ABSTRACT FROM AUTHOR]

Pumped storage is an important component of electrified wire netting. The safe and stable operation of pump turbines is extremely important. Among them, pressure pulsation is one of the main causes of pump turbine vibration. The characteristics of pressure pulsation are relatively complex, and it is difficult to directly observe their temporal changes using commonly used FFT methods. The division of frequency characteristics is often vague. Meanwhile, it is difficult to explain some phenomena such as frequency doubling. This article focuses on a certain model of pump turbine and uses SST model to numerically simulate the runaway condition of the pump turbine. And the Hilbert Huang transform method is used to analyze the pressure pulsation in the vaneless region and draft tube. The results show that the main characteristic frequencies of the vaneless region are blade passing frequency 112.5 Hz and rotational frequency 12.5 Hz. The main characteristic frequencies of the draft tube are vortex rope frequency near 3 Hz which energy ratio is up to 50%, rotational frequency, and blade passing frequency. The pressure pulsation characteristics in the vaneless region have changed from a complex composition of double blade passing frequency and rotational frequency to a distribution dominated by blade passing frequency. In the passage of the guide vane, the pressure pulsation is almost only characterized by blade passing frequency. The frequency characteristics of the vaneless region between the runner and the guide vane become complex again. Meanwhile, the results show that the characteristic frequencies of the vaneless region and the draft tube propagate upstream and downstream. [ABSTRACT FROM AUTHOR]

In this work, a surface-mounted circular cylinder with a fixed aspect ratio (ratio of height of the cylinder to its diameter) of 5 is subjected to a non-zero mean oscillating flow with a range of frequencies and amplitudes. Three-dimensional direct numerical simulations are then conducted on this finite-height cylinder. The mass and momentum equations are resolved using the finite volume-based Open Source Field Operation and Manipulation (OpenFOAM). A fixed Reynolds number Re = U o D / ν of 250 is used in this study, which is defined based on mean velocity at the inlet ( U o ) and cylinder diameter (D). Here ν is the kinematic viscosity of the working fluid. Non-dimensional velocity oscillation amplitude ( A ∗ = a / U o ) is varied from 0.1 to 0.3, while the non-dimensional oscillation frequency ( f ∗ = f / f o ) takes the values of 0.33, 0.5, 1, 2, and 3. Here a and f are the dimensional oscillation amplitude and frequency, respectively and f o is the vortex shedding frequency corresponding to a uniform flow at Re = 250. The three-dimensional vortex structures, presented with the help of iso- Q surfaces, show that the oscillating flow changes the size and shape of the hairpin-shaped vortices. Wake is found to be synchronized with the oscillation frequency at f * = 2 for each value of the A * and results in the maximum lift force on the cylinder. Hilbert Huang transformation analysis of the transverse velocity signals at a specific point in the wake reveals that the wake is more complex and aperiodic in nature for f * values of 0.33, 0.5, and 1, whereas it is periodic for f * = 2 and 3. In order to further disclose the nonlinearity associated with the oscillating flow, the degree of stationarity is discussed corresponding to each value of A * and f *. Dynamic mode decomposition is exploited to obtain information about the coherent vortical structures and their spatial and temporal behavior in the wake with a change in the value of f *. Effects of A * and f * on the dynamic characteristics are also investigated. [ABSTRACT FROM AUTHOR]

This research aims to provide the groundwork for smartly categorizing hand movements for use with prosthetic hands. The hand motions are classified using surface electromyography (sEMG) data. In reaction to a predetermined sequence of fibre activation, every single one of our muscles contracts. They could be useful in developing control protocols for bio-control systems, such human-computer interaction and upper limb prostheses. When focusing on hand gestures, data gloves and vision-based approaches are often used. The data glove technique requires tedious and unnatural user engagement, whereas the vision-based solution requires significantly more expensive sensors. This research offered a Deep Neural Network (DNN) automated hand gesticulation recognition system based on electromyography to circumvent these restrictions. This work primarily aims to augment the concert of the hand gesture recognition system via the use of an artificial classifier. To advance the recognition system's classification accuracy, this study explains how to build models of neural networks and how to use signal processing methods. By locating the Hilbert Huang Transform (HHT), one may get the essential properties of the signal. When training a DNN classifier, these characteristics are sent into it. The investigational results reveal that the suggested technique accomplishes a better categorization rate (98.5 % vs. the alternatives).

Pumped storage is an important component of electrified wire netting. The safe and stable operation of pump turbines is extremely important. Among them, pressure pulsation is one of the main causes of pump turbine vibration. The characteristics of pressure pulsation are relatively complex, and it is difficult to directly observe their temporal changes using commonly used FFT methods. The division of frequency characteristics is often vague. Meanwhile, it is difficult to explain some phenomena such as frequency doubling. This article focuses on a certain model of pump turbine and uses SST model to numerically simulate the runaway condition of the pump turbine. And the Hilbert Huang transform method is used to analyze the pressure pulsation in the vaneless region and draft tube. The results show that the main characteristic frequencies of the vaneless region are blade passing frequency 112.5 Hz and rotational frequency 12.5 Hz. The main characteristic frequencies of the draft tube are vortex rope frequency near 3 Hz which energy ratio is up to 50%, rotational frequency, and blade passing frequency. The pressure pulsation characteristics in the vaneless region have changed from a complex composition of double blade passing frequency and rotational frequency to a distribution dominated by blade passing frequency. In the passage of the guide vane, the pressure pulsation is almost only characterized by blade passing frequency. The frequency characteristics of the vaneless region between the runner and the guide vane become complex again. Meanwhile, the results show that the characteristic frequencies of the vaneless region and the draft tube propagate upstream and downstream.

The application of explosive blasting for rock breaking works was not possible in densely populated areas, which caused problems in urban tunnel construction. In the application of dry ice powder thermal shock breaking in tunnel excavation, a vibration monitoring system was employed to monitor the vibration response of the tunnel structure during the breaking process, and it was used to extract the signal characteristics by Hilbert-Huang transformation. Experimental results are as follows: 1) The peak vibration speed was below 50 mm/s for concrete structures safe as long as it was farther than 10 m from the burst hole in the support concrete. 2) Aggregate decay of vibration velocities caused by thermal shock breaking of rock with dry ice powder corresponded to the decay law of the power function. The range of influence of the vibration was considerably smaller compared to that of drilling and blasting the rock; 3) The new rock-breaking technology induced damage can be divided into three regions, with radial fractures generated by high-energy fluids dominating. 4) Hilbert Huang transform extracted signal features more objectively and accurately, with excellent reference for the safety monitoring of tunnel structures. [ABSTRACT FROM AUTHOR]

The communication between the human brain and the external devices can be established using Electroencephalograms (EEG)-based Brain–Computer Interface by converting the neural activities of the brain into electric signals. The EEG signals were isolated into an energy–frequency–time spectrum with Hilbert Huang transform that was used by the Deep Learning (DL)-based model to learn discriminative spectro-temporal patterns of the raw EEG signals of ten digits. This paper has two major contributions: first, create a novel dataset known as BrainDigiData of EEG signals of ten digits from (0–9) using a multi-channel EEG device. Second to propose a DL-based one-dimensional Convolutional neural network model BrainDigiCNN to classify the BrainDigiData of EEG signals of digits. The publicly available Mind Big Dataset (MBD) of digits was also used to evaluate the performance of the proposed model. The research done in this paper showed that the band-wise analysis of EEG signals in a complex scenario resulted in improved results as compared to the scenario used in the previously existing work for digit classification using EEG signals. The proposed BrainDigiCNN model achieved the highest average accuracy of 96.99%. The average classification accuracy of 98.27% was achieved for the MBD dataset of 14 channel device EMOTIV EPOC+ and 89.62% on the MBD dataset of 5-channel EMOTIV Insight. The statistical analysis of the proposed model on traditional Machine Learning (ML) classifiers using paired t-test resulted in a p-value less than 0.05 which shows the significant difference between the proposed model and ML classifiers. [ABSTRACT FROM AUTHOR]

Abstract In this paper, the Innovative Trend Methodology (ITM) and their inspired approaches, i.e., Double (D-ITM) and Triple (T-ITM), were combined with Hilbert Huang transform (HHT) time frequency-based method. The new hybrid methods (i.e., ITM-HHT, D-ITM-HHT, and T-ITM-HHT) were proposed and compared to the DWT-based methods in order to recommend the best method. Three total annual rainfall time series from 1920 to 2011 were selected from three hydrological basins in Northern Algeria. The new combined models (ITM-HHT, D-ITM-HHT, and T-ITM-HHT) revealed that the 1950–1975 period has significant wet episodes followed by a long-term drought observed in the western region of Northern Algeria, while Northeastern Algeria presented a wet period since 2001. The proposed approaches successfully detected, in a visible manner, hidden trends presented in the signals, which proves that the removal of some modes of variability from the original rainfall signals can increase the accuracy of the used approaches.

Automatic Speech/Music classification uses different signal processing techniques to categorize multimedia content into different classes. The proposed work explores Hilbert Spectrum (HS) obtained from different AM-FM components of an audio signal, also called Intrinsic Mode Functions (IMFs) to classify an incoming audio signal into speech/music signal. The HS is a twodimensional representation of instantaneous energies (IE) and instantaneous frequencies (IF) obtained using Hilbert Transform of the IMFs. This HS is further processed using Mel-filter bank and Discrete Cosine Transform (DCT) to generate novel IF and Instantaneous Amplitude (IA) based cepstral features. Validations of the results were done using three databases-Slaney Database, GTZAN and MUSAN database. To evaluate the general applicability of the proposed features, extensive experiments were conducted on different combination of audio files from S&S, GTZAN and MUSAN database and promising results are achieved. Finally, performance of the system is compared with performance of existing cepstral features and previous works in this domain.

In this paper, the Innovative Trend Methodology (ITM) and their inspired approaches, i.e., Double (D-ITM) and Triple (T-ITM), were combined with Hilbert Huang transform (HHT) time frequency-based method. The new hybrid methods (i.e., ITM-HHT, D-ITM-HHT, and T-ITM-HHT) were proposed and compared to the DWT-based methods in order to recommend the best method. Three total annual rainfall time series from 1920 to 2011 were selected from three hydrological basins in Northern Algeria. The new combined models (ITM-HHT, D-ITM-HHT, and T-ITM-HHT) revealed that the 1950–1975 period has significant wet episodes followed by a long-term drought observed in the western region of Northern Algeria, while Northeastern Algeria presented a wet period since 2001. The proposed approaches successfully detected, in a visible manner, hidden trends presented in the signals, which proves that the removal of some modes of variability from the original rainfall signals can increase the accuracy of the used approaches. [ABSTRACT FROM AUTHOR]

Psychotic disorders are mental disorders that negatively affect human life. Diagnosis of psychotic patients is usually done in consultation with the patient, and this is a time-consuming process. In this study, a Computer Aided Diagnosis (CAD) system that will support expert opinion with automatic diagnosis of schizophrenia (SZ) disease, which is the leading psychotic disorder, is presented. In this study, Hilbert Huang Transform (HHT) method was used to analyze the non-stationary and non-periodic structure of EEG (Electroencephalograph) signals in the best way. The data set we used in our study includes 19-channel EEG signals from 28 (14 SZ and 14 healthy controls) participants, and the second data set includes 16-channel EEG signals from 84 (45 SZ and 39 healthy controls) participants. First of all, HS (Hilbert Spectrum) images of the first four Intrinsic Mode Functions (IMF) components obtained by applying Empirical Mode Decomposition (EMD) to EEG signals were created. These images were then classified with the VGG16 pre-trained Convolutional Neural Network (CNN) network. With our proposed method, the highest classification performance was obtained as 98.2% for Dataset I and 96.02% for Dataset II, respectively, by training the HS images obtained from the IMF 1 component with the VGG16 pre-trained CNN network. In the next step, classification performances were tested with VGG16, XCeption, DenseNet121, ResNet152 and Inception V3 pre-trained CNN networks. The high classification success achieved by the proposed method in our study demonstrates the accuracy of the model in distinguishing between SZ and healthy control. [ABSTRACT FROM AUTHOR]

The occurrence probability of plasma irregularities has long been used to quantify that of ionospheric scintillations. We use the Hilbert-Huang transform to examine the occurrence probability and the instantaneous total amplitude of plasma irregularities in the low-latitude nighttime ionosphere observed by the low-inclination satellites, ROCSAT-1 during 1999–2004 and FORMOSAT-7/COSMIC-2 (F7C2) in 2020, as well as a high-inclination satellite, DEMETER, during 2006–2010. Generally, the longitudinal distribution of the occurrence probability is similar to that of the total amplitude, except that the probability of ion density irregularity observed by F7C2 and DEMETER yields nonphysical anomalous enhancements over the South American sector during May–August of 2007–2010 and 2020, respectively. S4 scintillations observed by F7C2 show that the anomalous enhancement results from the low ambient density in calculating the occurrence probability during the low solar activity year of 2020. We suggest the instantaneous total amplitude as a better indicator for describing ionospheric plasma irregularities. [ABSTRACT FROM AUTHOR]

Disabilities due to neurodegenerative disease are rapidly increasing in number. The need for rehabilitative devices to achieve a normal and comfortable life in the absence of biochannels has also increased. The activities of biochannels can be easily replaced by implementing rehabilitative devices. The electrooculography (EOG)-based human–computer interface (HCI) is one of the most important techniques for enabling disabled persons to enjoy a normal life. The technique of measuring the cornea-retina potential difference is called EOG. The technology for converting thoughts to different control patterns to activate external devices is called the HCI. In this paper, we carried out a study on ten subjects aged 20–30 years using a five-electrode signal acquisition system (AD T26). The subject performances were experimentally verified by implementing periodogram features with a feedforward neural network trained on a nature-inspired algorithm. The analysis was conducted offline and online to evaluate the achievement of the developed HCI. The study showed an average classification accuracy of 93.93% for four tasks, with 95% accuracy for the offline mode and 90.12% for the online mode. The participating subjects drove the mobile robot in all directions frequently and quickly with a recognition accuracy of 98.12%. The study confirmed that the four tasks (related to driving the external device) performed by the subjects were convenient to perform in real time. [ABSTRACT FROM AUTHOR]

Effective water resource management requires accurate estimation of the volume of water withdrawn through anthropogenic groundwater pumping. However, insufficient pumping information makes accurate estimation difficult. To address this issue, this study proposes a novel method for estimating pumping volumes that does not rely on pumping information. The proposed method uses high-resolution groundwater level (GWL) data from densely monitored wells and integrates Hilbert–Huang transform and empirical orthogonal function analysis to extract pump-associated signals, which can then be used to estimate total pumping. Our method was validated by analyzing GWL data from monitoring stations on the Choshui River alluvial fan in Taiwan. The results indicate the estimated pumping amount of 1. 71 × 1 0 9 – 2. 05 × 1 0 9 m 3 /year consistent with previous estimates obtained using survey and physics-based methods ranging from 9 × 1 0 8 m 3 /year to 2. 2 × 1 0 9 m 3 /year. In the areas with abundant groundwater head observations, the proposed method can efficiently estimate pumping rates with a high space–time resolution, and may serve as a useful reference for groundwater management. • This study proposes a method utilizing groundwater data to estimate spatiotemporal pumping rate. • Multi-year high space–time resolution of pumping rate distribution estimation was efficiently derived from hourly groundwater data. • The pumping rate distribution aligns with geological and agricultural water usage patterns. [ABSTRACT FROM AUTHOR]

This research aims to provide the groundwork for smartly categorizing hand movements for use with prosthetic hands. The hand motions are classified using surface electromyography (sEMG) data. In reaction to a predetermined sequence of fibre activation, every single one of our muscles contracts. They could be useful in developing control protocols for bio-control systems, such human-computer interaction and upper limb prostheses. When focusing on hand gestures, data gloves and vision-based approaches are often used. The data glove technique requires tedious and unnatural user engagement, whereas the vision-based solution requires significantly more expensive sensors. This research offered a Deep Neural Network (DNN) automated hand gesticulation recognition system based on electromyography to circumvent these restrictions. This work primarily aims to augment the concert of the hand gesture recognition system via the use of an artificial classifier. To advance the recognition system's classification accuracy, this study explains how to build models of neural networks and how to use signal processing methods. By locating the Hilbert Huang Transform (HHT), one may get the essential properties of the signal. When training a DNN classifier, these characteristics are sent into it. The investigational results reveal that the suggested technique accomplishes a better categorization rate (98.5 % vs. the alternatives)., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2024 The Authors.)

The present work investigated the microbial-induced pitting behavior of the 70/30 copper-nickel (Cu-Ni) alloy. Unlike the MIC of Cu by Desulfovibrio vulgaris , the alloy exhibited considerable pitting sensitivity and pit nucleation frequency under the 10%, 50%, and 100% carbon source (CS) levels. The metastable pit was more likely to evolve into stable growth in the full-strength medium. Meanwhile, a model was established to explain the alloy pitting mechanism induced by D. vulgaris. New evidence further verified that Cu-Ni alloy MIC was not only ascribed to the biogenic H 2 S but the electroactive D. vulgaris biofilms via extracellular electron transfer (EET) for energy harvest. ● The pitting corrosion of 70/30 Cu-Ni alloy by D. vulgaris is studied using Hilbert spectroscopy analysis. ● The alloy exhibits high pitting sensitivity for the 10%, 50%, and 100% CS levels. ● A model is established to explain the pitting mechanism induced by D. vulgaris. ● Metastable pitting is more likely to transform into stable in the full-strength medium. ● New evidence further proves that two types of MIC mechanisms coexist in the Cu-Ni alloys MIC. [ABSTRACT FROM AUTHOR]

• Novel hybrid CNN-MPL model-based bearing faults diagnosis method has high accuracy. • Mixed input processing allows increasing the detection accuracy for bearing faults. • Measured accelerations from the rotating shaft are tightly base for diagnosis. • Angular acceleration of the rotating shaft is highly sensitive to bearing faults. Rolling bearings are one of the most widely used bearings in industrial machines. The technical condition of the rolling bearings has a significant impact on the total condition of rotating machinery. Typically, the rolling bearing condition monitoring is based on signal processing from accelerometers mounted on the machine housing. AI-based signal processing methods allow achieving high results in the diagnosis of rolling bearings. Moreover, hybrid NN-based signal processing methods provide the best diagnostic results. However, the diagnostic value of each signal from the housing-mounted accelerometers is highly depend on the location of the corresponding accelerometer. On the other hand, mounting a sensor on the rotating shaft allows expanding and increasing the signal diagnostic value. This paper proposes a novel hybrid CNN-MLP model-based diagnostic method which combines mixed input to perform rolling bearing diagnostics. The method successfully detects and localizes bearing defects using acceleration data from a wireless acceleration sensor which is mounted on a rotating shaft of the machine. The detection and localization efficiency of the hybrid model with various size datasets (more than 27 k samples and about 1900 samples) for training was estimated. Moreover, the detection and localization efficiency of the trained hybrid model with a dataset of another shaft rotating speed for testing was estimated. The experimental results show that applying the hybrid model allows detecting and localizing the bearing faults with up to 99.6% accuracy. [ABSTRACT FROM AUTHOR]

Sensor-based human activity recognition (HAR) is a method for observing a person's activity in an environment. With this method, it is possible to monitor remotely. HAR can analyze a person's gait, whether normal or abnormal. Some of its applications may use several sensors mounted on the body, but this method tends to be complex and inconvenient. One alternative to wearable sensors is using video. One of the most commonly used HAR platforms is PoseNET. PoseNET is a sophisticated platform that can detect the skeleton and joints of the body, which are then known as joints. However, a method is still needed to process the raw data from PoseNET to detect subject activity. Therefore, this research proposes a way to detect abnormalities in gait using empirical mode decomposition and the Hilbert spectrum and transforming keys-joints, and skeletons from vision-based pose detection into the angular displacement of walking gait patterns (signals). Joint change information is extracted using the Hilbert Huang Transform to study how the subject behaves in the turning position. Furthermore, it is determined whether the transition goes from normal to abnormal subjects by calculating the energy in the time-frequency domain signal. The test results show that during the transition period, the energy of the gait signal tends to be higher than during the walking period.

The losses caused by fire are becoming more and more serious, with the continuous development of social economy. If fire is detected as early as possible, the losses can be reduced more than 80%. The detection accuracy of traditional single-parameter fire detection is low, which often leads to false and missing fire alarm. This paper proposes a multi-parameter fire detection method based on feature depth extraction and stacking ensemble learning model. Fire feature parameters, such as temperature, smoke concentration and CO concentration, are collected as the original data. The algorithms, based on XGboost, gradient boosted decision tree (GBDT), random forest and decision tree, are used to analyze and sort the importance of nine types of data, such as the slope of smoke concentration, the amplitude of CO concentration etc. Three types of data with high importance are selected to extract depth features using Hilbert Huang transform (HHT), one-dimensional convolution neural network (1D-CNN) and long short term memory neural network (LSTMNN). The extracted features are input into stacking ensemble learning model. Stacking ensemble learning model is a hierarchical combination learning classifier and improves the performance by learning the classification bias of the upper model. Experimental results show that the proposed method has higher accuracy, stability and efficiency against several single models (e.g., XGBoost, GBDT, and LSTMNN). [ABSTRACT FROM AUTHOR]

• HHT method can well process and analyze the EMR waveform signal of coal failure. • The EMR parameters extracted by HHT improve the identification of coal failure. • Compared with STFT and WT, HHT has more advantages in processing EMR waveform. Electromagnetic radiation (EMR) waveforms contain rich information on coal fracture law, which is of great value to the refined study of their characteristics. In this paper, the Hilbert-Huang transform (HHT) method is used to analyze and process EMR waveform of coal failure process. The results show that HHT method can adapt well to the nonstationary and nonlinear characteristics of EMR waveform. It adaptively decomposes EMR waveforms by empirical mode decomposition (EMD), extracts the intrinsic mode function (IMF) characteristics of EMR waveforms, and obtains the law of instantaneous energy changes of EMR waveforms. At the same time, EMR waveform energy is quantitatively expressed on the three-dimensional Hilbert energy spectrum (time–frequency-energy distribution), and the refined features of EMR waveform signal are effectively extracted. The instantaneous energy of EMR and the three-dimensional Hilbert energy spectrum processed by HHT method more easily describe the fracture law of coal body. Compared with conventional characteristic parameters, such as EMR amplitude and dominant frequency, the identification degree of the precursor characteristics of coal body instability is greatly improved. Among the short-term Fourier transform (STFT), wavelet transform (WT) and HHT nonstationary signal processing methods, HHT method has an improved adaptability and superiority and can reveal the characteristics of coal deformation and failure more carefully and accurately, which provides a new means for better using and improving EMR method to monitor coal instability and failure. [ABSTRACT FROM AUTHOR]