Mode Functions Research Articles

Fault diagnosis of wind turbine gears based on OCSSA-VMD and WOA-CNN-BiLSTM

Abstract The accuracy of wind turbine gearbox fault diagnosis will be compromised if the fault feature data is not adequately extracted during operation. To enhance fault identification efficiency and mitigate human interference in parameter setting, this paper introduces an optimized mode decomposition algorithm OCSSA-VMD, derived from variational mode decomposition (VMD) and further optimized by osprey-Cauchy-sparrow search algorithm (OCSSA). This algorithm offers two key advantages: (1) automatic optimization of parameters such as the number of modes k and penalty factor α; (2) reduction of feature dimensionality through mean impact value (MIV) algorithm based on minimum envelope entropy principle, resulting in a multi-fault feature vector set from 13 time-domain features in the intrinsic mode function (IMF) optimal component of wind turbine gearbox vibration data. Additionally, a fault diagnosis model WOA-CNN-BiLSTM is proposed based on whale optimization algorithm (WOA) and convolutional neural network-bidirectional long-short-term-memory (CNN-BiLSTM), which demonstrates improved fault classification accuracy to 98.3333% and diagnosis accuracy to 98.3853% under conditions of insufficient data when compared with other models.

Gearbox Fault Diagnosis Based on Adaptive Variational Mode Decomposition–Stationary Wavelet Transform and Ensemble Refined Composite Multiscale Fluctuation Dispersion Entropy

Existing gearbox fault diagnosis methods are prone to noise interference and cannot extract comprehensive fault signals, leading to misdiagnosis or missed diagnosis. This paper proposes a method for gearbox fault diagnosis based on adaptive variational mode decomposition–stationary wavelet transform (AVMD-SWT) and ensemble refined composite multiscale fluctuation dispersion entropy (ERCMFDE). Initially, the kurtosis coefficient and autocorrelation coefficient are presented, and the Intrinsic Mode Functions are denoised through the application of AVMD-SWT. Secondly, the coarse-grained processing method of composite multiscale fluctuation dispersion entropy is extended to encompass three additional approaches: first-order central moment, second-order central moment, and third-order central moment. This enables the comprehensive extraction of feature information from the time series, thereby facilitating the formation of an initial hybrid feature set. Subsequently, recursive feature elimination (RFE) is employed for feature selection. Ultimately, the outcomes of the faults diagnoses are derived through the utilization of a Support Vector Machine with a Sparrow Search Algorithm (SSA-SVM), with the actual faults data collection and analysis conducted on an experimental platform for gearbox fault diagnosis. The experiments demonstrate that the method can accurately identify gearbox faults and achieve a high diagnostic accuracy of 98.78%.

Diagnosis of incipient faults in wind turbine bearings based on ICEEMDAN–IMCKD

AbstractTo address the difficulty in extracting early fault feature signals of rolling bearings, this paper proposes a novel weak fault diagnosis method for rolling bearings. This method combines the Improved Complementary Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) and the Improved Maximum Correlated Kurtosis Deconvolution (IMCKD). Utilizing the kurtosis criterion, the intrinsic mode functions obtained through ICEEMDAN are reconstructed and denoised using IMCKD, which significantly reduces noise in the measured signal. This approach maximizes the energy amplitude at the fault characteristic frequency, facilitating fault feature identification. Experimental studies on two test benches demonstrate that this method effectively reduces noise interference and highlights the fault frequency components. Compared with traditional methods, it significantly improves the signal‐to‐noise ratio and more accurately identifies fault features, meeting the requirements for discriminating rolling bearing faults. The method proposed in this study was applied to the measured vibration signals of the gearbox bearings in the new high‐speed wire department of a Long Products Mill. It successfully extracted weak characteristic information of early bearing faults, achieving the expected diagnostic results. This further validates the effectiveness of the ICEEMDAN–IMCKD method in practical engineering applications, demonstrating significant engineering value for detecting and extracting weak impact characteristics in rolling bearings.

Design of EEG based thought identification system using EMD & deep neural network

Biological communication system for neurological disorder patients is similar to the Brain Computer Interface in a way that it facilitates the connection to the outside world in real time. The interdisciplinary field of Electroencephalogram based message depiction is gaining importance as it assists the paralysed person to communicate. In the proposed method a novel approach of feature extraction is done by Empirical Mode Decomposition on non- stationary & non-linear kind of EEG signal. EMD helps in the effective time frequency analysis by disintegrating the EEG signal in the form of six Intrinsic Mode Functions with help of the frequency components. In all nine features are extracted from the decomposed IMFs so as to predict the states or messages of the patient. The above computed features are then served to the Deep Neural Network to perform the classification. The performance of suggested method is studied through applying it to the acquired database generated by the designed hardware as well as also in real time message depiction. The maximum classification accuracy 97% for the acquired database & 85% in real time are obtained respectively by comparative analysis. The command messages generated from the proposed system helps the person suffering from neurological disorder to establish the communication link with the outside world in an efficient way. Thus, the proposed novel method shows better performance in real time message depiction purpose as related to other existing methods.

A hybrid model based on complementary ensemble empirical mode decomposition, sample entropy and long short-term memory neural network for the prediction of time series signals in NPPs

Accurate and reliable predictions are fundamental for the condition monitoring and maintenance of components and systems in nuclear power plants (NPPs). In this work, we propose a hybrid condition prediction approach based on the combination of complementary ensemble empirical mode decomposition (CEEMD), sample entropy (SampEn) and optimized long short-term memory (LSTM) neural network. Firstly, the CEEMD decomposes the time series signals into multiple subsequences called intrinsic mode functions (IMFs), by doing so, the complexity of the time series signals can be reduced, and this facilitates the accurate prediction of the original signals. Then, in order to reduce the calculation cost of prediction models for subsequences, SampEn is used to measure the complexities of the IMFs, and the IMFs whose values of SampEn are lower than the average are aggregated into a new component. Finally, the LSTM with the hyperparameters optimized by the Bayesian optimization algorithm (BOA) is used to perform the prediction of each component. The prediction results of the original signals are reconstructed by synthesizing the predictions of all components. The proposed hybrid prediction model is utilized on the time series signals collected from an NPP. The results obtained show that the proposed approach can capture the characteristics of the signals and has better performance in prediction accuracy than other models.

Detecting damage on engine mounts using hilbert-huang transform vibration analysis

Engine mounts, which are usually composed of elastomeric materials like rubber that can absorb excessive vibrations, are built to withstand vibration sources from engines. Engine mounts may eventually degrade from extended use. When rubber products age or sustain damage, they may grow harder and crack. Engine mounts need to be maintained regularly to ensure optimal performance. Visual examinations and the detection of excessive vibrations can be used to accomplish this. Vibration sensors are mounted on the engine mount in axial, vertical, and horizontal orientations using a vibration detection technique utilizing the Hilbert-Huang Transform (HHT) methodology. Matlab is used to examine the data that is gathered at three distinct rotational speeds: 750 rpm, 2000 rpm, and 3000 rpm. The HHT approach combines two key components: the Hilbert Transform, which analyzes the time-frequency signal of the first decomposition until only residuals remain, and Empirical Mode Decomposition (EMD), which breaks down the signal into Intrinsic Mode Functions (IMFs). According to test data, a damaged mount had an amplitude of 0.00005212 m/s² and a frequency of 8 Hz. On the other hand, in typical circumstances, the highest frequency was 7 Hz with the same amplitude. Five frequency increases were made in the damaged mount throughout this operation. In the damaged mount, the Hilbert Transform showed a frequency of 2124 Hz with an amplitude of 0.007594 m/s², indicating a significant resonance. This illustrates how the Hilbert-Huang Transform's capacity to handle non-stationary and nonlinear signal forms allows it to detect damage in components efficiently

Noise robust damage detection of laminated composites using multichannel wavelet-enhanced deep learning model

This paper presents a noise-robust damage detection framework for composite structures via a commonly used vibration-based non-destructive testing (NDT) method. Recently, deep learning-based models have shown promising performance in the autonomous damage detection of laminated composites; however, the poor noise robustness of these models has plagued data-driven damage detection. Moreover, none of the existing studies on damage detection in laminated composites focus on noise-robust deep learning models with high generalization ability. Therefore, this study proposes a hybrid deep learning framework called a multi-channel convolutional autoencoder-support vector machine (MC-CAE-SVM) based on empirical mode decomposition (EMD) and correlation analysis for noise-robust damage detection. This framework aims to first decompose the vibrational signal from multiple health states into intrinsic mode functions (IMFs). Secondly, highly correlated IMFs were extracted using correlation analysis to remove noisy IMFs. Finally, these IMFs were transformed into a time-frequency representation using continuous wavelet transform (CWT) and input to the MC-CAE-SVM model for feature learning and damage detection. Additionally, the accuracy and sensitivity of the model to damage are enhanced by optimizing the MC-CAE-SVM model hyperparameters. Moreover, anti-noise analysis is performed to check the noise-robustness of the proposed model by incorporating noise at various levels. The results showed that the proposed model can provide better damage detection performance compared to conventional models with excellent noise robustness.

The queuing system of a scientific journal

The results of the research difference of the calendar date between begining and ending editorial processes in the Journal of “Almaz – Antey” Air and Space Defence Corporation are given, and the counting of dates per month for beginning the processes (entering articles, transferring to literacy editing etc.). Based on the analysis it was concluded that the incoming article might be considered as an incoming application to queuing system, which the scientific journal is. For such incoming application can be estimate the distribution functions incoming flows, average and maximum term be in queuing system, also as the distribution functions of the service mode and another characteristics, which stay unchanged (stationary) in time for researched scientific journal.

A River Water Quality Prediction Method Based on Dual Signal Decomposition and Deep Learning

Traditional single prediction models struggle to address the complexity and nonlinear changes in water quality forecasting. To address this challenge, this study proposed a coupled prediction model (RF-TVSV-SCL). The model includes Random Forest (RF) feature selection, dual signal decomposition (Time-Varying Filtered Empirical Mode Decomposition, TVF-EMD, and Sparrow Search Algorithm-Optimized Variational Mode Decomposition, SSA-VMD), and a deep learning predictive model (Sparrow Search Algorithm-Convolutional Neural Network-Long Short-Term Memory, SSA-CNN-LSTM). Firstly, the RF method was used for feature selection to extract important features relevant to water quality prediction. Then, TVF-EMD was employed for preliminary decomposition of the water quality data, followed by a secondary decomposition of complex Intrinsic Mode Function (IMF) components using SSA-VMD. Finally, the SSA-CNN-LSTM model was utilized to predict the processed data. This model was evaluated for predicting total phosphorus (TP), total nitrogen (TN), ammonia nitrogen (NH3-N), dissolved oxygen (DO), permanganate index (CODMn), conductivity (EC), and turbidity (TB), across 1, 3, 5, and 7-d forecast periods. The model performed exceptionally well in short-term predictions, particularly within the 1–3 d range. For 1-, 3-, 5-, and 7-d forecasts, R2 ranged from 0.93–0.96, 0.79–0.87, 0.63–0.72, and 0.56–0.64, respectively, significantly outperforming other comparison models. The RF-TVSV-SCL model demonstrates excellent predictive capability and generalization ability, providing robust technical support for water quality forecasting and pollution prevention.

Financial Time Series Forecasting Based on Long-Short Term Memory, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise and Batch Normalization

Long-short term memory (LSTM) is a state-of-art and widely used model to forecast financial time series. However, primitive LSTM networks do not perform well due to over-fitting problems of the deep learning model and nonlinear and non-stationary characteristics of financial time series data. In addition, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is an outstanding data frequency decomposition technique that can decompose original time series into several intrinsic mode functions and a residue. Thus, this paper proposed a novel hybrid network CEEMDAN-LSTM-BN based on LSTM. Specifically, to avoid over-fitting, the modified LSTM-BN network consists of two LSTM layers, two Batch Normalization (BN) layers following each LSTM layer, and a dropout layer. Each of the intrinsic mode functions and the residue would be processed by CEEMDAN-LSTM-BN and the final prediction results are obtained by reconstructing each predictive series. The advantages of the proposed CEEMDANLSTM-BN networks are verified by comparing them to primitive LSTM, other hybrid models, and some famous machine learning models. Moreover, the robustness of the networks is assessed by numerical experiments on different stock indices datasets.

Magnetic Properties of a Nickel–Zinc Ferrite Powder with Different Degrees of Dispersion

The influence of the degree of dispersion of a nickel–zinc ferrite powder of a Ni0.7Zn0.3Fe2O4 composition on its magnetic properties has been considered. The material has been synthesized using the ceramic technology with preliminary mechanical activation of precursors. The degree of dispersion has been varied using different modes of its dry grinding in a ball mill. The patterns of the changes in saturation magnetization and the coercive force as a function of grinding modes and a specific surface area of the ferrite powder have been established. The changes in the pattern of the magnetic phase transition in the region of the Curie temperature of materials with different degrees of dispersion have been determined.

Integral predictive sliding mode control for high-speed trains: A dynamic linearization and input constraint-based data-driven scheme

A control scheme with high reliability and excellent tracking performance is essential for the automatic operation of high-speed trains (HSTs). In this study, a novel discrete-time data-driven predictive sliding mode control (DDPSMC) scheme is proposed for multi-power unit HSTs. Initially, a nonlinear integral terminal sliding mode surface was designed to replace the traditional linear sliding mode function, thereby achieving a rapid system error convergence and alleviating chattering. Then, receding horizon optimization was integrated into predictive control, which allowed the predicted sliding mode state to follow the expected trajectory of a predefined continuous convergence law. This scheme enabled the system to obtain higher output error accuracy and explicitly handle input constraints. Moreover, to enhance robustness, a parameter update law and disturbance delay estimation algorithm were introduced to calculate the control gain and total uncertainty, respectively. Finally, a comparative test of the proposed control scheme was conducted using a CRH380A HST simulation experimental platform in a laboratory setting. Simulation results demonstrate that the velocity error range of each power unit of the HST under the proposed control scheme is within [−0.176 km/h, 0.152 km/h], while the control force and acceleration are within [−55.7 kN, 44.8 kN] and [−0.564 m/s2, 0.496 m/s2], respectively, with stable variation, and other performance indicators are also better than other comparison methods. These results satisfy the safety, stability, and punctuality requirements of the train.

A rolling bearing fault signal denoising algorithm that combines a new adaptive information entropy with a new wavelet threshold function

Abstract Mechanical fault diagnosis is of great significance to industrial automation, and extracting vibration fault signals is one of the important tasks in mechanical health monitoring and fault diagnosis. However, due to the complex working environment of rolling bearings, a large amount of noise makes it difficult to extract vibration fault signals. Denoising the vibration signal of bearings can remove interference noise, simplify the early identification of signal features, and thus improve diagnostic accuracy and mechanical maintenance efficiency. This paper proposes a rolling bearing fault signal denoising algorithm, which constructs a new feature extraction and denoising function. This method first decomposes the noisy signal into Intrinsic Mode Functions (IMFs) by Intrinsic Computing Expressive Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN). Secondly, a new adaptive information entropy threshold function is constructed to extract the noisy IMFs from it. Then, the noisy IMF signal is denoised by a new wavelet threshold function. Finally, the noise-free IMFs are reconstructed to reconstruct the denoised signal. To verify the actual performance of the algorithm, comparative experiments were conducted on a self-collected dataset and a public dataset, the results show that this method improves the continuity of signal reconstruction and can remove various types of vibration fault noise signals more effectively and accurately, \hl{thereby improving the fault detection accuracy by 2%-9%.

Adaptive Feature Extraction Using Sparrow Search Algorithm-Variational Mode Decomposition for Low-Speed Bearing Fault Diagnosis

To address the challenge of extracting effective fault features at low speeds, where fault information is weak and heavily influenced by environmental noise, a parameter-adaptive variational mode decomposition (VMD) method is proposed. This method aims to overcome the limitations of traditional VMD, which relies on manually set parameters. The sparrow search algorithm is used to calculate the fitness function based on mean envelope entropy, enabling the adaptive determination of the number of mode decompositions and the penalty factor in VMD. Afterward, the optimised parameters are used to enhance traditional VMD, enabling the decomposition of the raw signal to obtain intrinsic mode function components. The kurtosis criterion is then used to select relevant intrinsic mode functions for signal reconstruction. Finally, envelope analysis is applied to the reconstructed signal, and the results reveal the relationship between fault characteristic frequencies and their harmonics. The experimental results demonstrate that compared with other advanced methods, the proposed approach effectively reduces noise interference and extracts fault features for diagnosing low-speed bearing faults.

Heart Rate Estimation Considering Reconstructed Signal Features Based on Variational Mode Decomposition via Multiple-Input Multiple-Output Frequency Modulated Continuous Wave Radar

Accurate heart rate estimation using Doppler radar and Frequency Modulated Continuous Wave (FMCW) radar is highly valued for privacy protection and the ability to measure through clothing. Conventional methods struggle to isolate the heartbeat from respiration and body motion. This paper introduces a novel heart rate estimation method using Variational Mode Decomposition (VMD) via Multiple-Input Multiple-Output (MIMO) FMCW radar. The proposed method first estimates human positions within the radar’s coverage, reducing noise by focusing on signals from these positions. The signal is then decomposed into multiple Intrinsic Mode Function (IMF) signals using VMD, and the heartbeat-specific IMF is extracted based on its center frequency. The heart rate signal is reconstructed using weighted addition of IMF signals for each radar cell, with cells defined by specific angles and distances within the coverage area. Peak detection is used to estimate heart rate from these reconstructed signals. To ensure accuracy, the method selects the heart rate estimate with the highest energy and periodicity for the first four time windows. From the fifth time window onward, it selects the estimate closest to the average of the previous four, minimizing extraneous variations. Experiments conducted with one and two subjects showed promising results. In case 1, with one subject, the method achieved a Mean Absolute Error (MAE) of 2.54 BPM and an exclusion rate of 0.94% using MIMO FMCW radar, compared to 4.72% with Doppler radar. In case 2, with two subjects, the method achieved an MAE of 2.28 BPM, confirming accurate simultaneous heart rate estimation.

Optimized variational mode decomposition for improved bearing fault diagnosis and performance evaluation

This research presents an enhanced methodology for diagnosing bearing faults using Variational Mode Decomposition (VMD) based on L-Kurtosis analysis. The proposed method focuses on selecting optimal parameters for VMD to extract the mode containing the most information related to the fault. The selection of these parameters is based on comparing the energy ratio of each mode and the absolute difference in L-Kurtosis between the Intrinsic Mode Function (IMF) with the highest energy and the original signal. The extracted mode is further refined using a specified kurtosis rate threshold to ensure the most relevant significant modes are captured. The proposed methodology was tested using real fault data from the CWRU, XJTU-SY, and a real-world wind turbine dataset related to electric motors and wind turbine systems. The results demonstrated high accuracy in fault detection compared to other methods such as the Gini Index, correlation, and traditional decomposition techniques like EMD. Furthermore, due to the simple computational nature of the improved VMD method, it is faster and more efficient compared to methods that rely on complex calculations or frequency band analysis, making it suitable for applications requiring real-time, reliable fault diagnosis

Method and Application of Spillway Radial Gate Vibration Signal Denoising on Multiverse Optimization Algorithm-Optimized Variational Mode Decomposition Combined with Wavelet Threshold Denoising

To address the noise issue in the measured vibration signals of spillway radial gate discharge, this paper utilizes the Multiverse Optimization Algorithm (MVO) to optimize the number of decomposition modes (K) and the penalty factor (α) in Variational Mode Decomposition (VMD). This approach ensures improved efficiency of VMD decomposition while maintaining accuracy. Subsequently, the obtained Intrinsic Mode Functions (IMFs) from VMD decomposition are classified based on Multi-scale Permutation Entropy (MPE). IMFs are divided into pure components and noisy components; the noisy components are processed with Wavelet Threshold Denoising (WTD), while the pure components are overlaid and reconstructed to obtain the denoised vibration signal of the gate. Comprehensive comparisons involving artificial signal simulations, gate flow-induced vibration model tests, and numerical simulations lead to the following conclusions: compared to other algorithms, the proposed combined denoising method (MVO-VMD-MPE-WTD) achieves the highest signal-to-noise ratio (SNR) in both the frequency and time domains for artificial signals, while yielding the lowest mean square error (MSE). In the gate flow-induced vibration model tests, the method significantly reduces noise in the vibration signals and effectively preserves characteristic information. The error in preserving characteristic information across model tests and numerical simulations is kept below 1%. Furthermore, compared to other optimization algorithms, the MVO demonstrates higher computational efficiency. The parameter-optimized combined denoising method proposed in this study provides insights into denoising measured vibration signals of hydraulic spillway radial gates and other drainage structures, and it opens possibilities for exploring more efficient optimization algorithms for achieving online monitoring in the future.

Tornado Occurrence in the United States as Modulated by Multidecadal Oceanic Oscillations Using Empirical Model Decomposition

Studies have analyzed U.S. tornado variability and correlated F1–F5 tornado occurrence with various natural climate oscillations and anthropogenic factors. Using a relatively new empirical mode decommission (EMD) method that extracts time-frequency modes adaptively without priori assumptions like traditional time-series analysis methods, this study decomposes U.S. tornado variability during 1954–2022 into intrinsic modes on specific temporal scales. Correlating the intrinsic mode functions (IMFs) of EMD with climate indices found that 1. the U.S. overall tornado count is negatively (positively) correlated with the Atlantic Multidecadal Oscillation (AMO) index (the Southern Oscillation Index (SOI)); 2. the negative (positive) correlation tends to be more prevalent in the western (eastern) U.S.; 3. the increase in weak (F1–F2) and decrease in strong (F3–F5) tornadoes after around 2000, when both the AMO and the Pacific Decadal Oscillation (PDO) shifted phases, are likely related to their secular trends and low-frequency IMFs; and 4. the emerging Dixie Tornado Alley coincides with an amplifying intrinsic mode of the SOI that correlates positively with the eastern U.S. and Dixie Alley tornadoes. The long-term persistence of these climate indices can offer potential guidance for future planning for tornado hazards.

Carbon price prediction in China based on ensemble empirical mode decomposition and machine learning algorithms.

The carbon emission trading market is crucial for reducing emissions, conserving energy, and enhancing the climate and environment. Studying carbon price forecasting can encourage China's involvement in international carbon financial instruments trading and promote the development of China's pilot carbon markets. This study employs ensemble empirical mode decomposition (EEDM) to convert the initial carbon price, which is a non-stationary signal, into several intrinsic mode functions and residual terms. Subsequently, hybrid machine learning methods are used to estimate the carbon price. For empirical analysis, the three main carbon markets with large trading volumes, Guangdong, Hubei, and Shenzhen, are selected from the eight pilot carbon markets. To achieve short-term carbon price prediction, a hybrid model combining a genetic algorithm (GA) and back propagation (BP) neural network is used. This model effectively addresses the issue of neural networks falling into local optimization. The results indicate that the hybrid algorithm is significantly superior to other algorithms for short-term prediction. Additionally, another hybrid model, combining the least squares support vector machine (LSSVM) and particle swarm optimization (PSO) algorithms, is employed to reduce forecast error while minimizing search parameters, which is not possible with traditional neural network models.

Interpretable multi-step hybrid deep learning model for karst spring discharge prediction: Integrating temporal fusion transformers with ensemble empirical mode decomposition

Karst groundwater is a critical freshwater resource for numerous regions worldwide. Monitoring and predicting karst spring discharge is essential for effective groundwater management and the preservation of karst ecosystems. However, the high heterogeneity and karstification pose significant challenges to physics-based models in providing robust predictions of karst spring discharge. In this study, an interpretable multi-step hybrid deep learning model called selective EEMD-TFT is proposed, which adaptively integrates temporal fusion transformers (TFT) with ensemble empirical mode decomposition (EEMD) for predicting karst spring discharge. The selective EEMD-TFT hybrid model leverages the strengths of both EEMD and TFT techniques to learn inherent patterns and temporal dynamics from nonlinear and nonstationary signals, eliminate redundant components, and emphasize useful characteristics of input variables, leading to the improvement of prediction performance and efficiency. It consists of two stages: in the first stage, the daily precipitation data is decomposed into multiple intrinsic mode functions using EEMD to extract valuable information from nonlinear and nonstationary signals. All decomposed components, temperature and categorical date features are then fed into the TFT model, which is an attention-based deep learning model that combines high-performance multi-horizon prediction and interpretable insights into temporal dynamics. The importance of input variables will be quantified and ranked. In the second stage, the decomposed precipitation components with high importance are selected to serve as the TFT model’s input features along with temperature and categorical date variables for the final prediction. Results indicate that the selective EEMD-TFT model outperforms other sequence-to-sequence deep learning models, such as LSTM and single TFT models, delivering reliable and robust prediction performance. Notably, it maintains more consistent prediction performance at longer forecast horizons compared to other sequence-to-sequence models, highlighting its capacity to learn complex patterns from the input data and efficiently extract valuable information for karst spring prediction. An interpretable analysis of the selective EEMD-TFT model is conducted to gain insights into relationships among various hydrological processes and analyze temporal patterns.