Combining Wavelet Packet Decomposition Research Articles

Internal defect detection quantification and three-dimensional localization based on impact echo and classification learning model

Accurately identifying, localizing, and characterizing internal defects is crucial for ensuring the safety and durability of concrete structures. While impact echo (IE) is a highly promising non-destructive testing method for detecting concrete internal defects, previous studies have primarily focused on defect identification with less emphasis on quantification and depth localization of internal defects. In this paper, we propose an intelligent detection method based on IE and deep learning to achieve intelligent identification, area quantification, and depth localization of concrete internal defects. The proposed method includes the following three components: (1) A one-dimensional model combining wavelet packet decomposition and the Gate Recurrent Unit network was proposed to achieve an automatic diagnosis of defect signals inside concrete structures. (2) A method incorporating defect identification probability heatmap and threshold segmentation is employed to quantify the concrete defect area and detect the defect area detection rate. (3) A two-dimensional model combining wavelet transform and convolutional neural network was developed to achieve defect depth localization. The proposed method has been effectively validated in laboratory experiments involving concrete slabs with artificial defects.

A deep learning method to monitor axial pressure and shear deformation of rubber bearings under coupled compression and shear loading

AbstractLaminated rubber bearings are the key components of base isolated structures. It is important to effectively monitor the axial pressure and shear deformation in rubber bearings. In previous studies, the axial pressure and shear deformation of rubber bearings have been monitored separately using wavelet‐packet‐based method. However, when the axial pressure and deformation of rubber bearings change at the same time, the prediction error of the above method is large. Therefore, this paper proposed a new deep neural network architecture, named P&D‐NET to monitor the axial pressure and shear deformation simultaneously. The P&D‐NET combines wavelet packet decomposition method and fully connected neural network. In this study, the coupling effects of axial pressure and shear deformation monitoring are studied by full‐scale experiments. The details of P&D‐NET are also introduced. Three full‐scale smart rubber bearings (SRB) were tested under different axial pressure and shear deformation to form a real measured dataset. The influence of feature extraction methods, network structures, and weighting coefficient in loss function was investigated to optimize the network. By using the optimal hyperparameters, the root‐mean‐square error (RMSE) of axial pressure prediction and shear deformation prediction in the test set is 1.41 MPa and 17.6%, which has reduced the error by up to 68.8% compared to the previous wavelet‐packet‐based method.

Wavelet packet energy-based damage detection using guided ultrasonic waves and support vector machine

Substantial work has shown that the modulation of structural damage on probing guided ultrasonic waves can result in wave components corresponding to different frequencies, causing wave energy transfer from central frequency to other frequency bands. To make use of the damage-induced wave energy transfer in different frequency bands, a data-driven method which combines wavelet packet decomposition (WPD), principal component analysis (PCA) and support vector machine (SVM), is proposed in this study for structural damage identification in both metallic and composite materials. Firstly, WPD is employed to decompose the original signal into different frequency bands, based on which the wave energy at each frequency band can be determined. A wave energy distribution vector is constructed according to the energy proportion of each frequency band. Then, PCA is recalled conducting dimensionality reduction for the energy distribution vectors, in order to improve the computational efficiency for subsequent SVM classification. The compressed energy distribution vectors are used as the input to train an SVM-based classifier for identifying structural damage. To validate the proposed WPD-PCA-SVM method, experiments are performed on both aluminum plate and glass fiber reinforced polymer (GFRP) laminate. According to the experimental results, the embryonic fatigue crack in the (aluminum) plate and the anomaly in the GFRP laminate can be identified by the proposed method, with a detection accuracy of 92.86% for aluminum plate and 95.45% for GFRP laminate, respectively, demonstrating the effectiveness of the proposed method for damage detection in both metallic and composite materials.

Research on planning optimization of integrated energy system based on the differential features of hybrid energy storage system

At present, it is very important to build the new power system with new energy as the mainstay. Some Regional Integrated Energy Systems (RIES) have installed renewable energy generation equipments with wind turbines and photovoltaics as the main power supply facilities. Due to the volatility in the wind turbine, photovoltaic output power and the customer's electricity load, there is energy waste in the RIES. In order to enhance the economic and renewable energy consumption rate of the park where the RIES is applied, an integrated energy system(IES) planning optimization model is constructed，which considers the mixed storage differentiation characteristics. The model expands the hybrid energy storage system(HESS) on the basis of the original RIES. A two-stage decomposition model combining wavelet packet decomposition method and power decomposition method based on discrete Fourier transform is established, in which the difference between power generation of renewable energy and customer load is decomposed. Depending with the difference in response speed of energy storage devices, the power of different frequencies is moderated. With the comprehensive energy system economy as the goal, the genetic algorithm based on elitist preservation strategy is adopted to optimize the planning model and verify the correctness of the model through simulation. The results show an improvement of 8.23 % and 23.79 % in economy and renewable energy consumption rate compared to the RIES before the HESS expansion.

A novel hybrid learning paradigm with feature extraction for carbon price prediction based on Bi-directional long short-term memory network optimized by an improved sparrow search algorithm.

An efficient carbon trading market can effectively curb excessive carbon emissions and thus slow down the pace of global warming, which heightens the necessity of improving the accuracy of carbon price forecasting. In order to overcome the weakness of previous prediction model that always trained data in one-way neural networks and propagated the data sequentially, this paper proposes a novel hybrid learning paradigm WPD-ISSA-BiLSTM combining wavelet packet decomposition (WPD), improved sparrow search algorithm (ISSA), and Bi-directional long short-term memory network for deep feature exploration of carbon prices. Firstly, WPD decomposes and reconstructs the original carbon price series into several independent subseries. Then, the input features of the all subseries are filtered with random forest to select the best input features for the prediction model. Finally, a Bi-directional long short-term memorynetwork optimized by the ISSA is employed to deeply delineate the intrinsic evolutionary trends of carbon prices, and the prediction results of all subseries are superimposed on each other to obtain the final carbon price prediction results. The actual carbon emission trading prices are collected as input to the model, and the experimental results show that the RMSE values of the proposed model are 0.2516 and 0.2962 under the mild and severe volatility scenarios, respectively. The proposed model has superiority and robustness compared to the comparison model and several existing models and better understands the intrinsic correlation between historical carbon price data. The results of this study can provide meaningful references for the carbon market development and emission reduction pathways.

A New Hybrid Forecasting Model Based on SW-LSTM and Wavelet Packet Decomposition: A Case Study of Oil Futures Prices.

The crude oil futures prices forecasting is a significant research topic for the management of the energy futures market. In order to optimize the accuracy of energy futures prices prediction, a new hybrid model is established in this paper which combines wavelet packet decomposition (WPD) based on long short-term memory network (LSTM) with stochastic time effective weight (SW) function method (WPD-SW-LSTM). In the proposed framework, WPD is a signal processing method employed to decompose the original series into subseries with different frequencies and the SW-LSTM model is constructed based on random theory and the principle of LSTM network. To investigate the prediction performance of the new forecasting approach, SVM, BPNN, LSTM, WPD-BPNN, WPD-LSTM, CEEMDAN-LSTM, VMD-LSTM, and ST-GRU are considered as comparison models. Moreover, a new error measurement method (multiorder multiscale complexity invariant distance, MMCID) is improved to evaluate the forecasting results from different models, and the numerical results demonstrate that the high-accuracy forecast of oil futures prices is realized.

Hybrid deep neural model for hourly solar irradiance forecasting

Owing to integrating photovoltaic solar systems into power networks, accurate prediction of solar irradiance plays an increasingly significant role in electric energy planning and management. However, the existing hybrid models ignore the influence of other factors except for the irradiance time series and adopt a single branch independent network structure, which may lead to decrease prediction accuracy. In this paper, a novel multivariate hybrid deep neural model named WPD–CNN–LSTM-MLP for 1-h-ahead solar irradiance forecasting is proposed. The novel WPD–CNN–LSTM-MLP model is based on a sophisticated multi-branch hybrid structure with multi-variable inputs, which the multi-branch hybrid structure combines wavelet packet decomposition (WPD), convolutional neural network (CNN), long short-term memory (LSTM) networks, and multi-layer perceptron network (MLP), and the multi-variable inputs include hourly solar irradiance and three climate variables, namely: temperature, relative humidity, and wind speed and their combination. The new model extracts the inherent characteristics of multi-layer inputs sufficiently, overcomes the shortcomings of traditional models, and achieves more accurate forecasting results. The performance of the model is verified by actual data from Denver, Clark, and Folsom, the United States. Comparative studies of traditional individual back propagation neural network, support vector machine, recurrent neural network, LSTM, the climatology–persistence reference forecasts method and the proposed LSTM-MLP model, CNN-LSTM-MLP model, and WPD–CNN–LSTM model reveal that the proposed WPD–CNN–LSTM-MLP deep learning model has better prediction accuracy in hourly irradiance forecasting.

Accurate classification of epilepsy seizure types using wavelet packet decomposition and local detrended fluctuation analysis

Electroencephalogram (EEG) signals are widely used in diagnosis of epilepsy. Accurate classification of seizure types based on EEG signals can provide vital information for diagnosis and treatment. Since visual inspection and interpretation of seizure types are time consuming and prone to errors, a novel classification method combining wavelet packet decomposition (WPD) and local detrended fluctuation analysis (L-DFA) is proposed for the computer-aided diagnostic system. The proposed method is able to classify a wide variety of seizures automatically and accurately. As the first step towards this goal, raw EEG signals are decomposed by WPD according to intrinsic frequency bands of human brain. Then L-DFA is applied to characterise the dynamical fractal structure of sub-band signals. Finally, EEG signals are classified by support vector machine based on the combined fractal spectrum features. The experimental results on Temple University Hospital database show that the proposed method achieves a total classification accuracy of 97.80%, outperforming existing methods based on the same database.

A hybrid deep learning model for short-term PV power forecasting

The integration of PV power brings great economic and environmental benefits. However, the high penetration of PV power may challenge the planning and operation of the existing power system owing to the intermittence and randomicity of PV power generation. Achieving accurate forecasting for PV power generation is important for providing high quality electric energy for end-consumers and for enhancing the reliability of power system operation. Motivated by recent advancements in deep learning methods and their satisfactory performance in the energy sector, a hybrid deep learning model combining wavelet packet decomposition (WPD) and long short-term memory (LSTM) networks is proposed in this study. The hybrid deep learning model is utilized for one-hour-ahead PV power forecasting with five-minute intervals. WPD is first used to decompose the original PV power series into sub-series. Next, four independent LSTM networks are developed for these sub-series. Finally, the results predicted by each LSTM network are reconstructed and a linear weighting method is employed to obtain the final forecasting results. The performance of the proposed method is demonstrated with a case study using an actual dataset collected from Alice Springs, Australia. Comparisons with individual LSTM, recurrent neural network (RNN), gated recurrent (GRU), and multi-layer perceptron (MLP) models are also presented. The values of three performance evaluation indicators, MBE, MAPE, and RMSE, show that the proposed hybrid deep learning model exhibits superior performance in both forecasting accuracy and stability.

Cavitation Detection in Centrifugal Pump Based on Interior Flow-Borne Noise Using WPD-PCA-RBF

Cavitation detection is particularly essential for operating efficiency and stability of pumps. In this work, to improve the accuracy and efficiency of identification, an approach combining wavelet packet decomposition (WPD) with principal component analysis (PCA) and radial basic function (RBF) neural network is introduced to detect the cavitation status for centrifugal pumps. The cavitation performance and interior flow-borne noise are measured under three different flow conditions. Then, time-frequency domain analysis is performed on the interior flow-borne noise signal using WPD, and the energy coefficient of each node is calculated to determine the optimal decomposition frequency band. Six-feature parameters are extracted based on frequency-division statistics, including three time-domain features and three wavelet packet features. After that, the PCA is applied for dimensionality reduction. Finally, three cavitation statuses of noncavitation, inception cavitation, and serious cavitation are identified adopting RBF neural network. The results show that the comprehensive identification rate of the proposed method for three cavitation statuses reaches 98.2% with low identification error. The method based on interior flow-borne noise analysis can be well applied for on-line monitoring and diagnosis of pump industry.

WPD and DE/BBO-RBFNN for solution of rolling bearing fault diagnosis

Abstract A new system of fault diagnosis which combines wavelet packet decomposition (WPD), radical basis function neural network (RBFNN) and a hybrid differential evolution with biogeography-based optimization (DE/BBO) algorithm is designed in this paper to improve the efficiency and accuracy of rolling bearing fault diagnosis. Biogeography-based optimization (BBO) is a new biogeography inspired algorithm, however it has the disadvantages of early-maturing and instable convergence. Combining BBO with differential evolution (DE) and making small modifications for the mutation strategy of DE, an improved DE/BBO algorithm is formed. Firstly, the fault feature vectors of original rolling bearing fault signals are effectively obtained by wavelet packet decomposition and reconstruction. Secondly, the RBF neural network is optimized by DE/BBO algorithm, the fault types of rolling bearing are trained and diagnosed next. The result of MATLAB simulation shows that in contrast to the traditional RBFNN, the fitness and precision of bearing fault diagnosis are higher and the root-mean-square error (RMSE) is lower because of the introduction of DE/BBO algorithm.

Fuzzy distribution entropy and its application in automated seizure detection technique

Abstract Visual inspection of Electroencephalogram (EEG) records is the conventional diagnostic method of epilepsy but it is expensive, time-consuming and tedious. Therefore, it is necessary to develop automated seizure detection technologies. In this paper, a new entropy named fuzzy distribution entropy (fDistEn) was first put forward and then a seizure detection scheme combining wavelet packet decomposition (WPD), fDistEn, Kruskal-Wallis nonparametric one-way analysis of variance and k -nearest neighbor ( k -NN) classifier was proposed. In the proposed scheme, WPD was firstly implemented to decompose the filtered EEG into several wavelet sub-bands. Subsequently, fDistEn values of all nodes in every level were calculated and followed by selecting significant features using Kruskal-Wallis test. Finally, k -NN was employed to classify ten kinds of EEG combinations. Experimental results show fDistEn can measure the complexity of signals and our proposed scheme is qualified to detect seizure automatically with not less than 98.338% accuracy in all cases. Compared with existing methods, our scheme outperforms most of state-of-the-art articles and it indicates the effectiveness of the proposed seizure detection scheme.

River Stage Forecasting Using Wavelet Packet Decomposition and Machine Learning Models

This study develops and applies three hybrid models, including wavelet packet-artificial neural network (WPANN), wavelet packet-adaptive neuro-fuzzy inference system (WPANFIS) and wavelet packet-support vector machine (WPSVM), combining wavelet packet decomposition (WPD) and machine learning models, ANN, ANFIS and SVM models, for forecasting daily river stage and evaluates their performance. The WPANN, WPANFIS and WPSVM models using inputs decomposed by the WPD are found to produce higher efficiency based on statistical performance criteria than the ANN, ANFIS and SVM models using original inputs. Performance evaluation for various mother wavelets indicates that the model performance is dependent on mother wavelets and the WPD using Symmlet-10 and Coiflet-18 is more effective to enhance the efficiency of the conventional machine learning models than other mother wavelets. It is found that the WPANFIS model outperforms the WPANN and WPSVM models, and the WPANFIS14-coif18 model produces the best performance among all other models in terms of model efficiency. Therefore, the WPD can significantly enhance the accuracy of the conventional machine learning models, and the conjunction of the WPD and machine learning models can be an effective tool for forecasting daily river stage accurately .

River Stage Forecasting Using Wavelet Packet Decomposition and Data-driven Models

This study develops and applies hybrid models combining wavelet packet decomposition and data-driven models for forecasting river stage and investigating their accuracy. The hybrid models are wavelet packet-based artificial neural network (WPANN) and wavelet packet-based adaptive neuro-fuzzy inference system (WPANFIS). Wavelet packet decomposition splits an input time series into approximation and detail components, and the decomposed time series are used as inputs to artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) for WPANN and WPANFIS models, respectively. The forecasting performances of WPANN and WPANFIS models are compared with those of ANN and ANFIS models. Based on performance evaluation indexes and graphical comparison, the WPANN and WPANFIS models produce better performance than ANN and ANFIS models in terms of model efficiency, and WPANFIS-coif18 model is also found to yield the best performance among all other models. Therefore, wavelet packet decomposition improves the accuracy of ANN and ANFIS models for forecasting river stage. The results indicate that river stage forecasting models combining wavelet packet decomposition and data-driven models can be used as an effective tool for forecasting river stage accuracy.

DEPRESSION DIAGNOSIS SUPPORT SYSTEM BASED ON EEG SIGNAL ENTROPIES

Electroencephalography (EEG) is a measure which represents the functional activity of the brain. We show that a detailed analysis of EEG measurements provides highly discriminant features which indicate the mental state of patients with clinical depression. Our feature extraction method revolves around a novel processing structure that combines wavelet packet decomposition (WPD) and non-linear algorithms. WPD was used to select appropriate EEG frequency bands. The resulting signals were processed with the non-linear measures of approximate entropy (ApEn), sample entropy (SampEn), renyi entropy (REN) and bispectral phase entropy ( P h). The features were selected using t-test and only discriminative features were fed to various classifiers, namely probabilistic neural network (PNN), support vector machine (SVM), decision tree (DT), k-nearest neighbor algorithm (k-NN), naive bayes classification (NBC), Gaussian mixture model (GMM) and Fuzzy Sugeno Classifier (FSC). Our classification results show that, with a classification accuracy of 99.5%, the PNN classifier performed better than the rest of classifiers in discriminating between normal and depression EEG signals. Hence, the proposed decision support system can be used to diagnose, and monitor the treatment of patients suffering from depression.

An approach to monitoring the thermomechanical behavior of a spindle bearing system using acoustic emission (AE) energy

Real time condition monitoring plays a key role in modern manufacturing processes. Critical processes, however, require systems capable of early detection and rapid adjustment of bearing conditions in order to ensure processing quality. Specifically, the spindle stiffness is significantly affected by the bearing contact pressure, which in turn is readily changed by thermal effects. A technique is proposed to monitor the spindle condition via acoustic emission (AE) signals. This method implements an analysis combining wavelet packet decomposition, Hilbert-Huang Transform, and screening processes (WPD-HHT) to extract the crucial characteristics from the measured signal to identify the running state of a spindle bearing. The implementation of the WPD-HHT envelope technique is also presented. Acoustic emission signals as well as the temperature growth measured from the spindle housing are used in the investigation. The results show that the proposed WPD-HHT and WPD-HHT envelope techniques enable the extraction of the most important signal features and their presentation in term of energy values. The displayed energy value is correlated to the preload changes since the temperature of the rolling bearing increases with AE energy.

Design of Online Monitoring and Fault Diagnosis System for Belt Conveyors Based on Wavelet Packet Decomposition and Support Vector Machine

Belt conveyors are the equipment widely used in coal mines and other manufacturing factories, whose main components are a number of idlers. The faults of belt conveyors can directly influence the daily production. In this paper, a fault diagnosis method combining wavelet packet decomposition (WPD) and support vector machine (SVM) is proposed for monitoring belt conveyors with the focus on the detection of idler faults. Since the number of the idlers could be large, one acceleration sensor is applied to gather the vibration signals of several idlers in order to reduce the number of sensors. The vibration signals are decomposed with WPD, and the energy of each frequency band is extracted as the feature. Then, the features are employed to train an SVM to realize the detection of idler faults. The proposed fault diagnosis method is firstly tested on a testbed, and then an online monitoring and fault diagnosis system is designed for belt conveyors. An experiment is also carried out on a belt conveyor in service, and it is verified that the proposed system can locate the position of the faulty idlers with a limited number of sensors, which is important for operating belt conveyors in practices.

Fault Diagnosis of Railway Axlebox Bearing Based on Wavelet Packet and Neural Network

A real time and effective axlebox bearing fault diagnostic method is significant in the condition-based maintenance. In the axlebox bearing fault diagnostic system, fault features extraction and fault patterns classification are two important aspects to identify whether a axlebox bearing is failure or not. This paper presents a method of axlebox bearing fault diagnosis based on wavelet packet decomposition and BP neural network. First decompose the vibration signal into a finite number of coefficients by wavelet packet decomposition. Then calculate energy moment of each coefficient and take the energy moment as an eigenvector to effectively express the failure feature. Finally BP neural network is used for fault classification. The experimental results show that combining wavelet packet decomposition with BP neural network could identify the axlebox bearing fault effectively. The average diagnosis accuracy rate is 96.67%.

Fault location in a series compensated transmission line based on wavelet packet decomposition and support vector regression

This paper proposes a novel transmission line fault location scheme, combining wavelet packet decomposition (WPD) and support vector regression (SVR). Various types of faults at different locations, fault resistance and fault inception angles on a series compensated 400 kV–285.65 km power system transmission line are investigated. The system only utilizes a single-end measurements. WPD is used to extract distinctive fault features from 1/2 cycle of post fault signals after noises have been eliminated by a low pass filter, and SVR is trained with features obtained from WPD. After training, SVR was then used in precise location of fault on the transmission line. The result shows that fault location on transmission line can be determined rapidly and correctly irrespective of fault impedance.