“…For long-term degradation prediction, assuming that the prediction starts at time t, the predicted RUL is defined as the difference between the predicted end-of-life (EOL) time and the current time [37], as shown in the following equation. In the comparison experiments, all models are trained with the first 450 h of voltage data, and the employed training datasets are the voltage data that has been SSA denoised and resampled.…”
An accurate prediction of the remaining useful life (RUL) of a proton exchange membrane fuel cell (PEMFC) is of great significance for its large-scale commercialization and life extension. This paper aims to develop a PEMFC degradation prediction method that incorporates short-term and long-term predictions. In the short-term prediction, a long short-term memory (LSTM) neural network is combined with a Gaussian process regression (GPR) probabilistic model to form a hybrid LSTM-GPR model with a deep structure. The model not only can accurately forecast the nonlinear details of PEMFC degradation but also provide a reliable confidence interval for the prediction results. The results showed that the proposed LSTM-GPR model outperforms the single models in both prediction accuracy and confidence interval. For the long-term prediction, a novel RUL prediction model based on an extended Kalman filter (EKF) and GPR is proposed. The GPR model is used to solve the problem that the EKF cannot update the model parameters in the prediction stage. The results showed that the proposed EKF-GPR model can achieve better RUL prediction than the model-based approach and the data-driven approach.
“…For long-term degradation prediction, assuming that the prediction starts at time t, the predicted RUL is defined as the difference between the predicted end-of-life (EOL) time and the current time [37], as shown in the following equation. In the comparison experiments, all models are trained with the first 450 h of voltage data, and the employed training datasets are the voltage data that has been SSA denoised and resampled.…”
An accurate prediction of the remaining useful life (RUL) of a proton exchange membrane fuel cell (PEMFC) is of great significance for its large-scale commercialization and life extension. This paper aims to develop a PEMFC degradation prediction method that incorporates short-term and long-term predictions. In the short-term prediction, a long short-term memory (LSTM) neural network is combined with a Gaussian process regression (GPR) probabilistic model to form a hybrid LSTM-GPR model with a deep structure. The model not only can accurately forecast the nonlinear details of PEMFC degradation but also provide a reliable confidence interval for the prediction results. The results showed that the proposed LSTM-GPR model outperforms the single models in both prediction accuracy and confidence interval. For the long-term prediction, a novel RUL prediction model based on an extended Kalman filter (EKF) and GPR is proposed. The GPR model is used to solve the problem that the EKF cannot update the model parameters in the prediction stage. The results showed that the proposed EKF-GPR model can achieve better RUL prediction than the model-based approach and the data-driven approach.
“…The specific parameters of simulation signal are described as follows: the sampling frequency is 1 MHz and the number of sampling points is 2×107; the coefficients of resonance damping B are all 900; the resonance frequencies f are 220 kHz, 440 kHz, and 650 kHz, respectively; the phases φ are all 0; and the impulse frequencies are 8 kHz, 14 kHz, and 19 kHz, respectively, and the slippage 18 is set as 2%. Figure 1 is the spectrum of the three fault impulse components.Figure 1.The spectrum of the three fault impulse components.…”
Section: Proposed Methodsmentioning
confidence: 99%
“…VMD can be regard as a series of wiener filter banks with narrow-band characteristics, which can adaptively divide the frequency domain of signal; thus, a series of mode components with sparse characteristics can be obtained through VMD, and both the construction and the solution of variation problem are the core idea of VMD. 17,18 Assuming a signal is decomposed to K modes, then the construction of the variational problem can be described as follows:…”
Section: Background Knowledgementioning
confidence: 99%
“…17 Different from EEMD and LMD, VMD is a completely non-recursive decomposition algorithm, in which the modes are extracted simultaneously, and it has been widely applied in the field of fault diagnosis. However, VMD performance is strictly subjected to the two key parameters (i.e., mode number K and balance parameter α), and some optimization algorithms such as cuckoo search algorithm (CSA) 18 and pigeon heuristic algorithm 19 are proposed to determine the key parameters of VMD. These methods can well extract fault features, but the parameter optimization process is too time consuming, which has certain limitations in practical applications.…”
To alleviate the difficulty of extracting weak fault features of hoisting machinery, a progressive decomposing and double screening strategy of variational mode decomposition (VMD) is presented in this paper. Firstly, the feasibility and effectiveness of extracting fault modes using progressive decomposition strategy is validated through numerical simulation, and it solves the problem of determining the mode number [Formula: see text] in traditional VMD. Secondly, a new index named energy fluctuation factor (EFF) is proposed. Specifically, EFF is more effective in detecting the signal periodicity compared with the kurtosis and the Shannon entropy (SE), and it is used to optimize the balance parameter [Formula: see text] of VMD. Thirdly, the criterion of double screening based on the kurtosis and the EFF is given to accurately localize and reconstruct the fault modes, and then Hilbert transform is utilized to demodulate the reconstructed mode. Finally, the numerical simulation and experimental and practical engineering applications verify that the proposed method can accurately extract the modes of weak fault and well solve the problem of determining the key parameters (i.e., [Formula: see text] and [Formula: see text] ) of VMD. Furthermore, the superiority of the proposed method is validated by comparing with other fault diagnosis methods.
“…Fuel cell degradation prognosis was performed based on Nonlinear Autoregressive Exogenous Neural Networks (NARX) and wavelet analysis in [32]. Degradation prognosis was also studied in [33] by using a degradation model based on the Wavelet Neural Network (WNN) and Cuckoo search algorithm. Energy management in PEMFC vehicles was studied using a MATLAB/Simulink model in [34].…”
The fault conditions degrade the performance of proton exchange fuel cells and reduce their useful life. The prolonged existence of a fault condition can permanently damage the fuel cell. This paper proposes four methods for fault detection and fault type isolation. These methods were based on the coefficient of variance, ratios of change in output power to change in voltage and change in output voltage to the change in current, fuzzy membership values and Euclidian distance, and wavelet transform. These methods are non-invasive to the fuel cell and involve non-destructive testing. These methods were experimentally validated.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
