Battery prognostics aims to predict the remaining life of a battery and to perform necessary maintenance service if necessary using the past and current information. A reliable prognostic model should be able to accurately predict the future state of the battery such that the maintenance service could be scheduled in advance. In this paper, a multistep-ahead prediction model based on the mean entropy and relevance vector machine (RVM) is developed, and applied to state of health (SOH) and remaining life prediction of the battery. A wavelet denoising approach is introduced into the RVM model to reduce the uncertainty and to determine trend information. The mean entropy based method is then used to select the optimal embedding dimension for correct time series reconstruction. Finally, RVM is employed as a novel nonlinear time-series prediction model to predict the future SOH and the remaining life of the battery. As more data become available, the accuracy and precision of the prediction improve. The presented approach is validated through experimental data collected from Li-ion batteries. The experimental results demonstrate the effectiveness of the proposed approach, which can be effectively applied to battery monitoring and prognostics.

Intelligent Prognostics for Battery Health Monitoring Using the Mean Entropy and Relevance Vector Machine

https://speechlab.eece.mu.edu/johnson/papers/johnson_speechcomm07.pdf

Speech signal enhancement through adaptive wavelet thresholding

Despite advances in the filtering techniques of partial discharge (PD) signals, an acceptable signal to noise ratio (SNR) remains elusive due to the presence of environmental noises. A wavelet packet transform-based method is proposed for effective denoising of on-site partial discharge data from a power generator and a gas-insulated substation. Taking advantage of the statistical properties of white noise, the method uses variance as the criterion to select the best tree. It is shown by comparative studies that the method restores the PD pulse from data with very low SNR. The denoised PD pulses enable more accurate diagnosis of insulation defects in HV equipment.

Denoising of partial discharge signals in wavelet packets domain

The diagnosis of brain tumours is an extremely sensitive and complex clinical task that must rely upon information gathered through non-invasive techniques. One such technique is Magnetic Resonance Spectroscopy. In this task, radiology experts are likely to benefit from the support of computer-based systems built around robust classification processes. In this paper, a Discrete Wavelet Transform procedure was applied to the pre-processing of spectra corresponding to several brain tumour pathologies. This procedure does not alleviate the high dimensionality of the data by itself. For this reason, dimensionality reduction was subsequently implemented using Moving Window with Variance Analysis for feature selection or Principal Component Analysis for feature extraction. The combined method yielded very encouraging results in terms of diagnostic discriminatory binary classification using Bayesian Neural Networks. In most cases, the classification accuracy improved on previously reported results.

Classification of human brain tumours from MRS data using Discrete Wavelet Transform and Bayesian Neural Networks

Noise reduction of nuclear magnetic resonance (NMR) transversal data using improved wavelet transform and exponentially weighted moving average (EWMA).

This paper introduces the principle of wavelet multiresolution analysis. Four kinds of threshold selection rules and three methods of threshold processing are given. According to the computer simulation, the best rule of estimating the threshold is found, which is combined with three kinds of threshold processing methods to denoise the same noisy signal. The best method of threshold processing is obtained by comparing the performance of three methods of threshold processing that are applied to denoising.

Dai-Fei Guo

Papers

A study of wavelet thresholding denoising