In guided wave pipeline inspection, echoes reflected from closely spaced reflectors generally overlap, meaning useful information is lost. To solve the overlapping problem, sparse deconvolution methods have been developed in the past decade. However, conventional sparse deconvolution methods have limitations in handling guided wave signals, because the input signal is directly used as the prototype of the convolution matrix, without considering the waveform change caused by the dispersion properties of the guided wave. In this paper, an adaptive sparse deconvolution (ASD) method is proposed to overcome these limitations. First, the Gaussian echo model is employed to adaptively estimate the column prototype of the convolution matrix instead of directly using the input signal as the prototype. Then, the convolution matrix is constructed upon the estimated results. Third, the split augmented Lagrangian shrinkage (SALSA) algorithm is introduced to solve the deconvolution problem with high computational efficiency. To verify the effectiveness of the proposed method, guided wave signals obtained from pipeline inspection are investigated numerically and experimentally. Compared to conventional sparse deconvolution methods, e.g. the l 1 -norm deconvolution method, the proposed method shows better performance in handling the echo overlap problem in the guided wave signal.

The reconstruction algorithm for the probabilistic inspection of damage (RAPID) is aimed at localizing structural damage via the signal difference coefficient (SDC) between the signals of the present and reference conditions. However, tomography is only capable of presenting the approximate location and not the length and angle of defects. Therefore, a new quantitative evaluation method called the multiple back propagation neural network (Multi-BPNN) is proposed in this work. The Multi-BPNN employs SDC values as input variables and outputs the predicted length and angle, with each output node depending on an individual hidden layer. The cracks of different lengths and angles at the center weld seam of offshore platforms are simulated numerically. The SDC values of the simulations and experiments were normalized for each sample to eliminate external interference in the experiments. Then, the normalized simulation data were employed to train the proposed neural network. The results of the simulations and experimental verification indicated that the Multi-BPNN can effectively predict crack length and angle, and has better stability and generalization capacity than the multi-input to multi-output back propagation neural network.

No abstract

Centrifugal booster fans are important equipment used to recover blast furnace gas (BFG) for generating electricity, but blade crack faults (BCFs) in centrifugal booster fans can lead to unscheduled breakdowns and potentially serious accidents, so in this work quantitative fault identification and an abnormal alarm strategy based on acquired historical sensor-dependent vibration data is proposed for implementing condition-based maintenance for this type of equipment. Firstly, three group dependent sensors are installed to acquire running condition data. Then a discrete spectrum interpolation method and short time Fourier transform (STFT) are applied to preliminarily identify the running data in the sensor-dependent vibration data. As a result a quantitative identification and abnormal alarm strategy based on compound indexes including the largest Lyapunov exponent and relative energy ratio at the second harmonic frequency component is proposed. Then for validation the proposed blade crack quantitative identification and abnormality alarm strategy is applied to analyze acquired experimental data for centrifugal booster fans and it has successfully identified incipient blade crack faults. In addition, the related mathematical modelling work is also introduced to investigate the effects of mistuning and cracks on the vibration features of centrifugal impellers and to explore effective techniques for crack detection.

No abstract