Comparison and analysis of multiple signal processing methods in steel wire rope defect detection by hall sensor

Highlights

• Experiment for wire rope inspection by hall sensor and MFL testing is conducted.

• Influence of the signal processing parameters to defect identification is studied.

• Interference signals could be eliminated when the parameters were suitably set.

• Comparisons of different signal processing methods are conducted and analyzed.

• A reference for selection of wire rope defect signal processing method is formed.

Abstract

Steel wire rope (SWR) defect signal processing techniques and methods are the key to exact defect detection and identification, which is important in guaranteeing human lives and property. But the inspection signals sensed by the commonly used inductive coil sensor are often susceptible to the detector’s scanning speed, wire rope shaking and vibration, whereas they are unaffected when induced by hall element sensor. Thus, the experiments for wire rope local fault (LF) of broken wire inspection by hall sensors and magnetic flux leakage (MFL) testing method are conducted, and different groups of original signals are obtained. Then, performances of multiple signal processing methods are compared and evaluated in the perspective of baseline drift elimination and signal denoising. Besides, the main impact parameters in these signal processing methods are investigated and revealed. Finally, comparisons and discussions for different filtering methods are analyzed and presented.

Introduction

Steel wire rope (SWR) is one of the most important loading and transporting tools in various application scenes such as the elevator, ocean platform, bridge cable, cableway, and coal mine. However, the internal and external surroundings of wear, shock and strike as well as the chemical corrosion have resulted in multifarious defects for SWR and brought huge economic and human live losses in recent years. Generally, these defects could be divided into two types [1], [2], namely, the local fault (LF) such as the wire broken, crack, corrosion and the loss of metallic sectional area (LMA) such as the wear and abrasion. The testing methods have been developed and applied from the simple visual method to electromagnetic nondestructive testing (ENDT) method such as the magnetic flux leakage (MFL) testing method [3], [4], eddy current testing (ECT) [5], acoustic emission (AE) testing [6] and ultrasonic guided wave (UGW) inspection [7] methods. Other wire rope testing methods may also include the metal magnetic memory method [8], X/γ ray testing method [9] as well as the machine vision based holographic testing method [10] and infrared thermal imaging method [6]. Besides, lots of finite element modeling (FEM) approaches were also proposed to improve the performance of the eddy current computations based on the Galerkin equations with a certain of novelty, such as the conjugate gradients squared (CGS) method with an optimized initial guess-the final solution from the previous frequency can significantly speed up the convergence of the CGS solving process and reduce the number of iterations particularly in the multi-frequency mode [11]. However, the most reliable and effective one is the MFL testing method [12]. Naturally, the testing sensors are invented and applied according to the inspection principles, and the frequently used sensors may include the inductive coil [13], hall element sensor [14], magnetic flux gate [15], magnetic resistance sensors [16] such as the TMR, GMR and AMR as well as the acoustic filtering sensor [17], the new proposed flexible printed coil [18]. Nevertheless, the testing signals can all be classified into the axial, radial and even the circumferential signal in the 3D MFL testing according to the spiral and columnar geometric structure of the SWR.

Actually, the detection signal can also be divided into LF and LMA signals [19] according to the defect types, which are featured with linear and drifting baseline, respectively, and a mass of wire rope signal processing strategies have been reported. Such as the notch filtering and wavelet denoising methods for strand signal reduction in MFL testing [20], and the time–frequency based short time fourier transform (STFT) through UGW testing [21]. Concretely, Jie et al. [22] proposed a morphological non-sampling wavelet method for online coal mine wire rope testing signal processing, which could filter out the baseline drift noise effectively by improving the signal-to-noise ratio (SNR). Owing to the intrinsic relation within the defect signals, the correlation analysis was also applied in elevator rope monitoring and fault detection [23], [24]. Zhou and Liu et al. [25] proposed a hall sensor array and MFL imaging based multi-channel signal fusion and oblique-directional resampling method for wire rope LF defect denoising under strong shaking and strand condition, which improved the defect detection performance effectively [26]. Zhang et al. [27] analyzed the MFL signals of SWR through 3D finite element modeling (FEM) and a signal conditioning and processing circuit to enhance the defect signal detectability, and the experimental results manifested that even a pit on half of a wire could be identified. Besides, they also eliminated the strand noise and achieved the accurate defect detection and location for wire rope by using a small device and the instantaneous phase solution in Hilbert transform and wavelet analysis [28]. What’s more, various of signal compensating sensors and algorithms have also been presented to eliminate the liftoff effect influenced by the corrosion and coating, for instance, the planar triple-coil [29], the combined inductive and capacitive sensor [30] as well as the optimized transmitter–receiver sensor and the modified Newton-Raphson algorithms for multi-frequency electromagnetic sensing system [31], which could all be effectively applied in reducing the measurement error and solving the inverse problems as well as the parameters influence of the permeability and conductivity. As the artificial intelligent techniques and quantitative testing requirements for wire rope defect inspection develop, various quantitative defect identification methods are proposed. Gao et al. [32] studied the relationships of the static magnetic field and the broken wire defect with different widths through the finite element analysis and experiments, where the wavelet transform was applied in the eliminating for background noises. Zhou et al. [33] made use of a hybrid data driven method based on the support vector machine (SVM), and the data mining experimental results indicated that the optimized algorithm was effective in various types of defect inspection. Other reported signal processing methods and techniques [34] in the quantitative defect detection may also include the adaptive filter, ensemble empirical mode decomposition, Elman and BP neural network as well as the convolutional neural network (CNN) [35], which all make the wire rope signal processing and quantitative testing more intelligent. It’s also a developing trend for wire rope defect inspection from the qualitative to quantitative testing. However, the hostile operation environment always makes the wire rope testing become challengeable, most of the new proposed methods still need to be verified, especially for the detection feasibility and accuracy in the practical applications.

Noteworthily, the LMA defects detected by the MFL testing method are usually acquired through the hall element and related sensors, which pick up the absolute value of the magnetic field, while the LF defect information or the leaked magnetic field is commonly acquired by the inductive coils or related magnetic sensitive elements, which mainly sense the relative magnetic field according to the faraday law of electromagnetic induction. Owing to the sophisticated testing environment as well as the online inspection condition, when the tested SWR was scanned by an additional electromagnetic detector, any shake and strike for the detector or the vibration of the wire rope would make big influence to the testing signals. Besides, other interferences including the background and the in-suite electromagnetic noise would all make the wire rope testing signals become more complicated. Apparently, the LF signal of SWR is frequently affected by the speed effect when sensed by various of inductive coil sensors. To avoid these unnecessary troubles, a hall sensor based LF defect detection and signal processing method is proposed and compared.

In consideration of the above-mentioned challenges and to eliminate the baseline drift and noise interferences in LF defect detection and identification for SWR, comparisons of multiple signal processing methods for SWR are conducted, and the main influencing parameters in these methods are investigated and analyzed. According to the experimental results for eight groups of typical LF defects acquired by hall element sensor, the signal processing performance and results for baseline drift and noise reduction are mainly investigated by the SNR calculation and comparison. Then, eight commonly used signal processing methods for SWR are compared and discussed. Finally, a suitable selection for SWR signal processing method is referenced, which not only provides a theoretical basis to the SWR signal processing, but also is helpful to the exact SWR defect inspection and identification under various application conditions.