Steel wire rope can be frequently damaged by various factors. Although weak defect signal detection is the key to guarantee safety, current classic stochastic resonance (SR) models and methods are unavailable due to the complicated interference of noise. Thus, in this article, a novel E-exponential SR model and weak signal detection method are proposed. The principle of this model is first introduced from the perspective of Taylor expansion and nonlinear system analysis. Then, the model performance is simulated and evaluated by output signal analysis and chaos characterizations. Afterward, model comparison with different system parameters, experiments, and case studies regarding three typical wire rope defects inspection and weak signal recognition is conducted to demonstrate the validity of the novel method. Finally, the advantages and limitations of the proposed SR model are discussed.

The quantitative defect detection of wire rope is crucial to guarantee safety in various application scenes, and sophisticated inspection conditions usually lead to the accurate testing of difficulties and challenges. Thus, a magnetic flux leakage (MFL) signal analysis and convolutional neural networks (CNNs)-based wire rope defect recognition method was proposed to solve this challenge. Typical wire rope defect inspection data obtained from one-dimensional (1D) MFL testing were first analyzed both in time and frequency domains. After the signal denoising through a new combination of Haar wavelet transform and differentiated operation and signal preprocessing by normalization, ten main features were used in the datasets, and then the principles of the proposed MFL and 1D-CNNs-based wire rope defect classifications were presented. Finally, the performance of the novel method was evaluated and compared with six machine learning methods and related algorithms, which demonstrated that the proposed method featured the highest testing accuracy (>98%) and was valid and feasible for the quantitative and accurate detection of broken wire defects. Additionally, the considerable application potential as well as the limitations of the proposed methods, and future work, were discussed.

The goal of continual learning (CL) is to continuously learn new tasks without forgetting previously learned old tasks. To alleviate catastrophic forgetting, gradient projection based CL methods require that the gradient updates of new tasks are orthogonal to the subspace spanned by old tasks. This limits the learning process and leads to poor performance on the new task due to the projection constraint being too strong. In this paper, we first revisit the gradient projection method from the perspective of flatness of loss surface, and find that unflatness of the loss surface leads to catastrophic forgetting of the old tasks when the projection constraint is reduced to improve the performance of new tasks. Based on our findings, we propose a Data Augmented Flatness-aware Gradient Projection (DFGP) method to solve the problem, which consists of three modules: data and weight perturbation, flatness-aware optimization, and gradient projection. Specifically, we first perform a flatness-aware perturbation on the task data and current weights to find the case that makes the task loss worst. Next, flatness-aware optimization optimizes both the loss and the flatness of the loss surface on raw and worst-case perturbed data to obtain a flatness-aware gradient. Finally, gradient projection updates the network with the flatness-aware gradient along directions orthogonal to the subspace of the old tasks. Extensive experiments on four datasets show that our method improves the flatness of loss surface and the performance of new tasks, and achieves state-of-the-art (SOTA) performance in the average accuracy of all tasks.

Support vector machine (SVM) has been applied in data classification and defect recognition in various scenes, the common kernels are not always suitable to satisfy the requirement of high accuracy and efficiency inspection. A new hybrid conditional kernel (HCK) and SVM model based on exponential functions is proposed in this article, the validity and feasibility of the basic characterizations are introduced theoretically first, and the numerical classification results for different datasets are calculated and compared with related algorithms. Then, the performance evaluation of the new model from the perspective of kernel parameter investigation, statistical data analysis, and experimental verification are conducted, and the wire rope defect inspection sample sets are built according to features extraction and eigenvalues computation. Finally, the comparison results where the new HCK based SVM can reach almost the highest classification accuracy of 91.70% and least running time of 0.1968 s for five different wire rope defects among 12 machine learning models not only prove the superior performance of the proposed method in wire rope defect recognition, but also show great promise in high precision classification in other application fields. The limitations of this model and future work are also discussed.

As one of the most important processes in steel wire rope inspection, defect signal processing is of great significance in guaranteeing safety and precision measurement. Aiming at the weak signal detection of steel wire rope with mixed strands and noise, a combined signal processing method based on magnetic flux leakage testing and multi-step filtering techniques are proposed in this paper. The experiments are first introduced and performed on three typical types of steel wire rope with diameters of 28 mm, 32 mm, 45 mm, and different wire broken defects detected under liftoff distances of 13 mm and 20 mm; the acquired signals are then analyzed both in time and frequency domain. According to the weak signal characterizations, the principle of the proposed methods and algorithm are given concretely. Afterwards, comparison of signal processing results between the traditional lowpass filtering, wavelet denoising, median filtering, and the proposed method are presented. Finally, the influence factors of smoothing types and moving average span of the proposed methods are investigated. The processing results of the proposed methods are shown through short-time Fourier transform and signal-to-noise ratio analysis, which not only demonstrates the validity and feasibility of the combined methods with the highest signal to noise ratio of 90.37 dB, but also exhibits a great potential of precision defect detection and practical application in steel wire rope inspection.