Blade bearings are joint components of variable-pitch wind turbines which have high failure rates. This paper diagnoses a naturally damaged wind turbine blade bearing which was in operation on a wind farm for over 15 years; therefore, its vibration signals are more in line with field situations. The focus is placed on the conditions of fluctuating slow-speeds and heavy loads, because blade bearings bear large loads from wind turbine blades and their rotation speeds are sensitively affected by wind loads or blade flipping. To extract weak fault signals masked by heavy noise, a novel signal denoising method, Bayesian Augmented Lagrangian (BAL) Algorithm, is used to build a sparse model for noise reduction. BAL can denoise the signal by transforming the original filtering problem into several sub-optimization problems under the Bayesian framework and these sub-optimization problems can be further solved separately. Therefore, it requires fewer computational requirements. After that, the BAL denoised signal is resampled with the aim of eliminating spectrum smearing and improving diagnostic accuracy. The proposed framework is validated by different experiments and case studies. The comparison with respect to some popular diagnostic methods is explained in detail, which highlights the superiority of our introduced framework.

This paper presents a “structured” learning approach for the identification of continuous partial differential equation (PDE) models with both constant and spatial-varying coefficients. The identification problem of parametric PDEs can be formulated as an ℓ1/ℓ2-mixed optimization problem by explicitly using block structures. Block-sparsity is used to ensure parsimonious representations of parametric spatiotemporal dynamics. An iterative reweighted ℓ1/ℓ2 algorithm is proposed to solve the ℓ1/ℓ2-mixed optimization problem. In particular, the estimated values of varying coefficients are further used as data to identify functional forms of the coefficients. In addition, a new type of structured random dictionary matrix is constructed for the identification of constant-coefficient PDEs by introducing randomness into a bounded system of Legendre orthogonal polynomials. By exploring the restricted isometry properties of the structured random dictionary matrices, we derive a recovery condition that relates the number of samples to the sparsity and the probability of failure in the Lasso scheme. Numerical examples, such as the Schrödinger equation, the Fisher-Kolmogorov-Petrovsky-Piskunov equation, the Burger equation, and the Fisher equation, suggest that the proposed algorithm is fairly effective, especially when using a limited amount of measurements.

Nonlinear Auto-Regressive model with eXogenous input (NARX) is one of the most popular black-box model classes that can describe many nonlinear systems. The structure determination is the most challenging and important part during the system identification. NARX can be formulated as a linear-in-the-parameters model, then the identification problem can be solved to obtain a sparse solution from the viewpoint of the weighted l 1 minimization problem. Such an optimization problem not only minimizes the sum squares of model errors but also the sum of reweighted model parameters. In this paper, a novel algorithm named Bayesian Augmented Lagrangian Algorithm (BAL) is proposed to solve the weighted l 1 minimization problem, which is able to obtain a sparse solution and enjoys fast computation. This is achieved by converting the original optimization problem into distributed suboptimization problems solved separately and penalising the overall complex model to avoid overfitting under the Bayesian framework. The regularization parameter is also iteratively updated to obtain a satisfied solution. In particular, a solver with guaranteed convergence is constructed and the corresponding theoretical proof is given. Two numerical examples have been used to demonstrate the effectiveness of the proposed method in comparison to several popular methods.

Leonurine (Leo) has been found to have neuroprotective effects against cerebral ischemic injury. However, the exact molecular mechanism underlying its neuroprotective ability remains unclear. The aim of the present study was to investigate whether Leo could provide protection through the nitric oxide (NO)/nitric oxide synthase (NOS) pathway. We firstly explored the effects of NO/NOS signaling on oxidative stress and apoptosis in in vivo and in vitro models of cerebral ischemia. Further, we evaluated the protective effects of Leo against oxygen and glucose deprivation (OGD)-induced oxidative stress and apoptosis in PC12 cells. We found that the rats showed anxiety-like behavior, and the morphology and number of neurons were changed in a model of photochemically induced cerebral ischemia. Both in vivo and in vitro results show that the activity of superoxide dismutase (SOD) and glutathione (GSH) contents were decreased after ischemia, and reactive oxygen species (ROS) and malondialdehyde (MDA) levels were increased, indicating that cerebral ischemia induced oxidative stress and neuronal damage. Moreover, the contents of NO, total NOS, constitutive NOS (cNOS) and inducible NOS (iNOS) were increased after ischemia in rat and PC12 cells. Treatment with L-nitroarginine methyl ester (L-NAME), a nonselective NOS inhibitor, could reverse the change in NO/NOS expression and abolish these detrimental effects of ischemia. Leo treatment decreased ROS and MDA levels and increased the activity of SOD and GSH contents in PC12 cells exposed to OGD. Furthermore, Leo reduced NO/NOS production and cell apoptosis, decreased Bax expression and increased Bcl-2 levels in OGD-treated PC12 cells. All the data suggest that Leo protected against oxidative stress and neuronal apoptosis in cerebral ischemia by inhibiting the NO/NOS system. Our findings indicate that Leo could be a potential agent for the intervention of ischemic stroke and highlighted the NO/NOS-mediated oxidative stress signaling.

Variable selection methods have been widely used for system identification. However, there is still a major challenge in producing parsimonious models with optimal model structures as popular variable selection methods often produce suboptimal model with redundant model terms. In the paper, stability orthogonal regression (SOR) is proposed to build a more compact model with fewer or no redundant model terms. The main idea of SOR is that multiple intermediate models are produced by orthogonal forward regression (OFR) using sub-sampling technique and then the final model is a combination of these intermediate model terms but does not include infrequently selected terms. The effectiveness of the proposed methods is analysed in theory and also demonstrated using two numerical examples in comparison with some popular algorithms.