Search

The size of datasets is growing exponentially as information technology advances, and it is becoming more and more crucial to provide efficient learning algorithms for neural networks to handle massive amounts of data. Due to their potential for handling huge datasets, feed-forward neural networks with random weights (FNNRWs) have drawn a lot of attention. In this paper, we introduced an efficient feed-forward neural network scheme (FNNS) for processing massive datasets with random weights. The FNNS divides large-scale data into subsets of the same size, and each subset derives the corresponding submodel. According to the activation function, the optimal range of input weights and biases is calculated. The input weight and biases are randomly generated in this range, and the iterative scheme is used to evaluate the output weight. The MNIST dataset was used as the basis for experiments. The experimental results demonstrate that the algorithm has a promising future in processing massive datasets.

The Lattice Boltzmann method (LBM) has been applied for the simulation of lid-driven flows inside cavities with internal two-dimensional circular obstacles of various diameters under Reynolds numbers ranging from 100 to 5000. With the LBM, a simplified square cross-sectional cavity was used and a single relaxation time model was employed to simulate complex fluid flow around the obstacles inside the cavity. In order to made better convergence, well-posed boundary conditions should be defined in the domain, such as no-slip conditions on the side and bottom solid-wall surfaces as well as the surface of obstacles and uniform horizontal velocity at the top of the cavity. This study focused on the flow inside a square cavity with internal obstacles with the objective of observing the effect of the Reynolds number and size of the internal obstacles on the flow characteristics and primary/secondary vortex formation. The current LBM has been successfully used to precisely simulate and visualize the primary and secondary vortices inside the cavity. In order to validate the results of this study, the results were compared with existing data. In the case of a cavity without any obstacles, as the Reynolds number increases, the primary vortices move toward the center of the cavity, and the secondary vortices at the bottom corners increase in size. In the case of the cavity with internal obstacles, as the Reynolds number increases, the secondary vortices close to the internal obstacle become smaller owing to the strong primary vortices. In contrast, depending on the sizes of the obstacles ( R / L = 1/16, 1/6, 1/4, and 2/5), secondary vortices are induced at each corner of the cavity and remain stationary, but the secondary vortices close to the top of the obstacle become larger as the size of the obstacle increases.

This paper presents a multi-state adaptive early warning method for mechanical equipment and proposes an adaptive dynamic update model of the equipment alarm threshold based on a similar proportion and state probability model. Based on the similarity of historical equipment, the initial thresholds of different health states of equipment can be determined. The equipment status is divided into four categories and analyzed, which can better represent its status and provide more detailed and reasonable guidance. The obtained dynamic alarm lines at all levels can regulate the operation range of equipment in the different health states. Compared to the traditional method of a fixed threshold, this method can effectively reduce the number of false alarms and attains a higher prediction accuracy, which demonstrates its effectiveness and superiority. Finally, the method was verified by means of lifetime data of a rolling bearings. The results show that the model improves the timely detection of the abnormal state of the equipment, greatly reduces the false alarm rate, and even overcomes the limitation of independence between the fixed threshold method and equipment state. Moreover, multi-state division can accurately diagnose the current equipment state, which should be considered in maintenance decision-making.

Thermal annealing is performed as a post treatment to tune the color generated by pulsed laser processing of a titanium substrate surface. A comparison of the reflectance spectra before and after thermal annealing shows the peak shift, as well as an increase in overall reflectance, which demonstrates that the color hue changes and the lightness of color increases. Microscope image shows that additional blue and yellow colors on the titanium surfaces are generated through the thermal annealing treatment. Further analyses show that the rate and area of the color shift depend on the annealing temperature and duration. Chemical composition analyses reveal that a TiO2 layer is generated after the laser processing of the titanium surface. Post-thermal annealing causes further oxidation and generates Ti2O3. The tuning process and mechanism behind it are discussed.